_LINC_Evaluation_Program_Meeting_Mar65 LINC Evaluation Program Meeting Mar65
_LINC_Evaluation_Program_Meeting_Mar65 _LINC_Evaluation_Program_Meeting_Mar65
User Manual: _LINC_Evaluation_Program_Meeting_Mar65
Open the PDF directly: View PDF 
.
Page Count: 517
| Download | |
| Open PDF In Browser | View PDF |
A Summary Description
and
Proceedings
of the
Final LINC Evaluation Program Meeting
March 18-19, 1965
Washington University
St. Louis, Missouri
The conference reported in this volume was partially supported by
the National Institutes of Health under Grant FR 00218. The work of the
various participants reporting Has supported by grants from the National
Institutes of Mental Health and th~ Division of Research Facilities and
Resources, National Institutes of Health. The LINCs used in these studies
Here funded by the National Institutes of Health ,vith the cooperation and
support of the National Aeronautics and Space Administration under NIH
contract PH 45-63-540.
Computer Research Laboratory
Washington University
TABLE OF CONTENTS
Conference
Partici~ants
Section 1
Commentary
Section 2
Report from Bowman Gray School of Medicine, Winston-Salem, N.C.,
by G. S .. Malindzak and F. L. Thurstone.
Section 3
Report from Brown University, Providence, R.I.,
byD. S. Blough and L. Marlowe.
Section 4
Report from Communication Research Institute, Miami, Florida,
by J. C.. Lilly.
Section 5
Report from Duke University, Durham, N.C., by C. A. Boneau.
Section 6
Report from Johns Hopkins University School of Medicine, Baltimore, Maryland, by G. F. Poggio and V. B. Mountcastle.
Section 7
Report from Johns Hopkins University School of Medicine, Baltimore, Maryland, by B. Weiss.
Section 8
Report from Northwestern University Medical School, Chicago,
Illinois, by F. S. Grodins and J. E. Randall.
Section 9
Report from Presbyterian Hospital, Philadelphia, Pennsylvania,
by E. O. Attinger and A. Anne.
Section 10 -
Report from University of Washington, Seattle, Washington,
by J. W. Woodbury and A. M. Gordon.
Section 11 -
Report from Washington University School of Medicine, St. Louis,
Missouri, by J. L. O'Leary.
Section 12 -
Report from University of Wisconsin Medical School, Madison,
Wisconsin, by J. E. Hind and C. D. Geisler.
Section 13 -
Report from Stanford University, Palo Alto, California,
by J .. Lederberg and L. Hundley.
CONFERENCE PARTICIPANTS
Representing Laboratories Participating in the LINC Evaluation Program:
ATTINGER, E. O. and A. ANNE, Research Institute, Presbyterian Hospital,
Philadelphia.
BLOUGH, D. So and L. MARLOWE, Department of Psychology, Brown University,
Providence.
BONEAU, Co A., Department of Psychology, Duke University, Durham.
GRODINS, Fo S. and J. E. RANDALL, Department of Physiology, Northwestern
University Medical School, Chicago.
HIND, Jo Eo, C. D. GEISLER and J. KEENAN, Laboratory of Neurophysiology,
University of Wisconsin Medical School, Madison.
HUNDLEY, L. Lo, Department of Genetics, Stanford University, Palo Alto.
LILLY, J. Co and Fo GRISSMAN, Communication Research Institute, Miami.
MALINDZAK, G. S. and Fo L. THURSTONE, Department of Physiology, Bowman
Gray School of Medicine, Winston-Salem.
POGGIO, G. F. and G. WERNER, Department of Physiology, Johns Hopkins
University School of Medicine, Baltimore.
WEISS, B., Department of Pharmacology and Experimental Therapeutics,
Johns Hopkins University School of Medicine, Baltimore.
WOODBURY, J. W. and A. M. GORDON, Department of Physiology, University
of Washington, Seattle.
WURTZ, R. H. and L. SIMPSON, Department of Neurology, Washington University
School of Medicine, St. Louis.
Representing the LINC Evaluation Board:
COX, Jo R., Washington University
DEMPSTER, E. R., University of California
PIPBERGER, H. V., Veterans Administration
ROSENBERG, M. D., Rockefeller Institute
SHIPTON, H. W., State University of Iowa
Representing Other Laboratories Instrumented with LINCs:
BOURIS, William, Stanford University
BROWN, Ro, Massachusetts Eye and Ear Infirmary
BRYAN, J .. So, National Institutes of Health
COULTER, No Ao, Ohio State University
DAVIS ,H., Central Institute f6r the Deaf
ELDREDGE, Do Ho, Central Institute for the Deaf
GERSTEIN, G. L., University of Pennsylvania
HANCE," A. J., Stanford University
HARTMAN, C., Air Force Cambridge Research Laboratories
LANGBEIN, D., Massachusetts Eye and Ear Infirmary
LATIMER, C. N., Lederle Laboratories
MOLNAR, C. E., Air Force Cambridge Research Laboratories
SHEFFIELD, Ho, Lederle Laboratories
SHERRIFF, Wo, National Institutes of Health
STEWART, D. L., Air Force Cambridge Research Laboratories
Representing the National Institutes of Health:
WAXMAN, Bo D., Division of Research Facilities and Resources
WIENCKOWSKI, Lo, National Institute of Mental Health
Representing Manufacturers:
FELLOWS, Go, Spear, Inc.
FOLEY,. J. P., Spear, Inc.
HINDLE, W., Digital Equipment Corporation
KLAUSMEIER, H., So Sterling Co.
POTTER, To, So Sterling Coo
RUDERMAN, Mo Eo, Digital Equipment Corporation
Representing the Washington University Computer Research Laboratory and
Biomedical Computer Laboratory:
CLARK, W. A.
ELLIS, R. Ao
ENGEBRETSON, A. M.
GERTH, V. Wo
GLAESER, Do H.
LEWIS, H. Co
LEWIS, K. Wo
McDONALD, Mo D.
NOLLE, F.
ORNSTEIN, S. M.
PAPIAN, W. N.
PASSERO, S. L.
SANDEL, T. T.
STUCKI, M. J.
TOWLER, C. F.
TOWLER, J. C.
WILKES, M. A.
Representing Washington University
School of Medicine:
BECKER, B., Ophthalmology
BEHRER; M. R., Pediatrics
BROWN; D., Biological Chemistry
BUTCHER, H. Ra, Surgery
LAATSCH, R., Anatomy
SATTERFIELD, J. H., Psychiatry
SLEATOR, W., Physiology
TERNBERG, J. L., Surgery
Other Schools:
BUNCH, M. E., Psychology
DAMMKOEHLER, R. A., Computer Sciences and Applied Mathematics
DEPIREUX, J., Botany
HEISE, J., Botany
KAYUSHIN, Dr., Botany
LVOV, Dr., Botany
TOWNSEND, J., Physics
WHITLOW, G., Computer Sciences and Applied Mathematics
WOOLUM, J., Botany
Section 1
Commentary
The final meeting in the LINC Evaluation Program opened on the 18th of
March in St. Louis, Missouri with clear skies and cold.
Ivas one of almost unprecedented harsh weather..
The preceding day
It speaks well of the motiva-
tion and perseverence of those participating and interested in the program
that only 3 of a possible 65 failed to arrive before the meetings convened.
Before the first day ended all Ivere present.
All sessions were held in the Washington University School of Medicine
Auditorium.
Various aspects of interest in the LINC were represented by
those attending the finale.
Among them were representatives from all the
laboratories participating in the evaluation program, LINC Evaluation'Board
members, LINC manufacturers' representatives, personnel from the granting
institutions within NIH, representatives of laboratories possessing and using
LINCs, local observers from Washington University, and staff members of the
Computer Research and the Biomedical Computer Laboratories at Washington
University.
A roster of those attending is appended to this report ~,
The general format of these sessions was the same as
~hose
at the first
inclusive gathering of participants in Portsmouth, New Hampshire, in June 1964.
During the first day participants in the progr"am made formal present,ations
describing for their colleagues and Board members the progress made in their
research since their last report.
strations with these talks.
A LINC was provided and used for demon-
The sessions were chaired by T. Sandel of the
Washington University Computer Research Laboratory.
Reports were heard from all
participa~ing
laboratories except the Johns
Hopkins University School of Medicinet~ Department of Physiology (Drs. Poggio
and Werner Ivere delayed in arriving by the' weather).
The written reports
submitted by the various laboratories appear in the following section.
1-1
The second day's activities were chaired by W. A. Clark of the Computer
Research Laboratory.
During these:sessions reports were heard from various
LINC users not formally a part of the Evaluation Program, but none-the-less
interested in sharing information and exchanging experiences with other
LINC users.
It is heartening to note that the use of LINC seems to provide
a solid bond between persons in very disparate disciplines;
a common
language with respect to computation bridges whatever interdisciplinary gaps
may exist.
The activities of the two days were not totally devoid of lighter moments.
We gathered from,various
informal remarks that merely being a resident at the
. .
'
Parkway House Motor Hotel
must'hav~
constituted a considerable adventure.
And, of course, true to our traditions as with the assembly sessions and our
meeting in Portsmouth a social gathering ."Has arranged.
In this case, our
host 1vas'Washington University who honored us with dinner following a pay-asyou-go cock-tail hour.
The meal was highlighted by a provocative talk by the
Provost. of Washington University, Dr. George E. Pake, who spoke on the role of
the university developmental and scientific laboratory as a pathfinder for
industry where ideas are so new and unproven as,to make it impractical for
industry to support new technical developments.
The Evaluation Program would
seem to:provide substantial credence to the viewpoint expressed by Provost
Pake.
Dinner 'ended on a light note with a reading by the program's unofficial
poet laureate, J. Walter Woodbury, of an original work (?) commissioned for
the occasion.
In addition to the strictly 'formal aspects of the meeting, there were
the usual cloakroom caucuses and other gatherings.
Perhaps the most note-
worthy of these was the spontaneous gathering of some of the participants
at W. A. Clark's apartment foll01vi ng the banquet.
At that time, a policy was
evolved concerning the Computer Research Laboratory's responsibilities in
future programming efforts with respect to LINC.
It is described later.
The appended reports suffice to describe the first day's activitiesj
the second day, however, did not consist of documented information and some
comment is of interest.
1-2
J. R. Cox and M. D. McDonald of the Biomedical Computer Laboratory
described and demonstrated their GUIDE
u~ility
program.
Copies of the
program on tape and descriptive materials were provided for those who wanted
them.
R. A. Ellis of the Computer Research Laboratory described and dis-
tributed tapes of various test programs he has written for LINC.
Included
,.,rere programs designed to exercise memory and to test instruction code
operation.
Recommendations concerning maintenance and improvement of the performance
of LINC tape units were presented by D. L. Stewart of the Air Force Cambridge
Research Laboratories.
C. E. Molnar of the Air Force Cambridge Research Laboratories led a
general discussion of engineering modifications.
Among the proposals dis-
cussed was a suggestion that the machines be modified to operate programs
out of the upper half of memory.
Other proposals included a modification to
allow' automatic interruption of -programs and a stated need for analog signal
outputs.
These and other proposals have been noted and are under study.
W. A. Clark commented on the CRL programming effort and announced the
intention of the group to provide in the not-too-distant future a final
double-memory assembly -program, now being written by M. A. Wilke's.
He
summarized the discussion of the previous evening and stated that the CRL
group would undertake to provide standard arithmetic subroutines for floating
point and multiple -precision operations.
The participants felt that routines
for addition, multi-plication, division, and the generation of square roots,
sines, cosines, logarithms, and exponential functions would be of greatest
hel-p to them.
They noted that logical and experimental operations in each
ap-plication were well enough understood so that each investigator could
relatively easily write his own programs, and expressed the opinion that the
lack of generaiity of such -programs made it pointless for CRL to assume any
responsibilities in that direction.
All, however, expressed a strong desire
to be informed of the existence, availability, and credibility of all programs written for the LINC.
Suggestions were made that certain machine
modifications might facilitate the -pro-posed arithmetic subroutines.
A final
1-3
comment of interest ·was made.' CRL has made the decision to divest itself of
further technical and engineering
with respect to LINC.
res~onsibilities
Such efforts, should now be continued by interested manufacturers.
In addition 'to the various technical details discussed above, progress
re~ortswere
heard from J. S. Bryan of the National Institutes of Health,
D. H. Eldredge o{'the Central Institute for the Deaf, A. J. Hance of Stanford
University, D. Langbein of the Massachusetts Eye and Ear Infirmary, and
H. W. Shipton of the State University of Iowa.W. Sherriff of the National
Institutes of Health showed a movie made with the use of LINC of a simulation
of various kinds of activity in nerve.
It'was the clear consensus of the
~articipants
that yearly meetings
ought to be continued.
Concurrent with the technical sessions on the second day, the LINC
Evaluation Board met with the NIH
~rogram
as it came to an end.
re~resentatives
to discuss progress in the
Some of the topics they discussed are of
interest'~
The Board recognized that the concept of the laboratory computer as
embodied in LINC and as demonstrated by the work of the participants was a
significant conceptual and technical addition to research in the biomedical
laboratory.
~rogram
They expressed regret that at the time of the initiation of the
that workers in areas such as biochemistry and molecular biology had
not submitted
pro~osals.
They expressed the hope, however, that the pioneer-
ing efforts of the LINC Evaluation Program participants would play a catalytic
role in future uses of computers in research in these other disciplines.
The Board noted with
ap~roval
the salutory character of the do-it-
. yourself training program
em~loyed
in the assembly phases.
They recognized
that within the temporal restraints of the program this was probably. as
efficient a use of training time as was possible.
1-4
The Board concurred that follow-up on use of the machines after three
additional years in the laboratory would be desirable.
They also suggested
that a continuing bibliographic effort to encompass all work done with LINC
be undertaken.
All the foregoing is provided to summarize the substance of our final
meeting.-
In the closing moments of our brief congress, J. W. Woodbury
graciously presented us of CRL with a properly executed and signed resolution
thanking us for our efforts.
In our closing now, we'd like to rejoin -- the
pleasure was all ours!
1-5
RESOLUTION
,
I
Whereas the good people of CRL nee CDO have labored long
fruitfully, cheerfully andUnstintingly and even nights to provide functioning LINC's to sundry physiologists, psychologists
and gadgeteers.
And whereas the aforesaid good people have carried on these
labors with stout heart" a' stiff upper lip and with a smile in
the face of insuperable obstacles, obstinate LINC-lunks and even
technical difficulties.
And whereas they have leaped these obstacles, soothed'the
savage breasts and solved technical difficulties with talent,
tact and technique.
Be it, therefore, solemnly resolved that the undersigned
LINC-lunks wish to hereby express their deep thanks and appreciation to these aforementioned good people of CRL not only for their
perseverance, tact and talent, for their hard work in a cause in
which they believed, for their unfailing good humor and good sense
but, also for supplying the undersigned with one of the most challenging~ interesting, and educational experiences of their lives
by introducing them to, an exciting and powerful new experimental
tool and technique and opening up a new world of ideas, p'ossibili-"
ties and expectations. With a rousing thanks and well done we
attac~ our signatures and appreciation this 18th day of March 1965.
4.
.
U~
_____
(J~ fl~
~
r;;;;)~\t~
~·i"
A~ r ~
I.?S-. (",,' __
, f- =
\_,~
\
I:"X
c
The terms necessary for the calculation of (cJ.) are given as:
J
0<..
1
=
tan
-1
c
J':::..x
(7)
w
(92 - 9 1)
- L
.uX
ol2
=
tan- l
l
..-£..;..
6-x
\.lJ+
c
~x
where,
(0=
A2
Al
~~
(8)
(9 2 - 9 )
1
2 i1 {'; 9 1 and 9 = harmonic phase angles measured at points xl
2
and x2, respectively; and Ax = distance between the two points
of measurement, xl and x2-
Since Wand ilx are known and (92 - 9 1) and (A2/Al) can be measured
exper~entally,it
is possible to calculate actual values for
~
as a function
2-6
- 5 -
of.£.
Again, however, .£ is an unknown quantity.
Fortunately, there are several equations in unknowns which can
either be measured c derived from each other.
If equation (4) and (5)
are combined to obtain an expression for (R/I), this expression has
th~.
form:
c 2 - Ac
c 2 + Ac
where
A
+B
+ B
(9)
=
(9a)
B
=
(9b)
Thus, if one assumes a value for £, equations (7) and (8) can be
used to calculate
(R/I) and (R/I)2.
~
, after which equation (9) can be used to calculate
These can then be used in equation (6) to calculate a
theoretical value for (A 2 /A I ).
The theoretical value can be compared with
the experimental value, and if there is a difference (error function) a new
value for .£
~y
be
~ssumed
and the process repeated.
This iterative process
eventually results in zero error, and the value of .£ at which this occurs
is taken as the "true" phase velocity for that particular harmonic.
The actual error function in this system of equations is not as simple
as one might have hoped, for the error crosses the zero axis at several
different values of.£.
In general, the smaller the value of .£ becomes, the
greater is the frequency of zero crossing.
Fortunately, in all the curves
we have analyzed, all zero crossings, except one, have occurred at .£ values
which are so low as to be out of consideration.
2-7
- 6 Therefore, the pne zero crossing yielding a reasonable value for £ is taken
as the "true" phase velocity.
It may be seen that this scheme of numerical analysis permits the
conversion of data from pressure pulses recorded at two points in the .blood
vessel to actual values for £, (R/I) , and (R/I)2, all of which are important
physical parameters which have not been determined heretofore.
The authors
are well aware that the assumptions made in this analysis reduce the sophistication of the approach.
Future development of the method undoubtedly will
include the perturbations associated with attenuation, non-linearities of
system properties, etc.
c.
Experimental Analysis
Our experiments were performed on dogs, anesthetized with nembutal
(30 mg/Kg.)
For the pressure recordings, special cannulae were constructed
of stainless steel spinal needles (20 gauge) for direct insertion into the
aorta.
These were connected to the ports of Statham P-23G strain gage
transducers.
The entire system was fluid filled and the linear ranges for
these transducers were
± 0-20
mm. Hg. and + 0-200 mm. Hg., respectively.
Phas~ shifts of the transducers were within 20 at all significant frequencies.
The overall frequency response of the system was determined by the Grass
Polygraph recorder used for monitor recorder
the FM Analog tape recorder.
5% at 40 cps.
and signal amplification for
The frequency response curve was flat, and down
It should be noted that our analyses were usually limited to
the fifth harmonic, which never exceeded 20 cps. in most experiments.
The data obtained from these studies consist of a series of simultaneously measured, arterial pressure pulses recorded from the aortic trunk of
'2-8
- 7 anesthetized mongrel dogs under various conditions descriged below.
The
special intra-arterial cannulae were designed for insertion directly though
the aortic wall so that the cannula tip came to lie securely against the
periphery of the internal diameter of the vessel with the orifice at rJght
angles to the flowing stream.
In this manner, the cannulae records true
lateral pressure and does not disturb the normal flow pattern.
Four such
cannulae were placed along the length of the vessel and recordings registered.
Originally, two simultaneous measurements were made; however, results from
. these studies suggest three or more simultaneous measurements would simplify
the method of solution for the model equations.
As many as seven simultaneous
measurement can be made with the cannulae tips spaced about 5 cm. apart for
information regarding the spatial distribution of the pressure pulse and its
reflections.
The pressure pulses were measured under various conditions corresponding to (1) transmission medium and terminal impedance alteration by use of
drugs (pharmacological vasoconstrictors and vasodilators), and (2) mechanical
alteration of the terminal impedance of the aorta (i.e., mechanical occlusion
of aorta at known distances from the catheter tips).
The pressure pulses from these experiments are recorded on FM
Analog Magnetic tape (Mnemotron 700/1400) for subsequent selection, digitalization and analysis by the LINe computer.
- 8 II.~
Past and Present Project Utilization of the LINe
Within the above framework, the LINe has been and is currently being
used for off-line and on-line processing of experimental data.
Prior to the
acquisition of LINe, ana1og-to-digita1 conversion was done by hand and
.
processing was accomplished on an IBM 1620 computer for which FORTRAN programs
.
had been written.
After acquiring LINe the decision was made to utilize the
LINe as extensively as possible without jeopardy to the on-going research
program.
A conventional IBM 026 Keypunch was then modified to accept BeD
information generated by the LINe and to punch this information onto standard
punch cards for use with existing IBM 1620 programs.
A description of the
design and use of this modification is presented in Appendix I.
This deve1op-
ment permitted stmu1taneous use of the IBM 1620 and the programs written for
it, and the LINe as an ana1og-to-digita1 converter until routines could be
written for complete processing.
Then began the huge task of writing the necessary arithmetic routines,
such as floating point, add, subtract,
cosine,
~u1tip1y
and divide, square root, sine,
arctangent, log, etc., required for project data analysis.
Thanks
to the efforts of several LINe users, the time required to finish this task
was limited to less than a year.
It has been only within the last two months
that the LINe has performed satisfactorily from the standpoint of providing
the required arithmetic output.
In terms of off-line processing the LINe is currently ,performing
Fourier Analysis of the arterial pressure pulse data producing an amplitude
and phase for each desired harmonic for each of four channels of information.
This information is transferred to IBM
pun~h
cards-and fed to an IBM 1620 for
further processing in terms of incident and reflected vectors, characteristic
and terminal impedance harmonic phase velocity, and related factors.
Programs
2-10
- 9 -
for
accomplishin~
this were written prior to the
commencem~nt
of the LINC
evaluation program and have not yet been completely programmed for the LINC.
Current on-line processing involves computation of harmonic amplitudes
and phase angles for each of four simultaneously measured pressure
followed by scope display of the data.
fun~.~ions
This need developed from bizarre
results following LINC processing of data after an experiment.
Specif.ically,
the results showed a non-linear (almost random) distribution of mean arterial
pressure and harmonic phase angle with distance along the aorta.
While
·this could be real, it is difficult to justify on the basis of arterial
dynamics.
Such unlikely distributions could be brought on by clot formation
within the cannulae, cannulae showing in situ, or other likely circumstances
difficult to detect with conventional monitor equipment.
When such data are subjected to the numerical analysis outlined in this
proposal, bizarre results are obtained simply because the model has not taken
these factors into account.
With the current LINC on-line processing facility,
we are able to monitor spatial distribution of mean pressures, and harmonic
amplitudes and phase angles for each pressure channel.
In this manner we
are able to detect prior to actual processing whether or not the data collected
during the experimental phase meet the criteria for the numerical analytical
scheme.
If failure is indicated, corrective measures are taken and subse-
quent recorded data are earmarked as "good".' Alternately, one could run
through a complete experiment without knowing until final processing is done
that the data do not meet the criteria
for successful numerical analysis.
A few of the programs written for· this project are presented in
Appendix II.
Included with each program are a routine description, a LAP
- 10 III listing of the program instructions with comments (in some cases), an
example of the output, and a flow chart where possible.
All such programs
listed have been checked to the best of our ability and function to the
extent shown on the output example.
More programs might be listed
they have not been thoroughly checked or completed to satisfaction
but~ither
a~
this
time.
2-12
- 11 -
III.
Future Research
~t
.has been only since our arithmetic routines have been working
properly that we have come to realize the full potential of the LINe in
regard to our research program.
What we have viewed as a sophisticated A-D
converter prior to this time has now become'a useful laboratory tool.
The research program on studies related to the hydrodynamics and the
transmission characteristics of the arterial system is scheduled to continue
for at least five more years.
It is obvious from this evaluation program
that the LINC or a machine with "LINC-type" characteristics is extremely
valuable if not necessary to achieve the desired end product.
)
With the present
investment in programming and trained personnel duplication of this effort
would be impractical if not fool-hardy.
With these thoughts in mind it is planned to carry this particular
research program to its logical conclusion with the aid of the LINC.
This
will involve a more elaborate examination of the behavior of passive hydraulic
transmission elements, such as rubber or teflon tubes as well as active
elements such as aortas and other large arteries.
The passive element study will involve excising a segment of aorta
and insertion of a teflon or rubber tube in situ.
The pressure parts will
be separated by a known distance in a tube whose physical properties are
either known or easily measured.
This maneuver would reduce by at least
five the number of parameters we must now take into account.
Once the
relationship between the passive element and the cardiovascular system is
worked out, this technique will be applied to an active aorta of varying,
length on different animals and on the same animal with varying degrees of
- 12 arterial stiffness from whatever cause.
The major objective will be to describe
in detail the spatial and temporal variations of pressure and flow in the
arterial system of mammals with special emphasis on the determination of complex vascular impedance patterns and in vivo arterial elasticity.
Our future plans include also the acquisition and utilization of a
Datamec tape recorder.
This machine has been ordered by another laboratory
within the medical school for a project involving pattern recognition of'the
phonocardiogram and its correlation with other events of the cardiac cycle.
The LINe is scheduled for use as a data reduction device prior to storage
of information on digital magnetic tape for processing by an IBM 7094 computer.
2-14
- 13 IV.
Training Program
During the past 16 months the LINC installation has, been associated
with several graduate and medical teaching functions which are likely to
recur each year.
These are summarized below:
Course Title
Student Classification
Data Processing and Computer
Techniques
..-"
Contact Hours
Graduate Students
45
Medical Physiology
Medical & Graduate
Students
48
Research Methods
Graduate Students
7
Medical Electronics
Graduate Students
45
In addition three graduate students have learned to program the LINC
and will most probably utilize the machine in processing their research data.
Last summer, three fellows worked on research projects in this laboratory
in which the LINC was utilized.
The LINC has not received widespread acceptance among the graduate
students for three very good reasons: '(I) most of the students are not to
the point in their program where processing experimental data is an important
consideration, (2) those who are aware of this need have been discouraged
by the amount of programming required to supply simple basic routines which
are easily available on any machine using FORTRAN, and (3) there is no
reference programming manual from which they could learn LINC symbolic programming techniques.
We have tried to train three programmers (other than graduate students)
on the LINC thus far.
The first had a nervous breakdown, the second simply
didn't want to learn and the third quit to take a better job.
In all three
cases, it required no' less than two months before these programmers could
write a new program with confidence (graduate students seem to learn much
2-15
- 14 faster).
Most of the graduate students currently using the machine feel the
display feature on the scope, the magnetic tape units and the debugging
facilities (I-STOP,'XOE-STOP, EXAM, I-by-I, and C-by-C) are among the best
features on the LINC.
The teletype has added a great deal to the respect
of many for the machine.
2-16
15
v.
Computer Performance,
A.
Maintenance
Of the nearly 2000 hours LINC has been operating since fabrication,
less than 4 hours has been spent in maintenance and repair.
"
This effort
involved replacing a transistor in one of the drivers, cleaning the air
filter, resoldering two broken wires in the fantail cables and adjustment of
the space bar return spring tension on the keyboard.
A considerable amount
of time was spent trouble shooting problems related to the tape units which
culminated in a visit made by Mr. H. Lewis.
During this visit tape tension
was adjusted, shoes were aligned, and a bent drive shaft was replaced.
Since
that time no malfunctions have occurred that could be directly attributed to
LINC hardware.
B.
Suggested Changes
The analog-to-digital converter has been adequate for most applications
encountered.
However, it is felt that 6-11 bit A-D conversion, switch se-
lectable would enhance the overall usefulness of the machine.
For this
particular project, at least 9 bits of conversion 'seem to be required.
It is our recollection that analog output was to be provided for on
LINC III.
In several instances we have wanted output on an x - y plotter, or
in some other analog form, ,especially in siuuationswhere the analog computer
was involved.
A digital-to-analog feature for our laboratory would be very
desirable.
The analog-to-digital preamplifiers are rated linear between ±Ivolt,
yet our system appears to be li~~fr over the range
± 3/4
volt.
In addition,
there appears to be unexplainable DC level differences and phase shifts introduced through the preamplifiers.
In our system, channel 10 is particularly bad.
2-17
- 16 The plug-in modular design of the data terminal box has facilitated
most of our interfacing problems.
At the same time they have caused considerable
congestion, at the console due to the many and varied cables that must
connection on the front face plate.
mak~
While this may be partially obviated by
"
physical separation the two units current space limitations dictate proximity
of all four units.
I~
would be advantageous if data terminal box connections
with external equipment could be made in the rear of the unit.
A design feature to allow for optimal lengths of interconnecting
cables between the main frame and component parts would be very helpful.
In
our particular setup, 30 foot cables are too short if the machine is to be
remote from the experiment and much too long if the machine is to remain nonp~ytab1e.
Provisions for more than two simultaneous display scopes in addition
to the one at the console would be helpful in many instances.
Also, an inter-
connecting cable greater than 30 feet in length between LINC and remote the
oscilloscope
C.
app~ars
to be particularly desirable in our laborabor setup.
Summary Remarks
With regard to the utilization and capabilities of any laboratory
computer, it can probably be stated with a certainty that the ultimate productivity of such a device, or its
contribu~ion
to research productivity, is
at least as much determined by factors other than machine capability as it is
by the specific machine design.
That is to say, the degree to which a
laboratory instrument computer can enhance the productivity of a research
program is influenced to a greater degree by the available
softw~~e
and
supporting information than it is by any specific hardware considerations.
2-18
- 17 There is little doubt at this time concerning the hardware aspects
of the LINe computer.
The machine has proved to be a 'reliable, fast, and
efficient computing device.
In practically no case has the actual
har~ware
limitation of the LINe computer imposed a serious restriction on its· use.
To
be sure, there have been specific machine malfunctions but these have not been
frequent and have not been more than one should expect with a system of this
degree of complexity.
The familiarity with the nature of machine function which
was obtained in the process of fabricating the instrument has proved to be
at a sufficient level to allow machine maintenance and interface design to
be accomplished by those who would actually use the instrument.
In the area of software and supporting information, however, a rather
severe limitation exists on the degree to which the LINe can be utilized in
research activities.
This limitation is exhibited most clearly in the lack of
available basic computer programs and in the insufficient communication along
those who are using the LINe computer and attempting to increase its effective'ness in their research.
It is perhaps unfortunate that the LINe was placed
into the hands of a number of individuals before a basic set of arithmetic
programs were available and,upon which a more uniform -programming system could
be built.
In addition, the lack of sufficient communication between individual
users has resulted in a great deal of needless duplication of programming
effort and also to very non-uniform programming procedures.
It would seem
advisable to establish at the earliest possible date a centralized programming
staff and office which would be charged with the function of alleviating these
difficulties.
Until a regular and useful programming system is established,
it is probable that many investigators can best apply their programming and
research efforts to machines for which greater supporting information is
available.
This may be true even though the most significant features of on-
line operation and rapid analysis which the LINe possessess may not be available.
2-19
- 18 VI.
Log Book
Our log book consists of 295 pages of information ranging from
registration cards for equipment ordered to all memos sent us from CDO.
This makes it impractical to send a copy for distribution.
Instead, copies
of related correspondence is attached.
2-20
- 19 VII.
Bibliography
A.
Publi'cations:
1. Malindzak, George S., Jr. and Ralph W. Stacy: "Reflections
of Pressure Pulses in the Aorta", Proceedings of the '2nd Annual Sympos;um on
Biomathematics and Computer Science in the Life Sciences, (In Press).
2. Malindzak, George, S., Jr •. and Ralph W. Stacy: "Dynamics of
Pressure Pulse Transmission in the Aorta", to be published by the N. Y.
Academy of Science.
3. Malindzak, George S., Jr.: IITransmission Line Characteristics
of the Mammalian Arterial System", to be publisheq by the 6th International
Conference on Medical Electronics and Biological Engineering.
B.
Talks:
1. "Use of a Computer in Physiological Research", 8th Annual
IEEE Symposium, Greensboro, North Carolina.
2. "Reflections of Pressure Pulses in the Aorta ll , 2nd Annual
Symposium on Biomathematics and Computer Science in the Life Sciences,
Houston, Texas ..
3. "The LINC Computer", Professional Group on Biomedical Electronics
of the IEEE, Winston-Salem, North Carolina.
4. "Engineering in Medicine and Biology", Western Electric Field
Engineering Force, Winston-Salem, North Carolina.
5. "Signal Enhancement of the Fetal Electrocardiogram by Statistical
Methods II , Department of Electr~cal Engineering, Rice University, Houston, Texas.
6. "Computers in Medical Science ll , Western Electric NIKE-X
Engineers, Winston-Salem, North Carolina.
7 • "Dynamics of Pressure Pulse Transmiss ion in the Aorta ", to be
given at the New York Academy of Sciences Conference on Advances in Biomedical
Computer Applications, June, 1965.
8. "Transmission Line Characteristics of the Mammalian Arterial
System", to be given at the 6th International Conference on Medical Electronics
and Biological Engineering, Tokyo, Japan, August, 1965.
2-21
APPENDIX I
LINC-IBM 026 Keypunch Interface
The purpose of this interface is to provide
t~e
LINC computer with
..~.
a punch card output.
The IBM 026 keypunch· was chosen for· this application
because of its inexpensive rental and its relative 'ease of modification.
A punch card output was deemed desirable because data can be permanently
stored and transferred to a larger machine with greater capabilities than
the LINC.
The operation and function of the keypunch unit is not altered
in any way by the addition of these modifications.
With this system it is
possible to punch the cards with any of the characters available in a
standard IBM format.
Under LINC program control, it is possible to format
information in any form onto IBM punch cards.
In order to proved the capability of punching· any IBM character
format, it is necessary to be able to punch as many as three holes in a
single column.
Numbers and plus and minus signs require a single punch,
letters require two punches including a 10, 11, or 12 punch together with
a digit punch, and punctuation requires three punches including an 8 punch
and an 11 or 12 punch.
The GA lines available in the data terminal box
have been designated to provide the necessary punching information as follows:
GA
GA
GA
GA
GA
GA
GA
GA
GA
GA
a}
1
2
BCD Decoded
'~igit
punches 0-9
3
4
5
6
7
8
9
11 punch
12 punch
Space
Extra 8 punch
Extra 0 punch
Release
,2-22
Appendix I
2
Figure 1 shows a component arrangement for our interfacing for
the LINC.
The DEC cards and all the control functions are combined in the
LINC data terminal box.
To avoid noise problems, relay drivers were built
into a separate unit and placed physically away from the LINC and supplied
with their own voltage supply which was located in the keypunch.
Figure 2
is the circuit diagram for the entire interface. It is noted that the GA
lines are all buffered with inverters for purposes of isolation.
labeled "one-shot" consists of two monostable multivibrators.
The block
These were of
our own design; however, any unit is capable of a delay from 5 to 30 milliseconds would be suitable.
A timing diagram illustrating two punching cycles is illustrated in
Figure 5.
The GA lines carry zero levels corresponding to the presence of
bits in the accumulator register.
With an OPR instruction, the OPR line
is enabled which in turn triggers the punch delay if a card is in position
in the keypunch.
line.
This is verified by a -3 level being present on the -3A
The end of the punch pulse triggers the second delay whose output
enables the XL line which allows the LINC to proceed.
The -3A line is
broken whenever a card is between columns or during a release cycle.
The
keypunch will require approximately 25 milliseconds to complete its column
transfer after the XL restart signal has been supplied to the LINC befor{ it
will again provide the -3A level.
The punch delay is approximately 10
milliseconds and the delay before restart is approximately 27 milliseconds,
thus a maximum rate of about 60 milliseconds per character is provided.
The relay drivers shown in Figure 4 obtain. their +15 volts from the
i.;-
supply located in the ki~unch.
The voltage levels on the base of the
·2-23
Appendix I
3
transistors swings between a +2 and a -1 volt to provide positive on/off
action of the transistor.
The relays and the relay contacts are shuttered
with diodes to provide transient protection.
, Figure 6 shows the rear of the IBM 026 keypunch and the location of
the items to be modified.
rele~se
cycle.
In the keypunch, relay 1 and 2 operate during the
For' this reason we have paralleled these relays with relays
lA and 2A (Potter and Brumfield type KA140, 110VAC, DPDT) which open the
-3A line during the release cycle.
Closeups of the wiring on relay 45 and
relays 1 and 2 are shown in Figures 7 and 9.
One of the largest problems in
using the IBM 026 keypunch is in the large amount of electrical noise
generated by this machine.
One of the major sources of noise was found to
be the program cam contact.
Placing a 0.1 mfd capacitor across these
contacts eliminated this noise source.
In addition, it was found to be
necessary in the data terminal box to run a direct ground lead from each
socket to a chassis tiepoint.
Thes,e connections are evident in the illustra-
tion of the data terminal box shown in Figure 10.
In addition, each of the
voltage supplies was by-passed with a 0.025 mfd capacitor to the chassis ground.
A 0.5 mfd capacitor on, the -3A line was also necessary for transient suppression.
To prevent false triggering of other enabling lines within the LINC, the
following enabling levels were tied off to ground:
~EL,
SNEL, UNEL, BEGT,
MONT, CLEL, MINT, and TNEL.
The punch cam modification is shown in Figure 3.
The metal jumpers
connecting 1, 2, and 3R and the lead from 2L are removed, 1 and 3R are then
jumpered and the -3 line connected to 2L and relay 1A line connected to 2R.
These are the contacts which open the -3A line when the keypunch is between
2-24
4
Appendix I
columns on the card.
In operation, the keypunch system is simple to sue and reliable in
operation.
It is only necessary to place the proper character code in the
accumulator, execute an OPR 10 instruction, and then execute a small delay
before the execution of the next OPR instruction.
.
A test program which can
be used to check the operation of the keypunch interface is included in
this appendix.
This test program also illustrates the punch
various characters which can be produced.
~ode
for the
This character code starts at
manuscript line number 21 with the number 0 and proceeds through manuscript
line number 124 and the code for release.
In this program, the necessary
small delay after an OPR instruction is produced by the XSK12 instruction
and its associated loop.
Wishing not to burden this report with further detail concerning
this LINe to keypunch interface, the authors would prompt anyone interested
in duplicating such a system to contact the authors for any further specific
information they might desire.
A test program is attached.
2-25
RELAY DRIVERS
LlNC DATA TERMINAL BOX
11M 026
KEYPUNCH
DEC.102
DEC 4161
DEC 4113
DEC 4102 ~
ONE SHOT
~
~
1------- - -_.
. +15V SWPLY
SYSTEM
BLOCKS
FIGURE 1
r~~;,:- --"l
I
I
I
I
I
I
':
GAl
~~
I
-£!j
1:..
-
T
L..r!1
•
r--;;'VUll;-
I
.I
I
I
I
r'I
~
L- _. _ _
w::-
.....L~JA~~
~ _'
I
!J
rn
I
I
I
I
-.I
I
II
::!
~
Ie....
A
I
l6I
I.
I
I
I
1,-----:
I
I
I
I
AC>:-'~I~
I
1
I
. _II
A
:.
1.&1
:.
~
M.'I.....
~,
rr-'
1:
":'
I
I
'+H-t-.........LO..
I
It!::dJ
-
I
L"
r---.l.,
r
++-\--t~<>=---f-:.~...:-o----. I~I ~
.0JI ~
(
I
r¢~~
I
.1
GMI\
"'·,,0...1 '" ]--!SIL,
-",1 '"
~.
__
j--i.6J
o!I-~
II. '
GAt
I
1~
. rl"'-'-
I
~
,.... -t
I'
'0
I,
TO
L"
I ..
DlC 4102
: ,...-----. ~
fA'! -t:j
IGAU
I
•• ..
~-l~--..
I
I
M
I ...
.. 'I =::
lu~.OC-l ~ , .tr-
1....1
I
.. :~:
_!...~'~
r- ~ ~ ._: : . .T~:~•• ---- IMTUFACl
FIGURE 2
-.
lilA'
?
lilA'
"(
2A
NO
-1- - --- -- -~ - - -- - -
I'
10"
.!!!L.!!l!.
::'1:::,"
s
~
IA
Me
'UNC. CAM No 2
C
~
e
r
1
l
l
llLAY 4'
lIP·,
~g
~
r
liLA"
"COIL
~
r
MODIFIED
BEFORE,
~~~~ TO IELAY IA Me
FROM -3
TO HANG LOOSE
'R
PUNCH CAM MODIFICATION
FIGURE 3
N
I
f\J
ex:>
L
R
r--~....- - - - - - - - - - - - - O FIOM 026
~ELAY 45 H/O-1
+15O-----------------------~~
IH20'I
+~o-------------~--------+-~----------------------------.
~~~-------------------------OTO
026
MAGHETS
101(
DUHCO
•• -IA
12V
560
2HI308
220
,O----"'"fV"~-__4I~-_t__t
0- 9,S,R
11,12
o----"'fV'I.Nv---....-----.,!--I
2HI308
THIS CIRCUIT IS
DUPLICAUD 14
TIMES
lit
lit
RELAY
DRIVERS
FIGIIRF
.t
o
G A LINES
-3
-3
OPR 10
o
-5
PUNCH(P~)~+_r_l_
-15
-_~
"".~~~."H
.•
~",<-..,.O' ..;.~
,.,,~ s~~/:~~~~~~~~l~~_:" .~
I
APPENDIX II
Selected LINC Programs
1.
Binary-to-decimal Conversion-V (with Fraction)
2.
Variable Sample and Display-IV
3.
Plot Routine-l
4.
Plot Routine-3
5.
Plot Routine-5
6.
Fourier Analysis-II , Generation of SINlI\I{t((:rtr/.ll\}and COS :~JCt\ t~Tt 1M:.}) Table
7.
Master III Routine
a.
Executive Control, Calibration, Sample, Display and Keypunch
b.
Fourier Analysis -17
c.
Fourier Analysis - 24 - Compute Cn
'd.
- Compute,
An
&
Bn
& Qn
Fourier 'Analysis - 25 - Display 1, Display 2, & Plot
2-36
MASTER III Routine
Program Description:
(1) Master III Control Routine, Start 20
(2)
Fo~rier
An
and Bn Routine, Start 400
(3) Fourier Cn and gn Routine, Start 400
(4) Display 1, 2, and plot Routine, Start 400
Procedure:
1.
All SNSs off
2.
ReG, 2/420, (location of Master III - binary)
a) Start analog tape
b) Read in calibration data
3.
Start 20
4.
1st, pause, strike "P" on keyboard to obtain calibration factor
5.
2nd, pause; SNS 0 on
6. ·Advance analog tape to pressure pulse, press P, then start 20 to samples
and display. Proceed to (7) only when they appear satisfactory.
7.
SNS 0 off to Scaling Routine
8.
3rd, pause; SNS 2 on; SNS 1 on to skip keypunch routine, press P to enter
FR. Analys~s Routine and compute An, Bn , Cn , 9n , Display 1. Press P.
9.
After display 1, use SNS 3 on to enter display 2.
routine.
ente~
TURN ON TTY for plot
10.
After display 2, turn SNS 3 off to
display 1 and repeat procedure (9).'
11.
After all the display, SNS 4 on to enter TTY, plot routine.
12.
Master III routine is re-read after plotting.
13.
All the SNSs off
14.
Repeat (1).
To skip calibration' factor routine, SNS 0 on.
2-37
Master III Routine Consists of the Following Principle Subroutines
(I)
Calibration Factor Routine
(II)
Sample and Display Routine
(III)
,Scaling Samples Routine
(IV)
IBM-026 Keypunch Decimal Output Routine
(V)
Fourier Analysis - 17
~
Compute
W\
An
1:1
2
M
2
Bn =-
'M
~ Y Cos (N(K(211' IM»)
K=l k
I\v\
1:... Yk Sin (N(K(211 IM»)
Fourier
Coefficients-
K=l
From 4 sets of Yk Pressure Pulse Data Input
From Channel 10, 11, 12, & 13
(VI)
Fourier Analysis - 24
Cn
0
en
an
•
tan- l
+ B2n
Compute
Fourier Moduli
)
An
HIi
Phase Angle
From 4 sets of An, Bn , Result of Routine (V)
(VII)
Fourier Anal sis - 25 - Display Cn & e8; For each channel, or display
6
Cn & 9n ; For each harmonic (9~ in degree, Cn 9n in decimal),
and Plot the four sets of Yk pressure pulse d~ta simultaneously.
Against Time Axis, each time interval being 48 milliseconds.
(VIII)
Recover original control routine
. SNS 0, 1, 2, 3, 4, are used during the computation and output
(from keypunch, oscilloscope, and teletype) in addition to the keyboard control.
With SNS 0 off, enter Calibration Factor Routine.
Calibration factor
is obtained according to the following procedures:
2-38
2.
(1) Obtain data from analog channels, 10, II, 12, 13, which are
connected to a MOdel 700/1400 Mnemotron FM Analog Tape Recorder
on·which has been previously recorded pressure pulse calibration
data.
(2) Average 2008 input samples for each channel and obtain ~ single
point on the calibration curve; do the same for the next calibration
level. Compute the slope between points to yield calibration
factors for each channel. With SNS 0 on; enter Sample and.Display
Routine. The procedure is following
.(a) By means of a Schmit-Trigger circuit connected to XL I
sample 200 points along a pressure pulse wave from each
channel. The Schmit-Trigger is used to fire a delay
which in turn provides a start pulse on XL 1; a stop
pulse is provided at the same point in the following
cycle. The data are then displayed on the Oscilloscope,
which now include 1 complete cardiac cycle.
(b) A potentiometer is provided to adjust the fixed delay so
as to obtain a complete cycle for each set of data, at
the ~esired start and stop points.
With SNS 0 off, enter Scaling Sample Routine which is simply a routine
to multiply each sample by its corresponding calibration factor.
Use SNS I
up to skip Keypunch Routine.
With "p" on the keyboard, enter IBM 026 Keypunch Decimal Output Routine.
This routine includes last octal digit round off and octal to decimal conversion.
The punch cards contain
(1) Total number of samples from each channel (on the first card)
(2) Ten data words with a decimal point before the last digit of each
number and a space between consecutive numbers on each card. After
each set of data are punched out, the program pauses until "p" is
struck on the keyboard. Four sets of data are punched out in this
fashion under the keyboard control.
After the last set of data cards have been punched out, the program
pauses.
With SNS 2 On, and striking "P" on the keyboard, Fourier Analysis -17
is read into QNl, QN2, and QN3.
2-39
Fourier Analysis -17 computes Fourier Coefficients for each of the
four sets of scaled samples.
Yk's are stored in the upper half of memory.
The comp.utations are done as follows:
(1) Fourier Cosine Coefficient
A
•
A
_
o
n
1
M
-
1
M
Cos [ N(K(21r
L
K=l
(2)'Fourier Sine Coefficient
M
Bn • 1M
/M»]
M
L
K=i!l
Where N is the number of harmonics (for experimental purpose, N is
set to equal to 5), M is the total number of samples from each channel,' K is
the increment such that l
~
K
~
M,
2 1r= 6.28 radian.
After the above computation, Fourier Analys is -24 is read:. into
QN1, QN2, and QN3.
2-40
4.
Fourier Analysis -24 computes Fourier Moduli and Phase Angles defined
as following:
. (1) Fourier Moduli
C
n
== ... 1 A2
V
n
+
B2
n
(2) Phase Angle
g
··n
=
tan- l
A
n
:Btl
Four sets of Cn and gn are computed from four sets of An and Bn.
After the above computation, Fourier Analysis -25 is read into
QN1,' QN2, QN3.
Fourier Analysis -25 consists of three main subroutines:
(1) Disp lay 1
(2) Disp lay 2
(3) Plotting 4 simultaneous pressure pulse waves against time .axis (in ms).
The subroutines are described as following:
(1) Display 1
Display 1 consists of three main subroutines
A.
B.
C.
Octal to Decimal Conversion
D. P. Multiply
Display Loop
Display format is as follows (in decjmal)
Where N'
= 0,
1, 2, 3, 4, or 5, CH represents channel, C represents
Fourier Moduli, GO represents phase angles, N represents the number of. .. harmonic.
Keyboard control is following:
A.
EOL to advance to next harmonic data, N' + 1
B.
R
~
N' •
, ],epeat display
(2) Display 2
Display 2 shares the subroutines A, B, C, with Display 1.
Display format is as follows (in decimal)
GO
N
C
0
XXX.XXX
XXX.XXX
1
XXX.XXX
XXX.XXX
2
XXX.XXX
XXX.XXX
3
XXX.XXX
XXX.XXX
4
XXX.XXX
XXX.XXX
5
XXX.XXX
XXX.XXX
CH
Where X
= 10,
=X
11, '12, 13, N represents the number of harmonic, C
represents Fourier Moduli, GO represents phase angles, CH represents channel
,from which data is obtained.
Keyboard control is the same as in Display 1.
2-42
6.
(3) Plot 4 simultaneous pressure pulse waves against time axis (48 ms
intervals).
This routine first print out the time increment in ms (octal) and then
plot four sets of pressure pulse data stored in the upper half of memory,.
Four
curves are plotted simultaneously using the numbers 1, 2, 3 and 4 to distinguish
the functions.
The abscissa is represented as an incremented time base.
Index
register number 3 in # 75 contains P-l, where P is the total number of curves
plotted; location 1744 contains the constant used to adjust base line; the
round off routine is in 7U, this may be varied to suit the data to be plotted.
The value C of the scaled samples varies as, 1400
~
C
~700;
2008 locations
are used to store a single set of data.
SNS 3 is used to select one of the display subroutine (either Display 1
or Display 2) and 4 ON is used to enter the plot routine; SNS 4 OFF provides
for continuous display.
After plotting the curves, Master I I I routine is recovered by reading
back its QN 1 and QN 2 portions.
The program halts and is ready for next sample and display by Start
20.
2-43
~c C,
p Atr1\v4.·,"''\..
'0 ··~f'~ .. ~~---:---------~
.
Bit'
NO
DIS
N.ee
N=
I
!
4--
- - - - - : - - -_ _-1
'\
RCcr
M... 4,.,..L..u...'\. .1il.
~
s~-
, l ( ' l ...
.
/
2-44
:~ acJ ~ I
e
R.
e.OMPU-r/E
A", A I '1
SIS'
An, IBn
n ::: 5· : If
-,
7
J
.sw&~
~\ 3
I ALc.,
1<::.1
1<+/..,1<..
r~ +y
NO·
NO
AD"DKl. S
of
= S 7"11y,- 1!
LA S,
I
~----Pi
il-l--:1 (){)~
Y!
1
2-45
1=" out<. ,e.
K
A '" A
(,-/
~ ,~
- ;1
i-I-
(' (.~ ~ ,...c. \..'t-{.
c,::.
J A..1.'-4- B~tI
t\
.
ND
h' ::. ,..,
o --,
,J'
(1-\ +1 ( (.\" f)")
.
,,'
CM ( It" 8",)
I
.
~t.S
f'i:- -----.---\
-S-~-..:JI.~
.D.tl.-~
P ~-:t
\
r----~
~~
2-46
4-
L-£.t
.0 CL~
"71
PI\.v4-~'--\.....C.
S ~~.t"a..-\ '-'Z..-tr'f.t.-~
.sJ.",~
4---------------·---·--···----·~""-----~t
I
t-f-f =- , 3
.
"2
NO
~'+I
->c.,.
8.:.." .....
tit'.'
2-47
Program Name: Binary to Decimal Conversion-V (with Fraction),
Author:
GEORGE S. MALlNO%AK. JR•• PH.D.
DEPARTMENT OF PHYSIOLOGY
!:lOWMAN GRAY SCHOOL OF MFnl,-1 .... J"!'
VINET"''''' C;". ....
-_.-.... .
Address:
Phone:
Subroutine:
AC-919; 725-7251 x 7177
x
Yes
----
----- No
If Manuscript;
No. Manuscript Lines: 424,
•
Origin: 400
Entry Point:
1A
Tags: 1 (A, B, C ,D ,E, F ,G, S , T , U ,X)
,6(A,B,C,D,E,F,G,I,J,K,M,S,Y,Z)
Equalities:
If Octal:
Memory Locations:
Entry Point:
~
Registers Used:
1,2,3,4,5,14,15,16,17
Analog Inputs Used (including knobs):
Operate Lines Used:
Tape Blocks and Units Used:
LAP III Mss-5 B1ks; Binary-1Qtr.
Description and Operating Procedures: Binary to Decimal Conversion~V: This routine
will convert a 4 digit octal (12 bit binary) number to its decimal equivalent and
display the 'value in the upper right quadrant on the oscilloscope. The program assumes
the number to be converted is in the accumulator at the time of entry. The x and y
coordinate is set at 415 and 465 respectively. The fraction conversion ,feature of
the program starting at 1X assumes the number to be converted is a positive fraction.
This routine was written for conversion of double precision numbers where the first
datum word represented the integer value and the second datum work represented the
fraction. In order to demonstrate these features the following utility program is
suggested for use with this routine.
20 tsW
31 000
21 JMP 400
32 LDA i
22 SET il
33 300
23 465
34 STC 415
24 LDA i
35 RSW
25 340
36 JMP 530
2,6 DSC i
37 JMP 20
27 0100
30 DSC i
Numbers enter through the LSW and RSW will be converted to their
integer and fraction decimal equivalent and displayed OR tae oscilloscope. Without the
utility routine only integers are converted and displayed.
2-48
Program Name:
Master III Routine (Fourier Analysis 17 )
Compute An and Bn
. A'uthol':
Gt!O"itGlt
Address:
s.
MAt.)NDZAK. :JR.. PH.D.
DEPARTMENT OF PHYSIOLOGY
bOWMAN GRAY SCHOOL OF MEOICINIf
WINSTON.SALEM, NORTH CAROLINA
Phone:
Subroutine:
_ _X_ _ Yes
_____ No
If Manuscript:'
,No. Manuscript Lines:
13748
Origin: ,8 400
Entry Point:
5Z- ll S
Tags: ,
Equalities:
3(A,B,C,D,E,F,G,H,I,J,K)
3(L,M,N,O,U,X)
4(A,B)
5(A,B,C,D,G,J,N,M,O,P,Q)
5(U,V,W,X,Y,Z)
6(A,B,C,D,E,F,G,I,J,K,L)
6 (:H, S, Y, Z)
If Octal:
7(A,C,M,V)
Memory'Loca.tions:
Entry Point:
.(3 Registers Used:
Anal.og Inputs .Used (including knobs) : .
Operate Lines Used:
Tape Blocks and Units Used:
Description and Operating: Procedures:
See attached description
2-49
Program Name: Master III Routine (Control Routine)
Calibration, Scaling, Keypunch
Author:
GEORGE S. -MAL.INDZAK.JR., PH.D.
Address:
DEPARTMENT OF PHYSIOL.OGY
BOWMAN GRAY SCHOOL. OF MEDICINE
WI NSTON.SALEM, NORTH CAROL.INA
Phone:
Subroutine:
---- Yes
x
_____
No
If Manuscript:
No. Manuscript Lines: 1367 8
Origin:
4F
1A
Entry Point: # LA
1 (A;P.,R, S ,X)
2(A,B,C,K)
3(A,B,C,E,F,K,T,U)
4(A,B,D,E,G,K,L,M,N,R,Z)
Tags.:
Eq'ualities:
5(A,B,C,D,E,F,G,J,K,L,~
5(N,P,R,S,T,U,Y)
6(A,B,C,D,E,F,G,I,J,K)
6(M,S, Y ,Z)
Channel 10,11,12,13
If Oc.tal:
Me.mory Loc.a. t ions:
Entry Po i:1.t :
(3 Regis·ters Used:
All
Analog Inp~::s .Used (including knobs) :
Operate Li:-_es Used:
Tape Blocks and Units Used:
Description and Operating Procedures:
See attached
des~ription
2-50
Program Name:
IBM 026 Keypunch Test Program-II
Author:
Address:
Phone: AC-9l9; 725-7251 x 7177
Subroutine:
---- Yes
x
No
-----
If Manuscript:
No. Manuscript Lines:
Origin:
20
Entry Point':
Tag~:
124
If 1A
l(A, B, T)
Equalities:
If Octal:
Memorx
~oca.tions:
Entry Point:
(3
Registers Used:
1, 2, 3
Analog Inputs Used (including knobs):
Operate Lines Used:
Tape Blocks and Units Used:
Description and Operating Procedures: Load binary version of program into QNO;
ready keypunch by registering card from hopper, and start 20. Output is of
the variety attached, and is recurrent until either the keypunch or the computer
is halted.
2-51
Program Name: Fourier Ana1ysis-11, Generation of Sin
I
Author:
GEORGE
s.
MALJND:tAK.
N(K(2nt~»
and Cos n(K(21'C1'1!f»
tab1E-
,.
·JR •• -l'H.D.
DEPARTMENT OF PHYSIOLOGY
BOWMAN GRAY 5(""1-11"\0' 0'" ~.u·nlr"'Nt='
Addres s :-"",!'=:.,.,.."', ~ ......
Phone: AC-9l9; 725-7251; x 7177
Subroutine:
---- Yes
If Manuscript:
No. Manuscript Lines:
Origin:
-------X
No
7728
20
Entry Point:
Tags:
3(G,H,I,J,K,L,M,N,O)
'5(A,B,C,D,G,J,M,N,O,P,Q,U,V,W,X,Y,Z)
6(A,B,C,D,E,K)
7(A,C,M, V)
Equalities:
If Octal:
Memory LQ,ca.tions:
Entry Point:
~
Registers Used: 1,2,3,4,5,7,10,11,12,13,14,15
Analog Inputs Used (including knobs) :
Operate Lines Used:
Tape Blocks and Units Used: LAP III
~s-10
Bloks; Binary-2 qtrs
Description a~d Operating Procedures: This series of programs consist of three basic
routines described blow:
A. Floating point (N(K(2 IL/M), argument for the trigonometric functions.
N represents the number of harmonics
K, 1 ~ K SM, is the number of increments where M is the total numbers of samp1ef?
in one period.
·
B. Sin (N(K(2 ~/M) is computed if SNS 4 is off. Answer enters sine table at
2001 + (2M-l)
C. COS (N (K(2 7t: IM:») is computed if SNS 4 is off. Answer enters cos ine table
at 2401 to 2401 + elM-l)
Each output value is in octal and is in double precision form occupying two
consecutive locations with the octal point assumed between the two locations.
To Use: (1) After the binary program is read into quarter
and :" put N into location,
11 and M into location 12.
(2) To generate a cos (N(K 2~/M) table at 2401-240l+(2M-l), set SNS 4 on and
Start 20.
(3) To generate a sin (N(K(2 TC/M:») table at 2001-2001+(2M-1), set SNS 4 off
and Start 20.
°
°
,. 2-52
Program Name:
Plot Routine-5
Author:
GEORGE S. MALINDZAK. JR .. PH.D.
CEPARTMENT OF PHYSIOLOGY
BOWMAN GRAY SCHOOL OF MEDICINE
AddressWINSTON.SALEM. NORTH CAROLINA
Phone:
AC 919;,725-7251; x 7177.
___
X"","-_
Subroutine:
If
Yes
----- No
Manuscrip~:
No. Manuscript Lines:
Origin:
20
Entry Point:LA-14
Tags:
274 8
8
1(A,B,C,D,G,J,I,Q,R,S,T,X)
'2(A,B,C,F,G,K,L,M,N)
Equalities:
If Octal:
Memory Locations:
Entry Point:
~
Registers Used:
1,2,3,4,7,10,12,13,14,14,16,17
An.alog Inputs Used (including 'knobs) : '
Operate Lines Used:
Tape Blocks
a~d
Units Used:
LAP III Mss-4 B1ks.; Binary-1 B1k
Description a~d ~)erating Procedures: This routine plots four curves simultaneously
using the numbers 1, 2, 3 and 4 to distinguish the function. The abscissa is represented
a~ an incremental time base.
Index register number 3 in 2R contains N-1, where N is
the total number of curves plotted; location 273 contains the constant used to adjust
base line; the round off routine is in 2G, this may be varied to suit t~e data to be
plotted. The value of C in this program varies as, 1400 S C S700; 200 S locations are
used to store a single set of data. Initial address of the first set of data is
assumed to be 3000. Terminal address is set to 3177
To Use:
See Plot Routine-1
2-53
Program Name:
Plot Routine-3
Author:
C;£CRGrt ~. Mil' "'--'1(, JR .. PH.D.
OEPARTMENT
. "'RIOLOGY
'(")"\'~~l\N OR"
sr.
'"'II" MEDICINE
Address~s, ·"l.r'
"4
"'r)R,H C':AROLINA
AC 919; 725-7251; x 7177
Phone:
_ _...X__ Yes
Subroutine:
----- No
If Manuscript':
No. Manuscript Lines:
Origin:
245
8
20
Entry Point:
Tags:
l(A,B,C,D,G,H, I,J, S, T ,X)
, 2(A,B,C,F,G,K,L,M,N)
Equalities:
See Plot Routine-I for additional comments
If Octal:
Memory Locations:
Entry Point:
~
Registers Used: 1,2,3,4,7,10,12,13,14,15,16,17
Analog Inputs Used (including knobs) :
None
Operate Lines Used:
Tape Blocks and Units Used:LAP III Mss-4 B1ks; 1 Binary-B1k
Description and Operating Procedures: This routine plots a single function on the TTY
according to the contents, C. With SNS 5 on, the routine plots a point representing
the function and its octal value along the ordinate, with SNS 5 off the routine plots
a point representing the function only. Memory location is plotted as the abscissa.
14008
~ C ~ 700 for this particular program bu~ this may be easily modified to
plot any function. Loaction 214 contains the constant to adjust the base line. The
round off of the last digit routine in 2G may be varied to make any data fit for p10tting~;
To Use:
Same instructions as Plot Routine-1
2-54
Program Name: Plot Routine-1
GJICRG~
Author:
Address:
Phone:
Subroutine:
s.
MA't.lNCZ'AK. ::JR•• PH:):).
DEPARTMENT OF PHYSIOLOGY
qOW~~AN GRAY SCHOOL OF MF.OICI~F
~". co.iI
"1"'I'"'Tq
C'1\~nl 1~1f
'"'''<:''1'''\'''
AC 919; 725-7251; x 7177
x Yes
----
_____ No
If Manuscript:
No. Manuscript Lines:
Origin:
8
20
Entry Point:
Tags:
232
iF 1A-148
1 (A, B , C , D, G,J , S , T , X),
2(A,B,C,F,G,K,L,M,N)
Equalities:
None
If Octal:
Memory Locations:
Entry Point:
~
Registers Used: 1,2,3,4,7,10,12,13,14,15,16,17
Analog Inputs Used (including knobs) :
Operate Lines Used:
None
None
Tape Blocks and Units Used: LAP III Mss-4 B1ks; Binary - 1 B1k
Description and ~)erating Procedures: This routine uses the TTY to print out a function
stored in memory with its location along the abscissa and its contents as a series of
dots (.) and octal value along the ordinate. The maximum number of d&t5 printed out
is equal to C, where 65 8 ~ C:)oi 0
To Use:
(1) Read function data into any quarter but QNO
(2) Read binary program into QNO
(3) Turn on teletype and start 20
(4) Type on TTY (P) @ , where P is the starting address of the function
data, and (Q) E, where Q is the terminal address of the function data
(5) Set paper hit "return", and away you go
Conventional delete feature from keyboard through the RUB OUT key is included.
2-55
Progra~
Name:
Variable Sample and Display-IV
Author:
Address
GEORGE S. MAt.~ND%AK. :JR•• PH.D.
DEPARTMENT OF PHYSIOLOGY
BOWtllAN GRAY SCHOOL OF MEOICINE
:N1NSTON.SALEM, NORTH CA~OLINA
Phone:
Subroutine:
AC 919; 725-7251; x 7177
---- Yes
If Manuscript:
------X
No
No. Manuscript Lines:
Origin: 20
Entry Point:
4ft lZ
Tags: l(A,B,C,D,E,F,G,H,I,J,S,T,U,Z)
2(B,C,E,F,G,S)
3(S)
4(D)
Equalities:
5 (D,N)
6 (D,N)
7 (A,D,M,N,S)
If Octal:
Memory Locations:
Entry Point:
~
Registers Used:
(1,2,e,11,12,13,14,15,16,17)
Analog Inputs Used (including knobs):
Operate Lines Used:
Tape Blocks and units Used:
Description and Operating Procedures: This program is designed to sample and display
analog information presented to LINC over ANCHIa at a sample rate variable from 15.6
to 1000/sec. There are several options within the program that permit one a good deal
of flexibility in data reduction and preprocessing of experimental information. The
delay between samples is determined prior to progr~m execution and set ~n the sense
switches (0-5). With all sense switch off the basic sampling rate is I KC; with SNS a
on, the rate is SaO/sec or 1 every 2 ms; with SNS 0, and SNS lon, the delay between
samples is 6 mSjetc., with SNS O~ SNS 1, SNS 2, SNS 3, SNS 4, SNS 5 on, the delay
between samples is 64 ms, and the sampling rate is about l5.6/sec.
Once the sampling rate has been determining start 20 will initiate
the program; once sampling is complete, memory is displayed 1000 locations at a time.
The location in memory from which the display starts may be set gy pot # 6 (course)
and pot # 7 (fine). The upper half of memory may be displayed. The x-axis may be
expanded by adjusting pot # 4. Sixteen possible expansion ranges are available. Pot
# 2 adjusts the length of the cursor which may be moved amng the data through manipulation of pot # a (course) and pot # 1 (fine). Once the cursor is in the display one
has the option of an octal or decimal display if the point beneath the cursor (upper
right quadrant of the scope). SNS 1 ort, the number display-ed will be dectmal; SNS 1
off the. number displayed wili be octal. In this octal display, the number may be
2-56
Variable Sample and Display-IV
Page 2
Description and Operating Procedures continued:
either signed
display there
displays) the
the cursor is
or unsigned depending on the positions of SNS 0; in the decimal
is no sign option; all numbers are signed. During all numerical
location from which a point is obtained for the display'·beneath
shown in the lower right quadrant of the scope.
2-57
Report on Use of LINC Computer Through February, 1965, and Proposal for Its
Continued Use
Donald S. Blough and Lloyd
~furlowe
Department of Psychology
Brown University
3-1
Report- on Use of LINC Computer Through February, 1965, and Proposal tor Its
Continued Use
Donald So Blough and lloyd ¥arlowe
Department of Psychology
Bro~m
I.
University
Past and present research
A. Nature of the researcho The LING has been used primarily for on-line
control and diia-cojj;ct1on ~r-operant conditioning experiments. Six pigeon
boxes are connected to the machine (see belotv). To date some half-dozen experiments have been run. These experinlSnts 'VIers concerned with stimulus control
(discrimination and generalization) processes, and with reinforcement scheduling
procedures. In most of the work, distributions of inter-response times were ot
central concern. Three of the most extensive experiments wil1 be described
briefly.
1) Inter-response times during ,.generalization testing. It is known
that rate of response declines when stimuli present in the experimenta~ situation depart from those values that are associated'Vrith reinforceln.ent. We have
been interested in the detailed nature of this change in rate. Our point of
view has been that responses are, to a degree, "chained" together, such that
the probability of a response being emitted at a particular moment is a joint
f'unction of current stimuli and of preceding behavior. To dea.l with this question, pigeons were trained to peck at keys illuminated from behind by lights
of fixed wavelength and intensity. Following a number of sessions of such
training, f'ood reinforcement ceased and the light on the key varied randomlY
in either wavelength or intensity. Responses emitted during these changes were
categorized by both stimulus and inter-response time. The LINe computed
conditional probabilities that a response would occur, given that a particular
stimulus was present and that a particular interval had elapsed since the prenOlls response. It was found that for IRTs under about '3/4 sec response
probability varied little 'tdth stimulus variation. Such responses, constituting
well over half of the total emitted, l-rere evidently chained to preVious
behavior and not under stimulus controlo
2) Inter-res12onse .:time~ durin~ .maintained .C&scrinlination testins."
The generalization procedure is unsatisfactor.y for quantitative analysis of
response probabilities, because the subjects are tested without reinforcement.
Relatively few responses occur before extinction, and those that do occur
take place during the transitory extinction process, lvhen lllTs are changing as
a function of the time since the last reinforcement. For this and other reasons,
a stea~state discrl.mination experiment 'Was run. Three pigeons were rUn
simultaneouslY, each pecking a key illuminated by one of 13 wavelengths spaced
at 1 mu intervals from 538 ll1U to 550 llIU. All stimuli tvere presented for an
equal number of brief trials during a daily session, and pecking went unreinforced to all. Interspersed among these unreinforced presentations were
occasional extra presentations of 550 mu; on these trials, pecking was reinforced
on a variable interval sChedule. A,relativelY stable discrimination function
3-2
Blough
developed in this situation, with rapid responding at 550 mu and gradually
less responding at wavelengths progressivelY farther away. The number of
responses given to each stimulus showed a highly regular relation to stinnllus
wavelength. These data were analysed as a joint function of inter-response
time and stimulus. A three-dimensional plot of such a function appears in
Fig. 1.
However, a plot such as that in Fig. 1 does not clearlY separate the action of
stimulus and IRT Variables. For this, the probability of response conditional
on the occurrence of any particular IRT must be computed. This; the so-called
"IRTs per opportunity" transformation, reveals the same finding that was
suggested by the generalization studies. As shown in Fig. 2 the stimulus-has
relative~ little influence on responses terminating short IRTs, ~~ile a regular
JRl S 11 z·
. . . .. . .....
31"-1
81M5
r
II r rr rr
1-51"
'SEC J
r
rrrrrrJ
r rrr r
.... _" .. r r ,
.....
r
I
sso
.. aUILI •• '" - _
decLining function "governs more widelY spaced responses.
3)
The reinforcement
s: least-frequent
interres;eonse times.
As the
Blough
above discrimination experiment continued, response rates increased but stimtllus .
control became progressive~ mora. variable. We f'elt that this might be due to
the phenomenon we had been investigating. That is, progressively stronger
interresponse "chaining" might gradually override the control exerted by exteroceptive stimuli. For this and other reasons, t-m devised a schedule of'
..'
reinforcement specifically designed to minimize -chaining and to favor the
emission of responses independent of one another. The appended report
(Appendix A), presentlY under review by PSIchonomic SCience, gives a brief
account of this schedUle and the early results obtained with it. Machine silmllation of "statistical pigeons" vms a helpful adjunct to this liork.
The above 'chree examples suggest the core of Blough's research with the
LINC. In addition, three graduate students have made extensive use of the
machine. Mro Jeffrey Sheff has conducted an investigation of wav$length
difference limens, using· the machine to control stimuli and to collect, analyze
and display data. Mro Charles Shimp has analyzed extensive data for his
. doctoral dissertation on discrete-trial discriminations. The experiments were .
run of.f-line, and the data fed to the computer through a paper-tape reader.
The LINC provided sequential statistics that could ba compared with existing
mathematical formulations of discrimination learning. Mr. Shimp also programmed
the machine to simulate pigeon performances, given certain assumptions about
the variables controlling choice behavior, with impressive results.
Mr. Lloyd Marlowe has done extensive programming on the machine (see below and
Appendix B) 0 He is also running an experiment on line, aimed eventually at
studying time discrimination.
.
Be
Inputs, outputs
~
au,g.lliary e9'llipment.
Addi.t.ions .t.g ~ ~ terminal~. A 32 bit external clock,
driven by machine time pulses, vlaS installed last fall. This is used for the
many timing jobs involved in on-line operant work. Three different 12 bit
time ranges may 00 selected via OPR 1, 2, or 30 ll-:alva auxilliary relays were
also installed, driven, by a 12 bit register installed in the DTB. These are
used to control the multiple stimuli in the pigeon boxes. Output to these is
parallel, 'tdth the left or right 6 bits independent~ selected via OPR "5 or 6.
Standard innutsc At present, 6 experimental chambers are connected
to the LIl\'fC. The 12 XL lines are used to sense responses and, in certain cases,
other experimental events o The analog lines have been used in a servo-type
control loop, by which the machine adjusts stitnulus wavelength and intensity
to pre-programmed values •. White noise on one analog line has been used as a
source of small random numbers. The OPR input lines, in addition to serving the
external clock, have been used to enter .stitm.1l:us codes.
Standard outp,!t.~. The relay register and the l2 newauxi1liary
relays are in constant use controlling stimulus and reinforcement events. They
switch both 28 volt 00 and no volt AC circuits. In the servo-type stimuJ.us
system mentioned above, the relays run wavelength and intensity drive motors.
The .Teletype, installed last fall, has been used very heavily to t:vpe out
programs, data tables, and data histograms. The scope serves for on-line data
3-4
4.
Blough
monitoring, program displays, and for graphing data, in addition to heavy use
with the IAP compiler. Experimental data are stored. permanently on magnetic
tape.
c.
Programs.
",
..,'
Perhaps fifty programs have been written, for a variety of purposes,
including: on-line experimental control and' data collection; simple data
analysis; display via scope or teletype of instructions for running programs,
data tables, graphs, and so on; simulation of animal behavior under a variety
of conditions; tests of various. input and output functions. Soon after we .
started using the ·machine it became evident that convenient systems for running
these programs 'Would be needed. First a "metaIt system was written, which permitted 'programs to be called systematically, and parameters to be typed into
program displays. Recently, Mr. Marlowe wrote the "Monitor" system. (described
in appendb: B) which enables us to· assemble and run at will long sequences of
individual programs •
.In addition to the n:eta-raomtor scheme ~le have a few programs that
other LINe users might find helpful. These'include several display programs,
including tabular and graphical display, and data packing and rearranging
programs. These programs are written to be run by l-1onitor. Copies of these
progralllS, as well as the 1f~nitor, will be available on tape at the ¥.arch, 1965
"LINe conference or upon request. It would be difficult to describe our on-line
experimental programs 1n detail, largely because they take care of so many
experimental events and contingencies that ~Ake sense only on exhaustive descr1~
tion of the experiu~ntal rationale and procedures. Appendix A includes a
copY' of the program used to run the experiment desct?-bed therein, as well as
a sample of the daily data printout obtained from the experiment.
II.
Future research
Our research now depends entirely on the LINO. It not only enables
us to do more easily what we formerly did with relay and solid-state programming
equ:tpment, but it has opened up entirely new possibilities. Most of the
research we are now doing and plan to do ..TOu1d be impossible without the machine.
A.
Nature
2!. future
work.
This will gro~v out of our past research, some of which is briefl.v
described above. We plan to continue to develop and study the random-response
schedules made possible by having. UNO on line. We intend to apply such
schedules to our stinru.lus control experiments, in the hope that they will
drastically reduce individual differences and make possible a rational qnanti. tative treatment of the rate measure. Quantitative functions will be sought
relating rate to reinforcement probability and several stimulus parameters.
We hope to continue discrete trial work and to begin work with response latency,
relating both of these behavioral measures to the rate measure. The disp~
capacities of the LINO encourage us to add some human perceptual studies to our
program, i f time permits.
s.
Blough
Though the machine is now uSed in our own program for from 8 to 16 hours
a day, there are tentative plans to use the machine for the analYsis of taped
electrophysiological data from Dr. Carl P:f'affuann' s laboratory here at Brown.
Since Pfaf'fmann' s recent acceptance of.· an appointment at the Rockefeller Inst ••
this use is uncertain. One or more of his students may do the work, or
Dr. Corbit, our new physiological man, may be interested in this application.
B.
Inputs, outputs.!!E auxilliaq eguipment.
Orders were recently placed for relays and logic elements to expand
o'Ur input and output facilities. The auxilliary. relay bank will be increased
to 48 relays,: .enough to control 8 independent events in each of our 6 experimental boxes. We have no plans to expand the number of simul.taneous4r run
animals beyond 6, as this already strains the machine's memory capacity. At
the same time, we will shii"t from XL to parallel reading of response events,
for reasons of speed, flexibility, and reliability. External registers will
hold "responses" until these are read in, at which time the response registers
will 1?e reset.
There is some chance, perhaps slight at present, that our ~C may
communicate via a teletype line with the regional computer complex at MIT.
The Brcnm Computation Laboratory at present rents such a line at low cost.
At the moment we have no pressing problems that call. for this tie-in, and, of·
course, the rate of data transmission would be low.
C.
Programs.
Marlowe hopes to begin work this summar on a special purpose compiler
for use in operant research. The hope woul.dbe to enable the e.xperiInenter to
write out expe~ntal contingencies in an approximation to English, LD.TC would
then compile a machine-language program that 'tvould run the experiment. Such a
compiler would speed up our programming a great deal, and would enable students
to run experilnents with minimal knowledge of the machine. Tentative plans have
been made for Marlowe to work out the compiler using SNOBOL and the existing
teletype ~ to the large computer facilities at M.I.T.
.
III.
Training program
I think the evaluation program has generally worked well as a training
device. Though I started witb no special lmowledge of any computer, I have
been able to make effective use of the LING, and to keep it going with relatively
little outside help. However, for. those of us with no programming background,
more speCific instruction 'Would have been helpful. It is my own belief that
making available a graded series of pre-written, annotated programs for newcomers
to study would have been the easiest, most helpfuJ. device. Detailed handbooks
on the nature of the machine and. its programming wouJ.d be especially valuable
to the many individuals now involved with the machines who were not in the
original group in the sunnner of 1963. The arduous and no doubt uninteresting
work of preparing such material might in the end be of more service than
equa~ effort on technical matters.
.
; ;-6
6.
Blough
IV.
C~uter
A.
per:f'or:mance
Maintenance.
We' have bad relatively little trouble 'With the machine. As of ....
1, 1965 it had run :f'or 2700 hours. 'rlv-o failures in the l.ogic ware
detected in the first 6 months, both apparently due to overloading caused by
faul.ty wiring (one, a wire missing from the frame, the oth~r an open clamped
load on one module). A transistor in one of the relay cases al.so failed.
Sticld.ng keys on the Soroban keyboard were fixed by 'caref'ul oiling.' The major
problem from the b6ginning was electrical noise; grounding, shielding, spark
suppression, and termination of all open lines to the ~B seem to have eliminated
the cH.fficul.ty. The machine is now grounded at one point only, through a heavy
strap attached to. the f a n t a i l . '
.
~!arch
Various bothersome tape dif:f'iculties have arisen from time to time.
Last summer, tapes tended to bind on the shoes. This may have been due to the
particular atmospheric conditions in our air-conditioned and humiditycontrolled· quarters. Mark clock and ACIP adjustments fixed some marking and
writing problems. A recent internd.ttent failure to R.:CC, due to a disagreement
between the data sum and the check sum, has not yet been traced. The tape unit
fuse has also blOlm tw:i.ce in recent weeks. This may be due to a recurrence of
prev.1.ous trouble caused by a short-circu:i.t between NO and NC contacts on a
tape control relay.
Burning on the scope screen ~ms a serious problem with the original.
Display channel separation liJ"aS also f'ailing on the old tube before it was
recently replaced ~rlth the ne.-: aluminized CRT. "lithin a vleek the soreen
J.ocation occupied by the L~ line number was noticeably burned on the ne'Vl
CRr. The problem arises from the nature of the LAP displays. If' the brightness
for a Single line is turned dO'tm to avoid burning, the 20 line display is not
bright enough to read. Using the button to brighten the soreen in the latter
case is quite awkvrard when one is shuffling manuscript pages. A possible
solution is the adjustment of display J.oops to oompensate for the size of the
display. The analog ladders controlling the displ.ay required adjustment after
about l.000 hours.
CRr.
B.
~ ~
E. oEerant
research.
In· many respects, we have found that the LIl\C is a very good l.aboratory computer. Rather than listing in detail its many good points and its few
bad points, -Wle shall discuss the ,LII\1C's' capabilities in relation to a coupleo:f'
problems peculiar to behavioral research.
1) Sn3eial ~ factors j.n. opera'l1t eondit.ioning exoerimants. In our
free operant experiments, the subject' (a pigeon) oan make a response at any'time
with a lllinir.nllll 1nterresponse ti..'Ue of about 50 milliseconds. Al.so there are
stimulus changes oontingent on these responses whioh shoUld appear to be an
immediate consequence of a response. In most situations, a delay of about 60
mil l; seoonds from response onset to stinml.us change is the ma:d mum delay that
3-7
Blough
should be allowad o If delays become very much greater than this, .the , delay'
before the stimulus change is datectabl~ •
.2) Adequacy.2! ~ Et:og£alU storaf£e capacity. As a result of these
special time factors it is frequently necessary to carry .out all ot the
computation associated with a response in less than 50 milliseconds. ConsequentlY,
the program storage capaoity·has usual~ been adequate for running our experiments. A good example of this is the program inoluded in Appendix A. AJ.though
this program takes almost 50 milliseconds to perform the necess~ calculations,
i f all :3 subjects respond simultaneously, it requires only :3 blocks of.: program
storage.'·Nevertheless, we can conceive of situations in which we woUJ.d be
pressed for program storage.
3) In§\bil1~.i2 .,Yse tapes during experiments. ·A second result ot
the special time f'actors mentioned above is the ;na.bility to use the tape units
during an experiment. Any tape operation will result in the possible delay ot.
a stimulus change lmich can not be tolerated in many exper:i..ments. Ofcourse
there are some experiments in wh1.ch there are times when those time factors are
not· in eff'ect. However, in most of our experiments, it is virtually impossible
to have these times coincide for each of 3 sUbjects being run simultaneouslY
. since the subjects must necessarily be run independently of one another. The
problem becomes worse if we consider ~ng mora than one experiment
si.multaneously•
The major problem produced by this restriction is the inability
t~ store all of the ra't·' data (over 10,000 responses' may be collected from
:3
subjects .in one session). Instead, the data must be at least partiallY reduced
as it is collected. This restriction foroes the experimenter to make a priori
guesses about what data he can ignore. In addition to losing data, valuable
programming time is lost in reducing the data. For 'example, about half of the
calculation time in the program in Appendix A is spent . in reducing and recording
the data. This time could be spent running a second experiment.
Another problemprodueed b.1 this restriction is the inability
to read in other programs while an experiment is ,being run. For example, in
the experiment in Appendix A, the average total response rate of all :; subjects
is less than 3 responses per second. In other ~ords, on the average, o~
50 milliseconds out of each second are used to run the experiment. This leaves
plenty of time for doing other things such as typing in progrruus and analYzing
data. . Ho'trever, due to the time factors in our experiments and the slow tape
.operations, it is not possible to.take advantage of this dead time.
.
~) Possible soluti~~ - input/output processor. A possible solution
is the addition of an input/output processor and a 256 wor~ core buffer storage
to handle the reading and writing of tapes. A LINC with. this addition would
~ook very much like two cOlnputers connected by a l2 bit parallel transfer
cable and a :re't~ synchronizing lines. The first computer would look like the
present LINC without any tape handling capabilities. The second oomputer would
look .like the present U1TC 'tdth only the ta.pe handling and operate instru.ctions.
When computer #1 needs information stored on tape, it executes a read tapa
..
3-8
Blough
8.
instruotion which tells computer i2 to read the specified block into the
buffer. In the meantime, ,computer #1 continues executing instructions and
periodically checks an external level which will be turned on by computer #2
after the block has been read into the buffer. When the external level is
detected by computer #1, it is resat. Then computer 41=1 executes a four "
register instruction which, in the last.:3 registers, states 1) the number ot \
registers to be transferred from the buffer to the main core memory, 2). the
first register in the buf£er to be transferred, and':3) the first storage
register in the main core memory. A similar instruction would be used to
transfer inf'ormati9n from the main core memory to the bu.:f'fer. A write tape
instruction would write the buf£er into the correct' block and set an external
level for computer II. By making the transfer instructions completelY general
and separate from the read and' wri'te instrtlctions, it is possible to read any
set of registers in any block into any set of registers in the main core
memory and vice versa for the write instruction. Also, it is possible to make
any number of transfers gOing either direction between successive reading and
writing instruotions. Since the LING can transfer information at the rate ot
125 words per millisecond, computer #1 can transfer information to and from
the buffer at its convenience and without producing intolerable de~s.
Quit~ clearly the addition of this sort of input/output.
processor would greatly increase the flexibility of the LINC. Data could be
periodically written onto tape without interrupting exper-lm.ents. Ex:periments
could be run simultaneously and could be started and stopped independently.
Data analysis and programming could be performed in the dead time between the
disposing of responses during experiments.
5) Changes ~ C?"nline inout/outpu:'c facilities. Generally speaking,
operant conditioning experiments treat a response as a digit signal - i.e.
it is either on or off. Consequently, we do not have much need for the analog-to-digital conversion units. Ho,,-"9ver, in the future, more interests may develop
in recording responses on a multivalued scale. Therefore, for our purposes, it
'Would be best to make the analog-to-digital converters an optional feature.
This is true for the knobs as well as for the online converters since the key~oard can do almost everything that the knobs can do, and do it more precisely.
In place of the digital-to-a,nalog converters, 't.ze wou1d prefer
to have a more elaborate digital input system such as the one we are currently
planning" .In this system there are two ban.l{s of 12 digital response detectors.
When the onset of a response occurs the response detector is turned on. It
remains on until the computer reads the 12 bit bank, containing that response
detector, into the accumulator over the TN line. After the response detector
is reset b.Y the computer, it stays off until another response is detected.
After the content of the response detector bank is stored, it is checked bit
by bit to determine which, i f any, responses have occurred. If' a bit is on,
a table look up procedure is used to determine the location of the subroutine
contingent on that bit. Although this approach may be slower than reading
consecutive SXL lines, it makes it possible to read all of the inputs !rom one
organism simultaneously. Also, the tabla look up system can be part of an
"interpreter" which interprets call sequences, stored in the second half' of
3-9
9."
Blough
memory, for subroutines stored in the first half of memory. Also, the response
detector holds the response until the computer has time to check it. This
can not be done directly with the current SXL lines.
,
It would be nice to expand the output facilities, also, since
llJal'lY events l1lU.St be under the control of the cOlnputer. For example, in each
of J of our eJ..1>erimental chambers there are 7 lights and a feeding mechanism .
which must all be under independent control. In some experiments, the nwnber
of relays required can be reduced by using combinations of relays to switch
a certain event., HOirever, this makes it dit.'£ioul.t to run mora than one
experiment in a chamber since dii'ferent experiments will frequently ~qui.re
different relay combinations. To cope with this problem, we have alrea~
added 12 relays and are preparing to install another :36 relays.
Another inputI output facility that vlould be very useful would
be an external shif't register to handle the timing and shifting required by
the teletype. This would :make it easier to run the teletype during .
experiments.
6) Automatic ~p.terrtl:pt system. If data processing and online
control are being performed simultaneously, it is necessary to go periodically
(at least once per 5 milliseconds) into the online mode to check all the
inputs. If nothing is found, valuable time is lost. Also, there is a variable
delay (up to S milliseconds) added to the usual delay from response onset
until stimulus change 0 Just as important, the data processing program nmst be .
written with special exits scattered through it, ,and considerable care must be
taken to make sure that an endless loop or soma other programming error does
not occur.
These problems could be solved by adding a very simple interrupt
system which l'1ould jump the computer to the input scanning routine whenever
the computer received an input from a response detector, from the keyboard, or
from the tape buf£er. The interrupt s.ystem would store a jump instruction in
a register, other than the :index registers, which would be used to reswue
data processing when the input was di.sposed of. The scanning routine would save
the accumulator contents be:rore doing anything else, and put this value back
in the accumu.la.tor before executing the "resume" j"WTlp. A more elaborate
interrupt system is not necessar.y Since, once in the online mode, ever,ything
has top priority, i.e. nothing can be interrupted.
~ ~~.
Most of the other problems connected with the
Although LAP has been an immense help, it does
not do the vlork of a compiler. At present, considerable time nm.st be spent to
per:rorm what appear to be rela.tively simple problems. This situation could be
allev.i.a.ted by writing arithmetic and experiment orlen-ced compilers. A more
feasible and flexible approach may be to write a macro assembly program and a
large library of arithmetic subroutines. For example it would be very nice to
have 1) subroutines for adding, subtracting, multiplying, dividing, and tald.ng
square roots for both fixed and floating point, double-precision numbers
and 2) a quick and simple method for assembling them into a program.
7)
LINe are programming problems.
3-10
Bibliography: Talks and papers on "Tork done with the LINO
IIIRT1s, Generalization, and the LINO" - Oolloquium, Harvard Univ. Dept. of
. PsychologyJ
Harch 18, 1964
-,'
IIInter-response times during stimulus generalization tests" - Paper-read
at Eastern PsyChological Assn. meetings, Phila., April 17, 1964
"stimulus control of response probability" Biophysics, 11. I. T.,
January
Talk, Lab. of Communication
14, 196.5
IlRecent research in operant conditioningll - Le:cture series,
The Rockefeller
Institute, Feb. 1.5 - 26, 1965
.
(
IIReinforcement schedules and stimulus control" - Talk, operant conditioning
group, Col~umbia University Dept. of Psychology, Feb. 17, 1965
"Reinforcement of Least-frequent Interresponse Times" - paper submitted to Psychonomic Science, l1a.rch ..... 1I 196.5
(copy attached) (~ppeV\J,\)i
A)
IIMomentary Response ?robability during Generalization and Discrimination" paper in preparation
3-11
APPENDIX
A
A sample operant experiment run by the LINe
Part 1 - Description of the experiment·
Part 2 - Sample of daily data printout
Part 3 - Co,y of on-line
pro~ram
3-12
1•
..
Re1ntorcemen~
of least-frequent Interresponse Times
Donald S. Blough
Brown University
3-13
A'Qstract
Pigeon subjects pecked a key for .intermittent todd reinforcement. An o~
line digital computer (the LINe). controlled the" experiment.
It reinforced only
-'
those responses that terminated inter-response times which had occurred least
frequent4r in the i:mmediate past.
This schedule discouraged. response-to-
response dependency, wb1le generating stable response rates and unitor.m behavior
among subjects.
, 3-14
Blough
The rate with which an animal emits a simple response can be an effective
measure of the influence of stimulus and other variables upon behavior.
Ho't'1ever, in most if not all operant conditioning situations, response rate'" is
also governed
qy
inter-response dependencies.
When behavior is maintained for
many hours by intermittent reinforcement, behavior patterns emerge and responses'
become "chained" together.
This influence is hard to evaluate and to control,
yet inter-response dependencies sometimes become so strong that rate is quite
insensitive to other variables.
Inte~response dependenCies can often be traced to stereotyped behaviors
that fill the interval between recorded responses.
MOst schedules of reinforce-
ment favor the growth of such behavior patterns by failing to control the
behavior that comes just prior to the reinforced response.
When 'such control
is lacking, subjeots tend to settle into "superstitious" patterns of behavior
[l} that generate varying rates and patterns of response.
Some schedules
specify that the subject must space its responses by a certain interval in order
to receive reinforcement.
chance, but are
Bere, interresponse dependencies are not left to
specifica~
generated by the schedule of reinforcement.
Whether arising by chance or by design, 1nterresponse dependencies
ordinari~
reveal themselves in no~random patterns of inter-response times, (IRTs).
~
Certain mTs occur more often than ,one would expect by chance, others less.
The present schedule of reinforcement [2J favors random resPonding by
forCing the subject to emit distributions of mTs that approximate those of an
ideal random emitter.
Such an ideal subject, emitting responses randomlY in
time, generates an 1nterresponse-t1me diB1?ribution described by
t(t)
=
,
i\e
,-At
3-15
Blough
4.
where
Arepresents the mean 'r~te per unit time,
between responses
and t represents the time
[3J. Physical processes that involve independent events
"(e.g.-radioactive disintegrations)produoe IRT data of this form.
The shape
of the funotion is suggested. by the data in Fig. 1To make this ideal distribution correspond even roughly to real behavior,
we must first recognize that subjects have a certain minimum IRT, and that
ver,y short IRTs have special charaoteristics (see below).
These facts raise
several difficult problems, but as an approximation we assume that responses
can be repeated after a brief fixed interval of time.
The mT distribution
effectively begins at this time, set at 0.8 sec. in the experiment described
Next, starting at this arbitrary origin, we mark off 16 intervals, '
below.
or. ubins", along the time axis.
rate
A,
These bins are chosen such that, for a given
they cH.vide the area under the mT curve into 16 equal parts,
(See ;thin spacing", Fig. 1).
responding at rate
.A , would
'Ihis means that the ideal random emitter,
on the average drop the same number of mTs into
each bin.
This ideal model controls the performance of the real subject.
,
.
computer, the UNC
[4J
A
senses responses 'as they occur, computes their IRTs,
and accumulates these IRTs in the 16 bins just described.
recent ~4 (or, recently, 152) mTs are saved, however.
Only the most
This distrlbution of
reoent IRTs is used to determine the next response to be reinforced.
reinforcement occurs,' the LINO
contains the .tewest IRTs.
sca~
When a
the 16 IRT bins and notes l'lhich bin
This bin beoomes nhotu :
a response is now reinforced
only i f it follows the preceding response by an IRT that falls into this bin.
That .is to say, the subject is reinforced for making his
"leastprobable~'
mT, .
3-16
Blough
s.
as estimated from his recent performance.
This schedule discourages interresponse dependence, expressed as
particularlY frequent IRTs.
a stereotyped ubow" after
If, for example, a pigeon subject starts out with
~ach
response, it will produce many mTs that equal
the time it takes to bOloT and peck.
The IRT bin corresponding to this
t~
thus be relatively fulJ., so lRTs of this length will not be reinf'orced.
will
As a
result, bowing should decrease in frequency until it becomes no more likel;r
than other movements.
Such reinforcement of least frequent IRT is the defining characteristic
of the present schedule.
Our application of the sche.dule has other details
,that arise from the realities of reinforcement and animal behavior, all of
which cannot be described in this brief report. Among them are the following:
(1) The first response after reinforcement is never reinforced, nor is its
IRT recorded.
reinforced.
(2) Responses terminating IRTs of less than 0.8 sec. are never
A reinforcement in progress terminates if such an IRT occurs.
Our
data suggest [5] that such short IRTs are relentlesslY non-random, and that
they are not.controlled by the same variables that affect longer IRTs.
(3) Frequenoy of reinforoement ~ be varied
responses otherwise scheduled to reoeive it.
b.Y omitting reinforoement of
~o
far we have used probabilities
of 1.0, 1/2, and 1/4 that a response will be reinforoed, given that it
terminates a I'least likely" IRT.
The following describes a portion of our work with the schedule just
described.
Three pigeons were used.
Two of the birds had been on a variable-
interval schedule of reinforcement for tna.nY hours previously, and they had
developed quite different rates and patterns of response.
A third bird was put
3-17
Blough
6.
on the present schedule after onlY about 2 hours o£ variable interval
reinforcement.
All birds were maintained at about 75% o£ their ad lib weights.
They worked daily :for 80 min. in standard experimental chambers, pecking back-
.
lighted keys that required 10 grams of pressure for electrical operation.
Reinforcement consisted of
4 sec. (later, 3.3 sec.) access to mixed grain.
The LINC computer ran the three birds simultaneously, programming reinforcements
~
from the special IRT bins described above, with
=
1/2.
It also recorded
responses into standard 0.1 sec. and 0.5 sec. IRT bins, and compiled a running
cumulative-response record for each bird.
.
Between sessions, the LINe
ana~ed
\
the data and produced displays, two of which serve as figures for this report.
Figure 1 shows mean frequencies of interresponse times in successive
0.5 sec. bins.
The means cover six daily 80 min. sessions for each bird;
mTs under 1 sec. are omitted in the Figure. . The distribution of IRTs within
the special reinforcement bins is not shown here, although the limits of these
bins are in~cated (a'bin spacing").
Figure 2 shows the data of Figure 1
. transformed by dividing each IRT value by the number of
mT had to occur.
~
"opportunities;~
that
This transformation estimates the conditional. probability that
response will occur, given that a particular time has elapsed since the
previous response
[6]. For the ideal random emitter this transformation
...
yields a horizontal line.
It will be seen that the data here approach this
ideal, though points under one second fail to conf'orm, and the curves fall
somew~t
with time.
IRT patterns within birds are much less marked than they.
were before the birds were placed on this schedule.
birds were very s1m1lar,with
programmed
.50.
A = .53,
The overall rates of the
.51, and .59, slightly higher than the
Cumulative records indicated that the birds responded with
3-18
Blough
uniform rates during the daily sessions.
Several general charaoteristios of this schedule are worth noting.
(1). It controls emissive behavior by basing reinforcement on interresponse'
time, while avoiding the reinforcement of speoifio IRT ranges either by
chance or by design.
(2) It uses a feedbaok prinoiple to concentrate rein-
forcements where they are needed to produce a preseleoted behavioral
eqUilibrium.
(3) It has thus far produced
mT
distributions that correspond
rather well to those expected from a response-independent system, i f very short
IRTs are excepted.
However, we cannot conclude that responses are therefore
independent. We can be fairly sure only that they are not tied in any simple
repetitive pattern.
(4) While behavi.or is variable from moment to moment, the
sohedule produoes stable average rates.
Despite their earlier differences on
other schedules, the subjects all emitted roughly the same number of responses
on a given day and from day to day.
Because
earlier.
Ais
fixed, the schedule precludes one of the objeotives stated
It does not let rate
experimenter may be interested.
reinforced IRT.
c~ange
freelY with any variable in when an
However, rate is not entirely controlled by
Deviations of obtained rate from programmed rate may show the
effeots of other variables.
The next step may be to put the parameter
itself' into the feedbaok loop that determines reinforcement.
past average rate were used to set the programmed
on a value dependent upon other variables.
1\,
A
I f the animal's
it's rate might "home"
It remains to see whether this
will happen.
rt-3-19
Blough
8.
Figure Captions.
Fig. 1 - Inter-responset1me frequencies of three birds by
0.5
sec.,
~~s,
.....
omitting IRTs less than 1.0 sec. long.
time "0".
The preceding response occurred at
The limits of the 16 reinforcement bins are indicated by
Ubin spacing".
Fig. 2 - The interresponse-times per opportunity (conditional probability
of response) for each 0.5 sec. period following a response, calculated
from the data in Fig. 1.
Values are included for all IRT bins except
the rightmost, for which the .value is always 1.0.
3-20
Blough
9.
Footnotes
l.
~.
Psychol., ,J§, 168-172 (1948).
1.
B. F. Sld.nner,
2. .
This research was supported in part by USPHS .grants MH-024S6 and
m-083SS.
:3.
W. J. IvjcGill, in Handbook
.2! Mathematical Psychology, R. D. Luce,
R. R. Bush and W. J. ~1cGill, Eds.
4.
(Wiley, New York, 1963), p. 316.
The LINe (laboratory Instrument Computer) is a digital machine l-dth
core storage of two thousand 12 bit words, and flexible input, output
and magnetic tape capabilities.
It was available through an Nm
sponsored evaluation program that placed LINCs in a number of bio-medical
and psychol.ogical applications.
The machine is now available commercially
from the Digital Equipment Corp., Maynard, Mass.
S.
D. S. Blough, in preparation.
6.
D. Anger, il.. ~. Psychol., ~, 145-161 (1956).
3-21
..
...
I
,
•
".
•
...
.I
PIli
If
.tt •
....
a •••••
.
,"nlll""
..•
I
•
1M
•
."
I.
-,
.
---
• • ..•e • • , , •
I
..,'
I
I
••• ".c...
I
IPti' ..."
H""II .
....•
FIG-. 1
.. _--_ .. _--_._-----
, .,.11
•••
I.'
.,
'1'
I.''S
•••
'.1
• • ."
IftS
••••
• ••• • ••••
•
••• •
• • • ••
•••••••••••
• ••
•
•
. ..... I.··
•
•
•
.-
•
-----------
•
•
,
,
I
..•e.,'• .....
"H
......
•
FIG-.
.2..._--_._-----
3-22
ASample of daily data printout for "Least
date
1.
Fre~uent
Interrer.ponse Time" study.
0201
birds 0007
,0593
0612
0051
mT 002i1
06i&9
Oil16
raw
0037
0055
freq. 0000
0000
in
0040
0000
0.1
0004
0011
and
0071
0063
0697
0.5
0324
sec
0034
0035
bins
0000
0000
0036
0000
0001
0017
0000
0033
072i1
0033
0057
0039
0000
0000
.0024
0000
0000
0014
0003
00il2
0010
0035
0157
0027
0197
0017
00il7
0032
001i1
003i1
0065
0029
0065
002i1
0031
0026
0262
0043
0145
0022
01i17
0011
0096
0016
0077
0020
0056
0012
0065
0009
0057
001i1
0051
0006
0000
0000
0000
0001
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0003
0000
0000
0000
0003
0001
0001
0001
0000
0004
0001
·0003
0002
0001
0001
0000
0003
0000
OOOil
0000
0003
0001
0033
0041
0017
0036
0125
00il7
0076
00-'i2
0016
0060
0032
0028
0053
0031
0024
0029
0033
0023
0156
0019
0231
0020
0171
0016
0065
0020
0095
0014
0089
0013
0091
0010
0070
0015
006 1
0010
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0003."
0000
0001
0000
0001
0000
0001
0000
0000
0000
0003
0002
0003
0001
0000
0000
0042
0001
0046
0016
0033
0014
0026
0014
0034
.0034
0026
0036
0029
0091
0026
0078
0014
0255
OOill
0162
0032
0131
0032
0077
0026
0069
0016
0069
0016
0061
0016
0064
0010
0050
0005
0000
0001
0000
0002
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0000
0001
0003
0000
0000
0000
0003
0000
0000
0001
0000
0000
0000
0001
0000
OOOil
0000
0000
0102
0062
0126
0043
0113
0065
0075
0043
0145
0042
0091"
0074
0067
0044
0092
0047
0062
0057
0060
0047
0070
0045
0069
0050
0062
0040
0055
0029
0054
0017
0062
0061
0066
0070
OOBO
0066
0104
0069
0057
0065
0075
0075
mT
0129
freq. 0071
by
006i1
special 0075
bins
0167
00B2
0000
. . 0003
3-23
. 2
o
1
2
3
4
5
6
7
8
9
10.
11
12
13
14
15
16
17
18
19
20
21
page 2
o
o
XXXXXXXXXX
XXXXXXXXXXXXXX
XXX
IRTs/O?
histoerams,
X
3 birds
XXXXXXX
XXXXXXXX
XXX
XX
sec. bins
XX
xxxx
xxxx
XXX
XXX
XXXX
.
XXX xxx XXX XXX XXX xxx xxx xxx xxx XXX xxx xxx xxx xxx xxx xxx xxx xxx xxxxxxxxx
3
X
4
5
XXXXXXXX
XXXXX
6
X
'7
XX
·S
XXXX
XX,
5
XXXXX
XXX
XXX
XXXX
'XXXX
XXXXXX
XXX
XXX
XXX
XXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
I·
2
3
5
0
0
4
0
X
5
X
6
7
XX
8
9
10
11
12
13
14
15
16
17
IS
19
20
21
3
XX
o
I
2
3
0.1
XXX
XXX
XXXX
XX
9
by
xxxx
2
10
, 11
12
13
14
15
16
17
IS
19
20
2l
5
xxx
2
1
·3
X
XXX
XXXXXXX
XXXXXX
XXX
XXXX
XXXX
XXX
XXX
. XXX
XX.X
XXX
XXX
X
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX~XXXXXXXX
3-24
o
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
2
xxxxxxxxxxxxxxxxxxxxxxxxx
page 3
xxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxx
IRTs/OP
xxxxxxxxxxxxxx
xxxxxxxxxxxxx
histocra7lls
xxxxxxxxxxx
xxxxxxxxxxxxxxx
3 birds by 0.5
xxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxx
sec. bins
xxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxx
xxxxxxx
xxxxxxxxxxx
xxxxxxxxxxxxxxxx
xxxxxxxxxxxx
xxxxxxxxxx
xxxxxxxxxxxxxxxxxx
xxxxxxxxxxxx
_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx,)(x)(xxxxxxxxxxx
I
o
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
2
5
1545
3
. 4
'
5
XXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXX
xxxxxxxxxxxxxxx
XXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXX
XXXXXXXXXXX
XXXXXXXXXXXXX
XXXXXXXXXXXX
XXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXX
XXXXXXXXXXXXXXX
xxxxxxxxxxxxxx
XXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
o
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
, 105
5
3
2
4
5
XX
XXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXX
XXXXXXXXXX
XXXXXXXXXX
XXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXX
XXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXX
XXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
0502
0438
(coddd total pause ti~eJ 3 hirds)
3-25
MONITOR PROGRAM FOR THE LINe
Lloyd Marlowe and Donald" S~ Blough'
Psychology Department
'Brown University
The MONITOR program is an elaboration of a programealled_META, which
was used with the 1024 word memory to read in and start, other programs ."
With the addition of the second 1024 words of core memory, the MONITOR
program was written so that up to 210 programs' could' be: . read in and ·
started without any operator intervention, by using a predetermined
MONITOR control sequence. However, the program layout and the manner
in which the programs are read in and started was not changed to take
advantage of the full core memory. Consequently, the reader will probably notice numerous cases of inefficient program'layout and storage. In
addition to reading in and starting programs the MONITOR program presents
a display that is associated with each program so that values for variables in the program can be conveniently typed in before the program is
executed.
FORMAT OF MONITOR CONTROLLED PROGRAMS
Any program which is going to be used under MONITOR control must be
written using the format described below. Each program has three basic
parts: a display, a control block, and a program proper.
Display
A sample display is shown below.
EXPERIHENT filA
STORE DATA BEGINNING AT
REGISTER __ _
BLOCK __ _
UNIT
The program display is made using the DISPLAY ~~KER program (see
Appendix A). The display produced by the DISPLAY MAl
nl ---------------,) 0
0
MIMIC ROUTINE
Restart
Pro1. #3
--------------->
D
(Duriiig mimic
routine)
"wrong" --------) (no answer)
nl --------------~ (nl) 1 ------------->(nl) 1
(-;'1) 1st X) ----~
(nl)2~
------i;.l) 2
(nl)12nd X)-----(nl)1 -----~Restart
(-;'1) + (nl)2---------------~Restart
.(;;1)3~
(1st XJ.1~(;
.
£\./.
~~~~-
~ (nl)3
(-;'l)l~
.
.
--------------(nl)l -----------------------------~mimic (nl)zx)
..."
n ---------------~(n + n)--------------------------------) "wrong" -----~wrong ans.)
n
---------------~
(n l ) 1
n4 (water)-----------------------------> "wrong"
-----~water
noise)
(n +
n)
)
(n + n)(2nd X)---------"wrong"---~R
n------------------
~correct'~--)R
n (any combination ------------JR
4
n ---------------~ "correct"
n
n ---------------~n + n)
(partially correct)
(nO"ready")n~ n) ------------------------------->
'''vrong''
(FOR PROB. BLK. DIA. #5)
----~correct)
-----~partially correct-2nd try fails)
nl)
nl) max. I req. for same cat.in a row (except when nl category is only category remaining
nl) that has not. been responded to 5 X correctly)
5 correct responses
= criterion
all correctly answered
for not asking for this category again (by itself)(see below)
single category
ir compound
(no---no) -------)"correct" -----Hcorrect sequential)
t
no
---------------->
no
----------------->
no
----------------~
no
---------------~anything ~ (Do--no)~'wrong" ---~(orig.
imperfect sequential)
C:CAT.
FLOW CHART CODE
TEACHING PROGRAM
CA'T.lECATEGORY
SEC.IiSECONO
NO.! NUMBER
OPT. ! OPTIONAL
N£G.1iN£GATIV£
'"n-·SYM80L
i\. !SIONAL.
"~-14
EVALUATION OF THE LINe COMPUTER IN COMMUNlCAXION RESEARCH INSTITUTE
Miami, Florida
Dr. John C. Lilly and Benjamin Locke
This report will be divided into three sections:
(I) the familiari-
zation pr.ogram given by the Center Development Office in July 1963, (II) .-,~he
initial setup at the Communication Research Institute and (III) the sUQsequent
operation there.
the amount and
In considering these three aspects, I will attempt to evaluate
va~ue
of the help and information supplied by the Center Develop-
ment Office and the ease of execution of each of the operations necessary for
the use of the LINC.
And) in conclusion, an.:.evaluation of the resultant
scientific yield as a function of the total investment by the Communication
Research Institute.
(I)
The Familiarization Program
Three features were emphasized in this program: (1) the programming
of the LINC, (2) the theory of how the LINC works, and thus how to trouble
shoot it, and (3) the technique of operating the LINC.
(1)
MOst of the necessary rudiments for programming were conveyed,
although I think considerably more supervised practice was called for.
(2)
The true test of onels knowledge of how the LINe operates is to
be able to locate or repair anything that goes wrong with it.
the program I did not feel at all fully
that might go wrong with the LINC.
c~petent
At the end of
to locate or repair
anything
This 1 think is the result only of insuffi-
cient information, the problem of which I think could be easily surmountable
in the future by the creation of a trouble shooting manual giving step-by-step
procedures for locating a trouble spot and containing pictures of the waveforms
that occur at key places during correct operation.
(3)
As far as the operation is concerned, the booklet entitled
~-15
- 2 -
The Console appeared to be quite adequate for explaining the operation of the
console, although description of the other units were less adequate.
Simul-
taneous with the instruction courses were the assembly, calibration, and
checkout of the LINC.
The assembly presented very few difficulties.
However,
"-,"
.
the absence, again of a procedure manual for calibration precluded an objective
analysis of the problems of calibration.
The same holds true for the checkout
procedures, and, as a result, I think too superficial a checkout was done on
the LINC and as a consequence it was shipped off without being thoroughly and
rigorously checked in all its "aspects.
(II)
Initial SetuE
The LINe was shipped by commercial van and arrived on tfme and
apparently unharmed.
We were able to unpack, hook up and accomplish a check-
out with the Killam and Hance checkout program, within one and a half hours
after arrival.
As far as our particular setup goes, we did not discover any
malfunction of the LINC directly attributable to our adverse environment, that
is, high temperature and high humidity coupled with a close proximity to our
relatively open salt water tanks.
However, since problems of this kind often
take quite a while to develop, and since we were constantly plagued by a
marginally malfunctioning LINC, it was very difficult to determine the effect,
if any, of temperature, humidity and salt water vapor.
The LINC was shortly
put in an air-conditioned environment, anyway,. as a precautionary measure.
(III)
The Operation of the LINe at Communication Research Institute
As mentioned during the entire period which we have had the LINC we
have been plagued by a large number of somewhat subtle and sophisticated
operational malfunctions.
Initially this consisted of erratic faulty
opera~'
,-'"
tions, characterized by an apparent inability to operate correctly until the
LINC had been on for five
~r
ten minutes, by a highly sensitive tape tension,
:4-16
- 3 -
and by associated magnetic tape unit problems.
As the weeks progressed these
problems became more and more severe until finally it was
a~ost
impossible to
read a program onto or from the tape and Lap 3 would not work at all.
During
this tfme a very large number of attempted fixes were tried, and the machine
was completely recalibrated to no avail.
Finally on the 23rd of May 1964,
Charles MOlnar came down and discovered that one of the problems was a faulty
fix which was done by Severo Ornstein while the computer was at MIT.
This
consisted of tying off part of the magnetic tape circuitry to ground via a
capacitor.
This capacitor effectively short-circuited a whole element of the
logic, with the result that tape transfers were never checked.
With the removal
of this capacitor the LINC was improved but still not fully operational.
The
problem then appeared to resolve into a complicated timing problem with the
whole magnetic tape circuitry.
The nature of this problem was recognized by
S. Ornstein at MIT and also by Charles Molnar down in Miami, however; and it
appears that the magnetic tape section of our LINe computer has not ever, and
is still not now performing at its
max~um
capability.
Another problem contri-
buting to this was the fact that all of our tapes had creased edges.
the use of brand new tapes did not eliminate the problem.
However,
In the interest of
fixing these difficulties we were supplied with a new and fully calibrated
magnetic tape unit from the Center
Developmen~
Office and this appeared to
work well for several months, however, in September again the marginal problem
with tapes began and another investigation was begun.
One feature that was
discovered was the severe abrasion of the underside of a new tape, used for Lap
3, to the extent that this tape would no longer work.
Several of these
abrasions appeared to coincide with the edges of the gulleys at the inner
margins of the tape shoes.
n~
This, however, appears to be a development of the
unit, and although ;t is an additional reflection
~~
this section of the
- 4 computer, it does not explain any of our earlier problems.
As a result of all this, the trouble shooting and repair techniques
as developed up at the MIT course were rigorously tested.
It soon became
evident that without a really thorough engineering knowledge of the events
that occurred during a given instruction and
their·opt~um
sequencing, and
thus without knowing what the proper waveforms ought to look like, it was
extremely difficult to do a repair on any sort of a really sophisticated
problem.
Faulty cords or connectors or the like could often· be very quickly
located but the ascertaining of maximum performance was quite difficult
without some sort of a reference as suggested above.
Since, in my opinion, the LINC is not yet perfected to anywhere
near its maximum capability, it has been almost impossible to determine
its dependability or its stability, that is, how long do the various components
last, how often recalibration is
nece~sary
and what is the effect of environ-
ment and line voltages variations.
'It is recognized that 1) the LINC has a small memory capacity,
and thus ltmited abilities as far as using a compiler program, and the
resultant more bulky compiled programs, and 2) many biological uses require
strict control over the size and construction of a program.
But "adva'ntages"
and "disadvantages," "possible" and "impossible" to the contrary, it was
found to be immensely difficult to create a large, complicated program in
assembly language) rather than in a compiler or algorithmic type language.
Assembly language programming, I think, necessitates a high and often unnecessary degree of knowledge concerning the operation of the machine, far
more than would be necessary for algorithmic programming and, in addition,
is a much more tedious, a much more lengthy and a far more complex procedure.
Large and complex,but 'constructionally uncritical programs must be continuously
·4-18
- 5 -
created and continuously modified for the biological sciences, and therefore
I feel that the lack of a compiler program facility to deal with this need
is a severe drawback.
No difficulties were experienced with the techniques of operating
of the LINC.
However, I think that this was probably to a large extent the
result of the thorough knowledge necessitated for trouble shooting and for
programming, and therefore the case for the naive operator might be different.
Two programs are presently under development, the first one is a
Click Generator
Progr~
This program. with the aid of a random number of
table occupying three quarters of memory, generates a random number of clicks
from one click to 30 clicks in a strict time format.
The click information
is supplied to the external world via the "Operate pulse," one pulse corresponding to one click.
The external apparatus generates the desired waveform
from the pulse, and at the termination of the waveform generation the
is restarted via the External Level input.
comput~r
This program was successfully run
and recorded on tapes for use in psychological studies on dolphins.
Also under development is a conditioning program which is of a highly
flexible nature.
The goal of this conditioning program, initially, is to
condition the dolphins to identify and differentially respond to various
categories of auditory stimuli.
This is done through a semi-random presentation
of the stimuli, an analysis of the response including an auditory prompter
mechanism for aiding in the improvement of partially correct responses, and
the resultant presentation of signals, rewards or punishments to indicate to
the animal the machine's final analysis as to the "correctness" or "incorrectness" of a response.
Associated with this program will be a statistical
analysis program to evaluate the dolphin's patterns of responses and to attempt
to deter.mine the degree of learning which has been aceomplished by the dolphin.
~-19
- 6 -
If this first part is successful then the long-range goal 1s to modify the
program and to use it in the measuring of cognition, i.e., the degree of
abstraction and communication of which a dolphin is capable.
is initially designed as an auditory experiment, the
Although this
connec~ions·
to'external
equipment can be easily modified to change to other kinds of stimuli.
The
development of this latter program has been hindered by the aforementioned
difficulties involved in using only an assembly language.
In conclusion, however, I feel that given the programming knowledge
and the necessary computer technology coupled with thoroughly documented
trouble shooting and calibration procedures, the naive scientist would be
sufficiently equipped to productively utilize LINC computer provided that
this computer is fully debugged and dependable.
Unfortunately, the experience
that we have had with the LINC is not especially convincing that a relatively
trouble-free and dependable state is easily attainable, and even at that state
the often unnecessary time and effort spent programming is so significant
that it detracts from the computational capabilities of the LINe, which
appear impressive.
~-20
CHRONOLOGY OF INSTRUMENTATION TAPE USED ON LINe
1.
Item #3, Information Bulletin #1 dated 14 February 1964 from S. Ornstein of
Center Development Office suggests use of 3M Tape Cat~ #(489-3/4-150-24066).
2.
3M cost; and spec sheet M-I-28 dated 1 February 1964 describes tape and.,quotes
price. ~: No warnings as to humidity or environmental conditions are
stated on this sheet.
3.
On 25 March 1964 20 reels of the above described tape were ordered by Communication Research Institute thru Magnetic Products Division, 3M Co., Chamblee,
Georgia. (P. o. #1400. Tape was delivered on 17 June 1964).
4.
LINC logbook (page 34) shows all programs were transferred to tape in question
on 20 June 1964. Old tape was discarded. Further: Perusal of logbook showed
no prior tape problems, other than frayed edges, as of this date.
5.
Next entry in logbook (page 35 dated 8 September 1964) describes problem in
that the tape would not store information. The tape was then sent to C.
Molnar for diagnoses. There is no record of a diagnoses available to us.
6.
A few days prior to November 23, 1964 I experienced difficulty and frustration
during an attempt to read an assembly program into the LINC. The trouble was
tracked to the tape, and an examination of the block in question with a dissecting microscope showed a regular flaking of the oxide.
7.
On 23 November 1964 David Peterson of Miami received the tape to submit for
lab inspection.
8.
On 17 December 1964, Mr. Don Tomasak of the 3M St. Paul test labs claimed the
flaking was due to ''high humidity. II Mr. Tomasak stated at this time that
subject tape was not reliable over 40-45% humidity. He also indicated that
an experimental tape spool for use in humidities over 40-45% would be transmitted to Communication Research for evaluation.
9.
On 17 December 1964 I spoke with the Miami representative, Mr. David Peterson,
who cla~ed that 55% was not too high and since we were using the tape under
65% conditions, he did not think the 20 rolls could be replaced.
GENERAL OFFICES • 2501 HUDSON ROAD • ST. PAUl,-MINN-ESOTA 55119· TEl: 733·1110
magnetic Produ ctsDlvlsl on
January 5, 1,65
Mr.
r.
Cri....n
C~nic.tion.·a.search
3430 Main Hi&hw.y
Mi_i, Florid.
Dear Mr. Gril.san:
Thi. letter will 8U111D&rize our lab findings andl"ecoqmDendationa on ,
your repol"ted problem with #489 Sandwich Tape.
The one roll of '489. 3/4" x 150', we received from )'ou for evaluation
of your problem" showed that the px.ide and sandwic.h layers hadb.,en
-pulled out in lection •• We reco~nize thi8 type of d.-ag~_and it
usually occur8 when the tape oper_ates in a high humidity ,environment--.
Moi.ture in thickness of only a few molecular layers act~ as an
adhesive. The' adhesion can take place between ~he sandwich layer and
epoxy fill spots in the head and it can oc~ur between the sandwich
layer and metal ~urfaces.
We generally conlider relative humidity to be on the high sid. when
it is above SOt. The 'exact RHthat begins to become problematical. is
Dot aalY to pinpoint because tape damage of th. type you experienced is
a function of tape-to-head (or guide) pr8S8 1;re. For f!xample buildupi
on the tape from scratched tape surfaces cause high preslure pointl
that aggravate the deleterious affect of high humidi~y. To sua up
then. higl1 humidity 1a to be a.voided. A clean machin~-tape systom is
essential.
.
J
Through our local representative, Mr. D. J. Peterson; you vere
supplied with a sample of '8972 that is lesa f..Iu8cept"'bl~ to the type
of damage you expert.~nceJ. ~e. (,6commend you ("(."'n91do'[' tha pr,')d~ct 0.1
future procurements. It. i. U14gneticall)' tt-,c 8-:lllle, as the #459 you have
been
UI ing.
I
l'va beenadviled by trae SA1~8 Departm~nt t:tat you wJ.ll be hearin& frOill
Mr. Peterlon in a weak or two on t_ha #~39 you have O{l complaint. In
the meantime, we hope this cla!'tfi~.g \.~.:;\ oolution to the problem.
------_.
truly,
..
(
(--~
'\
.... /"
YO)1I'S
.t
Itdt. (l
Donald F.
Technical
'f:
....
/t>~t.~~
~
~'
T~1sak
Se~ice
Supervilor
D~:jD
cc:
D. J.
P.'taraon, 3M
miN N E SOT A
m I ~ I N
(i
AND
m .f>, N U rAe T 1I f1 I N Gee m PAN Y
'[4-22
-
· Final Report
LINC Evaluation Program
C. Alan Boneau
Department of Psychology
Dulce University
5-1
1.
History of the LINe at Duke
The computer arrived in September, 1963, at a 'time when the
Department of Psychology ,was preparing to move into new quarters and
construction was still proceeding on the building in which it now"
resides. It \~s placed in storage briefly and then a temporary laboratory in which apparatus \"las being prepared for the research to be
described below. Early efforts to integrate LINe into the apparatus
revealed engineering problems of such magnitude that LINC was tempor,ari1y abandoned while the apparatus was completed on another basis in
order to continue the research program. By May, 1964, that work was
proceeding well and the ne\'1 quarters were complete. At that time a
move was made into the new laboratory quarters.
With more time available at this point to devote to LINe, the
engineering problems (involving shielding, control of ground loops,
and arc suppression) were solved enabling LINe to be utilized in the
~pparatus.
This basic system was continued, but the function of LINe
was gradually enlarged. It was soon placed in charge of controlling
all the timing and data gathering functions, as well as the contingent
'events in the operant conditioning setup consisting of two identical
pigeon boxes. The addition of an external clock at this time aided
,the process. In addition, wiring was completed to provide an additional
24 LINC-controlled relays as well as another 12 binary in~lt lines.
The new relay controls were used in part to control a surplus Flexowriter which was installed ana operating in September, 1964-. At about
this time LINe was on a 24-hour- a-day schedule, reliably running birds
overnight, collecting the data from some 12 to 15000 trials per bird,
and storingrthe latency data trial by trial on tape.
Later LINe was used to process and analyze the data gathered in
this fashion.
Meanwhile a new set of apparatus was devised and constrl1cted and
in February, 1965, LINe \"las hooked into it. The main features of the
,new appaaus are somewhat more f1exibi1i ty, an expanded span of attention (four birds at a time») and a provision for a computer-ini tiated
automatic change from one set of birds to another so that several different groups of birds might be run overnight on several different experiments.
2.
!!!! ~
present research
Up ,to the present time, the LINe has been used exclusively as an
on-line control and analysis instrllment for an operant condi tioning setup. I t has been utilized primar ily on a color- discr imina t ion problem,
but has been utilized for a number of pilot studies involving training
of responses of different. latencies, and the effects of various latency
contingencies.
Since the apparatus has been the same for all of these projects,
and the control program \~s general enough, we wil~ describe here the
main features of both, before proceeding to a more detailed description
of individual project~.
5-2
The essentials of the apparatus consist of boxes of approximately'
a cubic foot into each of ~lich is placed a pigeon. On one side of the
box are two holes, a small round one at head level and a larger (2-"x2")
at ground level. Behind the smaller hole is a piece of translucent
plastic hinged so that it moves when the pigeon pecks into,the hole.
The bottom hole contains a device for feeding the pigeon at stipulated
times in the presence of a light available only at those times. Onto
the plastic key, a monochromatic light is projected from .the rear. The
light source is a variable monochromator to whichhas been connected a
servo motor controlled by a set of twelve potentiometers, which can be
selected by feeding a 4-bit code into a relay tree.
Provision is made for a shutter interrupting the projected light
at specified times. In an early version of the appa~us, two such boxes
were used, the LINe relays being used to control the monochromators,
the shutters, and the magazines. The control Nas rather complex involvihg
52 of the possible 64 combinations of the six LINe relays, and a system
of external holding relays. To monitor the output of the pigeon; SXL
lines were in effect connected to switches on the keys.
The basic procedure consisted of a ser.ies of short presentations
of the various stimuli, food being made available for a short period of
time if several conditions were met. Typically, the shutter was opened
,to present the stimulus value and \vas closed again at the occurrence of
a peck or the end of the period, usually t\~ seconds. If reinforcement
(food) was to be given for a peck on that parti~11ar trial, the magazine
,~s opened for, say, three seconds, immediately following the occurrence
of a peck. At the end of a trial or of a reinforcement, \vhichever was
later, the monochromator code for the next trial was placed in the relays,
and a variable inter-trial duration averaging two seconds was introduced.
If a peck occurred in this period to the dark key, the delay \'/as started
again from that point in time.
The basic program consisted of a large set of subroutines for performing the various functions indicated. The core of the program was a
list of J!'1P instructions and NaP's serving as a central control. Conceptually, the computer stepped through the list jumping to the various
subroutines as required, closing the gate to them by inserting a NOP in
place of the JMP when the operation was no longer required for a trial,
(for example, after the first peck, the gate to the subroutine which
looked for pecks in a particular box was -closed) •
\\1hen timing operations were initiated, say, the reinforcement period, a ~~1P to the appropriate subroutine was placed in the table, it being replaced by the original N9P at its conclusion. A trial was over when the list contained
nothingA~e NaP's and the swi teh was made to a betl'leen trial .'routine.
Typically, the time from the beginning of a trial to the occurrence of a
peck (the latency) was measured in one of the subroutines. Quite frequently this time was placed in a location in memory corresponding to
its place in the sequence and to the color code prevailing at that time.
After 100 trials for each bird, the sequence of trials for"b6th birds
was placed on tape. In this inte~-block time, a ,new random schedule of
5-3
-3-
stimulus codes was generated, and placed in memory. The generation
program contained a table of the various codes to be utilized and the
number of occurrences of each in the lOO-tria~ block.
A random number
was computed by a simple addition process, the number having the property
that its range \~s the number of places still to fill in the random sequence. After ~ach selection, the number ~1.n the location was reduced
by one and the code corresponding to that location was placed in the
random sequence.
Much of this was done with the original half-memory LINe and space
,..,as pretty much at a premium. Consequently there '\\'as a considerable
amount of reading programs in and out of memory in the interblock time.
Controlling all the operations was a special program called the MasterEntry Exit program which read, in the appropriate programs in the prescribed quarters for them and then jamped to the starting location.'
Once the original program was written it became s1.lprisingly easy
to make major modifications in it with the addition of deletion of a
few steps. For example, a latency contingency \~s added simply by inserting a routine which added the complement of the criterion time to
the latency and then using an APO for the decision.
In the past several weeks, a new piece of apparatus has been completed and tied to the
computer. This apparatus consists of a rack holding 4 banks of pigeon
boxes in two levels. The two rows on each level are arranged otl~\\'ard.
Between them is a movable racl( containing a monochromator, and all the
paraphernalia, shutters, magazines, etc., for .t\ro opposing boxes. A
motor at one end of the rack when activated on command from LINe will
mOve the central.island from one position to the next. The two levels
are independently tied to the computer, and \"li thin each level, all
functions are independent with the exception of the color projected on
the key~ ~ince there is only one monochromator linked to the two keys
by a beam splitting device.
This setup is basically sUnilar to the pr~vious apparatus, except
that it has provision for running 4 birds at once, instead of two, and
can be programmed to move to a new set of birds or even a new experiment
after a stipulated period of time.
To handle all the control functions, a new set of relays had to be
arranged. To this end, t~IO tl"lelve bit registers and control logic ,,,ere
built into one of the data terminal boxes. These are connected through
invet!ers to external relays identical to the original LINC relays. The
external lines are sufficient at present to handle the input, consisting
of the output of several one-shot multivibrators triggered by the various external events, such as pigeon pecks, positioning switches operated
during movement of the bqxes, and photoelectric circuits monitoring the
lights in the monoch~omators.
5-4
-4-
Programming for the net\' apparatus setup was similar to the previous
programs in many respects, but of course, was somewhat expanded. We have
found it convenient for our purposes to have a concept of r.ather short
subroutines each of which has a specific function to handle, say record·
a latency, or/determine whether the conditions for reinforcement have
been met. These programs have been strung together by means of a central
control program consisting of JMPs and NOP's as before. Ordinarily, the
jump to a r0 1 ttine is made with a code attached so that the function is
assigned to the correct bird by virtue of the code.
The surplus Flexo\'/ri ter has been connected into the computer by
means of one of the twelve-bit registers feeding external relays through
inverters. The manner of connecting the Flexo\~iter to .the computeroperated relays may.be of interest. The particular model which is in
use is one which operates from an eight-channel tape, but upon which all
the normal functions are confined to seven of the channels. In effect
we. have paralleled the Flexowriter's tape reader furnishing relay closures
in place of the switch closures which normally would occur if a tape were
being read. The eighth channel tape switch has been disconnected from
the circuits in the Flex and an external line has been connected in.
\\,i th the Flexo\\Il"i ter operating, controllable by the computer through the
relays, all timing functions are controlled by looking at the external
line. When its switch closes, a new code. is placed in the relays; ''''hen
opened, the code is removed. A little difficulty was experienced with
arcing since the Flexowriter operates on a 90-volt DC system and has
several inductances in operation. This was very effectively countered
by placing ITT contact protectors across the 90-volt Itne? togrorlntl
wberethey entered the computer-operated relays.
Since the completion
of that system, absolutely no comP1-tter attributable trouble has been incountered. At one time we thought there was but it was ultimately traceable to some very delicate mechanical adjustments on the folexowriter keys.
Now that the art of adjusting has been mastered, we have had little trouble.
A number of programs have been written to handle Flexowriter output.
The most general is one. \'lhich writes from the LINe keyboard to the Flexo\'1ri ter and also into memory. The output ·format including spacing is typed
on the Flexowriter and at the same time into memory. Since much of the
out-put desired was data, a provision was made to provide for an output
matrix, the entries of '\fhich in the ini tial set up were tagged u.ocations,
the ta~, indicating that it \\Tas the contents of the location, not the
location which was to be typed out. In operation the data was read in,
,~s converted from octal to decimal by a simple program and stored one
character per word, was converted to Flexowr i tel:' code by a tf-..hle-l~ok
up procedure, the space, carriage return, and,other function codes were
inserted. This involved a considerable amount of tape churning as the
programs to perform the various functions and the data and characters
were read in and out. Once the location matrix for a given format of
data was prepared, the sets of programs could be shuffled together to
get out data in any format and tacked onto the end of any data analysis
program. We found this to be very satisfactory.
5-5
-5-
Other programs have been written for other kinds of read out before
the Flexo\~i ter \'las available, the most elaborate of \\hich displayed the
Tape Unit number, Block Number, Location, Contents, and an octal-decimal conversion of the contents of a particular location, with provision
to pick locations in sequence or jump to the next location with nort~
zero contents. One could in addition read any block from either unit
from the keyboard. °A less elaborate version permitted skipping through
memory one \'w'ord at a time with an octal version of the contents of each
location displayed. This program had provision for altering the contents
of the memory location from the keyboard and hence was and is our primary means of entering programs initially into memory.
In addition to these, we have had a number of pr.ograms g~v1ng a
graphical display of the ~esults of our experiments. It was possible
in the days before single trial recording to monitor the performance of
the birds by means of a display of points of light representing the
cumulative number of pecks to each stimUlUS. This was routinely done
at the beginning of a run to detennine whether everything was functioning properly.
3.
Future research
For about a year, mttch of my time has been devoted to integrating
LINe into my ongoing research project in an efficient way and attempting
to come to some closure on that research. This work was carried on,
ho,,,ever, with the idea of making greater use of LINC's capabili ties,
and the apparatus constructed was designed to be as flexible as possible.
It was obvious to us, also, that ~aving LINC around meant a different
look at the problems which \'Je were handling.
Of prime interest from the point of view of this presentation 1.s
the work which we would like to do involving the latency of the response
in our color discrimination task, and the way the interest in this work
was prompted by the availability of LINC. Our experimental procedure
had consisted of daily presentations of a random sequence of wavelength
values of monochromatic light, each for t,\,O seconds with a blackout in
be tween \\'hile the color was changed.
A peck to some of the values was occasionally reinforced by the
short period during which the pigeon wa~ permitted to eat grain. Other
values were never reinforced. After many hours of this training, the
birds would respond with highest pr.obability to those values for which
they had been reinforced, and responded little if at all to other values.
Each day the bird would receive several thousand ~rials, and the data
accumulated. It occurred to us that the lumping of the data was probably
obscuring short range tempor.al effects contingent upon individual reinforcer.lents and long range effects like changing drive level over time
as the bird gradually ate enough to near satiation. Consequently we
decided to record the latency on every trial, keeping an ordered list
of stimuli and the latency of the response of any to them. We would
5-6
-6-
then be able to go back over the data, pulling out various interesting
aspects of them as we saw fit. There gradually emerged out of these
data a relationship between latency on the one hand and accuracy on the
other. The longer he ~~ited before responding, the more likely was the
bird to give a response to a reinforced value. We measured such things
as the conditional probability of a response to a given value, conditional upon a response having not been made prior, since a response
ended a trial. This measure as well as simpJ,y· the proportion of responses to reinforced stimuli for various latency categories revealed
the relationship to exist no matter how the latency values arose •. If
the pigeon \'t1aS working hard, there appeared a characteristic distribution of latencies when looked at over a long number of trials. The
performance at the longer latencies was always better than at the
earlier ones. In fact, accuracy was a monotonic function of latency.
If the pigeon was not working hard, because of fatigue, or satiation,
or because the conditions were not propitious, he tended to slow down.
Thus, fewer peclcs per hundred trials, went alonG wi ~h longer latencies,
but somewhat paradoxically, the relative accuracy increased. If one
takes the human position that the pigeon's "heart just wasn't in the
task," the performance was better for these conditions than for the
hard working conditions in which there tended to appear a large number
of short latencY.lrelatively less discriminating; respon~es- From all we
can tell at present, the performance for a given latency is invariable,
no matter the circumstances from which the latency came. This has lead
us to speculate SOme upon the mechanisms underlying such a finding. In
particular we are interested in looking at the possibility that some
sampling process is taking place over time and that the increase in accuracy is simply a function of the fact that a larger sample was taken.
In the near future we plan to explore this' possibility, starting with
'an experiment in \'Jhich the wavelength stimulus appears for a variable
length of time, say 1/4 to 1 1/2 second and is replaced for the remainder of the t\'JO second trial period by a white light. We will then be
able to compare the latency distribution across stimulus values for the
various exposure times. t'le expec t that this procedure wi 11 interfere
with the performance at the shorter exposures, but if it doesn't it will
rule out the sampling hypothesis. We also intend to train pigeons to
respond with different mean latency distributions to determine Whether
we might increase discrimination performance by forcing them to wait
longer •
. It is quite likely that we would hever have become involved in these
problems without LINC. There is entailed an an extremely large amount
of data, at one time we were sorting through some one million relevant
trials.
The analyses tended to be somewhat complex, and only because
of LINC were we tempted to do them at all. It is possible that some of
the data sorts could have been handled by a bi~ computer if we knew \",hat
we "'ere looking for. Wi t~ the LINe, however, we were· in a position to
change the analysis slightly when something seemed to be coming out, and
we spent a fe\", hundred hours at this task zeroing in on hidden meaning,
a venture which ",ould have been prohibitive on the big computer, to say
nothing of inconv.enient.
.
.
5-7
-7-
For examp:l..e, we experienced a certain amount of difficulty in ascertaining the probability of response at different latency values under
different conditions. We were interested in the conditional probability
of.response, that is,the probability of a response in a trial giv~n
that a response had not occurred up to that point. 0uI: straightforward
procedure would be to divide the number of responses in an interval by
the number of opportunities. If the animal \~s working hard, there were
relatively few opportunities at the end of the interval, but the probability was high that these would be responded to for the reinforced
stimuli. When the animal was not working steadily, however, a problem
arose. Just what should be counted as opportunities, all trials for a
particular stimulus? There was a good chance that the animal was asleep
or '\'Vas not paying any attention to the colors on the key for many trials.
It was not unusual to have several thousand trials go by without a response. The more this occurred, the more likely was the opportunity effect to be washed out, particularly if it were to occur on following
periods of relatively steady working. A number of data analysis programs
were devised to deal with this problem, the most complicated being one
which looked at all trials and counted an opportunity for a response to
a stimulus if a response occu.rred wi thin x trials in the sequence from
it, \'1here x was a small number, say, 2 to 10. This was accomplished by
having two pointers looking at the sequence of trials, one at the present
and one x in the future. If a peck occurred in the future, a bit was
put in the rightmost position of a reference work. On each trial the
'\mrd was rotated one bit to the left. If a bit occurred in the last 2x
+ 1 bits of the reference work, then a response had to occur within x
trials of the now trial. This is in effect a sliding window. This procedure is mentioned because it points up the difficulty of making a
direct comparison of the relative accuracy at different latencies under
different conditions. We hope to get some data to answer this question
in part by using LINC to train slow latency responders.
In addition, we intend to examine the force of the peck in much
the same '\'/ay that \'1e have examined latency. The equirment has been prepared and the programs for L'lmning have been constructed wi ththis in
mind; the key as been perfected. (The key developed to measure force
use~ a strain ~lage as a transducer, the amplified output of which is
fed into the A-D inputs of LINC.)
As mentioned above, we have felt the need to train latency discrimination in the pigeon and have done some work along this line. We have
set up latency criteria and made reinforcement contingent upon them. A
somewhat elaborate experiment along these lines is planned for the near
future. It involves training the pigeon to peck on a schedule ,to which
he is reinforced, say, on the average every ten pecks. The distribution
of latencies will be obtained and the schedule \o,Til1 no\'1 be cbanged so
that the reinforcement is given every time a peck is made in an interval
of the latency distribution which contains one-tenth of the bird's responses. If he continues on as before, he will receive the same number
5-8
-8-
of reinforcements, but ~hey will all· occur in responses having
roughly the same latency. The LINC will monitor, the resulting distribution, altering the \ddth of the interval in order to keep the
probability of reinforcement exactly one-tenth. Under these conditions we predict that the pigeon will not learn to alter his distribution of latencies toward the reinforced interval as reihforcement"
theory would seem to demand. We have plans to complicate this procedure still further, by changing the relative reinforcement available
for different birds. In some the probability of reinforcement would be
less, in some higher, for responses in the selected interval after the
shift. \'1e are nurturing the hypothesis that it is not reinforcement
per.se, but an increase in reinforcement relative to some level which
is the critical variable in performance change and we would like·-to
see how much this change has to be in order to produce the change.
We suspect that it may be possible to train shifts away from the reinforced interval by programming the correct reinforcement contingencies.
4.
Training
pro~ram
One can most meaningfully evaluate the program as a training dedevice in terms of his own advancement. I feel that the month at
Cambridge, while hectic, was a very \'JorthwhlJ.e experience which netted
a tremendous amount of theoretical and practical skills which have
stuck ''lith me. At present I experience no qualms about running down
Some malfunc tion in the machine (\'lhich usually turns out to be something else) and have uncovered defective diodes and transistors with
no trouble. I have designed and built supplementary circ'l,litry following the DEC logic requirements for running ancillary equipment and
feel that I !{no\'l enough that if pushed I could des ign my 0't'JTl computer
(at least the logic) at this point. Admittedly, most of these skills
came about the hard way, the Cambridge stint furnishing guidelines,
but no details, and many the long hour I sweated over difficulties that
came up when I was trying to get it to do things that weren't in the
original plans.
There were times when I thought the machine was a monkey on my
back and resented the uncompromising demands for time. After spending
into the \\'ee hours several nights in a row trying to get something to
~~rk one qevelops a perspective verging on despondency which fortunately
usually vanishes with a couple of night's sleep.I suspect that if there is one difficulty ,\'i th the program it lies
in the fact that there just was not enouG'h time. Since I was doing most
of the work myself and had other responsibilities, I have not really
been able to capitalize on the machine as I hope to do in the future.
About all I could do was to tie the machine in as well as I could into
an ongoing research program and made the most out of LINe's talents in
that context.
I think at this point I need a little bit of time to
relax and think out the possibilities provided by the machine. The last
year and a half has been different because of t~e LINC. I have felt
pressed, harried and stressed, but I wouldn't trade the experience for
anything. There has always been a certain amount of amb~~olence over
5-9
-9-
the personal investment considering the possibility of having the
machine taken away at the end of the period. Many times I felt that.
the time and energy devoted to the LINC (which had to be pried loose
from other activi ties) \'lould go lIP in smoke if the stewardship were
only temporar~ but LINC demands and gets.
As for specific aspects of the training program I think we might
have profited by a little more 'enforced interchange, but realize that
this is something which we can: and will \\"Ork out among ourselves. Certainly with such a potentially powerful device with its capabilities
literally and figuratively unrealized, any opportunity for cross talk
to jog us out of our usual modes of thought would have been of help
to us and to the program.
s.
Computer performance
Astonishment is the only way I can describe my feeling about the
performance of the LINC. Except for a few tape reading difficulties
which were adjusted, the machine has g-iven no trouble whatsoever. It
has been completely reliable, and as far as I know in 4000 hours has
never made an error which wasn't attributable to something else when
,,,e finally understood what the difficul ty was. On several occasions t
we have had the ~achine running non-stop for as many as three weeks,
with no sign of trouble. If only the mechanical equipment were as
reliable, my maintenance troubles would be over.
Wi th respect to LINC capabili ties vis a vis- the problems encountered
in operant conditioning, it would seem to be the case that researchers
reared ori relay circuits are going to have some trouble probing the outer
1i~its of capability space of the LINC.
It is at least 3 orders of magnitude faster than the present problems require (not, however if one
wishes to utilize specialied transducers or monitor several'setups.)
There have been absolutely no procedural changes in the -cesearch
program necessitated by LINC. Rather, there have been a few (and undoubtedly will be more) \,-,hich ,\'ere inspired by the presence of LINC.
These have had to do mainly with the utilization of LIN~ capabilities
to squee~ more~'information out of the experimental si tuation. Whereas
we previously had measured only the number of pecks occurring in anextended number of trials,. we soon \Alere measurin~ the latency of the
peck a!ld storing the complete sequence of latencies together with the
stimuli and reinforcement info~ation, for example.
But herein lies one possible limitation to the LINC as presently
constituted, namely its capabilities of handling large quantites of
data. Our present four-box setup seems about the limit for the kind of
intiornation we are getting. The control program requires almost all of
the' first half of memory; a quarter per box of the second balf is allotted for the acclnnulation of data. Since our schedules are stored in
half-word format, we can get 100 trials in a q'uarte-c, counting 50 words
5-10
-10-
for schedule, and 100 each for latency and force measures. Thus data
is stored every 100 trials and then the process proceeds with a new
schedule, etc. If we are to continue as at present, it is obvious that
four boxes about exhausts the available storage capa.ci ty. The program
itself is capable of dealing with perhaps another dozen boxes on the
same problem, but there is simply no room for the storage.
6.
Specifications!.2!. ideal laboratory computer
My feeling is that the LINC is tr.e ideal laboratory computer. Certainly for operant conditioning setups, it has potential that \~ have
hardly begun to realize. There have been times when we have wished that
it was somewhat faster, had a somewhat larger memory, or a bigger world.
size. These difficulties were overcome quite readily, ho\~ver, with
but little sacrifice in efficiency. I think that one simply has to accept it for what it is. It \'las not intended to supplant the big computational machines. If one has a problem requiring facilities 'of that
kind, he is probably in a position to use'them instead of the LINC.
Herein lies one weakness of the LINe system I feel. It is designed as
,if it were a complete unit, but it is not. If one has a good bit of
data gathered on the machine and wishes to do some fairly complex analyses of it he has no recourse but resort to a ~ig machine but has the
problem of getting data from LINC into a large-computer-compatible form.
As standard equipment, LINC for completeness sake, (perhaps to emphasize
the fact that it is what it is) should have some form of output \'Ihich
is capable of easily being utilized on other machines. Admittedly it
is possible to do this, and I probably ,n11 providing I can scrape up
enough money.to convert to somethin like this.
7.
Publications based
~ ~
of LINC
Decision processes in the Pigeon:
performance.' In preparation.
8.
A model for color-discrimination
Latency and accuracy as variables in color discrimination.
preparation.
In
Generalization gradients and the discrete trial situation.
pre·paration.
In
Statement on future disposition of
~
By all means I wish to keep LINC in my possession as an integral
part of my research efforts. I would hope that Some procedure might be
instituted which would insure that LING would be free to move with me
wherever I go in the future. This hope is predicated upon an investment of a fairly concentrated few thousand hours including a year's
vacation period on my part in the LINe project.
5-11
Department of Physiology
The Johns Hopkins University, School of Medicine
G. F. Poggio, M.D.
V. B. Mountcastle, M.D.
REPORT ON LINC EVALUATION
and
PROPOSAL TO RETAIN THE LINC
6-1
TABLE OF CONTENTS
A. Letter Transmitting Evaluation and Proposal.
B. Proposed Research.
C. Evaluation Report.
1. Development of Input/Output Faci1i~ies.
a. Hardware.
b. Utility Programs.
2. Research Application of the LINC.
3. Computer Performance.
4. Comments on the LINC Evaluation Program.
6-2
A - LETTER TRANSIvfiTTING EVALUATION AND PROPOSAL
6-3
THE .JOHNS HOPKINS UNIVERSITY
SCHOOL OF MEDICINE
'7I1B N. WOLP'e: STREET
.ALTIMORIE, MAIIlVLAND • 2120B
DEPARTMENT DP' PHYSIOLOOY
March 13, 1965
Dr. T. T. Sandel
LINC Evaluation Board
Computer Research Laboratory
Washington University
700 South Euclid Avenue
St.Louis, Missouri 63110
Dear Dr. Sandel:
With this letter we wish to transmit to you the auached report on the use
of the Linc computer in our Neurophysiological laboratories. It is our intention
that this repOl1: sel"Ve as a formal proposal to you, and the National Institutes
of Health, that the computer remain in our laboratories after termination of
the Linc evaluation period. We make this proposal because we believe, as we
hope the report will document, that the availability of Linc and the associated
circuitry will allow us to undertake Neurophysiological investigations which
would be impossible, or at the best extraordinarily difficult, without it. We
emphasize that the capability provided by Linc is much more than that of
an elegant piece of hardware, which allows one to do a tedious job with ease;
for it is evident to us now that the integration of Linc into our experimental
arrangement influences the design and execution of experiments to a considerable
degree, and promotes the planning of an experime~tal investigation at a level
more sophisticated by a step-function than was possible before Linc became
available to us.
As you kno,v, Dr. Gerhard Werner joined us in our initial proposal that
a Linc be placed in our laboratories for evaluation. Dr. Werner has recently
accepted the pOSition of Professor of Pharmacology in the University of Pittsburg,
and will be leaving this department on June 30, '1965. We suggest, therefore,
that should our proposal meet with favor the award be made to Gian F. Poggio, M. D.
as principal investigator, and Vernon B. Mountcastle, M.D., as co-investigator.
The presence and evaluation of Linc has occurred at a time when extensive
renovations have beencarried out in the Department of Physiology. We have
taken this opportunity to make physical arrangements for the use of Linc by
either of our two laboratories. Linc and its associated equipment now occupy
a large "data-reduction" room, which is flanked on one side by Dr. Mountcastlets
laboratory in which studies of the neural mechanisms in somesthesis are under
6-4
3
Dr. T •T • Sandel
Page 2
March 13, 1965
way. On the other side we have created a new laboratory for Dr. Poggiots
studies on the neural mechanisms in vision. We thus believe we have made
physical arrangements for the most efficient use of our system.
We would like to emphasize the important role Linc has played in our
training program in Neurophysiology. During the past year two post..doctoral
fellows, DJ;s. Frank H. Baker and William Talbot, have become expert in
the use of Linc, and have in fact contributed heavily to the development of our
system and to the programs which are described in the body of our repo11:.
In addition, Dr. Robert DeVoe, a member of the staff of the department, has
purchased recording equipment and arranged to collect his own experimental
data in a form suitable for Linc analysis. We believe that the use of Line
can only increase, and its significance for the research programs in Neurophysiology is, we believe, recognized by all.
This letter is meant to be the introduction to our final report on the
evaluation of the Linc and our formal proposal that Linc remain in our
laboratories. There then follow, in the first part of the body of the report,
brief descriptions of the research programs underway and those proposed in
which Linc will be of conSiderable importance. Next we describe the inputoutput facilities we have developed, then the utility programs. Then we give
some specific examples of the application of Linc to actual experimental
problems. Finally we give our evaluation of computer performance and of the
Linc Evaluation Program itself.
In clOSing this letter 'we 'wish to express to you and to your colleagues"
of the Linc Evaluation Group our great appreciation for the help you have
given us during this evaluation. You have been untiring in aSSistance, and
ready with skill and foresight when called upon. Above all you have striven
successfully to understand our point of View, that any inst1'Umental system of
whatever complexity finds its greatest usefulness when designed to fit the
needs of the experiment, and not independently of it.
Sincerely,
~Bhu'
Gian F. Poggio, M.D.
GFPjVBM/mh
6-5
,B.
-.
,
PROPOSEDRES}!:ARCH
'Ail of the neurophysiological r~search in which we are now
engaged, and which we propose for the immediate future, involves the
use of the method of single unit analysis in unanesthetized monkeys.
The philosophical basis from which we work is simple, and old. It is
the premise that if the objective events evoked by sensory stimuli in
the brains, of experime'ntal animals can be observed under as normal
conditions as possible, they may then be compared with the sensory".
perfo:r;mance of human beings in response to sensory stimuli. The
expectation is that increasing knowledge of the former will lead to a
steady closure of the field of sensory neurophysiology and psychophysics,
and thus lead to further under standing of the neural mechanisms of sensation and perception.
The method of single unit analysis, which we employ, is a
powerful one, and its success in recent years in elucidating certain
aspects of CNS function has led to its widespread use. There is, however, to our knowledge no other method which for its success depends
so critically upon the way in which it is used, and upon the conditions
of the experiment. This is particularly true if one wi she s to study the
dynamic, the time-dependent aspects of neural activity. It is less so
if one's experimental objective is to study the geographic, or static
properties of central neurons: such things as receptive fields, modalities
of driving, etc. But ,if one wi she s to go further, to study for example the
spontaneous activity of central,.neuro'ns and' the way in which it is complexed
with evoked activity, or to tackle the probiern of whether information
translnission'in the centr~l nerv~u~. sytem depends to any degree upon
impulse-interval modulation, then the condition of the animal at the time
of the recordings become s of great importance.
It became clear several year s ago that studies of these timedependent aspects of the action of central neurons is useless in animals
which are generally anesthetized. The ,effect of the anesthetic drugs influences most profoundly the natural rhythm of central neuron discharge,
whether that occurring in the absence of or in response to specific sensory
stimuli. We have ther'efore recently devised an implanted double-chamber
method of recording. With it microelectrodes can be passed into the brain
of the unanesthetized monkey, who is under neuromuscular block and
artificially respired. The head is positioned in Horsley-Clarke space by
fixation of the implanted outer chamber, and no painful stimuli are delivered
to the animal. Fluid balance, CO 2 concentration, body temperature, etc ••
are regulated within normal limits .. and the animals remain awake in what
we believe is a cotn:£ortable condition for many hour s.
We believe that no one would deny that observations upon a single
neuron of the central nervous system are of more than anecdotal value.
What is required for the succe s sful application of the method of single unit
analysis is that for each nuclear grouping a very large number of single
cells be studied, ad seriatim .. under as identical conditions as possible.
With these data in hand, it is then possible to reconstruct the population
events;occu~ring in the nuclear region under the imposed conditions. Until
6-6
methods are developed for observing many cells simultaneously, this
remains the only valid application of the method.
If ones aim is to make some quantitative study of the relation
between sensory stimuli and the neural activity they evoke, a precise
control of the stimulus is required. For work ill the somatic' system we
have recently developed a stimulator which allows controlled indentation
of the skin (controlled to within about 21-1) for different distances, at
different rates, repetitions, durations, etc. For experiments on the
visual system, Dr. Poggio and his colleagues have developed a multiply.
beamed visual stimulator. This allows positioning of a total of four
individually controlled beams, two in each eye. For each beam, independently, there is control of position, size (down to l00l-L spots),
shape, contour, intensity, and color, as well as temporal pattern, of the
stimuli.
Our plans for the study of these two sensory systems are nearly
identical, so we shall consider them together.
1.
Studies of the spontaneous activity of thalamic and cortical neurons.
, Studies of this type have already been made of.neurons of the
ventrobasal nuclear complex. The temporal ordering of impulse s has
been described, and both the slow and rapid oscillations in the likelihood
of discharge characterized (Werner and Mountcastle, J. Neurophysiol.
26: 958-977, 1963; Poggio and Viernstein, J. Neurophysiol. 27: 517-545,
1964). Dr. Poggio intends to extend these studies to the latml geniculate
nucleus, and in particular to study the influence s upon thi s "internal
structure of the neural message" of variations in the intensity and color
of weak background illumination. Studies of the spontaneous activity of
cortical neurons» of both systems, remain to be done, though SOlUe
preliminary studies of this type have been made in other laboratories.
Our long--term aim in these studies is to be able to completely characterize
spontaneous activity, to correlate it with the level of awareness of the
experimental animals, and with the on- going spontaneous slow-wave
activity of the thalamus and the cortex.
2.
Interaction of spontaneous and evoked activity.
Our aim in studies of this type is to discover the laws governing
the interaction between spontneous and evoked activity. Evidence suggests
that the spontaneous a'ctivity itself depends upon a complex input from spontaneously discharging sensory receptors and input from those "intrinsic'I"
eNS systems concerned with lUaintaining levels of excitability. The question
iswhether the additional input evoked by a stimulus is treated additively or
multiplicatively. Some observations on thalamic cells suggest the former,
and this fits with the fact that precise estimations of stimulus magnitudes,
and intensity discriminations, are made by humans over a wide range of
awareness, from drowsine ss to agitation - which is interpreted as meaning
over a wide range of spontaneous activity levels. However, for other cells
this is less certain, and the possibility that mUltiplicative interaction occurs
is open. We hope to determine which is most common, and the circumstances
governing the occurrence of one or the:' other.
3.
The guantitative relations between sensory: stimulus and central neurl;.,:e7'0nse.
The aim here is to order stimuli along a scale of the physical
dimensions of the stimulus, the response in some reliable way .. by frequency. interval sequence. etc., and to discover the laws governing the
relation. This will then be compared with that controlling the human sensory performance. The result should allow some inferences about the
cascaded neuronal transformations which intervene between the input
(volleys of impulses in first-order afferents) and the response (subjective
estimate of stimulus magnitudes, etc.). For the somatic system studies
of the stimulus-response relation for cutaneous (Werner and Mountcastle,
J. Neurophysiol. 28: 1965) and joint primary afferents, and for third-order
joint neur9ns of the thalamus (Mountcastle, Poggio and Werner, J. Neurophysiol.1.§: 807 ... 834, 1963) have been completed. Studies of the thalamic
neurons responsive to cutaneous stimulation are under way, and those of
cortical neurons of the two types are planned for the coming year. Explorations of the stimulus-response relation in the visual system have
begun, and promise to be of particular interest, for there the stimulus
can be controlled most 'preci sely.
4.
Interaction of .'excitation and inhibition in the nervous system.
While studies of this subject have been made earlier in the somatic
afferent system, in a geographic and qualitative manner (Mountc~stle and
Powell, Bull. Johns Hopkins Hosp. 105: 201-232, 1959), it remains to study
this interaction under precisely controlled sti:mulus conditions in the somatic
and in the visual system. In the latter not only can the interaction depending
upon position and intensity be examined, but that between opponent color s
as well. The implications for the general phenomenon as well as for the
central neural mechanisms in color vision are obvious.
All these studies involve the handling and measurement of very
large amounts of data. In a typical successful experiment it is not uncommon
to record the electrical signs of 100, 000 to 200, 000 nerve impulses. We
require measurement of all the se impulse intervals, as well as an extended
series of subsequent statistical analysis. We wish to use pre-programmed
control of stimulus sequence patterns, and simple on-line analyses to guide
the subsequent course of experiments. We believe that we have achieved
this capability by incorporating the Linc into our present data analysis system,
and by producing compatible IBM tape with it when used in the data reduction
mode. That the system is of central importance for our entire experimental
program is certain.
·
G.
1.
EVALUATION REPORT
DEVELOPMENT OF INPUTOUTPUT F AGILITIES
The installation of the Linc:in the Department of Physiology of
the Johns Hopkins University School of Medicine was designed to allow for
on-line analyses from each of two adjacent neurophysiological laboratories,
and to make the instrument available for off-line computation to any member
of the department.
The ongoing research in those laboratories which utilize the Linc
is concerned with the study of some quantitative aspects of neural a.ctivity
"6=8'
in the central nervous system as described above. The experimental data
upon which these studies are conducted consist of (a) impulse activity from
single neurons of the central nervous system recorded with microelectrodes,
or from peripheral nerve fiber isolated by microdissection. Observations
are made on the ongoing activity ("spontaneous activity") as well as on the
activity evoked by physiological stimulation of peripheral receptor s, and
(b) activity from a population of neurons recorded with large electrodes
(electroencephalogr am, electroretinogram, II evoke d" re sponse s).
In order to utilize the Linc for our research purposes we found
it necessary to design and build suitable input-output interfaces and to write
a series of programs for experimental control and data reduction.
a.
HARDWARE
Introduction
Input-output facilities were built to perform the following
operations:
1. Input neural activity of the nature described above, together
with code signals used to define periods of stimulation, parameter s
of the stimulus, etc. Data processing and analyses may be perform.ed with the Linc either on-line with the experiment or at a
later time from records collected on analog magnetic tape.
2. Output signals for the control of external equipment and for
on-line selection of parameters of the stimulus delivered to the
animal in the cour se of the experiment.
3.
Input-output with Teletypewriter.
4.
Write-read IBM compatible digital tape.
5.
Output to high speed paper tape perforator.
The additional hardware necessary for these operations was constructed in three plug-in~.units (P. I. U. ) for the data terminal box, arranged
as follows:
INPUT P. I. U.
Input neurai activity and code signals
Input-output connections with Teletype M33TA
Output Relays
Exte:vnal Clock
DATAMEC P.l. U.
Connections with Datamec D 2020 Magnetic
Tape Unit.
PERFORATOR P. I. U.
Connection with paper tape punch set Teletype
BRPE.
Datamec P.1. U. and Perfo~ator P.l. U. can neither be used alone
nor simultaneously, but only in conjUnction with Input P. I. U.
9
6- .
In the following page is given a diagram.m.atic representation of .J?
.
StImulus
t
Preparation
r
I
-I
Response
the general arrangement of this system.
To obtain greater flexibility the system is designed to make the
input of data continuously possible (operation 1 above), while anyone of
three other operation (2,3,4) may be simultaneously performed after being
selected under program control. Each of these three operations utilizes
the Linc BUFFER REI;JAY. .J~(EGISTER outputs at the data terminal box.
These outputs are used:
1. to drive the external relays located in the INPUT P. I. U.
(BR _ 5) for operation 2.
O
2. to provide output code signals to the Teletypewriter (BRO)
for oper ation 3.
3. to ~rovide control levels to the digital tape unit (BR _ ) for
O 4
operatlon 4.
For each operation, logical gates are employed to connect the
relay register outputs with the external equipment. The selection is made
by executing
, an OPRn instruction. The hardware and the circuitry used
are detaildd in drawing #2 and #8. The following table summarize s the
operational arrangement of these logical gates.
Linc Relay Register Output Selection
R.egis~er'
outputs
Following OPR 4
(Relays Select)
Relay
Relay
Relay
Relay
Relay
Relay
0
1
2
3
4
5
Connections
Following OPR 5
Following OPR 6
(Teletype Select)
(Datamec Select)
Teletype write
n. c.
n. c.
n. c.
n. c.
n. c.
Forward drive
Rever se drive
l=enabled
Rewind drive
O=disabled
Write permit
Read threshold (O=low; l=high)
Parity select
(O=odd ; l=even),
Since no RELAY REGISTER outputs are used in paper tape perforation, when Input and Perforator P. I. U. s are us-ed together OPR 4 and OPR 5
operate as shown above while OPR 6 simply disables both the output relays and the
teletypewriter.
In the figure shown on the following page is an example of the
operation of the logical gates. At OPR 4. BOPR 2. 2 relay register outputs are
connected to external relays and disconnected from Datamec control. The opposite
occur s at OPR 6 • BOPR 2. 2.
Detailed drawings of the circuitry employed in the three plug-inuni ts may be found following a brief description of the main aspects of this circuitry.
6-11
OPR 4
Relay- output
OPR e;
Dat41!J.OOC
control
INPUT P .1. U.
The partially wired plug.. in"unit given to us with the original LINC
contained 8 input channels for A/D convel:sion and 6 output relays controlled by
the BUFFER RELAY REGI STERo As noted above we have introduced logical
gates between the relay register and the output relays. In addition we have made
provision for three external level input channels, for input-output with the
Teletypewriter and also added an external clock for program timing. No modifications were made to the Analog Input.
Level Input
At each of three level inputs signals in the range ± 10 V are fed
into a difference amplifier (DEC 1501 level standardizer) the output of which
is - 3V. 'whenever the input level is more positive than a fixed reference voltage.
The outputs of the three amplifiers are connected as follows: (see drawing #1)
0
i. each of them directl y to an XL line (XL 4, XL 5, XL 6)
and output 2 also to TN 10
0
lio the outputs for input 1 and 2 to flip flops (DEC 4209) in such
01
a way that one flip f10p is set by positive going level changes
at input 1 (event sense flip,,·t1op 1) and the other by either
positive or negative going changes at input 2 (event sense flipflop 2) The outputs of these flipHflops are connected through
an OR-gate to XL 3 (event sense line). The output of event
sense flipwflop 1 is also connected to TN 11
a
Q
It
Level output of OPR 3 provides TNEL causing the status of event
sense flip"lflop 1 and the condition of input 2 to be transferred into the two
leftmost bits of the Accunlulator After transfer has occurred the same
instruction clears both event sense flip-flops (OPR 3 + BOPR 2.2 thrau gh a
capacitor diode gate)
0
0
.
The logical design of the level input was dictated in large part by
our \vish to perform serial interval data processing at high temporal resolution.
The bu f ~ering flip"'flops, "hold" an input event-until it can be detected by the
program, the ORugate permits testing two inputs with a single SXL instruction,
and the OPR 3 connections allow the rapid j am transfer of a 2-bit coding of the
nature of the event detected.
A detailed description of the application of this logical design is
given under Utility Program #4, Serial Interval Reduction Program.
10
Teletypewriter
The Teletype input-output connections were made as recommended
in CDO Information Bulletin #6, May 26, 1964, with the addition of the logical
gates on the relay register output as previously described. (see drawing #2).
As an output device the Teletypewriter has performed very well.
We have experienced, however, considerable difficulties in the use of the
instrument for input to the Linc. Firstly, the 200 msec turn ... around time
needed to obtain a simultaneous hard copy of the input was found inconveniently
longo Secondly, and more important, the mechanically generated serial code
is sufficiently noisy so that input read errrns often occur A much more
satisfactory arrangement is to use the Linc Keyboard as the input device 'with
direct type .. out on the Teletypeo (see description of Keyboard to Teletype Routine
under Utility Programs) •
0
External Clock and QKRESTART control.
The output of a DEC Variable Clock 4401 is connected to a pulse
amplifier (4606) through the gated pulse input. This input is enabled by the
-3V level of the OPR 1 instruction (see drawing #3) • The output of the pulse
amplifier furnishes QKRESTART.
0
The use of the clock is primarily as a simple means to adjust
program loops to a standard but variable duration. It was found particularly
useful in slow rate sampling and counting operations. The clock is consulted
with an OPR i 1 instruction.
/
QKREST ART pulses are also generated in the Input P .1. U. in response
to signals from the Datamec P.loU. (drawing #6) or from the Perforato P.I.U.
(drawing #10)
0
The system we designed and built to write-read"'cheek IBM compatible
digital tape is very similar to that developed by Dr. Joseph Eo Hind at the University
of Wisconsin. Since a complete description of that system was sent to the
participants in the Line Evaluation Program (CDO Information Bulletin #7), May 7,
1964), a short description of our system will suffice.
The main· differences between the two systems sten1 from our use of
Datamec cards to generate and check the lateral parity bit, instead of a single
circuit built in the P .1. U. as in the Wisconsin system. This was done so that the
read after write mode of operation could be utilized. Details of the circuitry used
are given in drawings #3-#9.
11
Write: Bits 0-5 of the accumulator are written on tape tracks 1... 6.
Write clock command is given by the negative going transition of the level output
of OPR 13. Due to the time requirements of the Datamec electronics;, the bit
configuration of the character to be written must remain in the accumulator for
at least 26 microseconds after the onset of that negative level.
Read: SixMbit characters written on tape are read into bits 0 .. 6 of
the accumulator by executing anyone of the read instructions (OPR 10, OPR 11,
OPR 12 each of which provides transfer enabling level SNEL) The output of the
laterial parity check character is connected to bit 11 of the accumulator (SN 11)
and also to a flip ..flop (parity error flip ...flop)(drawing #6) the status of which is
tested with an XL line (XL 13). Lateral parity check bit is thus read into the
accumulator together with the six-bit character to which it refers. Test of the
parity error flip-flop will detect the occurrence of at least one parity error in any
given series of characters (one data record for instance) The parity error flip ..
flop can be cleared before beginning to read a record by an OPR6 instruction.
0
0
Datamec tape unit status is not tested with individual XL lines as in the
Wisconsin system~ but the five sense lines from the tape unit are connected to
bits 6-10 of the accumulator (SN6 u SNI0) and the status of these lines is transferred
into the accumulator with any READ instruction. Transport status is checked by
testing the configuration of the appropriate bits of the accumulator. In addition the
END OF TAPE sense line is connected to XL 10 in order to simplify the tape test
program subroutine
0
PERFORATOR P.loU.
This unit contains the logical hardware that allows transfer of the
configuration of the 8 rightmost bits of the accumulator through power relays to'
a Teletype BRPE High Speed Tape Punch (drawings # 10 and # 11).
The perforator emits a synchronization pulse every 9 milliseconds.
This pulse may be gated by the OPR 12 output line into a series of logical ~andzw
gates so that the contents of the accumulator ae transferred to the punch during
the 3 msec following the synchronization pulse.. Because it is necessary for the
desired character configuration to remain in the accumulator for the full 3 millseconds following the synchronization pulse from the perforator, programlning
practice is to load the accumulator with the desired configuration and then to
execute an OPRi12 instruction. The Linc then pauses until 3 milliseconds after
the next perforator synchronization pulse at which time a QK Restart pulse is
generated (drawing #3) and control returned to the program.
'. 1k.·.:J
\)-
12
Summary of LINC Input/Output Connections
OPR LINES
OPR 0
OPR 1
OPR2
OPR 3
OPR 4
OPR 5
OPR 6
OPR 7
OPR10
OPR11
OPR12
OPR13
OPR14
OPR15
OPR16
OPR17
- LINC output to teletype
.. External clock timing
-N.C.
.. PrOVides TNEL (to read into bit 11 of A the state of
event sense flip"flop 1 to read into bit 10 of A the
condition of XL 5)
Clears event sense flip ..flops.
- RELAYS select
- TELETYPE select
- DAT AMEC select; clear lateral parity error flip-flop.
... Read Reset
- Provides SNEL
.. Provides SNE L for reading digital tape.
- Provides SNE L
Paper tape perforator control
Write clock pulse
.. Write reset
.. KBD input control
- Right Switches into A
Retrigger 2.. 1/2 character delay'
.. Left switches into A
WI
XL LINES
XL 0 .. Teletype input to LINC
XL 1·· N.C.
XL 2 .. N.C.
XL 3 ... External event sense
XL 4 .. External input #1 ("3V = yes; OV = no)
XL 5 .. External input #2 (.. 3V = yes; OV = no)
XL 6 .. External input #3 (.. 3V = yes; OV = no)
XL 7 .. N.C.
XL10 - Eng of tape sense (.. 3V = end; OV = not end)
XLII ... 1"'1/4 character delay (-3V = expired; OV = not expired)
XL12 - 2... 1/2 character delay (-3V = expired; OV = not expired)
XL13 - Lateral parity error sense (-3V = Error; OV = no error)
TAPE STATUS AND PARITY CHECK (Left half of accumulator following OPR10, OPR11,
OPR12)
A6 ., (octal 100) .. Auto ... local sense - 1 = auto; 0 = local
A7 .. (octal 200) u Density sense
1 = high; 0 = low
AS .. (octal 400) .. Write enabled sense" 1 = not enabled; 0 = enabled
A9 (octal 1000) .. Rewind sense ... 1 = not re:winding; 0 = rewinding
A10 .. (octal 2000) .. End of tape sanse .. 1 :::: end of tape; 0 = not end
All- (octal 4000) ... Lateral parity error" 1 = error; 0 = no error
til
lilt
6-1&
13
DRAWINGS OF PLUG..IN ... UNIT CIRCUITRY
(Note: in all drawings the circled numbers indicate the pins of the P .1. U .. -DTB
connectors or of the connectoi·s to external equipment. The circled
letters refer to the pins of the specified DEC card.)
6-17
14
b. UTILITY PROGRAMS
1. General purpose display subroutine (KIN)
2. Keyboard to Teletypewriter I/O Routines (KBDTTY)
3. Program manuscript type-out (MANOUT)
4. Digiral tape program (DATAMEC)
5. Serial Interval Reduction Program (SERINT)
6-18
15
1. Subroutine KIN
Kin is a general purpose subroutine for keyboard input with display
of titles and numerical parameters; alternatively, it can be used for the
display of constants and parameters from the main program. In either
case, the main program which calls KIN in must contain a series of half
word keyboard representations of letter occupying two lines of title, and
the number of labels (n) which are to be filled by numerical parameters.
The main features of the routine appear in the accompanying flow chart.
a) Kin for display of keyboard input: Kin displays two lines of title
(obligatol.-Y) which can be followed by a list of labels (such as for instance:
Sl =
; etc). Numerical values for the labels can be entered from the
keyboard either in octal or in decimal numbers. They must not exceed
double word length and. must be entered in the order in wbCh the labels appear
on the oscilloscope screen. The typing of each value is terminated by EOL
if it is an octal number; a "0" followed by EOL is typed after decimal numbers.
The numbers are displayed as they are entered. The ·last line entered may
be deleted by typing DEL. When the number of numerical values entered
equals to n and when the last E OL is typed, Kin converts the numerical
values erered into binal-Y numbers and leaves them in an array of double
length numbers in KIN, starting with location 1350. These numbers can be
utilized by the main program. Control is then automatically returned to
the main program two locations after the point in the main program from which
Kin was entered
KIN + 2).
aMP
b) Kin for display of numerical values from the main program: Numerical
values are read from the main program and Be inserted into the appropriate
locations of the t~t1e-Iabel display.
6-19
SU'BROUTINE KIN
ENTRY FROM
MAIN
PROGRAM
DECIMAL
CONVERT
NUMBER
PLACE IN
INPT. ARRAY
a
DISPLAY
LABEL,S
INPUT
ARRAYS
CLEAR INPUT
S OUTPUT .........-..<
ARRAYS
DELETE
LAST LINE
ENTERED
YES
OCTAL'
CONVERT
NUMBER
PLACE IN
INP~T~ ARRAY
a
ENTER
CHARACTER
INTO INPUT
ARRAY
TALLY
NUMBER
OF LINES
CONVERT
INPUT ARRAY
PUT IN
OUTPUT ARRAY
a
6-20
.LV
2. KBDTTY·-Keyboard to teletype I/O routines.
This is a package of subroutines designed to make the keyboard and
teletypewriter an effective pair of devices for the initialization of programs and
the display of results. The following are included:
Teletype alphanumeric output routine: This routine takes a 6-bit keyboard
code character from the accumulator, translates it to teletype code and prints it
out. Upper-case characters are printed by entering this subroutine consecutively
with 23 and the character code. The routine is suitable for printing out characters
as they are typed in through the keyboard, or for printing lists of characters ,stored
in memory.
Octal loader: Thispermits keyboard loading of positive or negative
octal numbers without leading zeros. Digits are typed out on the teletype as they
are entered. When a space or EOL is struck control returns to the calling program
with the octal number in the accumulator. Only octal digits, spaces, and EOLs are
accepted by this routine. The type-in may be limited to 3 or 4 characters. Numbers
may be deleted at any time before the terminator is struck.
Decimal loader: This is like the octal loader except that a decimal-to-octal
conversion (single precision) is performed after the number is terminated and the
octal number loaded into the accumulator. Not more than four decimal digits
will be accepted. If the octal conversion exceeds 77778 an error indication is
typed out.
Octal typeout from the accumulator with leading zeros replaced by spaces or
with leading zeros printed. All nunlhers are treated as positive octal numbers. The
printout can be limited to the rightmost 2, 3, or 4 digits.
Decimal typeout from accumulator: This is like the octal typeout except that
an octal to decimal conversion is performed before the typeout. Two routines are
available, one of which converts octal numbers greater than 4000 8 to positive
decimal equivalents, the other to negative decimal equivalents. When used the minus
sign is moved next to the first printed digit
0
6-21
17
3. MANOUT ... ·Program manuscript typeout
This program uses the teletypewriter to print out a copy of Lap manuscripts.
Initialization from the Linc keyboard provides options for printing a 20 character
name on each 100 line page, for printing the binary conversion of each manuscript
line, for typing a ~ist of tags and equalities, and for restricting the type-out to
a specified span of lines. To simplify annotation of the manuscript the program
will pause at the end of the current line whenever a keyboard key is struck.
Thereafter KBD input results in teletypewriter output until a meta or case-G is
struck. The former restarts manuscript type--out after a line-feed and carriage
return; the latter causes return to Guide.
6-22
18
4. DATAMEC Program
This is a general purpose program to write, read and check IBM
compatible digital tape, low density, odd or even parity. For specific
purposes other digital tape programs are used in the laboratory (see,
for instanae, SERINT.)
Datamec program occupies two quarter memory (QN 2+3) and consists
of several subroutines and subprograms. The entire Datamec program uses
761 octal memory locations and index registers 0,1,2,3,4,5,6, 7, 10. A
brief descritpion of the program is given below, and its manuscript is appended.
Tape format used with Datamec program
Linc bit AO
=
" bit Al
=
bit A2
=
bit A3
=
bit A4
=
bit A
=
5
Lateral Parity bit =
II
II
II
II
Track 1 Datamec
track 2
"
track 3
"
track 4
"
track 5
"
track 6
"
track 7
"
unit
"
"
"
II
"
(ODD for binary; EVEN for BCD)
"
Inter character spacing, low density = 0.00544" (
184 CPI, nominal 200 CPI)
Inter character time = 120 Ils (fifteen, 8}.1s Linc cycle time)
Longitudinal check character (LRCC) gap = .0216"
Inter record gap (IRG) : .
write after write = :MIN • 720", NOM. 810", MAX .900"
write after read = MIN .750", NOM. 790", MAX .840"
write after backspace = IRG increases MIN 0, NOM .050", MAX .100"
End of File Mark (EOF) = 178 even parity + LRCC
Write gap with tape at load point = 3.65" 0.05" to first record.
Subroutines and subprograms
All subroutines and subprograms other than delay subroutines are entered
with tape at rest. With the exception of Re\:vlnd return to calling program will be
with tape motion halted.
a. Programmed time delays:
Initial Tag 1A = for 2. 7 msec delay
" I B = " 5.0 msec
"
"
"lC = " 5.9 msec "
"lD = "11.7 msec"
"
"IE = "17.0 msec"
"IF = "66.4 meec"
"
II
II
(free tape travel .123 ")
225 ")
("
"
"
• 265")
("
"
"
.525")
("
"
"
• 764")
("
"
" 3 .000")
("
"
"
•
b. Initial advance: Initial Tag 2A
Genera~~s initial write gap (erase) with tape at load point.
c. Backspace: Initial Tag 3G
Backspace one record of any length. When motion stops the tape is
properly positioned for any other operation to follow.
6-23
19
d. Rewind: Initial Tag 2R
Initiate reWind of tape to loadi point. Control is returned to calling program
about 17 msec after rewind command is given. Return configuration: c(A) = 20
e. Write record: Initial Tag 2B
This subprogram writes one record of 2000 8 six-bit characters taken
successively from upper and lower half of 7778 twelve .. bit Linc word stored in
QN 6 and 7. During writing process the running Longitudinal Check Sum is
computed using the character configuration loaded in the Accumulator. After
writing the LRCC the program waits (read after write mode) for the end of record
(first 2.5 delay expiration), retriggers 2.5 delay (OPR 16), then checks status of
lateral parity error flip-flop (XL 13). If no error the program waits for LRCC or
for expiration of second 2.5 delay and then checks longitudinal sum by testing
Computed Long. Sum = LRCC.
If no errors are detected tape motion is halted in the required way, for
generation of the after-write portion of the IRG (*), and control returned to
calling program. Return configuration: Accumulator = O.
If lateral parity error is present orlongitudinal sum does not check, motion
is stopped and the tape backspace-erased (WRITE PERMIT left on) the length
of the record just written. The 2000 8 stored characters are then re-written
over the same portion of tape. Re"writing may be repeated up to six times. If
after these six trials an error is still detected, the tape is backspaced once more
and then erased forward for about 3.0". Writing of the same record is then tried
again. The entire procedure may be repeated up to three times. If error is
still present, at the end of the third 3.0" advance erase the tape motion is halted
as above, and control returned to calling program. Return configuration: Accumulator
= 1.
An outline of this program is given in flow-chart #1.
(* - Note: Removal of WRITE PER:MIT generates a spurious character on the
tape stationed over write head. To eliminate it , at the end of a write
operation WRITE PERrvTIT is removed at the time STOP command is given. Once
the tape has stopped it is backspaced the length of one START+STOP motion crOG).
The spurious character is thus positioned in front of the write head and the
successive write operation will erase it before writing the first valid character.
f. Write End of File: Initial Tag 2K
This program writes an EOF mark (17 8 even parity + LRCC), checks it
in read after write mode, and generates the afterwwrite portion of the IRG
(STOP+JOG). Erros detection, automatic re-writing and return configuration
are similar to those used in the Write Record subprogram and they are detailed
in flo-chart #2.
g. Read Forward: Initial Tag 3A
This subprogram reads and identifies: (a) one record, defined as any
two consecutive characters separated by less than 2.5 intercharacter spaces.
6-24
20'
1 character record is considered as a read error. (b) one EnQ of File mark.
Characte);~ are read into bit 0 .. 5 of the Accumulator and store successively
in left and right 'palf of core registers QN 6+7. Record canIlot be longer than
2000 8 chal.-acters. If longer records are to be read set initial storing address
(location Tag 38+1) to beginning of QN 5 (for records up to 3000 8 characters) or
QN 4 (to 4000 8 characters).
(Note: With each character, the tape transport status senses and lateral
parity check bit are read in bit 6-11 of the Accumulator. See description of
Datamec P.I.U.).
The program tests for lateral parity error, missing characters, and
longitudinal check sum. If no error, the tape is stopped within the IRG
at a position that then allows execution of any other tape operation to follow,
and then control returned to calling program.
Return configuration: (a) Record: Accumulator = 0; Register 1 = no.
of characters stored (octal); unused portion of QN 6+7, if any, filled with
negative zeros. (b) EOF: Accumulator = 1; Register 1 = 3000 8
If any error is detected the tape is stopped, backspaced the length of
the record just read and the reading operation tried again. This procedure
is repeated up to 128 times, alternately with low and high read threshold. If
at the end of this :mries of trials an error is still present, the tape is stopped
within the IRG as above and control returned to the calling programo Return
configuration: for permanent read error: (a) Record: if lateral parity error,
Accumulator = 4002; if missing character error, Accumulator:;: 2002; if
lOngitudinal check character error, Accumulator = 1002. (b) EOF: Accumulator
= 4002, Register 1 = 3000.
An outline of this subprogram is given in flow-chart #3.
h. Skip EOF FOl.-wards: Initial Tag 3L
Backwards:lnitial Tag 3K
This subprogram will skip any given number of End of File marks
then enter READ FORWARD subprogram to check the last EOF mark identified and
to position the tape past it, to read or write the first record of the next file.
Return to calling program will be from READ FORWARD and therefore return
configuration will be the same as that of the latter subprogram. In addition
Register 10 will contain the number of records that constitute the last file
skipped. The complement of the number of Ena of File marks to be skipped must
be filled in register 5 before entering the program.
An outline of this subprogram is given in flow-chart #4.
6-25
21
5. SERINT: Serial interval reduction program
This progl·am is the latest of a series of systems used by this laboratory
for continuous data reductiop.. All previous methods have required that data
be recorded on analogue tape and replayed at greatly reduced speed for the
reduction itself. The present method can, when desired, be used for on-line
reduction or for off-line reduction in real time. It can even be used to reduce
data played back from analogue tape at speeds higher than recording speed
whenever a unit time longer than 240 microseconds is permissible.
The data for which this program is intended consist of one channel of
nerve impulses and a second channel carrying a three level code designating
the a,nalysis period (analysis gate) and time of stimulation (stimulus codes).
The program generates a list of time intervals between successive 'events'.
Significant events for this reduction are the onset times of nerve impulses,. in
the first channel and the onset and offset times of stimuli in the second channel.
The program produces a sequence of double word entries in which the duration
of an interval is marked by the ten low order bits in each word (in one 'word these
bits show a count of 240 10 microsecond time units; in the other they show a count
of 2000 8 X 240io microsecond units). In real time the longest possible interval
is slightly gl·eater than four minutes, the shortest possible interval is 240 10
microseconds, and the accuracy is +240, -0 microseconds.
As each interval is registered it is marked (in the two high order bits
of ~e word containing the high level counter) with a code identifying the nature
of the event which terminated that interval. These codes are brought into the
Accumulator via TN 10 and TN 11 (See Input P •I. U .). They are suitable for
direct use by LINC programs designed to operate on this kind of data; they are
also recognized by an IBM 1401 program which converts our double precision
octal expressions for the intervals to fixed pOint decimal multiples of milliseconds.
These converted expressions in turn serve as data for the laboratory's library
of IBM 7094 analysis programs.
This method of reduction is made possible by the characteristics of the
Datamec tape recorder and by the buffering flip-flops on the external level inputs
through which data are brought into the LINC. Although wl·iting data on the
Datamec requires a progl·am loop more than half of which is cO'mmitted to the
bookkeeping and mechanics of tapewriting, some time is left over in which an
input process can continue even while data are being put out. The LINC core
me mory serves as a buffer. The buffering flip-flops and their connections (see
ll\IPUT P.1. U.) 'were designed to reduce the time needed for examination of the
external levels sufficiently so that a check of external levels on two lines could
be made and intervals recorded within 240 microseconds when 6-bit characters
were being written on tape every 120 miGroseconds.
6-26
22
The general pattern of the program is revealed in the attached flow sheet.
The teletypewriter is used to generate a permanent copy of the name of the tape,
the file number, and the title of the data transfer as this information is typed in
from the Linc keyboard. (Up to ten 72-character lines of descriptive title can be
entered at this time. This title is associated with the data in all subsequent data
handling procedures.) The data transfer title is translated into IBM compatible
alphanumeric code. Our basic tape subroutines (DATAlv1EC program) are used
to write the title on the digital tape. Then the 240 microsecond loop program is
called in to make the transfer itself. To help insure reliability in transfers and
to minimize false starts, the routine begins with a triple pause; data collection
begins with the first event following the onset of an analysis gate which in turn
follows the striking of a keyboard key.
The initial data collection occurs in" a relatively simple loop from which
exit is possible only when the analysis gate is removed or half of upper memory
is filled. Counting of time is done in terms of 240 microseconds units because
this corresponds to twice" the character time in Datamec writing. When control
leaves the initial loop because half of upper memory is filled, it enters the loop
\vhere simultaneous input and output of data occur. All counting in this loop is
done in terms of pairs of character (6-bit) dumpted on tape. Two characters
are put on tape for each tally of the '240 counter' and therefore for each test
of external events. The only exit from the loop occurs when the entire half
of upper memory from which writing is occurring has been written on tape.
If the writing proves to have been successful, control returns to the simple data
input loop" until the other half of upper memory is ready to be dumped. If the
writing produced a lateral parity error (detected on the laterla parity flip-flop
before the LRCC has passed the read head of the Datamec) the tape is baskspaced
through the faulty record,. erased forward, and another attempt made to \vrite
the data. Throughout this process data continues to be collected. Alternate
writing and eraSing will continue until success occurs or the alternate half of
upper memory is filled. In the latter case the program halfs with an error type'"
out. When in the simple data collection loop the 'off' state of the analysis gate
is detected a final data 1:i>ck is written on the digital tape and control returned
to a sub-initialization routine which types out a transfer note \vith the name of
the tape and the new file number inserted into it. An exit to guide is provided
just before this type-out so that the run may be terminated at the end of any transfer.
Before taking the data tape generated by this pl·ogram to the IBM computing
center or before using it for LINC analysis the data are transferred from Datamec
tape to LINC microtapes and checked for errors in longitudinal parity and for
miSSing characters. (There is not time to check these during the data transfer,
only lateral parity then being checked.) The LINC tapes form the per~anent storage
medium for these data. During the transfer a permanent transfer note which includes
the transfer title is made as an index to the contents of the LINC tapes.
6-27
23
The basic SERINT program consists of 11 blocks of LINC tape which
are read into LINC core as follows:
MAIN DRIVER
TELETYPE ROUTINES
DATA:MEC ROUTINES
DATA TRANSFER
ROUTINE
END RUN
DIAGNOSTICS, etc.
SERINT
TIT LIN
KBDTTY
DAT:MEC.
QNO
QN 1
QN 2, QN 3
QN 2, QN 3
240 SR
SERENO
WORDS
QN 1, QN 2, QN 3
QN 1
QN4
6-28
. SERINT FLOW CHART
-- 1
Call teletype routines
\f
Initialize tape and file numbers
.
~
o(2)
Type in (and out) transfer note
w-
Call teletype routines
~
Pause for l(ey.
~
Is key struck "G"
I
1 - - - - - - - - - - yes ~
0
no
'J/
Subinitialize~
~
increment file counter
Type out transfer note
~<
Type in (and out) title. translate.
and store in core
'l
Call Datamec program
~
I
(
~
Is this file number 1 1 - - - - - - - - - n o -->~~
yes
~
Do initial advance
\1,
T~~~------------------------Write end of
file mark.
t
Was EOF written correctly ' l - - - - - - - - - n o
I
0)
4
>0
yes
~<
Write title block
~
Was title block written correctly 1 - - - - no
I
--->-0
yes
~ransfer
Call Data
routine
Pause on Keyboard
Pause on Analysis Gate (XL 6)
Pause on first event
(XL 3)
I pg.
Continued on
2
6-29
SERINT FLOW CHART - continued -- 2
t~---0
Is analysis gate off 1 - - - - - - - - - - yes
I
no
--.::;;-<2>
.
~
,
Tally 240 counter and store and clear when event
t
Is upper memory half full 1 - - - - - - - no ---.;;>~0
~
I.
~e:
.~
Dump two characters on tape, tally 240 counter
and store and. clear. when event.
t
.
Is the dump of half of upper memory ended 1-no ~
I
.
yes
~
Are we at the end of the tap,e 1 ' - - - - - - yes ~
I
no
'It
Have we filled one half of upper memory
before successfully writing the other
half on tape 1 - - - - - - - - - - - - - y e s
I
no
0/
5 6
26-7
'"
I
22
2 I
7 7_ II 6 4
25
9 6
4
5
435 7 5
.!!! 28-9
49
9 2
4 2
8 6 2
44
5 4
49
128225694
0
u 30-1
~ 32-3
o 34-5
5 6
7
4
6
2
4
2 6
2 I
40
7 5 7
3
I
7 II 9 7
5 2 5 9
II
5 I
4· 3 4
3
62
49
3 4
6
II
9
4
8
47
5 7 6 7
4
31
8 3
19
42-3
3
I
5
44-5
I
I
~
a: 36-7
I
I
38-9
40-1
~
-
8
52
I
2
5
32 31 23 27 19 20 34 16 24 18 23 20 27 17 21 21 20 26 23 19 24 24 31 30 27 27 20
(0 U(S) =-2p(S)· logp(S)
(4) Uc (SR) =Umax.(SR)
Uob.(SR)
J
Stimulus
[ Uncertainty
4
644
.
1
Contingent Uncertainty
TransmlttedJ
C=Information
6-40
~3
m
I
c:
o
"en
en
"een
§ 2
.=
c:
o
c
E
:;:
~
02
c:
2
3
4
Stimulus Uncertainty - Bits
3
en
t-
al
z
o
cn2
en
:E
' - - - - - - - - - - - -.... 100 m se c.
CI)
z
c:t
0:
tZ
o
~,
:E
0:
o
LL.
Z
1 2 3
STIMULUS UNCERTAINTY - BITS
4
6-411
Explanations: Blanking time (blkg. time) and Sam time define the observation
period, that is that phase of the a;esponse over which the discharges are to be counted;
it is referenced to stimulus onset.
Specification of SI allows the user to li~it the range of the random numbers
addressing the relay registers (e. g.: only stimuli above intensity 7 are admitted) •.
When the last value is typed, the program will store the
first acceptable random
/
number in the relay register, and write the identification block in1:ik. 1, unit 1. The
program will then wait for sense switch 0 to be put in the "up" position; from there
on, stimuli are applied at the rate determined by an external rate generator. When
the total number of stimuli is applied, the stimulus response blocks are written. In
each word, the first five (high order) bits contain the stimulus intensity, and the
other seven (low order) bits contain the response count, with a maximum of 1778 •
5: Ancillary programs: Programs are available for perforation of the saved data on
to paper tape. Figure 2 displays the print out of a stimulus-response matrix with
the responses categorized in classes of two responses each.
6: Results obtained: From stimulus"response matrizes such as that of Figure 1 (top)
information transfer as regards stimulus intensity was calculated for mechanoreceptive
cutaneous afferent nerve fibers, using the computational procedure outlined in
Figure 1 (bottom). For each nerve fiber, the computation was repeated for different
sets of stimuli, with different values of stimulus uncertainty. An estimate of the
maximal information transfer was thereby obtained (Figure 2 (top). In 14 different
fibers, this value averaged to 2.5 bits. Maximal information transfer was found to
vary with the "observation time" after stimulus onset, in a manner shown for one
typical example in Figure 2 (bottom). These results are in the process of publication
(G. Werner and V .B.Mountcastle, J. Neurophysiol., 1965).
6-42
b) ANALYSIS OF POPULATION ACTIVITY
Response averaging and amplitude and phase measurements of noisy signals
This program is Qsed in a research project on the biophysics of photoreception. One of the least understood aspects of vision is the way that absorption
of light by visual pigments causes membrane potential changes in the visual
receptor cell. This project seeks to quantitate the dynamic relations between the
amount of light illuminating the eye and the evoked visual cell potentials. These
potentials are measured as a massed response of many cells, the retinal action
potential, recorded from an intact spider in response to various modulations of
light intensity with time. From the simple ocellus of the spider, it is thought that
only electrical responses of primary visual receptor cells will be measured; thus,
these potentials may be considered the initial electrical events whose elicitation
by light. is being studied.
The dynamic relations between modulations of light intensity and retinal
potentials are most easily studied with small modulations (small signal approach),
whereby amplitudes of the retinal potentials are linear functi:>ns of light modulation.
In particular, it is possible to fit linear transfer functions to amplitudes and phases
of retinal potentials elicited by sinusoidal light modulations. From these transfer
functions, satisfactory predictions of transient (i.e., incremental flash) responses
have been made. More generally, present ·workindicates that large, non-linear
responses may be quantitatively predicted by a model whose parameters are, in
part, denied from the linear transfer function.
It is for the averaging of small amplitude (10 microvolts, or so) linear or
non-linear responses from much noise that this program was written. It permits
measurement of responses to sinusoidal l,ight modulations over frequency ranges
longer than would otherwise be feasible and it permits accurate measurement of
very small amplitude DC potentials, as in the responses to incremental steps of
light, that were hitherto not possible. This program also allows the amplitudes
and phases of these averaged responses to be determined, thereby making possible
the transfer function analysis described above.
Program Description: The program occupies the four quarters of lower memory
and is logically divided into four subroutines, portions of which may be skipped
under sense switch control.
1. ARC (Average Response Computation) samples the response waveform
at a preset rate while a Compute Gate (exte11lallevel) is present, adds the sampled
pOints of successive responses in double precision, and averages the preset number
of responses in single precision.
6-4~~
2. ASC ( Average Stimulus Computation) performs four functions:
a) It samples and averages in single precision a preset number of
stimulus waveforms; this may be skipped by SNS O.
b) It displays both the averaged and response and stimulus wavefo.rms
(see Figure 1).
c) Under contro.l o.f SNS 5, it allo.ws each display to' be ro.tated (see
Figure 2) and points added to' o.r subtracted fro.m the end o.f the (unro.tated) display.
All subsequent displays are affected.
d) It writes the averaged respo.nse and stimulus po.ints; each o.ccupying
o.ne blo.ck, o.n tape unit #1; the last wo.rd in each blo.ck co.ntains the number o.f data
points displayed o.n exit fro.m (c) above.
3. RAP (Respo.nse ~mp1itude and Phase)
a) allo.ws the amplitudes and zero.-cro.ssings o.f displayed respo.nses
and stimuli to' be measured by displayed ho.rizo.ntal and vertical lines, respectively~
mo.ved under POT co.ntro.l. (see Figures 3 and 4).
b) it computes the peak-to"peak response amplitude and the phase
o.f the (sinuso.idal) respo.nse with respect to.the stimulus; amplitude co.mputatio.n
alone may be cho.sen by SNS 1.
4. DDAP (Decimal Displ§ly o.f Amplitude and Phase) co.mputes the decimal
values o.f amplitude and phase co.mputed by RAP and displays their values (see
Figure 5).
6-44
32
Legends for Figures:
Figure1:
Display by ASC of averaged stimulus and response at '
20 cps. The upper trace is the stimulus. Number of
responses averaged = 256; number of stimuli averaged = 8.
F~re
Rotated averaged response displayed by ASC Vertical
~cale double that in Fig. 1. Beginning and end of response
do not meet, as the discontinuity near the center shows;
this is presumably due jitter in the length of the compute
gate.
2:
0
Figure 3:
Display by RAP of averaged response and a line set by
POT 2 to be at its positive peak. Had this peak been found
by the computer, the isolated point above the time peak: would
have been chosen.
Figure 4:
Display by RAP of the averaged response, its zero value
(horizontal line), and a vertical line controlled by POT 3
placed at the positive-going zero crossing.
Figure 5:
Display by DDAP of amplitude and phase computed by RAP.
6-l,t5
Fig. 1
Fig. z
Fig. :3
Fig. 4
Fig. 5
vv
3
0
COMPUTER PERFORMANCE
All modifications to the computer which have been recommended in CDO
equipment change bulletins #1 through #11 have been made and checked. The
installation of the upper half of memory was accomplished with no particular
difficulties in August, 1964. With the exception of the sense amplifier failure
noted belo'w the memory system has been working most satisfactorily.
Maintenance
Only two major maintenance problems have occurred during the entire time
we have had the Linc. In each case the trouble was localized quickly.
On March 3, 1964, a transistor of the bit-9 flip-flop (DEC 4205) of the
B-register failed. The card was replaced.
On October 12, 1964, bit-2 sense amplifier (DEC 1571) failed. The
cause of this failure is unknown to us. .i he card was promptly repaired by
DEC and reinstalled.
Microtapes units have given us quite satisfactol.-Y service. Initial
difficulties with the tension of the magnetic tape were greatly reduced by adjustment of the drive belts of the motors.
OccaSional difficulties with LAP3, particularly with insert and remove meta
commands disappeared when the first recommended change of 4221 DEC packages
was made.
Sporadic difficulties with the Keyboard input, in particular misreading of
the 'tag' key, occurred in September, 1964. They disappeared spontaneously
before becoming a serious problem.
General Comments:
We have no major suggestions to make concerning hardware changes in
the Linc. We feel that the instrument as it stands is an extemely able laboratory
computer in almost evel.-Y respect. Our use of thea:>mputer has been so greatly
expanded by the addi1i:>n of upper memory that ~ now consider the 2 K memol.-Y
to be essential. Further increase in memol.-Y size 'would probably be useful,
especially if it were to make available additional locations for program storage.
Our few comments on the limitations of the Linc refer mainly to software.
They are discussed belo'N. One member of our group does, ho\vever, feel that
it would be desirable to extend, if possible, the convenience of programming the
scope display to conventional analog X-Y plotters. He feels that for routine graphic
display the cost of the Calcomp digital plotter is excessive and the effort of
programming for it might be objectionable. He feels also that it would be desirable
l',~ ~7'
'
O-~t
to make use of the X-Y plotters often found in biological laboratories.
In our experience investigators with computational inclinations pick up
Linc programming quickly and with a minimum of personal instruction. However,
\ve believe that some didactic material on the art of programming the Linc would
be useful. Of particular use would be a manual describing the function and use
of Linc operation codes including illustrative annotated programs for some of the
more sophisticated instructions.
In spite of this relative ease of programming the Linc we share a
common experience of finding that a great amount of our time is spent in
programming and otherwise caring for the computer. Because of the pressures
of other considerations during our experiments, we must plan our programs 50
that they may be easily used by any member of our graup. Therefore initialization
procedures must be straight forward and, as much as possible, selfMchecking.
Writing programs to fit this set of needs is more time consuming than writing
programs for oneself alone. In the past year we have been able to devote time to
producing several large and fundamental programs because construction work in
the laboratory has reduced somewhat the numberof experiments we could carry out·.
With these as the principal working programs we hope to be able to use available
time to \vrite shorter special purpose programs which \vill add divers ity to our
Linc operation.
We have found the two major programs that CDO provided us with (LAP
and GUIDE) to be very useful. The further development of these programs along the
lines that have been discussed with us should make them even more useful. It
does not seem to us that there is any particular need for the development of
additional complete programs, because, by the very nature of the computer, the most
useful programs are those written to meet specific needs of onets own experiments.
It would be helpful, however, if some general purpose subroutines could be developed,
in particular a libral"Y of arithmetic subroutines. This recommendation is based in
part on the following reaction to the Linc made by one of the members of our group:
For the purpose for which I had intended to use the Linc, namely the
solution of statistical decision problems on line with the experiment,
I found the following difficulties and limi~ations:
1) The Lincts limited memol"Y and slow speed made it inp>ssible
to solve the problems in the form of conventional alga1i:ithms in the
time available.
2) I am not sufficiently inventive to replace the conventional
algorithms with approximation procedures which can be carried
out in the available time.
6-1~8
35
3) I believe that most attemps to replace conventional algorithms
with procedures in non-parametric statistics limit the scope of the
application and curtail severely the precision of the results.
In retrospect, we feel that it would have been more efficient and helpful to
us if we had had available certain general purpose subroutines that we had to write
for ourselves. These are not necessarily difficult routines, but they are basic and
widely used. They include: octal to decimal and decimal to octal conversion routines,
alphanumeric oscilloscope display, and keyboard to teletypewriter alphanumeric
translation. It has been' our experience with subroutines we ourselves have written
that programming time is greatly reducro when ade'quate subroutines are available.
36 .
4. THE EVALUATION PROGRAM
a. Initial training: The assembly phase of the Linc was an invaluable experience
for the two participants. Firstly, it led to a rather intimate acquaintance with the
technical aspects of the machine, its design, and the facts pertinent to check out
and maintenance. Secondly, and perhaps more importantly, it provided the
opportunity for intensive personal contact with the designer of the instrument
This greatly helped to give the participants a frame of mind and a way of thinking
about machine computation and data proceSSing which was previously considerably
less vivid and specific in their minds. This interaction with the personnel of
the COO fostered the development of ideas of ways in which computational methods,
in general, could be used in research, and of ways for their implmentation in the
laboratory. To that extent the initial assembly phase not only reflected on the Linc
alone, but more generally, on the use of cOrqJuters in biological research. Thus
this program served a wider purpose than merely setting the stage for the use of
the Linc per ~. This initial experience of the two participants in the assembly
phase was transmitted to other investigators in this laboratory, after the installation
of the Linc, with consequences similar to those described above. Some of the
contagious enthusiasm of the designers of the Linc was thus imparted on a larger
group, and with this a wider group of investigators was drawn into the fascination
for machine computational methods in physiological research. It is our opinion that
a program similar to that conducted during the Summer of 1963 in Cambridge is
to be recommended for any biologist who 'Wishes to use laboratory computer.
b. Concerning the period from installation of the Linc at home base to date:
At times we felt that during this period some advice and better communication with
the COO would have been very useful. In particular it was felt that some information
service concerning programs of o.ther groups would have h3en advantageous. Or
. else, some period distribution of more advanced examples in ,solving programming
problems that should have served as a follow-up teaching device. The idea here
is that of a kind of "correspondence course" to keep the particip.ants alert of certain
possible, typical "tricks" and general routines. ThiS, we believe, wuld have saved
us considerable "development time" which, valuable as it has been for better
understanding of the machine and its applications, had to be detracted from the time
we devoted to physiological experimentation. For these reasons we think it would
be profitable jf in the future, brief adjournment meetings were organized for Linc
users. These meetings, to be held no more often than once a year, would give the
opportunity for exchange of programs and ideas about laborato~'"Y application of the
instrument and help greatly to reduce duplications that are costly both in money and
time.
· 6-50
37
Publications of research in which the LING was used:
1) Gerard Werner and Vernon B. Mountcastle: Neural activity in cutaneous
mechanoreceptor afferents: Weber function and information transfer.
J. Neurophysiol., (in press)
6-51
.w ..__....
..-aft.U~;72..'~u
aUBJ- . tNplM'
P. I. t1-. . - - - "
~~.
¥.~ •••l :•• _•••••••••••••••••• _••• _
•_I~
1
. CH~. IIY___..DATE. .... ~ l~_~!.~~.~~e .!.~~!~~!..__ ..............
BV • .
.an
E\teU t . S~.p.~~· Fl Je-.rJ.2·11~ L.~
•••- . . . . . . . . . . . . . . . ••••
• • •·•
• . . . . . .. .
.
,
.
,
194~)
XL..;
11/1
~6)
~NIO
XL",
-
MeV
'18 ~I
•
I~
_.-_._--
XL 3
it',
~.~
'- - - T
I
I
I
I
..
'
''''PC/TO'
INPuT«
INPI../7
i
.
JOaN~••• "~•••---~~1-
................
I
Xl (,
.
~
SHalT -NO"::'-J."7-:~~OP:~":: '
'
...~,-
.' ~. !~'1 .
'
"I"
~y"'\'IL "----~~;~ 7/.~C~~;r./' ~~~-~CT
-- -------~--...
. INPllT"P ~I . 1 1 . ·
op?
----------..
....----.--.
_E_~I:;'1..1~__~!:'s1__ :L~£.t:.:r:l'..tL.?_~!.££!!'_~...f.UJ!:L~~pS ~OB NO. --~--~-------2--·
SHEET NO; ",.2/'
~------------------
,.'
CHKD. BY_____
', _DAT,E~~-~~---""
____________~_~ __....~ ___..:....:.. .ll!~sJ!..!~lL~$~
~
._.~~..~....-.....--_
.-----.~... ~ ---~-.
.....~--- .•••
---------
"'-,,
,W;ftFm
I
:- '--} ;"6fj ~.- i'//~,{·:
I
I
~
i
lIND.
I~'"
Ij !
II
II I
:.J I I
'
I J)f<'Y·
,
==.."7" ~vl--1
L __ \) .___ J IZ --
,-1-~ ~'- _
,
1,
L1:·,·
'1
1
p
~
-:--vr.---l
X
-, .
·j:I-~rl
I ,~I
\
I
I
I
i
.\
I
I
"I
I
L __ _ _ _ J
k H
__
,,'
I
I
r\~_
Lf 1/ a.
J-", .. - _ - - -1
'HI y:
i ':
! !
~~
,
I
"
~
I
J
~./
_i
ZI -. I .
- -:,
"
I
I
! I
~ ?,
,
,I
i
I
I
i
<
I
j
!
"'1'
I
L
I
i
I
!
I.
r
I
M- -
r----~---
'11/;( ,
I.
,I
:'
J
-1
I
'IN
'IT
~~-(~----4_----------+_---J
I
'I
I
r _L:. -. - -- "- - E _.- . " - I
I
·1
I
I
I
I
I
I
i
L __' _ _ _ _ ____. --'I
UV
I
I'
I
q
K:r
I"
J
I
0
1M
L- _ -...::_ '- - - - -'- ~!~~J
S R
~--------~--~--~
A"
H - f:" - X -- - T -
r
,I
3
I
J
I
I,
BI( S".I-/
;5
--.~
N
I
'"-r
n m
x~ ;
•
.:.::
'0 I1 vHr.1
;0
II n.~
1::1 1,'<
C/') . . .
':3 : rT I~'
.~
.rr.....
:0
In I.
9 IH
I.
Ie:
10
I.
':3
.~
""
~
o X
IU III
Z .~
_)"qT,I//-I.I"C'.:rc:(
/"/NC.
p[{.lC-
r0;1/7/:i:~' i~{..'6'
-
FL Cv
=
/~F
;?:1 g .;(1;
li'I-!,::>Pt!=IVe>L
.:,~ { .. 1/5'O1 -/65
o
I -~
!• ,ii
:1
:, '-t~
I
0
I ,;'l
I
I
'.I
~iQ"
i
.~IJ~/1)'$ElE.C7'~
o
.~fW/IVD
,
•.\>- 'FVEd!
op'j)
N
!,-r~'ii
z;.
.' ,,1-,
'I
e:6--.--.-"=T.'I
I
~-1:
I
I
~/-~Yl
.f
rill/a
I
./
W'
__
~
./
T-....: _ _ ~
U-·
W'
I
J
~
'*
_ ~ 012_
I •
J.
,I'
I
1
,
L.
. / --.,
.
~
"I"
H
..,
I
I
.1
I'
I
I
l
I
I
I
I
i
I
'
-xJl-
.~
,-o,'F·
r-
,I
i
f
r.
r
N
,
... -~
.
I·
'--;(-~~':!
_
.BI? 5"'
8lf! It
E/r 3
..
0f~
6 ..
·r-::-~-I
I .
"
';'(@
<"
0:
I "
OPlrloQ)-¢8
I
.1.
I
I 8/f
I ~A~'1
OPR/~(~
",1 roill
\.!!J~f)--:
I '
,
~:
·Be
~
..
"
~
'f3~'3 ~
r"
0"
- . --'·:';-_"r---·
I
,~
88
(8'Y
',,--'
cEG7
38
r;;;-.
(!) __h_dV.!,
88
~NgL
8])
@.:.I
{3B
(~3\
138 ,,-_,l-
-..::/
QB£.
MINP
I/Ntit..
CL,t.
"""!i#
°
3
!3f
®
er
1£
8F~.
0
:1"-"
5~IIF
-IS'V
"-'- ".
--.::r~)
. n)
() v,
.. :....
.
~'.
-
~
.. "
:
..,.
~
'~"':.,.
'-':~~ (!~r-;- _··;:2~~~~.;~~~~~~~~_~~~~-~~~~~.~~~,~
I "
'
;:,,~.~,.,,~>~,.,~J-:.};d·
,,;",;,':;0,.
::,;,'
.... :.;.
I
'.;
:.
•
.
'.',' ·.CHKD.·BY~D~TL.LL~·::::·:.~~£>~tie. u.:.1.<1Q__!:!iSJLl!ifJ!!....P..J.....U".l._-"-__
· -
<:.!.. ".~.•c-:\~,t,i,"·~t#j~~,·t;4".
I
.' ..
'::3'
I.
?OW
;;K;L,4yG
'. - - -
JOB
Nf'-~-~--1&'
---------
•••-----.
-~-----.--
I
, I
, .i
:
.. j
.: '. ';.. I
I
I·
. i.
I
': ~
s '
.,
,:r- -:-'- ~7,,-, '(
',I
"
.. :1"
'/'
'1"
"" .
I·
'"
"
I
IL
'+
-
_____
"
,
"
I
-1
F
'E
~:
- - _ _ _ ...I
V
F "E . , y .
r--t- - =-"--430/l
I
J'
I
1
I
I
,
.I
.
.
'
' :
I
r
.1
I
/"
I.
I
.____. . . ,-__z.....;;_---'
I,
i
"
··~tr··
··,iJ·····
:
--
IUfL
',',
RR.EL
~
.'
T--.~P
.
1
~
I
I
.I
OrR l.ol" . '.
'
16-61
',:'\..DATE__,,~ .. :;
'
PEH!.="ORf\TOR System
/'
BY ____........___
__/~j/
.:..I •. SUBJECT
__________________ ~---------------------CHKD~ BY.~..__ ~DATE_______ _
_!..~l"..c;,r_~~l.;'!.:tl'.._:__£~l!.ll~~_tJ.2.Q_!.sr_C2!~!.~!?_t..
~:
.
.
"
:,;
"
JOB· NO. -:---~~----,~.ll-
.R.:.J_:.~.:__t£'__Ts_L~llt?c.. P£lJl~.!..Ja pc .~_
------....---------........_--.;-_....-\
,
. SHEET NO.• -....;....:.;----OP'--~-·
___
~____
:'\
(..
,
;
I'
I'
f':
.
~
"
~
~J;/~r:" r"'!! /)
C.P. ci/.pi
I
_i
/O?: L. A
Y
(~,,:).··r:i!!CA6'U
•
~
b ..
.
6 5 I"'.' / OO,~
I-i E'S CW~ '/
·,,··.·.1-.. '.
~
,~
'Q-
.
~
. ~--,
SZ I
;
-,;
-"T II,
"
I
-,--. ---, -
:'~.-~
=1<
~:
I
: l\
'
~ l'
t-~ ~
I
I'
.
I
-1
1
i
I !I !
i
!
,I
-II
16
~
-0
--1,
~i
~,
.~ t;)
1~~
,~
;
I
'!)
C"(
II
'1--'-"
I
!I
-L)
~)
~---4)
; I
I i !!
i
i l!
SZi!
'-
I I;
I
~-~
~
Ii
Ii
1
I
j
lU;
.
.
-LL- ,I
i
r
--
u
I------J
I
___
r-:';:
. :
'1.:1'-
i ' .
t--: 13
I'~
I'
T
il
'
):
,j
t!
~"
: .--.. --~
.,
,I
1,
.'
I
I
,I
I
I
'I
I
5Tlllf,COR, ,
A
5~
~6/iIIMFH9L
4''1
/6 S
-
0 /-
C,
r- '- - -- ~r
1"'' Cl[1,
I
rt: f i T -L:;
~j
!'-f
;
._.
-:z-=r . r -..-_:':"---.'----'"--'---:"'--~--.t!-I
-,---_
!
__
:~ L~- ____________ -----------~ ~
~;-
, LJ-L)"
,
._-'----'-'_..
.
j
":
I',
I
~ ~
. ....
+c;:~ 11
-- -I-~-==-_'-_-_-~~=:~~_-'---==_--_----J, 1'------
6-62
Flow 'chart
'1.
I
:~~ITE RECORD (?OOOS char.)
, Ini t ia 1 TJg2I?
I
i
.I .
!
,~
Tape transnort status' -" "ini tialfze "
~<'
,
V
rC;::~I_-_O_d_d_p_a_r_ii:....tY_~i_,~_.'___... . ;.,ji_ _ _( ; )
Start
~~ri te 2,0008 .char •. from: QN, 6 &. 7 .. ).
Compute ru~in8 .longi,tudlnal che~k sum
l
i\ . !
.; ';
'.:
~ ,;'; , : , ! . .
;
.Generate.LRCe·gap
Wri t'e
f,\l.
. II ':
i'i "
URCC '
i
Delay 1A;
. Stop Tape
Backspaqe Erase
Stop Tape
Ret-lri te 'counter
I
.'.
Re-tri8~er
".~.
2.5 delay
_,~
Later~l
no
~'
3 .? -:--no .
yes
,
.V:
I
Advance erase tape
= 3.0"
IStoP Tape
Jog
r ~_l
I
parity error ? ------.>.> yes
, I
t~ai
,ID
I
i~
6 ? ' - no
\
. ~:
Advance. counter
·Wait tdr end of record. (rea~ after t03ri te
mode) d expiration of2~5 delay
.
;
yes
. !~
..
IICI
c
=
_ _---I
Return to Ca 11 ing Program
t for expiration of 2.5 delay or LR.CC
,~
Is longitudinal check sum correct ·?~no
I
. . . t
yes
~r:YT~;:'--J
t
c
=0
-'--r-~'
Q
Start Tape - odd pari ty
Acivance crase tai.,e = 3.0"
(,t
~~-')
Return 'to Calling Program
6-63
Flow chart '2
!>."dITE END OP FILE r~\m\ (EO'C')::
Initinl Ta~
2:,
.r··'.
Test transport
statu~
- initialize
. VIOIII!!E~'---------~----':(-l~"'-'\
\_. .
. Clear register,2M
-
l
Start tape- even parity
!~<-.----JI4)
Wri to EOF
1\1
178
\.----.
!
..' r .'
?
\,
I..
~
(
.'
i__ ·~ ...
' _ _ _-.
~
elC1Y
'lA
Stop Tape
Bact{space Erase
, S,top Tape
Rewrite counter
I
no ~
t'
.
't
Wait for EOF (read after write mode)
----;>
.~
I
--'l!>-
<$> .
'.J. ,
t, '
'--S-to-'p-T--'ape'
Jog
. Ace = 1.
no
yes
Advance Erase
~
6 1-
MVanCj counter =3 ? ' - . no
Write tRCC
Is c(A) .; 511781
5
yes·
,;
Generate tRCC Gap'
(
= 3.0"
.
yes'
~'
.
Return to calling program
Store c(A), in 2M
'l
Has 2.5 delay expired 1 ----=:>;;;...no
I
yes
t.
Re-trigger
Has
t.s
2.5
V'
d~lay
delay expired? ----=>"'!.
Clear register 2M
yes
I
no
·v
Is c(A)
III
----::>0 no <::[>-
c(2M) 1
0)
I
----------·----ye-g---- .-----. --...-----.-.... .'--_. - .
t .
Start tape· even parity
Delay'le
~
G:)
V
Has 2.5 delay expired '?--:>no ~
I
rn
y-~-~~·
Delay I C '
Stop Tape
.
Jog. '
.
AC'=-i
0 ______ .
,
Return to calling' program
6-64
READ FORt.JARD
:. Ini tia 1 Tag 3A
Test transport status - initialize
+<
..
~
~
Start tape - odd parity
Low-High Read Threshold alternately
t
Read and store first character
Start longi tucUnal sum
t
Is c(A) -
I
( 51178
( 5517 8
no
Write Enabled _
es
Write Disabled
yI.
Store c(A) in 3C
~ <~------------------~I
wait for next character
or 1.25 delay expiration
~
Has 1.25 delay expired 'l------yes-<&>
I
no
\I
Store character
Compute longi tudinal
<:<;;..-_ _ _ _ _ _ _
sum
f:'\
~
Wait for next character
or 2.5· delay expiration
t
Has 2.5 delay expired' 'I - - - - - - yes-~
I
no
0/
Count Missing Character
cb
·6-65
READ FORWARD - continued
o
pg.2
END OF RECORD
~,
Re-trigger 2.5 delay
,I,
Lateral pari ty error 1 ----------=>=--yes
I
no
~!
Wait for LRCe or 2.5 delay expiration
t
~6
Store LRCe or zero
.
~, """'o(;:----.--,;.-------:--------~
I
!Delay rA
IStoP T~:e
t
was there a lateral parity error 1---yes ~.
I
no
'V
Possible EOF mark 1 - - - - - - - - - y e s ~<~
I
no
~
Missing character 1 - - - - - - - - - yes ~
I
no
Is
t .
longitudinal cbeck
sum
correct· 1--no ~~
I·
yes
t
Store number of characters read
Fill unused QN 6 & 7 with negative zero
Accumulator .. 0
~
Return to calling program'
6.-66
READ FOR\wm •
.
","
cp
-
...
'.~
_
..
-_.._--_.--
pg.3
continued
.
wait for LaCC or 2.5 delay expiration
Backspace
Stop 'Tape
Sub Initialization
t
Set parity error indicator
\1,
o
Is
cr
this a 1 char. record '1 -no ~
I
yes
~
Set 'return configuration
A - 1
t
Return to calling
progr~m
Gp
Re-read counter - 12 '1 - ' - - - no ~
I
yes
~
Return to calling program
"
7 & 14)- Set return configuration
9
10
A-4002~
) - Set return configuration A - 2002
J-
Set return configuration A - 1002
11
,
..
6-67
SKIP EOF •
(
(
:
:
FOR1·1ARD
BACKWARD
Ini tisl Tag 3L
Initial Tag 3K
,Enter subprogram with number of !!OF,-:, counter set
Test transport status • initialize
Clear
t
t
-~---iC0
-E=-oE
sa
rQg1~t~r
Set record counter • 0
!
Start tape ~ odd parity
Low Read Threshold
!
~----------C0
Read first character
Is c(A),.
~
«
5117
5517
Write Enabled
Write Disabled
---;:>
yes
I
Store c(A) in SC
no
-
~<
Has
~
(record)'
2.5 delay expired 'I -------->~ no
I
yes
~.
.
-
Re-trigger 2.5 delay
~
-<£> (1 char.
Has 2.5 delay expired ? --------~yes
I
record)
no
t
I
Is c(A) - c(5C)
- - - - - - - - - - - . ; > no.~ (record?)
yes
'f
Has 2.5 'delay expired'? --------~> no
I
yes
~
-
-
>,0
Number of files counter • O------,.;;:>~ no
Forward. _~
Delay
r
y!s.~
U:J
I i=~~~~~;1
__~a~kyard
I~~!~y 10 ~
~
L Ta~~J
~
Enter. Read Forward Subprogr.anl
6-68
SKIP EOF - continued
pg 2 •.
G)
t
Wait for 2.5 delay expiration
t
"Re-trlgger 2.5 delay
t
!
Index record counter
. . Wai t for 2.5 expiration
G-)
6-69
F;:'•
~~~~~~~~
.
:::::
LINe Evaluation Report
Submitted by:
Bernard Weiss
'Ehe Johns Hopkins University School of Medicine
7-1
- 1 -
Past and Present Research
. In this laboratory, the LINe has been used to conduct and
experiments on operant behavior.
analy~~
~
A general purpose digital computer offers
a number of features which make it ideal for such experiments because of the
ease with which one can arrange complex schedules of reinforcement without
patching up new circuits for each new modification.
convenience.
But this is a matter of
It is much more important that the digital computer offers
new dimensions in beha.vior research; it allows one to devise experiments
unhampered by the present limits of instrumentation and also permits a detailed
examination of the fine structure of behavior.
For example, computers permit
one to synthesize schedules with both long and short-term serial dependencies
and with stochastic processes imposed.
On-Line Programs:
The experiments so far in which the LINe has been used on-line to
generate reinforcement schedules have mostly involved monkeys.
Several
. kinds of programs have been written and used.
1)
Autoregressive schedule.
On the basic schedule, the delivery
of reinforcement (the reinforcer is a small amount of fruit juice) depends on
the consistency of the Interresponse Times (IRTs).
An IRT is defined as
the waiting time preceding a/response; it is measured from the previous
7-2
- 2 -
response.
IRTs.
The LINe
pr~gram
evaluates consistency by comparing two adjacent
In the program, IRT + is divided by IRT •
i l
i
The closer these two
values are to one another -- i.e., the closer the ratio is to 1.0 -- the
~.
higher the probability of reinforcement.
This probability falls off expon-.
entially as the ratio deviates from unity.
The behavior resulting from this schedule is comprised of two
components.
One is a tendency to respond in a near-periodic fashion.
other is a significant serial dependency in IRT duration imposed on
tendency.
The
t~is
Thus, this program generates both a low overall variability and
a low serial variability.
Other modifications of this basic program have also been studied.
In one, reinforcement does not .occur on each occasion when it is due.
Instead it occurs every
~th
occasion.
With
overall variability is even more sharply
decreases.
In another variation, we
hav~
or 8, we have found that the
~=4
boun~
but that the serial dependency
compared, not successive IRTs,
but the current IRT vs. the mean of the last
i IRTs. This procedure also
reduced variability.
2)
Stochastic Reinforcement of Waiting.
be differentially reinforced is to set a
low~r
One way in which IRTs can
limit.
Every IRT longer
than the minimum duration is reinforced; those that ·are too short only reset
the time counter.
We have used the LINe to devise a stochastic analog of
such a schedule.
Instead of using a sharp cut-off, we have arranged it so
7-3
- 3 -
that the probability of reinforcement increases linearly with the duration
of the IRT.
The behavior typically generated by this schedule is characterized
by a bimodal distribution of IRTs.
One peak lies at about 0.5 secs.
The
other peak, which is much less in amplitude, lies close to about 20 seconds.
A more unimodal distribution can be achieved by evaluating, not single IRTs,
but the mean of the last
i IRTs •. Such a procedure punishes long trains of.
very short IRTs and they gradually fall in frequency.
In another variation,
the function relating probability of reinforcement to IRT duration is
Gaussian, rather than linear.
3)
Differential Reinforcement of Low Rate.
We have also studied
the traditional schedule described earlier, namely, the one in which only
IRTs longer than a specified minimum are reinforced.
to provide such training before shifting to the
Our original plan was
st~chastic
schedule just
described, but certain features of the behavior on the traditional schedule"
became apparent after analysis by the LINe and we continued to study it.
We found, particularly after a moderate amount of training, that quite strong
serial dependencies were present in IRT duration which often could easily
be observed even on a simple display of the qurations of successive IRTs.
4)
Other Programs.
In addition to the programs outlined above,
we have also written an on-line program that we have started to use in
studies of spontaneous activity in grouped mice given various drugs.
We
- 4 -
hope to obtain some information on the variables underlying strain
differences in the response to grouping.
We are also preparing a program
to study the behavior generated by a schedule involving various
transitio~
probabilities for specified responses.
As an example of the programming and hardware involved in the
execution of one of these on-line programs, consider the autoregressive
schedule.
Peripheral eguipment:
once.
lever.
Three monkeys are generally studied at
Each monkey, in a separate chamber, has access to a microswitch
These are connected to DEC 4410 pulse generators in the Data Terminal
Box (DTB).
Each depression and release of a lever produces a pulse from
the appropriate pulse generator.
This pulse is used to set a flip-flop
whose 1 output is connected to one of the external sense lines.
lines come from appropriate OPR terminals in the DTB.
The reset
Reinforcements are
delivered by means of the LINC relays, which actuate 24 v.
mercury~wetted
relays that close a circuit to a solenoid valve which controls the' flow of
fruit drink used as the
reinforce~.
The DTB also contains a DEC 4401 clock used to time the IRTs.
the ARG schedule, we used a clock rate of 100 pps.
For
At the end of an
experiment, an OPR pulse actuates a relay that turns off equipment such
. as stimulus lights, noise generators, and graphic recorders.
Program:
The program is so structured that several monkeys can
be studied at once with each IRT identified by monkey_
7-5
- 5
1.
Set up indexing for recording; store session time; store
address of first block to be used on data reel; store fixed ratio (KBD);
set up random number generator.
2.
Examine external
leve~s
~'
from clock and 3 monkeys in turn.
If
clock pulse noted, add 1 to each monkey's IRT counter and to session time,
,
and determine whether session is ended.
'
If no monkey has responded, search
external lines again.
3.
If a particular monkey has responded, store his identification
number for future reference; store the current IRT.
Divide the current
IRT by the previous IRT or the reverse, putting the larger value in the
numerator (octal division); use quotient to locate probability value table
(probability'is equal to entry divided by largest possible random number
value); store table value; generate quasi-random number; if table value
exceeds random number, go to reinforcement routine and see if ratio has
been met; if smaller, merely record.
4.
To reinfor:ce, actuate appropriate relay, the duration of closure
being set by a program loop; label IRT as reinforced by placing a 1 in the
siwbit.
Add 1 to reinforcement counter.
5.
Add I to response counter; store IRT; store monkey identification
number in adjacent word; if memory is filled w1th data, write tape; if not,
return to external level examination.
Because data are being gathered from 3 monkeys simultaneously, part
of the program is aimed at keeping their identities clear.
program requires about 600
S
The entire
instructions.
7-6
- 6 -
Analytical Programs:
1.
General display:
This is a program built up from one originally
~nd
provided us by Doctors Hance
the IRT data in various ways.
Killam.
It allows us to scan and transform
For example; display data as a bar graph;'"'
transform with 3-term moving average; transform with 7-term moving average;
plot deviations from moving
~verage;
plot successive IRT differences;
differentiate; integrate; plot mean; perform peak clipping; reshuffle serial
order randomly; expand or contract display; etc.
Transformed data
ca~
be
saved and studied.
2.
data.
Histogram display:
This plots an IRT histogram of specified
Options include; variable size; conditional probability transformation;
distribution of IRTs following IRTs above or below the median.
and time scale are variable.
Resolution
Axes are also displayed.
3.
Joint interval histogram:
4.
Sequence display.
Plots IRT. vs. IRT.+L •
J. '
J.
L is variable.
Plot, as a bar graph, sequences of 4, 3, and
2 IRTs, according to whether or not they lie above or below the median.
5.
Expectation density:
Plot.' this function, which is a type of
averaging, with variable resolution.
6.
Miscellaneous:
n
Include runs test, 2 his~ograms, histograms
of reinforced and post-reinforcement IRTs, mean square successive difference,
etc.
Utility Programs:
1.
Variable character display:
For making graphs.
7-7
- 7 -
2.
TeTe plot out:
Uses teletype to make graphs.
3.
Random number generators: Various kinds.
~.
Floating point conversion:
These use FLP routines supplied
by L. Hundley to enter and transfor data from various kinds of off-line
experiments as well as experiments performed with the LINC ..
- 8 -
Future Research
While we have made a
begin~ing
that we believe demonstrates some
~.
of the great potential of computers in operant behavior. research, a vast
~ange
of unexplored applications still remain.
report, we shall discuss
LINC.
60m~
In this section of our
of these and outline our future plans for the
These are to be taken, of course, as an indication of our wish to
retain the LINe for our laboratory.
Introduction.
It is clear, of course, that digital computers
are indispensable for perfonming a microanalysis of operant behavior.
with traditional
reinforceme~t
Even
schedules, such as the DRL schedule
(Differential Reinforcement of Low Rate), a detailed analysis of the timeseries structure of the behavior has revealed properties not observed before.
It is less clear, to many investigators, that digital computers are also
virtually indispensable for generating reinforcement schedules with
properties akin to those found in natural environments.
These are: the
stochastic character of the reinforcement contingencies and the prevalence
of short and long-term serial dependencies.
A digital computer can easily
simulate such properties.
One of the reasons that digital computers and operant behavior
. seem so well-matched is that operant behavior can .be easily translated into
digital terms.
One method of doing so'·is to shape behavior according to the
waiting times between successive· discrete
responses.
Such interresponse
7-9
- 9 -
times (IRTs) can be modified by the process 9f differential reinforcement
in exactly the same way as spatially-defined responses.
Their coding makes
them ideal for computers.
Methods.
In addition to the work performed so far, which we plan
to extend (see section on Previous Work), we plan to use the LINe to investigate several other kinds of behavioral phenomena.
a.
Schedule with periodic components.
On certain reinforcement
schedules, the.reinforcement is programmed to appear a constant length of
time following the previous reinforcement.
(FI) schedule.
This is called a Fixed-Interval
If the intervals are not uniform in length, it is called a
Variable-Interval (VI) schedule.
Typically, FI schedules are used to study
behavior controlled by periodic events and VI schedules are used to produce
relatively constant rates of responding.
These two schedules become most
logically related when they are described in terms of their
waveforms--
compone~t
When they are described, that is, as an engineer might describe them.
Once such an analysis is made, we can then plan to study behavior
maintained by interval schedules from a more uniform point of view.
Thus,
any interval schedule would.then be described as a waveform with known
component frequencies, and we could then analyze the resultant behavior from
the standpoint of how these component frequencies are reflected in the
temporal structure of the behayior.
7-10·
- 10 -
Using the LINC, we can then not only perform the proper analyses,
but also generate new variants' of such schedules.
generate
a schedule
as a sine function.
on
wh~ch
For instance, we could
the probability of reinforcement rose and ;e1l
We could then impose more and more randomness on the
function to determine the degree to which behavior reflected the underlying
pattern in the presence of "noise."
Interesting questions. spring from such
an approach in relation to drug effects and the stimulus control of behavior
(Weiss, B. and Laties, V.G.
b.
Fed. Proc., 23:801-807, 1964).
Schedules with state changes.
Changes in the sequential
patterns of behavior may occur, not only with IRT measures, but with
discrete responses, and may be shaped accordingly.
For example, suppose
that in a chamber containing three levers, A, B, and C, one arranged the
following transition matrix:
Response i +l
Response.
~
A
B
C
A
.01
.02
.04
B
.02
.04
.01
C
.04
-01
.02
That is, given that response. was made on lever A, the probability of a second
~
re~ponse
on lever A producing reinforcement is .01; the probability of the
succeeding response producing reinforcement is .04 if lever B is selected.
Humans find sequential dependencies greater than simple Markoff chains
extremely difficult to learn.
We plan to study such chains in the free-operant
7-1'1
- 11 -
situation and then determine whether it is possible to go on to higherorder chains.
Perhaps, by employing ratio schedules, it will be possible
to find evidence of control by higher-order dependencies by looking at
tb~
distribution of responses on the various levers.
c.
Adaptive schedules.
Schedules e'xist on which a function or
value is varied, in accordance ,with some rule, by what the subject does.
have been called 'by various names:
adjustin~,
These
titration, and conjugate
schedules.
In our laboratory, such a schedule has been used to study pain and
analgesia.
The subject,
~
electrodes, receives an electrical signal
whose intensity rises in steps, an increment occurring every few seconds.
Edch response, e.g. a lever press, drives the current down by a step.
thi~
In
manner, one can train monkeys to trace out an aversive threshold that
displays remarkably little variability.
This is a psychophysical method which has proved relatively sensitive
to analgesic agents.
of the LINC.
We are interested in pursuing it further with the aid
One may think of the system as a proportional controller;
the probability of a decremental response is a function of shock intensity_
Would it be possible to produce a more stable
example, derivative control?
~hreshold
by adding, for
This would be accomplished by making the size
of the step that decreased the shock a function of the rate of response; the
greater the rate, the greater the size.
By viewing the monkey as a sort of
psychophysical control system we might learn something about the aversive
control of behavior as we!.l as make the technique more sensitive.
7-12
- 12 -
d.
Traditional schedules.
Our understanding of how the more
traditional types of reinforcement schedules exert their effects is still
elementary.
We therefore plan to examine some of these with the LINC as
~.
we have done with the DRL schedule of reinforcement.
In addition to the research discussed above, we plan to continue
ou~
studies with drugs.
So far, we
ha~e
completed only one drug experiment,
mainly because we felt it important to enhance our skills in the use of
computers and to perform purely behavioral studies first.
Now, however,
we are ready to pursue our interests in behavioral pharmacology.
Facilities.
Beginning June I, 1965, both the principal investigator
and the co-investigator, Victor Laties, will be on the faculty of the
University of Rochester School of Medicine, with primary appointments in the
Department of
Radiat~on
Biology.
In addition, the principal investigator
will have a joint appointment in the Center for Brain Research and the
co-investigator will have one in the Departnent of Pharmacology.
Radiation Biology is the largest department in the Medical School
at Rochester, and we will have access to all of its many facilities.
These
will include a computer center partly devoted to on-line work, so that we
hope to be able to expand the scope of our experiments by connecting lines
to their computers.
The great variety of disciplines represented in
Radiation Biology, coupled with our joint appointments, will undoubtedly
extend the range of behavioral problems studied with the LINC.
We are now
in the process of designing interface logic to be used with Dataphone
7-13
- 13 -
transmission between laboratories in order that we may conduct on-line
experiments from our laboratory in the Department of Medicine.
At
Rochester, this system will enable us to conduct collaborative work with
the Center for Brain Research.
Because of the move to Rochester, we request that the renewal
of the LINe grant, if approved, be made.through the University of Rochester.
7-1"4
- 14 -
Training Program
lbe training program that we underwent at .M.I.T. was excellent in
/
most respects.
At the time,I felt that too much emphasis was being
placed on the maintenance, and' design of the LINC and not enough on the
programming.
Since, then, I have changed my mind.
Programming skills, I'
found, once a computer was available to use as a teaching machine, were
,more readily acquired than I anticipated, although I could have used .
guidance when I began to try to write complex programs.
The material on design and maintenance proved valuable for two
reasons.
In the first place, a considerable amount of trouble-shooting
remained to be done even after the machine was delivered to Baltimore.
Although I had a great deal of on-the-spot assistance from the C.D.O. staff,
the information I was able to give by telephone often sufficed to provide a
diagnosis.
In the second place, the design of reliable interface circuitry for
our particular needs proved to be no simple matter, and the education I had
received during the training program plus my trouble-shooting experience
was an immense help in evolving an appropriate system.
Discussion of
interface problems during the course would have been extremely useful.
Naturally, a less hurried program would have been easier to cope
with.
Perhaps more important, if we had had more access to a LINe during
,the course, and could have spent considerable time learning its features
7-15
- 15 -
by actual operation, many of the points made would have been easier to
grasp_
The LINe itself is an excellent teacher and it is unfortunate that
its potential could not have been exploited because.of the circumstances;
7-16
- 16 -
Computer Performance
. a.
Maintenance.
A number of early difficulties arose from the ....
prototype nature of our computer.
As can be seen from our log. entries l
some of our early problems resided in the power supply.
These were solved
eventually after the redesigned circuits were incorporated.
Other problems
were found to be due to circuit cards that had been hand-soldered, and that
contained po·or. connections.
Still others were simply the result of what·
turned out to be faulty components, such as transistors.
During the first
three months of operation, therefore, a large percentage of time was spent
on maintenance, althougn the machine ran well for periods sufficiently long
to give me practice in writing programs.
During the next few months, my major difficulties were due to
interfacing logic.
Perhaps this is one area in which the training program
could have made a greater
not experts.
contribu~ion,
at least for those of us who were
Our problems were mainly those of electrical noise produced
by relays, solenoids, and motors.
After making numerous changes in our
designs and former methods, and consulting with the C.D.O. staff, we
finally developed a system that has worked rather well.
For about a year,
now, the LINC and its interface equipment have operated with very little
time lost to dysfunction.
The largest problems arose from an oscilloscope
malfunction and from a'nicked cable.
The upper half of memory was installed
by C. Molnar on 10/3/64 and has been operating satisfac·tz·orily since.
7-17
- 17 -
b.
LINC design.
behavior research.
The LINC is extremely wel,l-suited for operant
Its speed makes it easy to devise very complex programs
without worry about time resolution.
Its interface flexibility allows
variety of paripheral equipment arrangements.
fe~tures
a~ide
I will comment on some
that might make it more convenient for behavioral work in my laboratory.
(1)
Memory size:
is generally sufficient.
For data analysis, the present 2048 word memory
The LINC would be a more efficient machine for
on-line use if the upper 1024 words were programmable, so that more experiments
could be conducted simultaneously.
If four different programs could run at
once (each involving, say, three monkeys) the cost of the LINC would be
competitive with the cost of the conventional equipment required to conduct
this many experiments.
(I am estimating $3000 per monkey.)
The flexibility
of the LINC and its analytical power would then be a bonus.
(2)
Tape transfers:
When 3 monkeys are all responding at a high
rate, and fine resolution (10 msecs, say) of the IRTs is needed, tape
transfers can introduce a small error because of the time required for tape
motion.
If the transfer were autonomous, so that block searching might go
on at the same time as the central program, part of this problem would be
avoided.
Perhaps the·buffering. required could be incorporated into the DTB.
(3)
Compared to the present D.E.C. relay packages (e.g., no.1803),
the LINC relays are rather slow.
Also, I would prefer to have the outputs
which drive the relays available independently for other purposes should I
7-18
- 18 -
need them.
I would be in favor of changing these to -3 v. level outputs to
be used as the experimenter wishes, which can be done now if the relays
are not 'used.
This feature would also mean more room in the Data Terminal
Box.
(4)
The Data Terminal Box is a good design for those experimenters
who concentrate on one kind of experiment.
Those who, like myself, conduct
several different kinds of studies, might find it more convenient, I think,
to have the
DTB
output available vie a plugboard to which can also be
connected a wide variety of logic elements as
etc.
~ell
as lights, switches,
We are experimenting with' such a system now, using the Mac Panel
plugboard.
We shall inform the Computer Research Laboratory of our progress,
if any.
7-19
- 19 -
Bibliography
A.
Published papers and papers in Dress.
1.
Weiss B. and Laties, V.G., Drug effects on the temporal patterning
of behavior.
2.
Fed. Proc., 23:801-807, 1964.
Weiss, B. and Laties, V. G., A reinforcement schedule generated by
an on-line digital computer.
3.
Science (in press).
l;eiss, B., Laties, V.G., and Siegel, L.
A computer analysis of
serial interactions in spaced responding.
J. exper. anal.
Behav. (submitted for publication).
4.
Sieg~l,
L.
A method for converting voltage levels.
J. exper.anal.
Behav. (in press).
B.
Symposia
1.
Weiss, B. and Laties, V.G.
Programming reinforcement schedules with
an on-line digital computer.
American Psychological Association
meetings, September 6, 1964 (Symposium on the Experimental
Analysis of Behavior).
c.
Talks
1.
Johns .Hopkins University School of Medicine.
2.
Institute for
3.
Albert Einstein College of Medicine.
4.
University of
Beha~ioral
Research.
Roch~ster.
·7-20
LINC Computer Evaluation
Grant FR 00146-01
June 1, 1963 - March 31, 1965
Principle Investigator:
Dr. Fred S. Grodins, Professor of Physiology
Co-principle Investigator:
Dr. James E. Randall, Professor of Physiology
Northwestern University Medical School
303 E. Chicago Avenue
Chicago, Illinois 60611
FINAL REPORT
March 1, 1965
8-1
1.
THE NORTHWESTERN UNIVERSITY GROUP WISHES 'ID KEEP THEIR MACHINE
The LINe has\become a wa:y of life in this laboratory and its loss
would be considered a near-catastrophe. Two graduate students have Ph.D •.
dissertation problems directly related to the machine.
One faculty
m~~
ber has a grant-supported research problem which directly depends upon
the LINe for data analysis.
Other faculty members are working in
gener~
areas which are almost certain to generate problems which can utilize the
LINe's capabilities in the future.
We therefore request that we be allowed to retain this LINC in our
laboratory.
This is a reflection of the most important conclusion of our
evaluation program,
i.e.~
we have found the LINC to be a very practical,
very useful, in fact an essential laboratory instrument and we want to
keep it!
8-2
2.
PAST AND PRESENT USES
A.
Classes of Programs
The following kinds of programs have been written for use in the
research problems described later.
1.
TeletyPe subroutines -- This general-purpose package, taking up
one memory quarter, includes typing the contents of the accumulator in decimal.
This has greatly facilitated the interpreta-.
tion of output data.
Other operations include typing a stored
table of teletype characters and returning to the mainline
routine when the character for
*
(252) is ·read.
These teletype
characters can be stored as two 6-bit numbers per address since
those ending between 00 and 37 are preceeded by a 3 and those
between 40 and 77, by a 2.
Line feed, return, and space are
separate instructions in the subroutine (JMP 2F, JMP 2R, and
JMP 2S, respectively).
2.
Double-precision subroutines -- These permit addition, multiplication and scale right operations for 23-bit numbers stored in
two consecutive addresses.
3.
Autocorrelation
1008 lags.
~
variance spectrum -- 4,0008 6-bit samples with
These are displayed on the scope.
If these are of
interest, a block number is typed on the keyboard and the correlation and spectral values are saved on tape in that block.
may be recalled and typed as graphs on the teletype.
This data
The mathemati-
cal expressions are given in APPENDIX III.
4.
Cross-correlation and cross-spectra -- Cross-correlation of 2,0008
data pairs, with 1008 lags.
Program also computes and types out
graphs of in-phase. and quadrature covariance spectra.
These will
later be expanded to compute the transfer fUnction (gain and
8-3
phase as a function of frequency) between the two statistical
inputs.
The mathematical expressions ar<7 given in APPENDlX III.
5. Miscellaneous -- Other more specialized routines, to be described
elsewhere, included:
random pulse generator, synchronization of
analog tape with LINe, reading movie images projected upon photocells, and reading amplitude histograms.
B.
Finger Tremor Variance Spectra
In terms of data collection, the major laboratory use of the LINe has
been for computing the variance (power) spect'ra of finger and hand tremors.
Robert N. Stiles is evaluating some of the mechanical factors which determine the spectral peaks of the finger tremor as part of his doctorate research problem.
The speed of data reduction which the LINe makes possible
has allowed us to determine what factors must be controlled to get reproducible spectra and to appreciate that this is a distributed mechanical
system with many vibration modes.
Programming for cross-spectral analysis
between tremors at two locations is just about completed.
When the position of a finger is being maintained there is a statistical fluctuation about a mean position.
This tremor has been· widely studied
in descriptive terms because of the interest in its origin and purpose and
because it becomes exaggerated in certain neurological diseases.
tions have traditionally used many
differen~
tent and of the amplitude of the tremors.
Observa-
measures of the frequency con-
Quantitative tests of specific
hypotheses concerning the source of tremor have been limited by lack of a
unifYing scheme which is mathematically precise but which will allow for the
random nature of the fluctuations.
Our present purpose is to put some
order and meaning into the measurement and quantitation of this fluctuation
by treating it as a statistical time series.
Variance is used as a
8-4
measure of tremor magnitude and variance spectra indicate the frequency
information averaged over a sampling period.
between tremor and other physiological signals
Cross-spectral analysis
m~
give clues regarding
the relative contributions of several possible factors to the total variance.
Tremor is sensed by a 2.5 gram variable reluctance accelerometer
taped to the finger of a subject.
The output of the
~ccelerometer
amplifier
system is a ± 1 volt signal which can be applied to the analog input jack
on the LINC.
Sampling rates of about BO/second are used and the samples
stored in memory.
Data has been analyzed in batches corresponding to a
few seconds of the tremor.
Longer samples are needed to be certain that the
statistical parameters have reached. some sort of a steady state.
Methods
of doing this are described elsewhere in this report.
Our attention has been focused on the center frequencies of the spectral
I
peaks
of acceleration tremor.
We have preliminary data which are presented
here only as illustrations of how the LINC has been used, and not as evidence
for or against any particular hypothesis of tremor mechanism.
peak, near 9 to 10 c/sec, is particularly consistent.
One spectral
Since this peak is
relatively sharp, often with a bandwidth of about 1 c/sec, the 10 c/sec
component is the one most obvious in a time tracing.
Much of the literature
on tremor has been focused on this particular frequency band.
However, by
sensing acceleration (which is more sensitiye to higher frequencies than is
a measurement of displacement) and using a kind of spectral analysis which
~s
sensitive to random, non-periodic fluctuations, we have revealed additional
frequency components which are significantly greater than the background
noise.
These other components are more variable and often have broad spectra
which would not be revealed by many of the methods of analysis used in the
past.
The center frequencies of these bands appear to be particularly
8-5
associated with mechanical factors associated with the hand and finger.
When all of the hand is supported except for the forefinger, the
broad spectral peak occurs at about 25 c/sec., as in Figure 1.1.
Adding
weights to the finger lowers the center frequency of this band, as in ....
Figure 1.2 where a 10 gram weight had been taped to the finger.
As more
weights are added the center frequency of this band falls, the relationship
approximating the smooth curve to be expected for the resonant frequency of
a linear mechanical system with lumped elasticity and mass.
This spectral
band also corresponds to the undamped natural frequency of the vibration
observed when the finger is externallY forced with a mechanical tap
(approximating an impulse).
When the arm is supported, leaving the hand and fingers free to move,
the broad spectral band centers near 15 c/sec (Figure 1.3).
This frequency
is lowered when masses are added to the hand, and corresponds to the frequency of the damped vibrations which an externally forced hand undergoes.
On some measurements where only the finger was free to move there was evidence of the 15 c/sec component associated with the hand.
One of the more interesting aspects occurs when sufficient mass is
added to bring the high-frequency spectral peaks down to below 9 - 10 c/sec.
In Figure 1.4 the hand was supporting a 450 gram mass and the two spectral
peaks (9 c/sec and 15 c/sec) appear to have'merged and moved. to about 6
c/sec.
But upon increasing the mass added to the whole hand the two spectral
bands separate again into one at 10 c/sec and one at 5 c/sec.
This suggests
some interaction between a specific load added to the hand and the origin
of the 9 - 10 c/sec component.
C.
Tremor Cross-spectra
The programs for cross-correlation analysis and in-phase and quadrature
cross-spectra are just being
completed.
These will be used to rest for
8-6
L'-8
BLOCK 0122
F'RE(~
0005
0010
0015
0020
TOTAL POWER
=
LSW
0225
PO WER/BAND WI OTH
x 0016
• •• ;.. 0014
•••••••• 0026
••••••• 0023
• •••• 0014
• •••• 0014
•••• 0018
• • • • • •• 0021
••••••• 0023
• • • • • • •• 002'7
• ••• :•••••• 0034
• • • • • • • • •• 00'34
• ••••••••• 0033
••••••••••••••••• 0061
•••••••••• •••••••••••••• 0089
••••
• •••••••••••••• 0078
••••••••••••••• 0053
••••••••••• 0039
• • • • • • • • •• •• 0041
•••••••••••• 0043
•••••
0030
• •••••••• 0028
• •.• • • • • • • •• 0038
••••••••••• 0037
••••••••• 0031
••••••••• 0030
•••••••••• 0034
••••••• ~ •••• 0041
•••••••••••••• 0048
••••••••••••••• 0055
••••••••••••••••• 0061
•••••••••••••••••• 0067
••••••••••••••• ·••••••• 0081
••••••••••••••••••••••••••• 0100
• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• 0131
• •••••••••••••••••••••••••••••••••••.• 0143
••••••••••••••••••••••••••••••••••••• 0142
t'Xj
.....
OQ
C
Ii
ro
t--'
t--'
t:dtll
~
I-rjI-3
Ng~
ro :s::
II II
ON
t-hO
.
..... :::0
::s
OQ<:
"'~
ro
n t--'
....... 0
I
til
ro
n
>
Ii :::0
+:-OQ
N
.)!
~
(")
tE:I
Ul
I-d
tE:I
(")
1-3
~
• • • • • • • • . • • • • • • • • • • • • • • • • • • • • ! • • • • • • • • • • • • • • • • • 0176
•••••••••••••••••••••••••••••••••• ~ ••••••••••• 0179
••••••••••••••••••••••••••••••••••• 0132
0025 •••••••••••••••••••••••••••• 0104·
•••••••••••••••••••••••••• 0098
••••••••••••••••••••••••
~
•••• OliO
•••••••••••••••••••••• ~ •••• '0102
•••••••••••••••••••• 0075
• •• • • • • • •.• • • • • • •• 0063 .
•••••••••••••••• 0059
•••••••••••••••••• 0064
OO~O ••••••••••••••••• 0060
••••••••• ~. 0038
••••••.•• 0025
••••••••• 0028
••••••••• 0029
••••••••• 0031
••••••••• 0030
••••••••• 0028
'0035 ••••••••• 0029
•••••••• 0026
••••••• 0020
•••••• 0017
L
•••••• 0016
•••••• 0018
nn",
0004
-
Figure
TREMOR VARIANCE SPECTRUM
Forefinger '+ 10 grams
s2 = 2 x 10- 4 g2'
1.2
C,\l
o
q".
o
o
o
·tn
BW = 0.6 c/sec'
·-
• 0\.
- ..
• •
.-II
•
0
00 •
-
o • •
• • •
• •• ••
\ ()\()
- . ..
q" .•
l/') _. • •
00
..· .•'.
-
•
• 00
•
•
• •o
o
00
\()
'0
•
o
\()\()
("')
--tq-t-t-O
•
.O\()("')
\()OOO
. ·'...
:lr::
l/')
..
•
I
o
o
o
0
•
• o •
• • ••
•
o
co
•
· '..
0
• • •
o •• •• .'•
• • •o •
• •
•
• •
(.\J
•
•
\()O\-tq""
•
•
("') ("') ("') ("')
0 '.' • •
000("')-0-0·· • •
\()\()\()OOOO("')("')("')
•• • •
(.\JC,\l(.\J
0000·'- • •
000'.'
000'·' • •
• •
0 0 0 •• •
• •
• •
• • • • •• •
• •
• • •
•• •• •• •• •• •• •• •• •• •• •• •• •• •• ••
• • • • • • • • • • • • • • •
• • • • • • • • • • • • •o •• •
'
•
• • • • • • • • • •
• • • • • • • • o • • • •• •• •
•
• • • • • • • • •
•
• • • • • • • • • • • • • • •
·0
..J
o
•
0_
• • • •
("')
(.\J
!II
(.\J
o •• •• 0
• • • •
Q\
o
("')
· .-
0\'
x
U
•
• •• •• ••
• •
• •
o
M
0"
•
o
(.\J
'
• •
"00
'000
• • 0 0 \()
.,
• ••
•
•
•
•
Q\ 0
OO~t.nO
0 l/')
••
•
0
•
• • • ••
0
0
•
•
•
•
•
0
•
•
•
•
•
•
•
•
•
•
•
•
•
0
•
•
•.•
C,\l
•
0\
•
o •
o ••
•
•
•
•
•
•
0 0
•
•
•
•
0
•
•
•
•
•
•
•
0
•
•
•
•
•
•
•
•
·•• ..•• ·
0
•
•
•• ••
•
·.. .
•
• • • • • • • • • • • •
• • • • • • • • • • • •
• • • • • • • • • • • •
••
o
o
•
•
• • •
• '.
•
• •• ••
• • •
•
• •
••
• • •
• •o •• ••
•
• •o •• ••
•
• • • •
• • • •
• • • •
• •• ••
•
•
• • • • • • • • • • • • ••
• • • • • • • • • • • • • •
-
• 00
'•• 0
•
•
• .....
••
•
•
•
• ~•
·• .••
.0.• •. •. •. •. •.•. ••
• • • • • • • • • • •
o
o
o
•
·· ..-•
• 0
: ....
• •
•
.0
0
•
e
• •
• •
• •
••
•
•
•
•
•
•
•
•
•
••
•
•
•
•
•
•
•
•
•
·.'....
· .. • ••
~
o(.\J
o
o
~
IX)
'000
·00.q'
·· .... •• •• • · ..
... '.
...
·
• • •
0
• •• ••
0 • • • • • •
••••
••••• • •
• •
•
..
•
•• ••
• • • • • • • • • • • • •
• • • • • • • • • • • • •
• •
•
•
•
•
·.· ... .......
··· ...' • •
•
Q\IX)IX)
·t.ntntn
'000
:000_
•
•
•
•
•
•
•
''f.
•• ~
•• 0
."
0
••
t.n
IX)
("').~
0 ('I)
\()OC
• • ··C,\l0C,\l
o·
0
0-
0-
.. • • 0 C,\l
0\ \() C\J C,\l IX) \() \()
00 '" ~
(\J C\1 '" \
····OOO··--OO---l/')l/')--l/')~oo~tn
•
·• .• .• .• .• .• .• .· ..' ...............
.
.
.. .. .... . .\... . . '. ....
, ..
·
.
.
.
.
.
.
, ..
• • • • • • •• . . . . . . . . . . . .
• · .• • • •
• .
• .
• .
• .
•.
•.
• .
• .
• .
•.
• .
•.
• .
• .
• I.• .
•.
•
• • • • • •• • .
••• · 0
"0000000--00--000-'
•••••
·'·00
000000000
00
• •.• • • • • • ..
• •
00
00"
0
'
~
• • • • •
If)
C,\l
o
o
• •
o("')
o·
o
l/')
("')
o
o
00
6~8
BLOCK 0205
F'REQ
TOTAL PO;WER =
LSW
0045
POWER/BANDWIDTH
0001
• • 0002
• • 0001
• • 0002
• • 0001
0002
• • 0002
0005 • •• 0004
•• ·0003
• •• 0005
• •• 0006
• ••• QO,1 a
••••••••••.•• ·0042
·.
X
0016
·.
.......................
........
~
0123
•••••••••••••••••••• ~ ••••••••••••••••• 0144
00.1 0 .••••••••••••••••••••• 0079
• ••••••••• 0033
• ••••• 0019
• ••• 0011
• ••• 0008
•••• 0009
0013
•••••• 0017
0015 • •••••• 0020
• •••••• 0023
••••••• '0023
• ••••• 0016.
• ••• 0011
• ••• 0008
••• 0006
• •• 0005'
0020 •• 0001
0001
• .-0000
• • 0001
• • 0000
• • 0000
• • 0001
0001
00~5 •• 0001
.• • -0000
•• -0000
•• 0001
• • 0001
• • 0000
•• ' 0000
• • 0000
oopo
0030
•• -0000
•• -0000
• • -0000
•• -0000
•• -0000
· •• -0000
•• -0000
0035 • • 0001
•• 0001·
••
-0000
6
•• -0000
• • 0000
• • 0000
•• 0000
......
~
.•
...
·.,
....
"'1j
()Q
c:
'1
(t)
.
t-'
w
::Il
1-3
nl,~
E..
~
~
~
S!
~
t%j
0004
'l}·'"'\,T~~, 8
. """'"
BLOCK 0266
TOTAL POWER
=
0231
LSW
w
F"REQ
00Q5
0010
0015
POWER/BANDWIDTH
X 0004
• •• 0004
• •• 0005
• •• 0007
• •• 0005
• •• 0007
• •• 0007
•••• 0010
••••••• 0023
••••••••••••• 0046
•••••••••••••••••••••••• 0091
•••••••••••••••••••••••••••••••••••••••••• 0162
• ••••.•.•••• • ••••••••••••••••••••••••••••••.•• 0164
••••••••••••••••••••••••••• 0101
•••••••••••••• 0051
••••••••• 0030
•••••••• 0025
•••••• 0017
•••• 0010
• ••• 0010
• •• 0007
• •• 0005
• •• 0006
0004
• • e
• • 0001
• • 0002.
0002
0001
••
• • 0001
• • 0001
• • 0002
~
....
• • 0002
OQ
c
• •• 0004
Pi
ro
0004,
• ••
I-'
• •• 0004
+:-••• 0005
• • 0003
:::r: t-3
• • 0001
III
~
::s
• • 0001
~
6
0000
••
+ ~
•• 0000
+:-VI
~
• • 0000
0
~
H
0000
••
OQ
>
Pi
!2l
• • 0000
n
III
tzj
!3
• • 0001
CIl
0000
••
• • 0000
• • 0000
0001
•• -0000
•• -0000
•• 0000
• • 0000
• • 0000
• • 0000
• • 0000
• • 0000
• • 0000
• • 0000
• • 0000
•• 0000
• • 0000
• • 0000
..
'
0020
.
0025
0030
0035
01
...
0002
..~ •. .,-
f'
ll,~~~~
BLOCK 0222
F"REQ
0005
0010
0015
.OO~O
TOTAL 'POWER
=
0066
LSW
0004
POWER/BANDWIDTH . X 0016
•• 0003
•• 0002
•• 0003
•• 0003
••• 0006
••••• 0015
•••••••••• 0032
••••••••••••••••••••••••••••••• 011'S
•••••• ~ ••••••••••••••••••••••••••••••••••••••••••••••••• 0216
••••••••••••••••••••••• ~ ••••••••••••••••••••• 0175
•••••••••••••••••••••• 0083
•••••••••••• 0042
• •••••• ~ •. 0031
•••••••• 0027
••••••••••• 0037
••••••••••••••.••••• 0070
•••••••••••••••••••• 007~
•••••••••• 0034
•••• '0008
•• 0002
t'1j
.....
•• 0001
OQ'
c::
•• -0000
ti
(t)
•• -0000
•• 0001
~
•• 000 1·
VI
'•• -0000
•• 0.00 1
.::t: t-3
Pl
•• 0000'
~
::J
0.
•• -0000
6
+ ~
•• -0000
VI'<
•• 0001
VI
>
0
~
•• -0000
H
OQ .>
•• -0000
ti
z
Pl
n
.
~.-OOOO
•• -0000
•• -0000
• •. -0000
•• -0000·
•• 0000
0025 .•• 0000
•• 0000
•• 0000
•• 0000
•• -0000
•• -0000
•• -0000
t!.-OOOO
0030 •• ~OOOO
•• -0000
•• 0000
•• 0000
.... 0000
•• -0001
•• -0000
•• -0000
0035 •• -0000
•• -0000
•• -0000
11
•• -0000
•• -0000
•• -0000
• • -0000 .
atil
tz:I
correlations between tremor and other physiological fluctuations as clues
of the sources of the tremor spectral peaks.
A few preliminary observa-
tions are given here as illustrations of the LINe performance even though
we are not in a position to make any physiological interpretations as yet.
The cross-spectra can be illustrated for pairs of signals of known
phasic relations.
The cross-correlation function for a signal compared to
itself is the same as its autocorrelation function.
are no quadrature spectral values
spectrum of the signal.
For this case there
and the in-phase spectrum is the variance
This is illustrated in Figures 2.1 and 2.2 which
are the in-phase and quadrature spectra for a 10 c/sec sine wave crosscorrelated with itself.
Figures 2.3 and 2.4 show that for the correlation
between a sine wave and a cosine wave there is only a quadrature spectrum.
One accessible signal to be compared with the finger tremor is the
. electromyogram, the complex summation of the action potentials arising
the many motor units which fire during muscular
a~tivity.
fro~
Figure 2.5 shows
the variance spectrum of the electromyogram measured at the muscle which
appears to be contributing to finger tremor.
The variance is spread out
over a broad spectrum peaking near 75 c/sec and showing very little activity
at the frequencies of the tremor spectral peaks.
If the electromyogram is
rectified and passed through a low-pass filter the resulting waveform is
related to the envelope of the
index of muscular activity.
in Figure
electromyogra~
and is often considered an
The power spectrum of this envelope, as shown
2.6, has peaks at frequencies corresponding to those in the
tremor recording.
This then raises the questions whether there is a consis-
tent phasic relation between electromyogram envelope and tremor.
The cross-
spectral analyses will eventually measure the consistency of any relationship
and perhaps the dynamic nature of it.
8-12
I
, <,a-
0
'C\l
0
o
o
o
'0
0
0
0
0
0
X
:c
q'
0
.bl
l)
Z
....
0
Figure
2.1'
CROSS-SPECTRUM OF SINE WAVB
Spectrum of in-phase components of
10 c/sec sine wave cross-correlated
with itself.
~
to
0\
f-o
0
....
3
-..
•• 0
•
•
v ,( rms )
· ...
- ··• ....•• ..•
• '.• . . . . . . . . . . . . . . . . . .'. . . . .
• ••• • ·
o
o
o
• •
o
o
o
o
o.5
s
• • •
• • •
• • •
• • •
••••
• • •
M
•
•
•
••
••
•• 0
•
BW
o
o
o·
•
•
o
-0-0
0000
0000
0000
.. I..,
1.2 c/sec
~
0 '- 0 0 0 -
00000000000'0000
.00000000
00000000
0 0 0 0.0 0 0
00000000000000
00000000000000
00000000000000
t .1 '. 1 I I I I 1 ., • I I I
I
•
I
I
I
'.'
•
0
Z
,0,
,00
(\J
3
Z
0
0
Q
....
3
(\J
(\J
- . . . . . . ... . . . ...••
0
0
0
0
0
0
O·
Spectrum of quadrature components.of
sine wave cross-correlated with cosine
~ave of same frequency.
0
0
0
0
,
I
•
• 0
·'0
:
•
•• •
• • • • ' . : t "• . • • . • • • • : • . • ':
.~.
0-0-00
000000
'000000
000000
tlll'l
WAVE
•0
co
·
0
0
0
0
0
2.4
CROSS-SPECTR.UMOF SINE
•
.3
co-
--•• •••
.'.-••
0.0
0
'~igure
..
co- (\]
•
OCOC?
0000
0000
0000'
I'"
•.•.•.• - • : • .,
•. . _ .... - -(I) - '
C? coco - 0\ CO 0 0 0 0
; co- .. co- - - 0 0 - 0 0 0 0 0 0
- 0 - 0 0 0-000'00000
0 0 0 0 0 0 0 ' 0 0 0 'f"'"
.• M
•
o, •
0
I
&I)
•
'
,
.
••
,
••
.
.
I
Q
Z
I
........
'
~
.
&I)
w
b.
0
0
0
0
0
&I)
0
0
.. • • (:)•
(\]
0
0
• • • • • • • 0• • • • '&I)•
&I)
'OJ
-0
0
M·
0
0
M
0
0
.. .
.
•0
~
0
0
.-
.....
0\
\0
o
o
o
Figure 2.5
I
"
VARIANCE SPECTRUM OF ELECTROMYOGRAM
(\)
'tn
1
01"-
~e
BW
~M~
....
·. .
• · ..
- •
o·
•
.i
oI"o
o
II
o
~
CX)MO
Z
bl
..1
0
t.)
If)0...i''f
0000
0
00
0 •
0
-(\)
q'
• • • • M~
• •
MM
00
00
00
00
00
00
•,
•
~(\)
t-_
0-000
000
00
00
00
00
00
00
• •
• • •
• • • • "f• • • .. q'• •·M-t-M
• • • • "f'r• • • • • • • • • • • •
00
00
00
• •
I
t-OO-q'~
0
0
0
0-0000
0-0000
000000
0 I I
,,
•
•
,
000000
000000
000000
I
•
,,,
•
::3
0
Z
00
'-
.0::
W
:::
0
a..
w
0::
:::>
f-
.as
(\)
0
~
t.)
0
.J
00
CO
I
f\J
)-A
OJ
W
0:
b..
• • • If)• • • • o• • • •:If)• • • • o• • • • If)• • • • o• • • • If)• • • • o•
o
o
o
o
o
o
o
C\J
o
O·
C\J
o
o
M
o
o
M
0
0
q'
o
o
A preliminary cross-correlation function obtained between tremor and
the electromyogram envelope is given in Figure
2.7. This shows evidence
of a correlation between these two quantities with a period of about 0.1 sec.
When
th~s
is subjected to a cross spectral analysis the frequencies at
....
Figures 2.8 and 2.9 show that
~pr
this particular case most of the correlation was at the 10 c/sec band.
It
which correlation exists can be seen.
is these kinds of records we wish to collect for a variety of experimental
conditions.
D.
Quantitation of Microcirculation
An important factor in determining the supply of nutrients to the cells
in a vascular bed is the
resi~tance.
to blood flow for that bed.
Flow and
pressure are usually measured in large arteries so that the estimates of
resistance are macroscopic.
Actually the total resistance is a complex
summation of a large number of elemental resistances, those of the arterioles
and capillaries.
Microscopic observations of this microcirculation indicate
that the flow through different vessels may be independent, the individual
vessels opening and closing apparently at random.
Also the flow through
any one path does not appear to be continuously graded, but more nearly
of an all-or-none nature.
This suggests that the total resistance may be
graded (to some extent) by changes in the number of paths open at any instant
and by the duration of the open periods for any one micro-vessel.
Thus a
description of the mechanism of change of the total vascular bed resistance
may amount to a shift of a statistical distribution in such a direction as
to change the number of paths open at anY one instant without necessarily
saying what will happen in any one path.
There is no quantitative experi-
mental evidence by which one can evaluate the statistical nature of the
microcirculation~
8-22
For many years the microcirculation has been viewed and, filmed,
,
particularly in the bat wing.
streaming or stationary.
Movies show the individual red blood cells
However, because of poor contrast, observation
-
of individual film frames does not distinguish between ,a cell and its back.
ground. 'The movement of cell with respect to background is what the eye
can detect.
For this reason there have been no manual measurements to
determine whether the flow is
grad~d
or whether· the flow, no-flow states
of different vessels are independent or c,orrelated.
The question of how to measure the statistical nature of the microcirculation had been discussed in our laboratory group for a nwnber of years.
a direct result of becoming
f~lliar
As
with the LINC and seeing those seven
analog input jacks, one graduate student in physiology, Patrick D. Harris,
had an inspiration about how such measurements might be done automatically.
He is working on the technique
~s
part of his doctoral research problem.
As the technique stands now, (with its unresolved imperfections) movies
are taken of the microcirculation in the bat wing using a microscope with a
power of 645X at film speeds of 16 frames/second.
After development, this
film is projected on a translucent screen giving an overall optical magnification in which an 8 micron red blood ,cell projects as about 8 millimeters.
High-speed photoconductive cells are placed
~n
vessel and on a background (non-vascular) area.
the projected image of a
Differences in transmitted
light intensity are sensed by the photocells and fed to the analog inputs
of the LINC.
The differences in transmitted
l~ght
with and without a red
blood cell are only slightly greater than the fluctuations in the nonvascUlar
background.
The major task at the moment is to obtain positive evidence
that the differences in photocell
volt~ges
sampled by the LINC are the re-
sults of a passing red blood cell.
8-23
Once the sensing of the presence of a red blood cell by one photocell
can be perfected, the measurements of several photocells can be combined along"
with computer logic to sense
1) spacing between individual cells
2)
velocity of individual cells
3)
the duration of flow, no-flow states
4)
probability of flow state for any particular vessel
5)
joint probabilities for the states of a vessel versus states of
other vessels.
If these latter two probabilities are identical, it is evidence that the flow
states in the different vessels are independent.
Since casual observations have placed the durations of capillary states
in the range of 1 to 30 seconds with a majority falling into a 6 to 10
second range, it is felt that 400 to 500 feet of film of a particular area
would be needed in order to have confidence in the probability determinations.
For 40 frames per foot taken at 16 frames/second, a minimum of 100 state
changes would be recorded for 400 feet of film.
The computer program has
been designed to take one photocell sample for each projector flash.
The
fact that there are three flashes for each film frame provides some redundancy "
for evaluating the variability of all instrument components beyond
the
film, e.g., projector vibration, projector light intensity changes, and
photocell noise.
For the present program, which stores a photocell sample
for each flash (3 per frame), a conventional reel of LINe tape will store
the
inf~r"'': "'~"
on for 500 feet of film.
8-24
3.
FUTURE PLANS
Thesis problems for two graduate students and one grant-supported
research problem are dependent upon the continuing use of the LINC for
the next one to three years.
In'addition, a former member of the LINrr'
group at Stanford has recently joined the Northwestern Medic'al School
faculty and wishes to make some use of the LINC here.
His plans are
given in detail in this section.
A.
Student Predoctoral Research Problems
Robert Stiles will be using the LINC to obtain variance spectra and
cross-spectra in the next year as part of his study of the mechanism of
finger tremor.
During this period particular attention will be placed
upon cross-correlation of finger tremor and other fluctuations and their
in-phase and quadrature spectra.
PatriCkHarris will be using the LINC to sense movement of red blood
cells as projected on photocells.
After development of the technique,
statistical information about the flow and no-flow states of the microcirculation will be collected, as described in a previous section.
Other students anticipated for future years, 'will very probably choose
research problems which can take advantage of this computer facility.
B.
Computation of Transfer Functions Between Statistical Signals
James Randall has plans to use the LINC as the primary means of data
analysis for his research grant "Stochastic Properties of Physiological
Time Series!', USPHS Grant HE 08516-02, which will run through 1967.
The
purpose of this study is to develop techniques of obtaining the transfer
function between two time-varying quantities using naturally occurring
variations as the forcing and the response.
The tremor studies constitute
8-25
one specific application of these general techniques.
A discussion of
the mathematical relations is given in Appendix III.
Measures of coherence (phase consistency) between two signals may
be considered as a correlation coefficient which is a function of frequency.
A conventional correlation coefficient may not sense the high correlation
between two signals within a narrow frequency band because of the presence
of large uncorrelated fluctuations at other frequencies.
Coherency is an
index of the consistency of the phase between a given frequency component
in two signals.
When the coherence is high it may be possible to represent
the relationship between the two signals by a, transfer function with gain
and phase as a function of frequency.
The cross-spectral techniques which
are being developed on the LING will permit this to be obtained even in the
presence of large amounts of incoherent noise in the signals.
C.
Neuropharmacology Studies
Dr. Charles Berry, Assistant Professor of Pharmacology, has had previous
training on the LING at the Stanford University Medical School.
He is inter-
ested in using the Northwestern University LING to reduce and analyze electroencephalographic and evoked potentials obtained in his studies of drugs and
neural system interactions.
His present effort in neuropharmacology is directed toward measurement
and evaluation of "recovery curves" derived from 'paired responses evoked
from the cortex of unanesthetized cats.
If pairs of identical stimuli having
interstimulus intervals of 16-400 msec are presented to the cortical surface,
characteristic alterations in the waveform envelope of the second local cortical response occur when compared with the first response.
When the ratio
of "second response"/"first response" is plotted against increasing
8-26
interstimulus time intervals, recovery patterns emerge which are interpreted as a measure of the excitable responsitivity of the local cortical
area.
Prior experiments have shown that visual, auditory, somatosensory,
association, and motor cortexes each exhibit distinct area-specific
r~9overy
patterns.
Until the present time the data upon which these. results are based
has been recorded photographically from a cathode ray oscilloscope and
subsequently projected from the film for measurement purposes.
It has
been proposed that use of a magnetic tape recorder in conjunction with LINC
might materially reduce this data reduction bottleneck and for this purpose
an Ampex SP-300 recorder has been acquired.
By utilizing magnetic tape
facilities and proper programming it is anticipated that LINC could be
used for the following purposes:
1.
Constructs of averaged response data could be generated for display
and computational purposes.
In the experiments described above
only amplitude measurements have been utilized as indicators of
neuronal recovery functions.
However, other measures such as
latency changes, the area within the response envelope, or alterations in the waveform pattern may be more appropriate.
Analytical
comparisons of this nature are very tedious without programmed
computational assistance.
2.
The construction of recovery curves would be based upon the appropriate data generated in 1) above.
Additional anticipated uses of LING not directly related to the research plan
outlined above would include auto- and crosscorrelation analyses of electroencephalographic data recorded from focal discharing epileptogenic lesions
8-27
and "mirror foci" prior to and following pharmacological manipulation.
It
is also planned to utilize the flexible capabilities of LING in attempting
to correlate characteristic waveform responses generated in the EEG with
the stages of behavior which develop in cats during the successive stages
of learning a problem in conditioned behavior
---'
8-28
4.
TRAINING PROGRAM
The strongest aspects of the program are the intangible ones which can
not be measured in terms of numbers of routines or the length of the biblio.....
graphy.
The single outstanding feature was the one month spent at M. I. T.
The intensiveness of this period, without outside
distraction~and
the
dedication of the C. D. O. staff to the task combined to make this a unique
experienc-e.
The first-hand contact with computer logic and machine language
broke down a barrier which assembly and compiler programs have always erected.
Having the computer in the laboratory has stimulated our own quantitative
thinking, and that of our stUdents, to a greater extent than any amount of
computer center service could ever do.
Looking back on the one-month training period we can realize that many
of the methods served purposes that were not then evident.
The holding off
of the LAP 3 until the last week forced the group to become familiar with
machine language to such an extent that this has helped in debugging programs.
On· the other hand, the hours and hours of reading timing diagrams for computer logic operations has not been productive.
Since this information was
so detailed and was not put to use, it soon slipped away.
If it had been
needed to keep the computer running this past 18 months no one would have had
the time to master these intricacies.
One
~an
not do good programming with-
out some knowledge of machine logic, but there is a limit to how much detail
will be useful.
The digital input-output capabilities of the machine are still not fully
appreciated.
There were many programming tricks and formats that have become individualized which complicates the interchange of programs between LINC users.
8-29
5.
COMPUTER PERFORMANCE
During the 18 months that we have had the LINe there has been only
one period of 3 to 4 weeks when time was spent on servicing the machine
rather than on normal operation.
The problem then (June 4, 1964 through
June 27, 1964) was eventually minimized, but never completely eliminated.
In retrospect it is now appreciated that the difficulty was marginal
operation of the tape heads.
This was manifested as a failure to get check
sums to match data sums on the WRC instruction.
At times one could not
use the LAP 3 because of a continual rocking on one block.
Tl1e~'
single
operation of cleaning the heads at the time of marking a new tape proved
to make the most difference.
Other factors as head alignment, grounding,
deburring the tape guide shoes, etc. contributed marginal improvement in
the tape signals.
At the present time tapes are always marked on the unit
which gives the greatest signal level.
Even with these precautions there
are many times when the tape must make several passes to get a correct check
sum on WRC.
After filling the tape with OIS with WRI instructions, the
checking is frequently hung up.
Once this difficulty was understood it has
not been a serious limitation.
The original input preamplifiers in the data terminal box were non-linear,
but this was corrected by a modification suggested by Bill
Simon~
This and
the above tape troubles have been the only difficulties that we have experienced.
In fact, since installing the
secon~
part of memory in November of
1964 the dust covers have not been off the rack nor have the power supply
voltages been adjusted ( or even checked).
has been used for a total of over 450 hours.
During this period the machine
This suggests that the design
is "physiologist-proQf".
8-30
We are not confident of the analog-to-digital ladder adjustments and
fear that they may introduce systematic errors in our data collection.
We
would be more confident if there were some objective, simple method of testing
and adjusting this.
The potentially high sampling rate fills the memory in a short time,
limiting computation to batches.
Our method of overcoming this, described
in an appendix, requires that the signal be recorded on analog tape.
haps there is some other
Per-
so~ution.
Since purchasing the teletype machine the camera has not been used
8
The camera is awkward to use, the film is expensive, and the manuscripts
too small to edit.
On the other hand, the teletyped manuscripts are very
popular since they can be written upon and also give an explicit memory
location for debugging programs.
Similarly, graphs of data which are tele-
typed are easy to file and read.
So far we have had no mechanical difficul-
ties with the teletype as an output device.
We have had no difficulties with the keyboard reading in wrong characters as some have had.
The keyboard has always been sluggish when the machine
is first turned on, but this has not gotten worse over the past 18 months so
it is not considered serious.
The teletype is not reliable as an input de-
vice and is too slow for convenient entry of LAP 3 manuscripts.
. At this writing the light behind the
but this button functions normally.
ex~
push button is intermittent,
Also there is an occasional momentary
intense whiteness to the 6-bit light of the accumulator.
This is most often
seen when entering the LAP 3 from the left switches and occurs just before the
accumulator reads the block number 300.
Only one or two of the input potentiometers are ever used at a time in
our applications.
It is possible that the rest of them could find use as
8-31
additional analog inputs, though this has not been a limitation for us.
Because of the nature of our computations, and the number of products
being summed, II-bits has not been sufficiently large for handling temporary
sums. 'Is there any possibility of automating some of the double precision
arithmetic steps?
The input preamplifiers in the data terminal box have an adequate input
sensitivity.
However, since laboratory sources have wide ranges of output
voltages, it would be advantageous to have more flexibility in gain adjustment at the computer.
At the present time this range is from about' 4.51 to
5.5X.
There should be a service organization backing up the LINC.
We question
whether it would be efficient use of our time in trying to trace down a subtle
malfunction.
Sooner or later something is going to wear out, and it frightens
us to be so highly dependent upon the machine and to be cut off from service
help if we need it.
In our experience, which has been fortunate, there would have been no
difficulty in handling computer operating costs from existing research grants.
There is same question whether one could justify the LINC's initial expense
as part of a single-investigator's operation.
6.
BIBLIOGRAPHY
There have been no publications.
8-32
APPENDIX I
SYNCHRONIZING ANALOG TAPE WITH LINC SAMPLING
The combination of LINC sampling rate and memory size means that
analysis may be limited to batches of a few seconds of data.
To be sure
that the batches were long enough, i.e., that the data statistical variability would not be much different for larger batches, a scheme was developed for transferring data on analog tape into the LINC memory and then
onto LINC tape by making two passes of the analog tape.
A total of 1.3
x 105 8-bit samples, corresponding to 20 minutes of sampling of a single
channel at 100 samples/second., can be transferred to one spool of LINC
tape.
In principle this could go on indefinitely by switching from one
tape unit to the other after all blocks are filled.
Our scheme consists of putting marking pulses on one channel of the
analog tape at the time that the data is recorded on that tape.
These
pulses are supplied by a coder-decoder designed and constructed by the
Northwestern University Biomedical Instrumentation Laboratory.
mounted on the mobile Ampex analog tape rack.
It is
Output of the coder consists
of alternating sequences of 256n pulses of +1.5 v and 256n pulses of -1.5 v,
n corresponding to 1, 2, or 4 at the option of the operator.
The pulse
rate can be selected at 50, 100, or 200 samples per second or can be driven
by an external oscillator.
analog tape speed.
data
samp~es
The maximum pulse rate was limited by the
These pulses determine when the LINC will take and store
from the analog tape channels.
During playback the analog tape channel containing marking pulses is
fed through the decoder which stores them to clean pulses of +1 v and -1 v
with equal off and on intervals.
These pulses and the analog data channels
are fed into the LINC analog input jacks.
A program was written which
stores data values in the LINC memory each time a mark pulse goes from
8-33
o to
+1.
After 256n of these samples, the contents of n quarters of
memory are transferred to n successive blocks on the LING tape.
The
program then ignores the negative marking pulses and takes its next
sample on the next positive pulse, the first of a new train of 256n.
This continues until one-half of the LINC tape is filled with samples
taken for each positi ve marking pulse.
The analog tape is then rewound
to the same starting place, and the LING program changed to take samples
on the change from 0 to -1 v in the mark channel.
The samples are trans-
ferred to tape during the positive pulses.
The LING program was designed to resolve certain possible ambiguities
which might occur by starting the analog tape up at some arbitrary point.
For example, it has to ignore starting the sampling in the middle of a
train of 256n pulses of the appropriate polarity and must start the sampling
during the second pass in phase with that of the first pass.
The test to
see whether data samples have been missed has been to display two successive
blocks of tape on the scope and look for discontinuities between the blocks.
8-34
Figure 3.0
Marking
pu~ses
recorded on 1 channel of analog tape, data on other channels
.
2
3
4
----
-
-
--
---
!
1
256
i -
_ ~l: ime -- -- ~-t ~-.~ ~-----
- lv
1
2
3
4
255
256
-I
I
+
FIRST pASS OF ANALOG TAPE (SNS SW DOWN)
LINC
--ignores
LINC samples data ChannelS+ LINC transfers memory f 'LINC samples
at start of each pulse,
to LINe tape then waits
and stores, etc.
samples stored in memory
.
-1__
SECOND PASS OF ANALOG TAPE
_
co
I
\.>J
V1
~I+-
LINC
ignores
____
LINC
ignores
---:---:-'- -.-
LINC slIMP. LES and
stores data .
-.
-
-+..
.
.
.
LINC tra.nSfers_
memory. toLINC_
tape, waits
.
APPENDIX II
SIMUIATION OF FORCING A DAMPED SYSTEM WITH STATISTICAL RJISES
One explanation for the 10 c/sec component of the tremor spectrUm has
been that it corresponds to the resonant frequency of a damped mechanical
system·which is being forced by the statistical manner in which muscle.....
units may fire.
It was not intuitively evident how changes in the
sta-
tis tical distributions of the forcing would be reflected in the response
from such a mechanical system.
The LINC was used to generate a statistical
forcing which was applied to an analog computer model of a second-order
system with a natural frequency of 10 c/sec and damping coefficient of 0.15
critical.
This is not presented here as a serious model for the tremor, but
as an illustration of the flexibility of the LINC.
The LINC generated a sequence of pulses of 20 milliseconds duration of
amplitudes which followed some statistical distribution and at pUlse-topulse intervals which followed some statistical distribution.
Two different
distributions were used, a Gaussian and a pseudo-random one with a flat
distribution.
The numbers determining pulse amplitude and pulse-to-pulse
interval according to a Gaussian distribution were obtained by sampling
the input noise of two Tektronix 122 preamplifiers connected in cascade.
The output of the second amplifier was about ± 1 volt, the amplitude was
Gaussian distributed, and the variance spectrum was flat over the amplifier
pass band.
Figure 3.1 shows the forcing pulses with such a distribution
and the simulated output of the mechanical system.
The numbers determining pulse amplitude and pulse-to-pulse interval
as distributed according to a flat distribution were obtained from a pseudorandom number generator supplied by William Simon.
and found to give a flat distribution.
This program was tested
Figure 3.2 gives the appearance of
the forcing pulses with these statistical characteristics and the resulting
8-36
output of the simulated mechanical system.
Figures 3.3 and 3.4 show variance spectra of the outputs of a simulated 10 c/sec damped system driven by statistical pulses with flat and
Gaussian distributions.
Note the difference in the center frequencies of
~.
the spectral peaks.
This illustrated to us the possible significance of
the forcing on the output spectrum.
This was further illustrated by
shifting the resonant frequency of the damped system to a lower frequency
and measuring the output spectra for the same two forcings.
In Figure 3.5,
in Which forcing was by pulses with a Gaussian distribution, the spectral
peak at about 12 c/sec is a manifestation of the mode of the pUlse-topulse intervals which was near 1/12 sec.
the damped system.
The other spectral peak reflects
However, for the flatly distributed pulse-to-pulse
intervals, in which there is no mode, the only peak is at the resonance of
the simulated mechanical system (Figure 3.6).
Once again, the point is
not that these are simulated tremor spectra, but that the LINe allowed us
to program a rather special pulse generator and to experiment mathematically.
8-37
.
Figure 3.1
,
GAUSSIAN DISTRIBUTION OF'AMPLITUDES AND INTERVALS
-----4-------~----~----'!~--~--------~~~----~----------~~-------
-
1-
-
-- -- --
- -
---
----- - ~----- --
------.~---~---~~--~-_l_----
-
-
----H-fH---iJ-,-----+~ft_tt____H__+_.,---I!_+-H__-
--- --
--·-~tr--1~-~-
-- --- r---I-I---+HI
~
f-
~~~
f"-
r- ~ ~ r--.. ~ ~ r- ~ r--- ~r-- ~
t
- - .-.--- --- .-------.- - ------.-------------t--
r- ~ ~~:~ ~ ~ ~ ~
I'
'
;.
"
r: ~ ~:r--- ~ r ~~ ~ ~
-.
- - - - . - - - .- ' - - - ' -
"
I
--~.----------.,...--------
---------~-------
v v
V
II
I:: ;
v
I
•
co
I
\.N
00
--
•
J
•
~-.------~--
~-
--__________ ~______ ~ _~ ____l~_,______---'-_
I
I
; I I . I!\,: 5TH 1,\1 I: ~~.....----,-OI-VI-Si..,..O~---,-C,F"..,C"""l"""Ev"""'ITE"...C=-:O,-:"R=-:PO,-:"RA...,T...,IO..,..N--.~t-::C~lE~VE:-:-l..,..,AN.",D-O=-:H~IO::-----::P-=-:RI.,."NT""~O::-I,.,.,N~U~S..,..A.;..,~.....-----------..,i---JI-------------~
w
co
I
I
!
•
!
I
'!
t
i
I
I
:
ii,
!
I
! ;
j
\
i
!,
i
:,
1
I
,
DisTRI~U~IOk
OF ~MPL1TubE AND! INT~RV1LS
,
:FLAT
: Figure! 3 2
I \
1
I
--
~~
r - I-
I
I-
i
I
!
I
.
I
\ !
I
I
I·
i
I
!
!j
i
i
~I--_ _---~ UUL
i
t
I
I
i
!
II
---~-'--------
..-. -.- -
-
.----.----- ---- --- --- -
t----.----~-----
,
I,
I
l i"
I
I
-------------. ------------------------t -I
OUT~UT
:'
II
,
!.I
I
I
!
I
OF
~)ECOND-batiER' S~STEM:
I!
I
. -(Nat ur ar-f~e(r~ ~ :c 7sec-;--:-aamp rn'g--c-o-et. -:=-q;"':-.1:"..,;,5...-)~:----+---~-I---;\--:----+-....,!--:---;---+---
'1\
-7Y-' ~-,
,
-
IIi 1 ;
1\
--~-
V'- -VVv-~
1\: .
~
fl
_\1\
r\
('v
,v :
A
,
11:' 1
: ~\
:f
j
-
'
i
I
A
i
:v:
Y:
-i.
II
I
'A
; f\ I
{\ :/ \!
Y:
\
VV'
.:/\
~
I
i
I
.1\
:
N
nA A
1"
I
I
I
I
I
1J
I
w,.
I
V;
,
Y,
v
I'
I
I'
i
I
iii V
1
i !
---+---t-!,---r-----t-
----- - ---- --- - --------------~-~__t__t__t_t__t------'l------'-----If--t+---f-+----++--- ~·--.J__¥__-_____v__+___tt______-ttl·r-j,
+-1
,
co
I
\..N
_':0 .
'---
---------
-----.--.----~----------
-----.--. r----- ---I
i
I
I
!
I
! !
I
I
t
1
I
:' i '
i !
~----~~~-------------------.~r------------------~~~-------------------.~~------------------~~
ii-
W
\0
M
o
o
o
3
Figure 3.3
en
.J
VARIANCE SPECTRUM OF OUTPUT OF 10 C/SEC DAMPED SYSTEM
(Flat distribution of forcing pulses)
00
0
0
0
X
'"
loO
M
0
II
0-0
-00
:c
MO
0
'"
.f-
....3
o
bl
3
• •
• •• •.
::;~ if Ior·<
FINAL REPORT
LINe EVALUATION PROGRAM
I.
II.
University of Washington
Department of Physiology and Biophysics
Seattle, Washington
J. t-lalter Woodbury
Professor of Physiology
Albert H. Gordon
Instructor in Physiology
III.
Laboratory facilities and personnel.
A. Facilities. The laboratory is equipped for doing research on the
electrical properties of excitable cells--notably cardiac muscle. The
most applicable experimental technique is the use of intracellular or
extracellular microelectrodes for measuring transmembrane or extracellular
potentials and for applying currents. The requisite instrumentation, e. g.
micromanipulators, stereo microscope, dual beam eRO's, wave form generators,
calibrators and preamplifiers with electrometer inputs, and stimulators
are all available and have been set up to enable data collection and
experimental control by the LINC in the laboratory.
B. Personnel. The laboratory is used by the principal investigator and
the co-investigator, by Dr. D. R. Firth, a postdoctoral fellow in Biophysics,
and by Mr. H. L. Hardy, a graduate student in Biophysics. In addition the
LINC is used by remote control from another laboratory by Hr. fll. H. Calvin,
·a graduate student in Physiology.
IV.
Present research in which LINe is an integral part
The present research of the principal and co-investigator will be
described in detail. The other projects using LINC in which others are
principally involved will be listed below and described in detail in a later
section.
A. The determination of the electrical ~quivalent circuit of cardiac
muscle during diastole; to investigate cell to cell electrical interconnections.
Data acquisi tion and preliminary analysis. J. H. V100dbury and A. !o1. Gordon.
B. ~easurement of the fluctuations of the interspike interval of the
crayfish stretch receptor as a means of investigating the excitability
mechanism. D. R. Firth.
C. Analysis of correlations between interspike intervals and synaptic
noise in spinal cord synapses. W. H. Calvin.
10-1
2
D. Theoretical calculations on receptive field characteristics.
W. H. Calvin.·
E. Dependance of conduction velocity in frog myelinated nerve
fibers on external sodium concentration. W. L. Hardy.
A. Equivalent circuit of cardiac muscle. A long thin cell such as a
nerve or skeletal muscle fiber is equivalent, electrically, to a lossy
cable. An intracellularly or extrac'ellularly applied current produces
changes in transmembrane voltage (V) which vary both in time and distance.
Records of V (x,t) and extracellular potential can be analyzed to give
the basic parameters of the cell. However previous research in this
laboratory has shown that cardiac muscle cells are electrically interconnected by means of closely approximated patches of low resistance
membrane, probably the tight junctions seen in many electron micrographs.
Hence the equivalent circuit of a block to cardiac tissue is much more
complex than that of skeletal muscle or nerve. A first approximation
is that the tissue is a one, two, or three dimensional equivalent (depending on the relative dimensions of the tissue) of a lossy cable. Preliminary analysis based on steady state distribution only has shown that this
approximation is a reasonable one fitting the data'with some accuracy. A:
more complete analysis requires the knm07ledge of both the time and spatial
variation of voltage caused by a step of applied current. The LINe is
an integral part of this present research. Voltage is measured as a
function of time and distance following the application of a pulse of
current either inside a cell through an intracellular electrode or
outside a cell through a suction el~ctrode. The later method for current
application has heen used in the present study because of the difficulty in
keeping two microelectrodes in cardiac ~uscle cells near one another and
of forcing enough current through high resistance rnicroelectrodes. He are
still in the data collection and preliminary analysis phase of this
problem. The detailed analysis will be described below in the proposed
research. The data collection involves the following:
1. A subthreshold step of current is applied to a bull frog atrial
trabecula through a suction electrode. The intracellular or extracellular
potential is recorded l07i th a microelectrode at a distance along the
trabecula. Records are taken both with the microelectrode outside and
inside the cell. The LINC is triggered to spmple both the current and
voltage at variable time intervals during the applied current pulse.
Both records are stored in memory and the pulse repeated. A predetermined
number of records are included in the average to get around the problem
of noise in some of the low level voltage signals. The usual sampling
is 32 microseconds interval (250 points per record) and 96 microseconds
for current (63 points per record). The records can then be displayed
with the record information and stored on magnetic tape or rejected.
10-2
3
Ancillary information stored along with the records are the resting potential
recorded if the microelectrode is inside a cell (both before and after all
records included in the average)t the injury potential measured by the
suction electrode used for passing the current pulse, the time scale for
the particular sample rate, the voltage calibration signal which is measured
with each record, and the. distance from the suction electrode to the
microelectrode (ca.n be set on a potentiometer).
The control features of the LINC a.re utilized in these experiments. The
LINe determines when the microelectrode is inside a cell as indicated by
a resting potential large enough to indicate little or no da.mage. The
senSing of a sufficiently large resting potential initiates all the sampling
procedures. This feature obtains records which would otherwise be lost
from cells in which the microelectrode stays for a very brief time (seconda)~
In addition the relays of the LINe are used to obtain calibration levels
or pulses at appropriate times. Since the voltage change resulting from
the applied current pulse is much less than the resting potentia.l, the
LINe also is used to increase the gain of the voltage mea.suring system
before the current pulse is applied.
2. After the voltage and current records have been stored on magnetic
tape, they can be read in at a later date for editing.
3. To obtain the cha.nge in the transmembrane potent ial in response
to the applied current pulse, the difference between voltage records
taken with the microelectrode inside and just outside the cell membrane
must be obtained. The appropriate records can be read back into the LINC
memory, the difference ob?ained, displayed on the LINe oscilloscope, and
stored with the proper values extracted from the difference record. This
procedure largely cancels ca.pacitative and reSistive artifacts due to the
current pulse and as~ociated circuitry.
v.
Future research in which LINe will be an integral part
A. The determination of the electrical equivalent circuit of cardiac
muscle during diastole, stea.dy state and transient analysis.
B. The use of this equivalent circuit to calculate the effects of
transmembrane current denSity on membrane potential from applied currentvoltage curves obtained throughout the prolonged cardiac action potential.
A. Equivalent· circuit of cardiac muscle. Steady state a.nd transient analysis.
We are writing programs to compute the solutions to the steady state cable
equation for the one, two, and three dimensional cases ( e-X,HO(jX), and
xH!L2(jx) respectively). This theoretical curve (one of the above) will
be displayed on the oscilloscope of the LINe along with the experimental
measurements of V(x,GO). The sum of the squa~d vertical differences
between the experimental points and the theoretical curves will be computed
and simultaneously displayed. A fit will be obtained by varying the vertical
and horizontal scales by means of the parameter knobs of the LINe AID
converters. These will be a.djusted until the variance is minimized.
The basic membrane parameters can then be calculated from the scale factors
10-3
4.
and other data on current strengths both outside and inside the cells.
Once the general form of the spatial decay of voltage is determined,
an attempt can be made to compute the transient response for each distance
and to compare these to the observed curves. From all this analysis, one
should be able to make reasonable estimates of the cell-to-cell resistance
in cardiac muscle.
B.
Ionic conductance changes during repolarization in cardiac muscle.
1. The experimental procedure is somewhat the same as described
above in the determination of the equivalent circuit of cardiac muscle.
An exception will be that the currents will be applied at various times
throughout the prolonged cardiac muscle action potential (ca. one sec.
in frog ventri.cle). The LINe can be programed to apply the current pulses
at various times.
2. The equivalent circuit developed above will be used in an attempt
to calculate membrane current density in the imlnediate region of the stimulating
electrode. Such calculations will give memhrane current density-membrane
voltage curves from which specific membrane conductance can be calculated.
VI. We would definitely like to retain the LINe in our laboratory. Much
time and effort has been expended in writing programs for the particular
problems mentioned and in making the LINe an integral part of the experimental
equipment and procedure. The data collection and storage) experimental control
and data analysis features of the LINC have been found to be extremely
advantageous in the experiments we have wanted to do. Our major complaint is
in the time taken for programming and this is almost completed for our
present problems.
We have been particularly impressed with the use of the LINC in experimental control. Experiments that would be marginal or impossible because
of difficult procedures involved, are made possible with the control
features of the LINC. The ability to collect and process easily large
~ounts of data is another feature of LINe that is particularly important
to our research. Averaging of our low level noisy signals is essential
to obtain good data. The 'LINe has become such an integral and essential part
of our research program that it would be difficult to do without it.
10-4
5
S~PPLEMENTAL
I.
INFORMATION
Past and present research
Several investigators have used the
were mentioned above in section 4. Only
co-investigator will be discussed here.
other investigators) Dr. D.R. Firth) ~lr.
are attached at the end of this report.
of the remote control use of the LINC by
LINe in this laboratory.
These
the research of the principal and
The reports on the research of the
W. H. Calvin) and Mr. W. L. Hardy
Of special note is the discussion
Mr. Calvin.
The use of the LINC in the determination of the electrical equivalent
circuit of cardiac muscle was described above in rather general terms.
More specific details of the data collection and storage program will be
given here. This program involved use of much of the input-output) external
sensing and control features. In all the program utilizes:
4 external AID channels
8 potentiometer AID channels
6 sense switches
4 external level lines
5 relays
both magnetic tape units
oscilloscope, 2 channels
The programs have been written for the 2048 word memory.
from a control program under key board control.
All are entered
A. Data Collection Prograln. The current pulse used in measuring the
equivalent circuit of cardiac muscle is applied either through an extracellular suction electrode or through an intracellular microelectrode.
In either case the potential measured by this electrode must be sampled. One
of the SAM lines samp~es the output of the preamplifier \~ith electrometer
input. A 50 mV. calibration signal is sampled on this line when a
calibration relay is closed. This same relay puts a 50 mV. calibration
signal on the input of the microelectrode used to measure the voltage change.
This calibration signal is used as a comparison level by LINC to determine
when the microelectrode is actually measuring a "respectable" resting potential.
While the investigator is manipulating this microelectrode trying to impale
a cell, the voltage measured by this microelectrode is sampled and compared
to the 50 mV. calibration signal. When the 'measured resting potential is
large enough) the LINC then proceeds to the voltage-current sample program.
This can be by-passe4 when the extracellular potential is being measured
with an SXL. The calibration signals and the potentials measured by the
current and voltage electrodes before the current pulses are stored with the
record for display later. After the current-voltage data is collected,
these potentials are sampled and stored for comparison with the initial
values. This enables the investigator to check that conditions did not
change during the data collection.
The main data collection program samples both the current and voltage
while the pulse of current is being injected. Both current and voltage
10-5
6
are displayed on a Tektronic S02A oscilloscope which is .used both for
monitoring and amplification. The output of the dual-beam S02A vertical
amplifiers are then sampled by 2 AID channels. Because the voltage response
to the current pulse may be in the millivolt range while the resting potential
is about 70 mV. the gain must be changed.before the current is applied.
This is done using one of the LINe relays to switch the output of the
voltage amplifier from a variable attenuation position to AC coupling.
The automatic performance of all these routine functions releases the experimenter for other functions and permits the collection of data which would
otherwise be missed; intracellular electrodes are easily dislodged from
the small cardiac muscle cells and records must be obtained rapidly.
The data collection program samples the current and the voltage at
rates which can be set from the .. keyboard. To make the program applicable
to both current pulses in the quiescent heart and in the active heart t
the sample rate can be changed from an initial fast rate of either one
voltage sample every 32 or 64 microseconds to a one voltage sample every
96 microseconds to every 200 milliseconds. This change:over of rate can
be set to occur at certain fixed intervals from the start to the completion.
This is done so that the rising phase of a cardiac action potential could
be sampled at the fast rate needed while the plateau an", ':cpolarization
phase coul", ue ~ampled at a much sloweL rale.
Tue start of the sampling is triggered by a pulse on one of the external
level lines. The delay from this trigger can be set to any particular
value up to about 16 seconds under keyboard control. To attain the maximum
sampling rate while sampling both current and voltage, the program is set
up to sample two voltage readings, one current reading, t~vo voltage readings,
etc. This causes no confusion of "holes" in the important voltage data
since the LINC averages a number of records before the sampling is concluded.
The starting of sampling is change in a cycle of three so that there will
be a voltage sample on one record where there was a "hole" in the previous
record. Thus to obtain two voltage data points at anyone time from the
trigger pulse, three voltage traces must be recorded. The number of
voltage records in the average is set by 2 sense switches with a maximum
of 24 or 16 "complete" records. The averaging is necessary because of the
noise in the 1m·, level signals. Much of this noise comes from the high
resistance (lO-80M~ of microelectrodes. In a complete record there are
252 voltage pOints and 63 current points. The usual sampling rate is once
every 32 microseconds.
A calibration signal for the voltage records is provided by closing
a LINC relay which puts 2mV. at the voltage input. The LINC relays are
slow to close, hence the AC coupling time constant must be fairly long so
that the calibration signal will not have decayed too much before it is
sampled
To get around this problem we have had the LINe relay close a
fast relay closing in less than 1 msec. Activation of this relay triggers
the LINe to proceed via an external level line. Thus the delay can be
much less. The calibration signal is stored along with the data.
0
The distance that the voltage electrode is from the current electrode
is set on a ~elipot which is sampled using two LINe relays to give the maximum,
zero, and actual setting. This information is stored along with the record.
B. Data Display and Store. Once the data has been collected, it
is displayed while LINe waits for a key to indicate storage on mag tape
or rejection. The data display includes. the voltage average, the last
voltage record, the last current record, a voltage axis generated by the
calibration signal, a time scale set by the sample rate, the cross over
10 _6
7
point where the sample rate changed the value of the vo~tage calibration
signal, the values of the time scales for the two sample rates, the delay from
the trigger pulse, the value of the distance from the current electrode in "
microns, the initial and final potentials measured by the current and
voltage electrodes with their 50mV. calibration signals and the block number
,·rhere it will be stored. All this is shown in Figure 1. In this figure,
the upper trace is the output of "the voltage amplifiers; the middle trace
is the average of 16 complete voltage traces; the lower trace is the current
pulse. Along the left side the voltage calibration increments are two
mV. as indicated in the upper left hand corner. The time scale is one
millisecond per division along the abscissa as indicated in the lower left
hand corner. The distance reading is 95 microns, upper middle. The
initial block number of storage is 6 as indicated in the lower middle.
The figures in the upper left hand corner give the delay from the trigger
pulse. The horizontal lines along the right side indicate the potential
that would be measured by the voltage and current electrodes with 5OmV.
calibration signals. The flow chart for this program along ~-lith the data
collection program is included in the appendix.
c. Display for Editing. Once the data is stored on tape, it can be
retrieved and displayed. The block number of the data read in is controlled
by the RS"l. This data can then be stored in the appropriate blocks under
keyboard and LSH control.
D. Transmembrance potential-voltage difference program. To obtain
the transmembrane potential record, the difference between the voltage
measured outside and inside the cell must be taken. This can be done under
keyboard, RSH, and LSH control. Figures 2 and 3 show an example of the
action of this difference program. Figure 2 shows the two records to be
used in the difference. The upper record is the same as in Figure 1,
corresponding to the intracellular record. The lower record corresponds
to the extracellular record. Both are averages of 16 records. Figure 3
shows the difference between these two records (averages) displayed with
all the record information as that mentioned for Figure 1 above. The
transmembrane record can be stored under keyboard control.
II.
Future research
A. Of most immediate interest is the calculation of the equivalent
circuit of cardiac muscle using the data collected with the programs just
described. This has been described in some aetail in the first section of
the report. It basically involves fitting of theoretical curves to the
experimental points. lThe theoretical curves will be of the for~m of
e- x , or H6(jx) , or xH- (jx) depending on the geometry of the cardiac
muscle section used. Itfte parameters used in these functions will be taken
off the knobs with some measure of the least squares fit to the data
printed out on the screen along with the parameters. This program is in
the planning stage. Data can be plotted out on a Moseley X-Y plotter by
a method we have previously communicated.
B.
Once this equivalent circuit is obtained, it can be used to calculate
10-7
8
the effects of transmembrane current density of membrane potential from
applied current-voltage curves obtained throughout the prolonged cardiac
action potential. This work is described above in the future research section.
Most of the programs written for the equivalent circuit study can be used
here with minor modification. LINC will be particularly useful in controlli~g
when the current pulse will be applied and also in calculating the effects
of the applied current pulses from the equivalent circuit. These experiments
are in the planning stage.
C. A couple of engineers here have wanted to set LINC up to obtain
and store transistor characteristics to assist in circuit design. This
seems to be feasible. The engineers are at present planning the sircuitry .
needed.
III.
Training program
The LINC evaluation program has been of great value in demonstrating
to many investigators the value of a digital computer inddata collection,
experimental control, and data analysis. Although digital computers have
been in use for some time here, \ve have benefited immeasurably by the LINC
program. \le have given two courses on digital computers and the LINC in
particular, several graduate students have made LINC an integral part of
their experiments and many others have learned more about digital techniques •.
The LINC is set up to demonstrate clearly many of the basic principles
of digital computers.
During the evaluation program, it bacame clear to us that one of the
main drawbacks was the amount of time taken in programming the LINC. More
assistance in programs of a very general nature and in encouraging and
promoting program sharing could have been helpful and still would be to
a limited extent.
IV.
Computer performance
A. ~ve have experienced no serious breakdowns with the LINC. Minor
troubles have come up with the keyboard picking up bits, but this has been
remedied by the new keyboard sent to us from St. Louis. Several of the
computer buttons stick occasionally and sometimes one of the switches does
not function properly, but these are once a month occurances and are minor.
Little time was lost in the changeover to the 2048 word memory.
l-IDEL 2 setting was found to some\vhat critical with the final setting
somewhat shorter than anticipated, 1.56 microseconds.
The
It is felt that the relays used in the LINC are too slow and variable
in their closing and opening times. We have as a result used these relays
to drive faster ones, ~~gnecraft magnereed dry reed encapsulated relays)
and sensed when these were activated using an external level. We recommend
faster relays.
B. In our present experiments, we are operating well within the limits
of the LINC computer. It is conceivable however that in the near future we
will feel the need for a computer with more powerful calculational capabilities.
10-8
9
In this case i"t would be convenient to be able to transfer the data to a
larger computer. This would require the production of·IBM compatible tape
or a direct communication system. The use of the expensive tape transport
producing IBM compatible tape is certainly a possibility as has been sho~m
at a number of labs. We hope to take another route by coupling the LINe
directly with a larger computer, the Raytheon PB 440. At present the two
computers are too far apart spatially. In the near future the two computers
will be in adjoining rooms so that direct communication will be possible.
For this procedure, the LINe is ~lell equipped with the OPR instruction."
which has a wide variety of synchronizing levels and pulses.
There may be some problem in attaining high speed in this data transfer
between the two computers. This comes because the LINe is not fully buffered
on the input and output. Thus the whole computer is tied up while the data
is being transfered. This is really not much of a problem in our
laboratory, but it does point up one disadvantage of the LINC. However, the
fully buffered operation is somewhat lvasteful of the computer and costly
since it does not make use of most of the available registers as do most
of the LINC input-output operations. In our experiments we have not yet
come upon situations where we would need this fully buffered capability
but this is not to say that we will never need it.
In some experiments) a larger memory is essential. We have not had to
face this problem. For those people with this problem it l'10uld be
helpful if the LINC vlere somel.,hat more modular in construction so that
extra memory, extra input-output devices could be incorporated with little
difficulty •
•i
l~e have gradually come to realize that the most powerful feature of
the LINe is the ability for experimental control. Our programs and experiments have evolved in this direction. The relays) e}::ternal levIes, A/D, etc.
make it a powerful tool in experi@ental control. These features should
certainly be retained. At times we have desired an output voltage (D/A on
sOme register) that can be held for some time while the computer is
sampling, etc •• As both the A and B registers are changing during this
time, some sort of buffer register is needed. Of course we could wire
this in externally. Hould such a register be feasible if it could be utilized
for some other task in the LINC (divide or fast storage and retrieval)?
'·lith the price of transistor circuits falling daily) it is conceivable
that a computer with greater computational capabilities could be built
for about the same cost as the orginal LINC~s. Modular construction for
easy expansion of memory, input-output, could be of great advantage. The
big question centers around how well this has been done already and whether
this can be improved on.
10-9
.LV
v.
Log Book
Attached are the important pages of our log book. TIle other
pages contain copies of your letters to us on equipment changes,etc.
VI. Bibliography
The
on~r
published research which has used
Firth, D.R., Biophys
i..,
LING to date is
Abstracts, Annual Heeting, 1965.
10-10
Plov/ Chart of Vol ta·G9-Current
Sarnp 1 ins Pro f.r arn
---- ._----_.-.....__.------_.----.
Set Sa~n~)linc RD. te and
cross over - fast
SA~ Key 1, sloITer
SAI·!~ ~«e·~r
2
SET tri~~er delay
i
~(ey
384
r
I...;w......w-~.....:...
_ _ _ _ _ _ .____
J
it
.s:ST if records in avera:3 9
[SHS 0, 1
I'~~:---------
!Cl~-;-;:'-;~~~;;~~~=!
!VaH fo!' ;(~'y for custancei
Control read in ; <.:__ Q.. _____.____________..0:E:¥._~.~~~.~.!'j ---)..IGe t
B:fQ
lch~-~~-t~--;~-~-if-J.o6o----I.&,",1icl'lon vnl1J.e is 0
I,
; ~; tor e J.~!1I.L~- if c1:l stan c e !
!store 1111 if not
I
t-. ------------....
-.--.--.--.
.
.
v
I
IGet
~'.
zero
zero
\Get lCrna
8ettinr::-~_.§.~li:._~iJJ
'v
mic-ron--~
Isettlng '¥R
1
Get actual
~ ~ L~, R 5
5
distanc·e--~
i--.-:..-._ _ _ _ _ __
:r:or_B.!_~.~]·--7'I;·~sxI~~.'
,~-----__
r'
I' -.£rlQ-----..-~~~--~ --7\ Get zero for ... ~_~.J
.__ -;). . .rte la-y 3 ,:)O".nV cal. IP
SOmV
i
'W
r~l1m1~
._ _..J,-_ __
i
I
Cal. RP l -1 Skip
.1:_ __
~·l~ N~, -----~>
I
RP :7
Yes
..-
RP>
50lliV-,
50mV I
1';0 i
t
.--~--
•
Chan. r.;o ;I'.,ain AC I <:::
.._.'._" -----..1
I
'i,l
R2
_
Icou~lins
--+.----.- '\: ... -... '. ".... -.-----------. "-_·· . ·. . . ,--._._1
t·;J:l~·i~i·r~-~_-~·.-(~e lay S;{L~~6·-~------------.v._ _ __
t~$A:·:;.pM t'! Pus t e s f-8~··-s··foreV&I·J
.------_.~._'---------
LS!\I\: variable &.
storey'&I
J .'
10-12
linclude~!~i§orc~ In .!ve~
.
..__
~.
.~_~.J
f~~~ex-anr~~ef;::renO·Uf,h-r·~~~:~:
~hanr;e
__
1
TOt"'ii~
c;ain back to ori~;inal, R. 2
.J
-SA101 R.I., I. P.
SAN 10 & 12
I
after I
I
1!3t~~~~~~~:~~~~d3et
c~
J
tSAT:1 lO-zel'o ca.~ .•
1
rclose-' Ri, 112, .~:~~~al-t~·
I
---------\:y'--.-_._... ....... .
C01.;1-:;al"s new cal to old
within 3?
10-13
Apr:f.l 13. 1964
To: S. M. Ornstein
From: J. W.
Subject:
~loodbury
Plottir~
and A. li. Gordon
data from LINe with an analog X. Y plotter
Your memo of March 24, 1964, item 3 prompts us to report a simple
and (if 'We do any so ourselves) clceant means of m~king X. Y plots on
any an~log X. Y plotter that ltas provision for raining and lowering the
pen. The method requires no ~~:tra equipment in addition to the X. Y
plotter and no altc~ations to the LINC. Further, tha plotting rate is
controllable from the LINe console and can be adjusted to fit the nature
of the plot. i.e. several different curves can be plotted at once if th~
rate' is mad<3 Gufficiently 010\\1.
Proc~du~:
The X value to be plotted io put in a beta rC3ister
(Gay 1) and the Y v.:l1uc put in the accumul~tor. (2) The plottin~~ Qrcl.(·r
is SET 2, OnOl~ ~~c(.r~ 2 is any b(.~ta resister !l..Q.£. ~ elsc':'7h£!,£ in the
program. (3) The left. switches .are sct to 0002 a...··Hl the ~{OE button
pushed. (4) T11e st~rt prog~~ button is pushed; LINe will stop at 0002
with Y in A and X in B. ~hio is the first point to be plotted and analog
_
values of these numbers appear at pins A (Y value) and 3 (X value) Oll. the f."V\
right 1u1nd display scopa pluz-in"unit plug. (5) Tbe Auto-restart button
',JD .
io pushed and the prog~am rcsu=es until the p~o3ram reaches the XOE stop . ~ .
at 0002 and the next X and Y values arc in the B and A registers rcs~act- .
ively. Thereafter. successive valucsappcnr aut~tically. The Autorestart froqucncy can be varied according to the plotting job being done.
if the points a~e close together and the pen docs not need to bo lifted in
between successive points) the rate can be rapid, etc.
e ":"
~ rai9in~ and loucrin~.
The pen can ba raiDed tlnd lO'tvered by m~c.l'lS
of the ATR order and connection from the propC4 relay terminals
the data
terminal box to the pen lifting circuit on the X, Y plotter. A sample
. progr~ for lo~ding a point, lowarins the pen) then plotting tho valu~s in .
successive memory loc~tions with the pen do'C·m nnd raising the pen at the cnd
of the program is attached. Clearly, the points can bo calculated between
plotting o~dc~s. but this becomes too clow if the calculation ttmc p~r point
is appreciable. This method has bean used successfully on the Ho~y Hodel 2D
nnd the EAI 1110 X-Y plotters. A Dpiral, plotted on tho ~fosoltly is enclosed.
on
The distanco bQ~vCBn points is too larac. account1ng for the slightly
nature of tho lina.
~avy
10-14
s.
M.-Ornstein, page 2
Scope display. The data to be plotted can be chccl~ed on the scope
unit with the same program simply by increasing the Auto-restart rate
(with X) Y plotter turned off) and increasing the intensity of the scope
beam by mc~ns of the screwdriver intensity control. The program onclosed
cnn be made repetitivo by cl1anging 0062: JMP p-2 to. 0062: JMP 20. _ This
could be done with a SNS switch.
Additionally, only slight modifications are required to plot out any
data displayed on-tha scope.
Sincerely yours,
J. W. Woodbury, Ph.D.
Professor
J'WW: ch
Enclosure
10-15
Program for plottingY against X
on an analog X, Y plotter
X values are stored in 100 to 777.
values are sto~ed in 1100 to 1777.
Y
Left switches set to 0004.
Start 20.
0020 CL..lt
0021 ATR
0022 Set i 1
0023 0077
1 11.
0024 Set i 2
0025 1037
0026 LDA i 1
~~
0027 STC 3
0030 u)A i 2
- 0031 STA i
' -it'"' v
0032
0033 SET 4
0034 0003
X value in 3
Y value in A
X vAlue in B
XOE Sl'OP
---------------------~----------0035 LDA i
0036 0001
0037 ATR
Close relay. pen down
0040 LDA e;' 140)
:DiAL
])'''L
~
/77 -
Sn"(ICU'I3~
-r;,Nd40')~[
"P.-W5<'- (0.
n~) {oJr--------"
- - _..- -....,"'----.........--,,_ _
,!'---y-----J'--of'---I'--'--~
7 MU$\:!"
:)...., ,,.c:.tt T()
D;AL.
IV~" t>,~~
K \!L\J~
w,,.. . . . B~L.c
Program:
('0,1"\(..
"2. .. )
:biA'-
1?~T'N<:' "7..-.: 'TjC.~;
g10C..-n...-r--
#lD LDA
~
TP
e"... liS.
o
STC 2D
SETi12
o
SA11 11 ~
ADAi
set pulse counter to zero
control channel
-40
APO
if V is not above Pause region (0-37>,
it exits \'lith ACe set to -0
Jl1P 3D
ADAi
-!~O
APO
JHP p-10·--l
SETi15
1700
XSK115
~
JMp p-l .....
if in dial tone region, waits for dial
step or eeturn to Pause region
time delay
10-29
page
Dial Reader Subroutine
#5D
SAM 11
ADA!
-100
APO
JMp 4D
ADM
dial step completed?
yes
-40
APOi
at pulse level? (140-177)
yes
no, stiil at step. Is this return fram a
/yes/ pulse (is pulse switch up)?
no
set pulse length to zero
JMP p+6
XSK 13
JMP 6D
SETi14
o
J11P 5D
SETi13 ~
o
set pulse switch up
count on pulse length
XSKi14
JMP 5D
JMP 3D
set oupput to -0 if pulse length reaches 1777
1777
get here at end or a pulse (with switch up),
so set switch down
14
get length count
-100
length criteria (about 10 msec.)
#6D SETi13 <:LDA
ADA!
APOi
XSK i12
J11P
long enough, so tally pulse counter
not long enough
if ever get 1777 pulses (1), exit with -0.
SD
JMp 3D
#$D LDA
12
get pulse count
AZE
Jl1P P4
4
#3D t~!i iDa
-0
SAEi
12
JHP
IJ2D
CLR
JMP
J
NOP
not zero
if zero, returns 'be start (!X;-t5
entry for -0 exit
<;
P+2J
~
ten pulses? If so, set output to 4- 0
out to p+l of originating program
10-30
Flow Chart for Dial Reading Subroutine
_. _-- ._ J___~
C_.. _. . . _... j .............._.
set count to ~~~
c::2~nma 1
~
yes
Tone)
_~.
r--6>£:>--
~
(~~-d~1a!)
____ 1_____
t---------~~~-----------Cn dial
·1
1/
!:
~p~~ level?
------<---~
.~.--.-..
~
(Goes-t~' LINOTYPE
(lAf\}-@.:\
-f-S>---=~---1\ subroutine to type in '_
'-three line legend
_
:~ i'-
-1-r-
lQ
Y4
(DisPlay
LEGEND~)
fa
type three digit o~t~l\.
BN. Reads BN into €P . .t>. ~
and ;ON 1 i n t o .
....-//
other charaoters
have no
effe~
L@
Flow Chart for HISTOGRAM Proper
~
?G~
"-
~
~-'- - - - - .
~ ce.lcule. te bar
: width from
\J.?1ob 7
i
1\
I
calculate
vertical scaling
factor 2 n from
knob 5
ISNSi~
Ie--
I
•
step x jcount on width
test for end of
screen (777)
L
~cl>
J}1P p-l
stall for
photography
.
.J.
set first bin location
"
set first x-coordinate
J
display bar from
y dOrm to -337
.-1 ..
/Count
8-
~
on
'\
( bar width;
)
{ last column '
"'- of bar don.a?,/
~ ~,
set bar width
set bin width
I~-·
[
!
space one
to leave
\
room betwj
bars
---
scale ,dth vertici\l
scaling factor 2
.~
/.l
~I----------------------------------~~--------~
SNS 2 turns off coordin~
lines; cycle thru them
6 times
Coordinate \
Lines
Coordinate Lines Flow Chart
~~
'y..",,-~
__
get coordinate
i -_ _ _ _--i:3>4-....,_ _ _ _ _ line from table
previously typed~.------~------------~~
U
I in Hi th 11 N
J
.
v
last ~-~~dinate
line
dono?
,
I-i>-!
df~plaYn
L. ___.
luBec.
'-r---~
L~-8
Horizontai/
or Vertical~
/
HORIZONTAL
[
.
t
VERTICAL
calc~"--l-a-t-e'-l--x-position
of line
.,
Calculate
dot density
' - - .--.-
D
+:---.-. ·.~tLl
calcu13te I
y-position .
of line
I
I
F==-==-:="---Y'~~:;=~':::-::==::::::;,
1
11
:
Calculate .
dot den~ity'
I
I
r ; : - -J,.
- ' - - '".--''' ..
- '-'"
I
. in horizontal1j
'___._______ ._2J
l
. line starting
at 5L~ and
I
ending at
"----------~
'Di~!liaY-~b;~issa\
value in msec. I
beneath line
~
-~
~'Di sp~ia Y:·"·dot"fil
Display dots ;
in vertical
line starting;
at -340 and
ending at 237;
L
....-"-,, ..
t.i
I
1--1___~D * r~
I
I
Octal-to-decimal
I
---I
Display ordinate
value in spikes ,
'Oer bin at left !
end of line
,_
~l
I
777
j
I
I
I
I
Ii
i
L---+r;.:..::.-----J
\
an~u~;.~~t;!~
lC
V b.
. (~~~~~1~~'--~-~-e-Ri-z-e---
Sum~ary of Operating Instructions
HISTOGRAM with Invariant Coordinates
!n
knob 5 scales amplitude by 2
knob 7 ~ets bar vTidth and thus horizontal scale
SNS 0 stall~ for taking time exposures
SNS 1 up turns orr legend
SNS 2 up turns off coordinate lines
nAn integrates bins 3001-3777
"B U sets bin width (type one octal digit)
"en reads and and transfers control to Control Program (0/0)
"Du delays first bin displayed: type in 3 digit msec. delay
"1-111 jumps to LINOTYJ?E to type in legend; get back after 3 EOLs.
UN" type in coordinate line (4 digits)
.
first digit sets for vertical or ho~izontal
and sets dot density:
number
a
line
vertical
If
1
II
2
3
~6
11
horizontal
II
II
11
7
dot
sEacin~
E:Dt allo':,:ed
20
10
4
40
20
10
4
next three digits are location of line in
milliseconds or spikesjbin (type indecimal)
a vertical line of minimum density at zero
spikes/bin is not allot.J'ed since this would
come out 0000, vThich serves to terminate
displaying of coordinate lines.
UN" followed by ,"del" deletes laRt coordinate line.
IIT" reads tHO blocks from tape (type-three digit octal BN) unit 6.
BN will. go into fjp; BN-rl into 'tl.
Data is kept in quarters 6 and
7t
10-43
HISTOGRAM with Invariant Coordinates (Mark III)
tv. H. Calvin
University of Washington
Department of Physiology & Biophysios
origin 14
3X-l
NOP
RDG
#5K
3010
location of progrrum on tape
KST
J}1P 2G
KBDi
SAEi
24
JMp P~2J
JNP 2F
SAEi
26
A for integrate
J
C for oontrol blook
J11P p+2
JlIP fjrt 3F
SAE!
47
JMP p+16
JMp 2K
T for tape read
3 digit ootal keyboard subroutine
ADA!
6000
places data in Q)
STA
p t5
o
ADAi
1001
STC p+4
:J
RDe
reads BN-f-l into r:tl
RDC
~1P
SAEi
40
2G
J11P ~X
JMP lA
#2F SET.i2
3000
LDAi2
out tohlstogrrum proper
M for LINOTYPE su broutine
LINOTYPE
~ii~ <)
JMP p-2..l
Integ~te
starting at 3001
~~dfng at
3777
Jl-1P 2G
#3F SETi17
Out to Control Program
~1P 16
#2X SAEi
Read in oontrol progrmm at 0/0
o
41
JHP
KBDi
P+231
N for ooordinate line
SAEi
13
delete?
10-44
page two
HISTOGRAM with
Invariant Coordinates
p+lO
JMP
eLR
STA 14
LDAi
-1
ADM
delete last line
14
set back counter
ROL 11
STC ~+3
~4P 4K
ADAi
shif~
JNP p+7----;rt;
first digit into left position
three digit decimal keyboard subroutine
STAi14
~
JMP 2G
SAEi
25
B for bin width
JHP p-1-13
KEDi
AZE
P:3]'
JMP
LDAi
1
if type zero, set to one
STA
lY
CO}!
STC 4B
j
J}1P 2G
SAEi
<.--1
27
D for delay in first bin displayed
JMp 2G
change this if ever add additi6ngl letters to test
J11P 4K
:Bhree digi t decimal keyboard subroutine
MULi
1
for 1 msec bin Width; change to 5 if use 200 umsec.
C011
basic
notifies Coordinate Lines of/bin width if not 1
STA
6F
C011
ADAi
2777
STG IS
location of first bin displayed
JHP 2G
#2G SAl"\! 7
get bar width from knob 7
ADAi
200
make positive
2
range is 2-9 (change to 1 if eliminate bar spacing)
SCR 5
ADAi
STA
p+6 -"I
COM
f
'.)/
lO-~-5
J
ADAi
~ggo3B
LDAi
#lY
"
NULi
1
reciprocal of basic bin Hidth (5 if 200 usee bins)
tells Coordinate Lines where to locate, given msec.
get vertical height scaling from knob 5
5H
SAl1
APOi
makes actual bar width one less th~ spec1f1ed,
to allow for one space between bars. Change
to 1777 if eliminate spacing.
bar width
multiples of basic bin width
~fULi
SPO
:Qage three
HISTOGRAM wi th
Invariant Coordinates
5
JMP p+6
OOM
z. -
SOR 4
ADM
SORi
cTI1P
SOR
ADAi
ROLi
STA
p~J
4
23
STC IG
SNSiO
~
JNP p-1J
#3G SETi2
#IS 2777
"1
manufactures shift instruction for later use
for use in Histogram pDoper
for use in Coordinate Lines
serves to halt computer after one complete
loop thru the program for uniform time exposures
first bin location -1
SETi4
1054
x-coordinate of first bin displayed
1f4G SETi5
#3B
#4B
bar width iii:
SETi3
-1
LDAi2
XSKi3
J}1P
J}1P P'.j.2~
2S
ADAi2 (-.t'
add suceeding bins (if any)
J11P p-2 J
S CRi n "«
or ROLi n. (set earlier)
XSKi3
1123
bin width
get bin
count on bin width
°1
LZEi
cTMP P~3J
ADAi
1
ADA!
-737
APOi
CLR
ADAi
400
STA
<=
round. up
net effect i9 -337, but keeps bar height from
exceeding screen height and thus going around
again. If bin contents are so large as to
exceed ~, this obviously doesn't worle.
4737
4H
10-46
page four
HISTOGRAM with
Invariant Coordinates
#5G XSKi4
<:-
Jl1P J>T 2]
JMP IF
LDAi
#4H
<.
~t.4 ~J
x-coordinate of bar
last x.
Go to legend
& coordinate line control.
y-coordinate
-1
SAEi
-340
J}IP
fills in bar down to zero
p-.5
XSKi5
count on bar width
5G
XSKi4
JMP
JHP
..... J!.!L'O
(-337)
Change to XSK 4 if
you wish to eliminate spacing between bars.
_:xmz:qUct~
4G
JMP IF
--.~
Legend & Coordinate Line Control
sense switch i up will turn off legend
#IF SNSil
J11P lL
SET! 10
g~e~x~gR.Edxd~~~~~~~~
l7IDm
XSK1l0
JMP p-J..2
J11P p .. 3 1_1
JHP 1Q
.J1-1P p-4
. #lL SNSi2 ~
I
JMP
<11
J'
5K
SETi13
1770
XSKi13
JMP IH
J1-1P 5K
sets number of times legend is displayed, so that
it will appear the desired brightness in time
exposure photographs •
jumps to legend display ~ubroutine
rettmn from II
"
II
sense switch 2 up will turn off coordinate lines.
back to start
sets number of times that Coordinate Lines are
displayed relative to Histogram proper.
jumps to Coordinate Lines. Return is to p-l (XSKi13).
back to start.
lO-~·7
Coordinate Lines
(part of HISTOGRAM with Invariant Coordinates)
page five
Table of coordinate lines besins at 3X.
Each coordinate line is specified by one word. The b1gh~three
bits speciry the line direction and the dot density. The low
nine bits specify tp.e spikes per bin or the milliseconds.
#lH SETilO
table of coordinate lines
3X-l
LDAi 10
AZEi
7G+2l
APO
Jl1P 3H
#2H SCR 11
ADAi
SCR
STC P....
LDAi
40
J
SCR
STC 6H
LDA 10
BCLi
zero denotes end of table.
Go to print "msec. U
Indicates horizontal line (4,5,6,7)
get highest three bits (will be 0,1,2,3)
VERTICAL LINE t,msec)
manufaoture shift instruction
widest dot spacing
get lowest nine bits
7000-
#6F
ADAi
HULi
#5H 1
ADAi
reciprocal or basic bin width
(5
if 200 usec. bins)
54
x-coordinate or first bin displayed
x-coordinate of vertical line
-340
start veryical line at -340
STC 4
LDAi
DIS 4
1ihli ADAi
#6H
add delay in rirst bin displayed
add dot spacing
ADAi
<40,
20, 10,
4)
-237
APOi
J}1P P'~4J
ADAi
stop line at 237
237
JMP p-ll LDAi .t;
-4
ADlwf
4
SETill
-377
LDA 10
BCLi
7000
JMP JB
JMP IH+2
subtract 4 from x-coordinate of line so as to
place absmissa value under line.
IR 11 holds ~-coordinate, IR 4 holds x-coordinate
and ACC holds octal value of abscissa for
DSC SUbroutine., with octal-decimal conversion.
page six
Coordinate Lines
#3H BCLi
4777
gets dot density (0,1,2,3)
SCR 11
ADAi
SCR
m~ufactures
shift inRtruction
STC P.J.3 ]
LDAi
40
SCR
widest dot spacing
<-
STC 7G
LDA 10
BCLi
7000
set earlier (multiplies by 2±n)
#1G SORi n
ADAi
-340
y zero at -340
CLR
clears link bit
STC p ... 6
LDAi
5L~
~tarts
4
STC
horizontal line at x=
54 8
LDAi
DIS
LDA
4
4
#7G
ADAi
add dot Rpacing <40,20,10,4)
ADAi
Atop at end of screen
-777
APOi
JMP P+4.~
ADAi
777
JMP
-16
SETi4
o
<.-
SET 11
IG+l1
\.
ordinate label
set x-coordinate of first number to zero
y-coordinate of horizontal line & of ordinate label
LDA 10
BCLi
7000
Jl1P ID
cTHP IH+2
SET i4
700
SETil1
-377
SETilO
5Q,..1
SETi6
-5
to DSC subroutine with octal-to-decimal conv.ersion
tf
msec ." label
page seven
Coordinate Lines
LDAi 10
JMP 10
LDAi 10
J}fP 10
LDAi
3
ADH
4
XSKi 6
JMP p-ll
JHP lL 4
#5Q 3077
7730
5121
4651
5177
4151
4136
2241
0001
0000
DSC subroutine.
space of 3 between p~~f~tl letters.
count on number of letters
M
S
E
C
•
10-50
Octal-to-dec1malling, zero-surpressing
Display Decimal Numeral of reduced size
Subroutine
Displays decimal numbers from 0 to 511 with leading zero supress.
X-coordinate must be in 4, y in 11, number in ACC.
#lD STC 2D
ADD 0
STC 7D
SETi5
o
set hundreds register to zero
LDAi
#2D
CO}!
ADAi
144
APOi
Jl1P p,l).
tallies hundreds register
XSKi5
J}lP p-5
ADAi -
-144
SETi6
o
ADAi
12
APOi
set tenR register to zero
~
Jlv1P POI-~
XSKi6
JMP
ADAi
p-5
tallies tens register
-12
COM
STC 7
LDA
set ones register
5
AZEi
JMP 3D
if hundreds is zero, do not display a zero
x2
HaL 1
ADAi
6D
STA
.
first word in list of code words
pr5
ADM
1
STO por5'1
LDA
JHP 10
LDA
JHP lC
JMP
<--.
p-.r5 1,
first half of character
DSC subroutine
second half o~ character
DSC subroutine
10-51
page
Display Decimal
Numeral of
Reduced Size
#3D LDA
6
AZEi
~IP
LDAi
tens
4D
zero surpresa
3
ADM
4
spacing of 3 between numerals
LDA
6
ROL 1
x2
ADAi
6D
first word in list of code words
STA
·p·~5
ADAi
LDA
~
JMp 1C
LDA
DSO subroutine
~
J11P 1C
LDAi
3
ADM
4
spacing of 3 between numerals
7
do not zero surpress ones
#4D LDA
x2
HOL 1
ADAi
6D
STA
,p·~5
ADA1
LDA
..
JMP 10
LDA
<\-
J'HP 10
#7D JMP
#6D 4136
3641
7721
0001
4321
3145
retunn to p+l of originating progrmm
o
1
2
10-52
page
Display Decimal
Numeral of
Reduced Size
4122
2651
3
21t-1J.~
04~1
5121
4651
1506
4225
4443
6050
5126
2651
5121
3651
4
4
5
6
7
8
9
10-53
Reduced Size DSC Subroutine
W. H. Calvin
page
y-coordinate in IR 11 (left unchanged)
x-coordinate in IR 4 (lett stepped by 6)
Code word in ACC
#lC STC 20
ADD 0
70
STC
SETi1
-2
#3C SETi2
#20
-6
LDAi
code vlord
#4C BeLi
7776
AZEi
JHPTTlp 0;4
LDA
11
DIS
4
LDAi
~
j
save lowest bit only
do not display dot it bit is zero
y-coordinate
x-coordinate
(change this to change character size)
3
ADM
11
LDA
step y by 3
2C
SOR 1
STA
2C
XSKi2
J11P
LDAi
3
ADM
40
shift code word right one
count on number of dots in vertical line
(change thi~ to change character size)
4
step x by 3
-22
reset y (6 times dot spacing)
LDAi
ADM
11
XSKil
#70
J11P 3C
JMP
do vertical line twice
out to p "·1 of
originatin~
program
10-54
Three Digit Decima1-to-Octal Keyboard Subroutine
W. H. Calvin
page
#4K. LDA
o
STC p-t14
STC p+l0
zero count
SETi2
-3
SET-ll
p-i~7
KBDi
lwlULil
ADMi
XSK12
JHP
JMP
144
p-5
out to P+l of originating program
12
1
10-55
Three Digit Octal Keyboard Subroutine
Calvin
\0[. H.
#2K KBDi
ROL 3
STAi
KBD!
page
hundreds
tens
ADD p-2
ROL 3
STC p-4
KBDi
ADD p-6
JMP 0
ones
out to p+l of originating program
10-56
LINOTYPE
for Histogrrun with
vI. H. Calvin
Invar~nt
Coordinates
....
#lA
'
SETi~
2A-l
list of locations of first character in each line
SETilO
-3
LDAi7
STC 2
#3A KSTi
JMP
~1P
three lines
get location ot first character in line
P-+3]
lQ
JMP
KBDi
SAEi
3A
Legend display subroutine
<-
13
JMP p+15
delete?
LDAi
-2
ADM
2
CLR
STAi2
delete last character
STAi2
LDAi
-2
ADM
2
backspace
JMP 3A
SAEi
12
~.
JI·IP p~4
XSKilO
J"HP
JMP
j
IA-r4
5K
ROL 1
ADAl
~
7A
STA
EOL?
count on lines
jump out
x2
first location in list of character code words
P4,5
ADAi
1
STC
LDA
Pol-5]
STAi2
,LDA
<_
STAi2
cThIP IQ
JMP 3A
10-57
page
LINOTYPE
f}2A
fi~
4X-l
5X-2
6X-2
-
#7A 1.~136
3641
7721
0001
4321
3145
4122
2651
2h14
Q47.tl-
5171
4651
1506
4225
4443
6050
5126
2651
5121
3651
0
0
0
0
0
0
1212
1212
0
0100
0400
0404
0400
0416
0000
0077 .
7777
7777
0
0
4437
3744
5177
3651
l~136
2241
4177
3641
5177
4151
list of code words for keyboard characters
0
1
2
3
4
5
6
7
8
10
9
11
EOL
12
del
13
space 14
--
15
period
~u
16
p)
minus 17
plus
/
20
(origin)
0
21
(brackets)
case
23
A
24
B
25
c
26
D
27
E
30
22
10-58
page
LINOTYPE
5077
4050
L~136
F
31
G
32
H
33
I
34
J
35
K
36
2645
1077
7710
4100
4177
0102
7601
1477
4122
0177
0301
3077
7730
3077
7706
4136
3641
4477
3044
4536
3743
4477
3146
5121
4651
L
37
M
40
N
41
0
42
p
43
Q
41t-
R
45
S
46
4077
T
47
7601
0176
7402
0677
7701
14·63
6314
0770
7004
u
50
V
51
W
52
X
53
y
54
z
55
4040
ID176
~.543
6151
10-59
Legend
HISTOGRAM ld th Invariant Coord1na tes
page
Displays three lines at top of screen (240-371) using DSC instruc~n.
It wish smaller characters, can use JMP 1C in place of DSCi3, changing
x-coordinate from IR 1 to 4, and y from 3 to 11.
#lQ LDA
o
STC 1Q
SETi5
-3
sets for three lines
340
y-coordinate of first line at 340
SETi6
SETi3
4X-l
SETil
beginning of table of characters
1771
initial x-coordinate is zero -1
1746
2510 characters per line
#4Q SETi4
#2Q LDA
6
DSCi3
DSCi3
xsmu-
on number of characters
JMP
JNP P~21
3Q
LDAi
4
'ADM
between characters
1
J1.1P 2Q
#3QXSKi5
Jl1P p
#1Q JMP
LDAi
on line number
+21
~
jump back to p-rl o.f originating program
..
spacing between lines
-40
Al»1
6
JMP
#4X
4Q..2 ~
5X is at
6X is at
3Xis at
4XJ~62
4X"l44
4X.;.-226
empty spaces inbetween
10-60
Histogram with Coordinate
Lines Invariant to Changes in Scale
and Origin
C16~-15
e FEBRUARY
SPON1ANEOUS AC1IUI1Y
1~65
SP~NA~ I~lE~NE~RO~
. .' .... ... : ... .......... .
.··. . . ... . . .. . . .... . . . . . . . . . .
.· .. '.'. .. .: .. '.'. ......... .
. . .
500! ••
'
'
400~ .•
300i·
'
200; .
•••
'.'
• • • • • • ,.
•••••••
e' • •
100l.
o :
o
10
20
30
40
50
"SEC
William H. Calvin
Department of Physiology & Biophysics
University of Washington
School of Medicine
Seattle, Washington 98105
(c
10
M~rch
1965
',.
10-61
IN1ERSPIKE
HIS10GRAM
IN~ERUAL
40 . . . . .
20 . . . . . .
.
o
eo
100 120.ttSEC
ISIs are clustered about 100 msec.
ORIGIN DELAYED ~O 80 t1SEt
BY lYPING D080. SCALING
JJNGHAt~GED
.
40 .
o
i:iI)
100 120
ttSEC
Shift origin to 80 msec. Coordinates shift also.
ORIGIN SHIFTED. VERTICAL
CO"PRESSED AND HORIZON1AL
SCALE E:-:PAt~DED.
o
80
100
120
ItSEC
Coordinate shift accordingly when change scale.
10-62
FLUCTUATIONS OF THE INTERSPIKE INTERVAL OF THE CRAYFISH
INTRODUCTION
STRETCH RECEPTOR By David R. Firth
Depa~tment of Physiology
and Biophysics
University of Washington
Seattle, Washington
The crayfish stretch receptor emits a very regular series
of nerve impulses at a rate dependent on the degree of stretch.
It was thought that fluctuations of the interval between impulses
(interspike interval
~)
would reflect fluctuations in the membrane
potential and in the excitability mechanism.
With these basic
sources of membrane noise in mind a study was made of the standard deviation (s) of the interspike intervals as a function of
mean interval (T).
RECORDING OF DATA
For reasons mainly connected with the evolution of the experiment from simple beginnings, the interspike intervals were recorded
on moving film instead of being timed directly by the LINe computer.
The particular display
u~ed
on film is somewhat unusual and is
made necessary by the small size of the fluctuaton
20 ros, SrvO.l5 ms.
(~.g.,
when
T
=
This contrasts greatly with the frog muscle
spindle, where the fluctuation at
T
= 100 ms is about
± 25 ms.
10-63
"
(See Hagiwara 1954-J Buller, ~ ale 1953).
The film moves contin-
uously but quite slowly, and the oscilloscope beam rapidly sweeps
perpendicular to the direction of film motion and once for each
-,,'
impulse.
The sweep is triggered at ,the end of a constant time
delay) which is itself initiated by each impulse.
By adjusting
the length of the time delay and the sweep it is possible to arrange
that every nerve impulse arrives during the sweepJ and on arrival
it triggers the beam brightener.
The result is a band of short
blips at right angles to the length of the film and lying anywhere
across its width.
~
(See Fig. 1.)
-
USEFULNESS OF DISPLAY OF DATA
By means of this device the two dimensions of the film both
represent time scales (though of different size» the small fluctuation is "magnified," a vivid display results, and film is used
economically.
It is clear that the LINe scope display could be
used to achieve the same effect, thereby giving the "eye-brain"
co~puter
a chance to do what it best can) to recognize patterns and
unusual effects in the sequence of intervals.
10-64
TRANSFER OF DATA INTO LINC
The positions of the blips on film represent the interspike
intervals whose values are entered into LINe in a semiautomatic
manner.
A film projector was modified in
su~h
a way that, by
and slides a potentiometer,
moving a lever which turns a mirrorA a D.C. voltage proportional
to the height of a blip is produced.
By A-D conversion of this
voltage the intervals are stored in LINe and on tape in digital
form.
About one or two intervals can be read each second with
this device.
THEORY OF ANALYSIS TO REMOVE DRIFTS
If the standard deviation of the intervals is calculated in
the usual way from the sequence (Tl'
T~,
. . . ) a serious error
arises ov]ing to the fact that the "true" mean interval at any
time is subject to drifts, partly due to adaptation, but also
due to less regular effects.
The drifts are often much larger
than the fluctuation from interval to interval so that the fluctuations must be calculated about some kind of drifting local mean.
We have foreknowledge as well as afterknowledge and must clearly
10-65
give distant intervals less weight.
As a result of mathematical
trials using various local means it appeared that the use of differences was a powerful way to remove the effect of drifts.
Thus if Tr is the r'th interval in a sequence we define
first difference
second difference
third difference
If s is the true standard deviation of the
TiS,
it is a very good
approximation)even in the presence of drifts)to write:where sl' s2' s3 are the estimates
or
of s.d. by means of the various
or
differences.
In the absence of drifts or correlations between intervals
LIt\C PROGRAH
A program was written by Mr. Kurt Beam, (an undergraduate
summer student),'to deal with the input ,storage ,calculation, and
output of the necessary quantities.
10-66
5
The input to LINC consisted of the sequence of interspike
intervals of a given run as described earlier, along with calibrations of the time axes of the film.
The intervals were stored
on tape in blocks of 255.
The following quantities (defined above) were calculated:
The output form '-las somewhat inconvenient, and consisted of a decir,1al
number multiplied by a power of 2.
A histogram of ~2 was read off
sequentially in binary form from the accumulator.
CRITIQUE OF DATA HANDLING
The system described above, for data acquisition, analysis
and display appears rather makeshift and inefficient.
The real
reason for this is that the use of LINC was grafted onto an earlier
system and the whole experiement was intended to be a preliminary
and exploratory one.
In addition there was a purely personal time
factor; it was more important in this case to achieve a limited
aim quickly than to do a fuller experiment in a more efficient way.
The amount of data to be handled was not so large that time would
10-67
o
have been saved by redesigning the system.
Another reason for this
mixture of old (photographic) and new (computer) systems was the
importance of having the visual display, described earlier, in order
to increase the chance of finding unexpected effects.
The idea of
using LINC to achieve a similar display did not occur until later.
A valid (but post hoc) reason for using film may actually lie
with the smallness of the interval fluctuations, (e.g.)
at
~
s~
0.15 ms
= 20 rns) for in order to avoid adding appreciable measuring
error to s, the graininess of the
unaaxiR~
measurer should be
considerably less than the smallest values of s; this may pose
complications O\ving to the cycling time.
RESULTS
Some of the results were presented at the Biophysical Society
meeting (Firth 1965) and consist essentially of the curve relating
sand T, shown below in Figs. 2 and 3.
The main features of· the
graph are its upward curvature, non-zero slope at the origin, and
the size of the effect as a whole.
lO-6r9
·7
DISCUSSION
It is believed that a simple model consisting of a constant
source of noise voltage (independent of the mean interval) super- _,'
posed on the generator potential can explain the shape of the
curve.
If the generator potential were to rise linearly from start
to firing level a linear
S-T
curve would resulc.
cellular recordings (Kuffler and Eyaguirre
However, intra-
.1955) indicate that,
at larger T values especially, the generator potential rises in a
very non-linear fashion, approaching the firing level much more
slowly than if it had risen linearly.
This allows a given fluctua-
of firing
tion in voltage to cause a larger fluctuation in timeAthan it would
have done on a linear generator potential.
A source of noise voltage is at hand in the
the membrane resistance.
of
IO~
Johns~n
noise of
Calculations show that in an infinite axon
diameter having typical membrane constants, the fluctuations
in voltage would be large enough to explain the observed fluctuations
of interval.
In fact the crayfish axon is about
20~
diameter, the
cell is somewhat larger and the dendrites are smaller.
It appears
thoug'4
possible, therefore, (:tSHgR the calculation is·rather crude) that
10-69
8
another source of noise (such as spontaneous release of inhibitory
packets) is not needed to explain the fluctuations of interval.
Intracellular recording of the generator potential is necessary to
pursue the question further.
This certainly offers a good oppor-
tunity for the use of LINe in analyzing the fluctuations of the
generator potential and of the firing level.
10-70
REFERENCES
Hagiwara, Jap .•
Buller,
~
ale
1..
Phisiol.
1..
Physiol.~:409,
Firth, Biophys.
1..
~:234,
1954.
1953.
Abstracts, Annual Meeting),1965.
Kuffler and Eyaguirre)
7
1..~.
Physiol.12:102, 1955.
10-71
-
."
11
*
f
r
.
2.0
Experi ment 7
Run # 795-818
Temperature =18°C-19.5°C
Plotted. curve is
S2 =(3.82 x 10-5)T 2 +(5.9
x 10-3)T
0
0
0
0
+
1.5
~
~
~
+
1.0
0
0
0.5
50
F,&..2.
100
-r(ms)
150
10-73
"
9
i c.)(pstirnent
~ · 0 -;
RL~ ~l
-tt: 105 9 -
,
:,\J.
; -. m
i 0"" ,,-- .'"
,
.
•
.' -..j
Ii
o
r....a :r
i\"'"
u 'M ~
~
i i9 ?
o
ISo -:-
'f" """"
r ~ '. !.zd
··.t....:.v
I
I
I
I
0
I
oI I "
. I
I
I
I
r-
r-rog
I
I
d
,I
I
I
I
I
,I
I
n
S I rr~~J/"i.;·'r.
• ,""...,
Crayfish
~,
)
ft
'f
r,;~ r-:~
, '"'" v
\oj
•
6"\'YO
rti j .
~J I
I
\
~,
,
//00
.f
r- '::'"
\. ; \;; /\ t.l
':'!:\
'
\/
Ii
I
I."'
i
}/
I
I
I
I
o
I
I
~
.,
1.01
,
~
i
,
,.
I
I
I
I
I
I
,,
/0
II
~
/
o.t>~
I
I
'0
,
0
0
o
~'o
I
.-'.-~'---
;
;. . . ,!'\
'-.,..;1
V
I
-~
~ ~.~~
iJV
10-74.
'.,.
UNIVERSITY OF WASHINGTON
SEATTLE, WASHINGTON 98105
School of Medicine
Department of Physiology
a"d Biophysics
CONDUCTION VELOCITY EXPERIMENTS - W. L. HARDY
LINe is being used by a graduate student, William L. Hardy, in an
attempt to determine the quantitative effect of external sodium concentration on single fiber conduction velocity in a ~rog sciatic nerve.
The experimental set up consists of a sciatic nerve preparation held
in a perfusion chamber, a set of three equidistant stimulating electrodes
and a microelectrode with necessary preamplifier and electrometer input
for recording the action potential intracellularly from a single fiber.
With a given perfusing fluid sodium concentration, LINe is programmed.
to sequentially select using the relays a stimulating electrode. The
nerve-is stimulated from an external pulse generator a designated number
of times (set by the sense s~~itches). The intracellular potential is
sampled every 32 microseconds l-1i th a total of 150 points per record. The
record is averaged with previous records at that stimulating position.
Averaging is necessary because of the high noise level. The procedure is
then repeated for each of the other two stimulating electrodes and the
three average curves are stored on the magnetic tape under keyboard control.
Using LINC to control the experiment through selection of site and time of
the stimulus has made this experiment more possible. Because of the very
small size of these myelinated nerve fibers, it is difficult to hold the
microelectrode in the cell for an extended period of time.
It is planned to use the LINe to calculate the conduction velocity
directly from the stored data. This information can then be stored with
each record along with data on the sodium concentration in the perfusion
solution. In addition it may be of use to record the time since the
perfusing solution was changed. For this a real time clock would have to
be installed.
Details of the sampling program will not be included as there is
nothing very unusual about the method of changing stimulus positions with
the relays, sampling at 32 microseconds between samples after a trigger
pulse, storing and averaging data, displaying data,' and storing on magnetic
tape under keyboard control.
10-75
FINAL REPORT
Washington UniversitY'School of Medicine
Department of Neurology
LINC COMPUTER EVALUATION - March 3, 1965
I.'
PROPOSED RESEARCH PROGRAM
(Supplenlented in detail in subsequent section).
A.
Background and Introductory Statement: Two years
ago our group, then containing Sidney Goldring, M. D., now
Professor of Neurosurgery at Pittsburgh, was invited to join a
program to evaluate the LINC Computer.
In the summer of 1963
Dr. Goldring and our electronics engineer, Mr. Lloyd Simpson,
participated in the MIT program and assembled the LINC
computer assigned to us there.
For the several subsequent
fall months we worked with the computer on test runs
establishing an averaging program and modifying our own
electronic equipment to its needs.
Commencing in winter 1964
we moved to the neurosurgical operating room in Barnes Hospital
and there did a series of studies upon the potential field of the
somatic evoked response.
Whereafter we returned to the experimental·
l'aboratory to establish further details and re -work the program in
studies upon experimental animals.
Meanwhile, Dr. Robert Wurtz
was starting a new program on aversive and approach problems in
stimulation of the caudal diencephalic
.~nd rostr~
mesencephalic territory
11-1
2
of the rat with implanted electrodes.
In winter and spring 1964 Drs.
H. G. Schwartz, Professor of NeurosurgerYI ,Dr. William Coxe,
Associate Professor I and Dr. Wurtz took the W. U. courses given
on the LINC computer.
In fall 1964 we added a second mex:nory and
Dr. Roy Wright took the LINC computer course.
people working on the LINC who
hav~
We now have two
constructed their own programs
with the aid of the LINC computer group and are now engaged in active
daily research using that instrument as a primary tool of the research.
B.
Objectives and Program
It is our purpose to study the interaction of communities of
neurons using whatever electrical signs of their activity are appropriate
to the particular needs of an experimental design.
The working
group has had experience with ultraslow potentials
l
extra- and
intracellular unit potentials, and evoked potentials of somatic
auditory and visual types.
We have excellent laboratory facilities
dark rooms, animal quarters
l
experimental operating rooms
, and access to neurosurgical patients.
l
l
The present shielded rooms
which house the LINC were set up with the view in mind of doing
implanted electrode studies upon the human thalamus with natural
stimulation and percutaneous electrical studies.
We also have a
scheme in mind for programming the computer to differentiate
between the electromyograms of muscle units recorded percutaneously
11-2
3
which are indicative of neuromuscular disease as compared with
normal muscle.
If this proves feasible we will have' established
a practical clinical us'e for the LINC.
Similarly.. in our studies
upon the human thalamus we hope to provide a program which
will differentiate between spontaneous activity at various depths
and positions with the ultimate view of being able to decide more
effectively where to place lesions for alleviation of Parkisonian
symptoms. '
Dr. Sidney Goldring who commenced with us has moved
to Pittsburgh and there applied for anteceived a LINC~omputer.
.
,Thus we already have one satellite l~boratory in operation .
. One of the most perplexing problems we have met is that
of providing a sufficiency of skilled operators who are capable
of adapting the LINC to their uses.
We now have a schedule of
training for those who will take future LrnC courses and we hope
gradually to make our laboratory 1000/0 LINC competent.
Details of this generalized proposal are offered in the
ensuing lengthy supplement which follows the headings outlined
by the Evaluation Committee.
11-3
4
II.
1.
. SUPPLEMENT
Past and Present Research.
a.
. Average' evoked somatosensory responses.. Averaged
evoked responses in animals and man were studied utilizing
three different averaging, programs.
Program #1: The program causes a relay to activate a
Grass stimulator once per second.
Sampling is begun after a 7
msec. latency and continued during the entire interval between
stimuli.
During the first 150 milliseconds sampling frequency is
rapid to capture the fast response components.
slower sampling
f~equency
Thereafter, a
records the later and slower components.
The response is displayed on two lines; the upper on showing
the initial 150 milliseconds of response, and the lower line, the.
remainder (850 msecs.).
A sense switch is used to either continue
display or clear the memory in preparation for recording the next
response.
Stimulation and samplings are started by depressing
the Start 20 button.
Program #2 is basically the same with these additional
features.
After the last response is added, the .averaged response
is automatically stored on tape and then, using a Sense 'Switch,
either the memory is cleared or responses stored on tape can
be called back inte memory and subtracted from each other.
In
11-4
5
addition, two vertical cursors which are under knob control are
displayed and used for comparing response latencies.
Also, two
horizontal lines, also under knob control, are displayed as base
lines from which response deflections can be measured.
The two programs described above were used predominantly
in the experimental laboratory.
For patients a more elaborate
J
program (written by Severo Ornstein) was used.
.It records and
recalls data via keyboard request and displays a table' of contents
upon request.
The program is capable of recording on one or two
channels (traces) and at the end of a sampling run writes the
'data on tape.
Records are labeled from the keyboard and once
recorded may be recalled by label.
One can record on one trace
.. while holding a previous recording on the oth.er - or record on
both traces - or show post records on both.
The basic process consists of typing a label and then asking
to record ("R") or display history ("D").
When labeling, the
desired trace must always first be indicated by
"LII (lower).
"u"
(upper) or
A label consists of a station number (from 0 to 39
decimal) followed by an "S", a
"w"
or a "Tit (to designate surface,
white matter or transcortical recording) followed by a dash followed
by a version (repeat) number (from 0 to 9 decimal).
11-5
6
After labeling, striking "R" will cause recordingon'any trace
with a label not previously used.
That is: . (l) A trace holding a
history display will not be affected.
(2) A duplicate of an already
used label will not be accepted for recording action.
unlabeled trace will be unaffected.
(3) An
-,'
The upper trace accepts input
via the SAM 10 line; the lower, via SAM 11.
Recording can be done with either of two sampling intervals
controlled by sense switch 2. Up
= 600 psec;./ sample;
down
=
304 }lsec/.sample.
This should never be changed in the middle of
any sampling run.
The initial dot on a trace after
sampling indicates the sample frequency used on that run.
above the trace indicates 600 usec.
304 )lse(!. sampling.
hit the "XI1 key.
A dot
Sampling, below indicates
If during recording it is desired to reset,
This erases the data for trace(s) that were. being
recorded but leaves the label(s) so that a new "R" action can be
taken when ready to go again.
At the end of each sampling run the data are recorded on
tape for future
re~erence.
The first record is stored in block 6
and "subsequent records are stored in subsequent blocks. A table
of the labels for recorded traces is kept updated in block 1
(backup duplicates of this block are also in blocks 2 and 3).
For redundant.recording.on"unit 1, sense switch 0 should be
11-6
7
left down.
This should hold for a complete experiment.
The data
as well as the label table will. then be redundantly recorded.
Each record consists of the sum of 50 sets of samples.
Each sets consists of 256 10 samples where sampling starts
immediately following stimulation (via relay 0) .. Stimuli are
approximately 1 second apart.
Individual samples are 304
o~
600 psec. apart (depending on the setting of sense switch 2).
To call for display of an earlier record, merely type the label
on the desired trace and hit liD.
II,
single trace or both both traces.
Requests may be made on a
Requests for non -existent traces
will be ignored.
At. all times except during actual recording a pair of
vertical display lines are accessible if sense switch 1 is put up.
The X positions are controlled by knobs 0 and 1.
A maximum of 58 10 records can be handled on anyone
. tape.
An ordered tabular display of the labels of all records
. entered on a tape so far is obtained by hitting the case key.
This
display is terminated by the EOL key.
To study a tape recorded on an earlier occasion read in
.block 1 rather than block 200 at the
outset~
(If for any reason
block 1 will not read and check, blocks 2 and 3 are duplicates of 1. )
Starting at 20 in the usual fashion gets the program started and
then one can inspect the table of contents and call up histories
11-7
8
(or add new records if desired) in the usual manner.
To make a copy of the' tape for other experiments, copy
blocks 0 to 5 inclusive and blocks 1 77 to 210 inclusive onto the
, corresponding blocks of the new tape.
A tape may be reused after
\'
an experiment by rereading block' 200, etc., but further recording
wilL then write over old records.
The program consists of 4 separate ma.t?-uscripts which
are located at blocks 240, 250, 260, and 270 respectively. ' The
tapes also contain LAP3 in the standard place.
, b. ' , Steady potential correlates of reinforcing intracranial
stimulation. This experiment was designed to determine whether
,the steady potentials recorded from cerebral cortex with direct
coupled amplifiers change consistently with electrical stimulation
of two kinds of subcortical structures, one producing aversive
behavior and the other approach behavior (self -stimulation). Initially
rats are tested behaviorally over a period of several months to
determine whether they will either (1) press a bar to turn on current
through an implanted electrode (self -stimulation) or (2) press a
bar to turn off a current pulsed at one per second through an
implanted electrode (escape).
Once the behavioral significance of the
electrical current application has been established non -polarizable
electrodes of a type developed in our laboratory are iPJ.planted and
11-8
9
the changes in steady potentials following stimulation are recorded
using a cortical surface electrode and an indifferent one on the nose.
The potentials are amplified by a Grass model V polygraph with a
chopper stabilized preamplifier.
Using a "reverter" supplied by. .·
Grass the signal from the driver amplifier is converted to a
signal of 3 volts above or below a zero-volt level.
A I-volt signal
is obtained by attenuation of the amplifiers.
LINe both averages the responses and controls the reinforcing
stimulation.
The steady potential described above is averaged
(added) before and after each stimulus over a course of 100
successive stimulations.
Two programs are used.
In one the
stimulus is triggered by the computer and in the other by the rat .
. Where the computer triggers the stimulus a sample is taken for
1. 28 sec." the stimulus is given, and a sample is taken at 10 msec.
intervals for another 1.92 sec. and additionally for 5.12 or 15.36
succeeding seconds depending on the setting of the sense switches.
Where the rat does the triggering, the stimulus is given immediately,
and samples are taken over a 3. 84 second period.
LINe is also
sampling and storing between bar presses and when the bar is
pressed LINe saves the
~ast
1. 28 sec. before the bar press so
that a sample is averaged both before and after the stimulus in
this pr.ogram" also.
11-9
10
In both programs l sampling is done continuously between
bar presses and a weighted average is updated as these samples
ar~
taken.. This weighted,average reflects
t~e
.constantly changing
d. c. level arising from movement of the animal drift in electr6aes
l
l
. the "reverter" or the sample amplifier· of the computer. ' In
I'
each. period where samples are taken for' averaging thew¢ighted
averag'e is subtracted from .each sample so that what is averaged
is a difference between the pre-sample period and the sample period.
LINe is triggered after a bar press by a pulse -former on peripheral·
. equipment - a' Massey-Dickenson transistorized programming unit.
The 60-cycle stimulation was initiated by a relay pulse from LINe
and duration was controlled by the peripheral programmer.
The
equivalent of the programmer could have been built into. the
. data terminal box and would have been had the programmer not
already been available and compatible with LINe.
c.
Plastic properties of synapses. A change in the
functional connection between presynaptic terminal and postsynaptic
neu:ron is assumed by most microphysiological theories of learning.
One ,variation of such theories predicts that if the postsynaptic
neuron fires after a presynaptic, terminal has been active, the
connection between 'the pre - and postsynaptic neuron is facilitated
while other inactive synaptic ,junctions are unchanged.
In the
11-10
11
future this EPSP produced by the presynaptic activity would have
a higher probability of discharging the postsynaptic neuron.
The
paradigm is much like operant conditioning in the intact animal
and it seemed possible to interpose LINe in the pre - and postsynaptic
system in a manner similar to its use in other operant conditioning
situations. LINe has been. programmed therefore to recognize
a particular EPSP waveform and either record the frequency with
which it is followed by a spike discharge or pass current through
the postsynaptic neuron to produce a spike.
The neurons chosen for this experiment are in the visceral
ganglion of the sea- slug, Aplysia· californica.
They are of large
size, and consequently are easy to record from intracellularly over
several hours.
Glass micropipettes are used to penetrate the cell
and a bridge circuit permits current to be passed through the sa.m.e
electrode used for recording.
Input to the sample amplifier ·of
LINe is via an oscilloscope amplifier and is capacity coupled.
OUtput f,rom LINe is via an operate pulse which triggers a pulse
generator which in turn passes current through the microelectrode.
LINe is programmed to recognize a particular EPSP and
respond to this particular EPSP but not to any other \lsing a program
similar in general but not in detail to one written by Simon.
The
program is divided into two segments:. a selection mode involving
11-11
12
a selection of and 'EPSP as the standard for'recognition, and a
running mode in which EPSPs are compared to the standard and
either current passed through the microelectrode or the frequency
of spontaneous cell discharge after the EPSP determined.
In selecting the appropriate EPSP, samples are taken at
4 msec. intervals with a continuous one-line moving display of
the most recent samples.
Pressing a keyboard button stops the
sampling and displays four lines of data with eight consecutive
intensified points which may be moved along the lines under knob
control.
A standard EPSP is selected by a keyboard instruction
which transfers the eight points which are intensified to a catalog
and this catalog is then displayed.
The catalog may contain up to
six possible standard EPSPs and one of the six standards in the
,.catalog is picked ,as the standard.
The continuous display.. may
then beres1arted and each EPSP identified as equivalent to the
standard is marked.
The acceptable difference between standard
and current sample is urider right switch control.
Once the
threshold is decided upon, the current r~quired for spike production
is set by
an.ot~er
subroutine and the' experimental mode of the
. program is ready to be r,un.
At any time a subroutine may 'be
,interrupted and any', other subroutine: called up by a meta command
given via the keyboard.
11-12
13
The standard sample actually used for comparison is not
the sample displayed but a normalized value which permits recognition
of the shape of the EPSP independent of any slower baseline fluctuations
on w'hich it might be found.
The normalized value is obtained by .....
taking the mean of the last eight sample points, and subtracting
this mean from each of the samples.
As each new sample is taken,
it is added to a list of the eight most recent samples, and the oldest
sample is discarded.
A new mean is found (by adjusting the old
mean) and a normalized value for the last eight points is obtained
by subtracting the new mean from each point (actually by adjusting
the previous seven normalized points and adding one new normalized
point).
These eight normalized points are stored in a second list.
Comparison is then made to the eight points of the normaliz ed
standard~
the absolute difference between each point is' taken, the
differences summed, and the sum compared to the threshold value.
If the value is less than an acceptable threshold, a recognition is
made; if -not, no recognition is made" another sample is taken and
the comparison cycle is repeated.
When the experiment is started the frequency of spike
discharge following the first 16 recognized EPSPs is recorded
and current is then passed after the next 124 recognized EPSPs.
The 16-EPSP control period is then repeated and the experimental
11-13
14
(current passing) and control periods then alternate.
Data output
is currently in the form of a- bar graph of the number of EPSPs
followed by a spike (1 to 16) in successive control periods.
Programming of this experiment" particularly the-normalizing
procedure and cursor program has been greatly benefited from the
aid and suggestions of Cox, Ornstein" Sandel and others at the
Computer Center.
The experiment and.program have been in the
process of refinement since mid- January" 1965.
2.
Future Research
a.
Electrophysiological Studies on the Cingulum and
Cingulate Gyrus. Two projects are currently active.
The first
of these involves the use of averaging to trace the cingulum conduction
spike anteriorly and posteriorly at conduction distances beyond
, -which. a conduction spike is clearly discernable with standard
techniques.
This kind of approach is most valuable in elucidating
both the origin and distribution of the cingulum bundle as well as
providing information about the conduction rate of its component
.fibers.
These same experiments will provide data which will
permit a much more detailed electrophysiological analysis of the
cingulum conduction spike in its supracallosal portion than was
possible in earlier computation by this investigator.
The second project.is underway inasmuch as data taken
for study of the cingulum conduction spike includes much useful
11-14
15
for it.
This project involves a study of the instantaneous electrical
fields for given cross sectional positions in the cingulate gyrus.
These electrical field patterns will be' correlated with the cross
sectional configuration of the gyrus in an'attempt to elucidate
the virtual generators and the sequence of events in' the' cortical
evoked response which follows cingulum stimulation.
A program for sampling an evoked response in the
cingulum bundle and cortex of the cingulate gyrus has been developed
with assistance of Severo Ornstein.
512 iJsec. are used.
Two sampling rates, 32 and
The number of samples at each rate and/ or
the rates are variable.
A total of 777 octal samples are possible.
This program permits averaging of 32 responses and includes a
feature for saving the average on tape with appropriate run number
and indication of regular or special mode of recording.
A display
subroutine permits observation of the last response sampled as
well as the build up of the average.
This display subroutine ruso
includes sense or right and left switches selected amplitude controls,
switch selected horizontal (time) gain controls, amplitude calibrations
for both current and average traces, and vertical cursors.
, latter
ar~... cou.pled
The
to the traces in such a manner that both the
time difference between any two points and the amplitude at a given
I
point (with respect to -the calibration voltage) are computed.
.
~,
11-15
16
In summary a program is in use wru.ch permits saving of an
I
evoked potential with sufficient on-line data analysis for optiInal
direction of the experiment.
,
Although designed for this particular
'
purpose this program is readily adaptable to a wide range of
similar experiments in neurophysiology.
A data analysis :program utilizing the above data saving
program and its display subroutine is partially complete.
This
includes features permitting keyboard selection of data on tape to
be read in for comparison of the response from different recording
. positions. (even from different experiments) with the option to add
. or subtract the two responses and display the result.
A feature
permitting selection of the potential at any given instant of time
for each of a large number of responses is included and will
permit use of the computer in plotting the instantaneous potential
vs. recording electrode position either as a single or a family
of curves.
A special program using this data to plot the electrical
field contour lines on the computer oscilloscope has been
. discussed with Dr... Cox.
When it becomes available this program
will markedly facilitate the correlation of the electrical fields with
the anatomy of the generators involved as well as permitting studies
of the changes in the fields with respect to time.
For stimulatio]J. use 1n made of the OPR i instructions
which provides a pulse sequence.
I
Only the first pulse is used here.
11-16
17
This is amplified via an indicator light driver and used to trigger
a Tektronix waveform generator-pulse generator combination.
This pulses a Grass stimulator which delivers the stimulus to
the preparation. In order to have the sampling timing uniform
with respect to the stimulus a pulse which occurs at the same
time as the stimulus is taken from the Grass stimulator and
inverted with a Tektronix waveform generator and led to the
computer via external level line (XL).
This ends the computer
pause state and sampling begins within a few microseconds.
The input to the data terminal box for sampling is derived
from special vacuum tube amplifiers in use in this laboratory.
These are fed by cathode follower amplifiers adjacent to
the preparation.
b.
Use of LINe in the Analysis of Steady Potential
,Changes in Chronically Implanted Animals. Two series of
experiments are planned at this time.
Relation of steady potential shifts to unit responses during
wakefulness and sleep. Work in the laboratory on cats with
chronically implanted electrodes for d. c.' recording indicates that
steady potential shifts occur
dur~g
sleep referred to as "deep sleep.
1J
wakefulness and a phase of
Using LINC it will be possible
11-17
18
to relate these spontaneously occurring changes quantitatively to
unit responses using the following general approach.
of data will be compared- a d. c. shift sampled
.per second and unit firing frequency.
~bout
Two types
50 times
As the cat awakens or goes.
into deep sleep the unit firing frequency will be taken in a series
~
of time periods each lasting five to 15 seconds.
The length of
each time period will be determined by the rate of the d. c. shift
which will be sampled continuously.
Thus when the shift has
reached 100 p,V, the first time period will end, at 200 p,V the
second time period will end, and so on.
Over a number of
periods of sleep the frequency histogram for each time period will
be summed.
The significant correlation will be between successive
. d. c. amplitude changes and successive changes of unit firing.
The same program methods for elimination of drift used in the
'. previous experiments on steady potential will again be used.
Unit responses will be discriminated by amplitude.
Cortical d. c. potential changes during -conditioning. The
purpose of these experiments is to determine whether consistent
d. c. potential changes occur in cortex during conditioning and
if the shifts are similar to those arising spontaneously- in the sleep ., .
wakefulness cycle.
Cats with chronicaJly implanted electrodes
for d. c. recordingwlll be conditioned using a fiashing light for CS
11~18
19
and a shock for UCS.
Potentials will be recorded over six areas
of cortex including sigmoid, suprasylvian, and posterior lateral
gyrii before, 'during. and after the onset of the CS-UCS period
using the Grass polygraph amplifiers and "reverters." CS, UCS;sampling periods, etc., will be controlled by the program and a
data terminal box.
Samples will be averaged over blocks of ten
trials. Since sampling will be of the order of 50 to 100 samples
. per second there would be enough time to sample all six available
polygraph. channels on each behavioral trial.
A rough program has
·been drawn' up as a class exercise at the Computer Center.
c.
Plasticity of single neurons.
This work will be
continued with further effort concentrating first on such variables
as initial probablity of an EPSP firing a spike or ratio of experimental
to control periods and later on the changes associated with different
EPSPs and changes in several neurons.
The general format of
the present LINC program may be simply modified to encompass
many of these additional experiments.
3.
Training Program
a.
(S. G.) The best feature was the month course at
Cambridge.
This was the best period" of intruction I have eve:r;
experienced.
The remainder of the program
was essentially,
,
learning by doing.
Tre fact that we were able to ask for and
11-19
20
receive help from Clark" Cox and associates was instructive and .
valuable in carrying out our own program.
b . ( R . W.) I did not participate in the month at MIT
but the 'course I had with Dr. Cox proved adequate to understand ,.,0
.and begin programming of LINe; it did not intend to nor did it
)
provide a basis for repair.
Direct contact with the group in.St. Louis
has been a unique and very profitable advantage.
Other people in the Department desiring to use LINC have
also taken the Computer Center course(Drs. Kelly and Wright}
but an individual. must wait until the course is taught; teaching
. only one person at a time is very time-consuming. Consequently"
I
c~ot
pass up the opportunity to suggest that a program could
be constructed using the standard question and answer routine
on the oscilloscope so that LINC could instruct on the use of
LINC - a very expensive teaching machine.
Coupled with the
console function" binary numbers" and order code write-up and
some practice programs" the rudiments could be taught on an
individual basis in any laboratory at aqy time.
Once the individual
started to program" instruction could easily be continued by questions
to the more LINC-knowledgeable people in the laboratory.
11-20
21
4.
Computer Performance.
a.
Maintenance.
Maintenance of LINC in our situation
was made at practically no loss of time, thanks to the Computer
Center and its personnel, in particular, Drs. Jerome Cox and
· Severo Ornstein.
..,'
Overall the computer required much less
maintenance and upkeep than we anticipated as our log will verify.
Defects were usually of a minor nature, easily repaired once
. located.
Our power supply would kick out the +18 volt breaker but
this occurred over a short period of: time and eventually corrected
itself.
No cause was determined.
If the computer was allowed to
· remain overnight in an air-conditioned area the LAP 3 program
would not perform until the computer had been turned on for at
least 30 minutes.
This was solved by simply turning the air.-
conditioner off at night.
The tape units gave us trouble failing to
· perform tape instructions occasionally.
a bad diode located on the tape unit.
This was determined to be
The computer failed to put
data into memory on one occasion and this was
found to be a defective
T
I
transistor in the E P 105 series write card.
This same transistor
had previously been replaced at MIT during the assembly phase of
the course there in August, 1963.
memory a transistor was again
During the installation of second
fo~nd
to be defective in a series
11-21
22
read card.
No other difficulties of
~onsequence
were experienced.
~
It is our opinion that the computer has performed exceptionally well
for us insofar as maintenance is
b.
con~erned.
' Performance in experiments. At the time we received
LINe, no one in the laboratory had had any previous experience
or contact with
co~puters
or digital computer techniques.
Consequently,
a substantial amount of our time has been spent in developing our
skills in the use of the machine and our appreciation of its
capabilities rather than in the development of more extensive
input-output connections. In addition, our LINC is physically
situated so that it may be conveniently used on-line and all of our
experiments have been done on-line with LINC analyzing the data
and controlling the experiment.
Fer our purposes, the performance
of LINC has been superlative and we would not suggest any basic
. modifications. The comments that follow we feel are relatively
minor and arise when comparisons are made to an "ideal. "
The order code has been adequate and we have actually used
or seen the need for all the instructions with the possible exception
of RTA.
An instruction that would skip if the contents of a beta
register equal Y would frequently save an extra counter or the
time required for an LDA and an'SAE and would be more useful
than RTA if instructions could be .so· juggled.
For our purposes,
word length has been entirely adequate"and with the addition of
second memory core memory size has been satisfactory.
11-22
23
)
All of our recording has been analog data using the sample
amplifiers ~ either capacity-coupling from osciUoscope amplifiers
or
dir~ct
coupling from the postamplifier of a Grass polygraph
(after demodulation of the choppered preamplifier stage).
, sample amplifiers have been found to be adequate.
The
Any'drift in
, them has been small compared to that of the preparation and.
electrodes when d. c. recording was used and has been 'taken out
in programming.
Tiggering LINe from other equipment by the
use of an external level has been no problem and is an essential
featur~
of the machine for our purposes.
To control other equipment from LINe, the pulses produced
by the relay or operate instructions have been preferCiLble to the
use of relays.
The one exception is the control of alternating
,current. We now trigger from LINe a variety of laboratory
instruments with various
trigg~ring
requirements.
Grass stimulator,
+50 volts; Tektronix pulse and waveform generators, +5 volts; stimulus
isolation units, ±5 volts, or +20 volts; Massey-Dickinson transistorized
,programming apparatus and other behavioral apparatus, ±12-2.8 volts.
Indicator-drivers and Tektronix pulse generators have been used to
convert the 3-volt pulse to the required amplitude and we anticipate
doing all such triggering using onlyD. E.~. cards with
in cost (from tied up pulse generators).
reductio~
Since this problem appears
11-23
24
completely and inexpensively solvable by the use of appropriate
D. E. C. cards, it would have been useful initially to have a list
of cards, including transistor substit\ltions, that could be used
to produce low current
voltages.
pul~es
at various positive and negative
It would also be ideal to have
a. higher voltage source
available in the data terminal box (about 't30-50 voltsr to avoid
the ·use· of batteries or' an external supply.
The data terminal box has been admirably suited for our
purposes; the standard box would be more useful if only the
sample' amplifier connections were made and other terminals
installed but left unconnected.
Because LINe has been compatible with existing equipment,
we have added very little input-output hardware.
The teletype
. has proved valuable for hard copy of programs and numerical
data.
The oscilloscope has been adequate to date for displays of
.,-analog data.
We have not had available an adequate tape recorder,
but hope to acquire one in the near future.
One of the greatest advantages of LINe, its flexibility, also
"is related to what has been its greatest disadvantage, the time
r~quired
for programming.
While this has decreased with
proficiency, the same proficiency has also permitted 'construction
of more elaborate programs, and the time consumed for 'a program
11-24
'25
beginning to utilize LrnC capabilities has been substantial.
Time
spent programming, of course, relates in part to the clarity of
the individual's experimental ideas and methods and his ability
to translate the program format, and no hard- or software seems""
likely to change this.
The software tools now available for
constructing and storing programs, control of binary programs,
and standard routines such. as manuscript print.-out
ar~
necessa!y',
to use of the LINC so that, of course, we never program without
these tools.
adequate.
At this time, these assembly devices' seem to be
What we have felt a need for is not more trowels but
more brikcs, that is, program subroutines.
These have accumulated
in the laboratory as more programs have been written although many
core parts of programs are not of much transfer value.
There
remains, however, a category of routines which are used or could
be used if they did not involve construction each time, for example,
teletype print, octal and decimal conversion, display of numbers,
and·storage routines.
have been' in
th~
All' can be constructed and man! of them
laborato'ry, but it would be idea1.to have them
'available on tape and debugged.
As important as having the program
would be the format: preferably a print-out of the manuscript,
anotated so that it could be' understood (which in turn greatly helps
the LINC l1ser l s programming abilities).with the lines for the various
parameters labeled.
11-25
26
Even' with this aid, programming time will remain substantial.
This limits the number of experiments in the laboratory/which,are
'programmed for LINe but it also motivates discrimination between '
..,'
. experiments which would greatly benefit fromLINC and those
which would benefit but in relatively trivial
ways~
In net), where
. LINe is deemed appropriate the results produced more than
"compensate for the 'programming time.
The programming simply
is an added experin;tental "cost" as are the other instrumentations
and techniques.
Evaluation of the cost of LINe is difficult since it has been
, used on experiments which would not have been done without it
and so cannot be assigned a dollar value.
Because of its versatilitYI
the cost need not be assigned to one specific experiment but to a
variety of experiments even over a number of years.
In this light,
, the estimated commercial price seems to us to make LINC a
, reasonably priced laboratory instrument.
This is particularly true
in our case since we have been able to use it without large investment
, in other hardware,.
In net, w.e have found LINC to be a remarkable machine and
, .a Powerful-laboratory tool.
It has been reliable, has connected to
our laboratory equipment with little difficulty or expense, and
has produced any analysis
desi~ed 'with, sufficient
programming effort.
11-26
27
If there is any inadequacy, it is in software rather than hardware.
As our ability to use LINC has increased, we have been even more
impressed with its flexibility and range of application and this has.,
in turn, suggested experiments not possible without its capabilities.
5.
Log Book.
A"Xerox copy of "our log book is enclosed.
6.
Bibliography.
a.
Papers.
(Copies are enclosed)
(1)
Goldring, S., Kelly, D. L., and O'Leary, J. L. :
Somatosensory Cortex of Man as Revealed by Computer Processing
of Peripherally Evoked Cortical Potentials, Transactions of the
89th Annual Meeting of the American Neurological Association.. 1964"
pp. 108-111, Springer Pub!. Co ... Inc. New York City.
(2)
Kelly .. D. L. Jr ... Goldring, S., and O'Leary, J. L. :
Average Evoked Somatosensory Responses from Exposed Cortex
of Man, Archives of Neurology, 1965.
(3)
(In press)
Wurtz, R. H. : Steady Potential Correlates of Reinforcing
"Intracranial Stimulation (In preparation)
b.
Talks and Abstracts. Work resulting from the studies
on somatosensory evoked responses was presented at the 1964
11-27
28
Annual Meeting of the Southern Neurosurgical Society and.also
at the 89th Annual Meeting of the American Neurological Association
in AUantic City in June of 1964.
7.
Continued Use of LINe.
'As time has passed other members of the laboratory staff
await training in the use of the LINCcomputer and. have ideas
formulated for its use.
For example.. one of these concerns the
" possibility "that LINC could differentiate between units in normal
and diseased muscle and thus markedly speed up the laborious
analysis of EMG traces.
There are several potential field studies
which it is predicted one should be able to conduct' efficiently and
rapidly by the LINC.
It is earnesUy hoped that we will be permitted to continue
the use of this instrument since more possible uses arise daily.
The instrument is so versatile in meeting laboratory needs that
in the future 'we would be decidedly handicapped if deprived of
it as a readily available .laboratory .instrument.
11-28
:,-
FINAL REPORT ON LINC COMPUTER EVALUATION
LABORATORY OF NEUROPHYSIOLOGY
UNIVERSITY OF WISCONSIN MEDICAL SCHOOL
Partially supported by
NATIONAL INSTITUTES OF HEALTH GRANT MH-08354-01
March 11, 1965
J. E. HIND
C. D. GEISLER
I.
INTRODUCTION
This is the final report of the participation of the Laboratory of
Neurophysiology of the University of Wisconsin in the LINC Computer Evaluation
Program, July 1, 1963 to March 31, 1965. Our participation has been of immense
value to us. Not only have we received considerable training in computer
technology, but our laboratory has had the use of a" powerful new tool. We are
grateful to the National Institutes of Health for their support of this program.
We wish also to express deepest appreciation and admiration to the members of
the LINC design group who made this development possible.
We have utilized the LINC extensively (for more than 2650 hours) and we
are indeed enthusiastic about its performance both as a computer and as a
laboratory instrument. The LINC has become an indispensable tool in our
laboratory and many of our plans for future research are built around the
machine (see section III of this report). We respectfully request that the
LINC be assigned on a permanent basis to the Laboratory of Neurophysiology.
1"1 .
PAST AND PRESENT RESEARCH
A.
Temporal Analysis of Single Neuron Discharges
1. General Development. Prior to the arrival of the LINC in our laboratory,
the Control Data Corporation 160 and 1604 computers in the University's Computer
Center were utilized to process data from auditory microelectrode experiments.
The 160 was arranged to time the occurrence of neural discharges and stimulus
events in multiples of 75 microseconds as these data were reproduced from analog
magnetic tape. The digital tape produced by the 160 was then processed in the
1604 computer under Fortran programming. Thus the analog data required only one
playback in real time into the computer system, after which an almost unl imited
variety of analyses could be carried out on the digital ized data. The programs
developed for the 1604 include the following: creation of inter-spike interval
histograms and calculation of the mean and standard deviation of these distributions
together with cumulative conditional probabil ity curves; serial correlation
coefficients among successive intervals; relation between mean and standard
deviation of collections of interval "values under varying stimwlus conditions;
and post-stimulus histograms including an analysis of the successive inter-spike
intervals in a train of discharges evoked by repeated, brief stimul i. This
system was Just beginning large-scale use wh~n the LINC project was initiated.
12-1
-2-
Upon arrival of the LtNC at Wisconsin, initial effort was concentrated
upon arrangements to permit determination of post-stimulus and inter-spike
interval histograms on the LINC through introduction of the tape-recorded unit
and stimulus pulses into the XL (external level) 1 ines. A bank of Tektronix
pulse generators (recently replaced by DEC digital modules, see p. 56 of log
book) was inserted between the tape recorder and the LINC to shape the pulses
and to set their duration equal to approx. 150 microseconds. Programs were'
developed which permit the time bins for the histograms to be selected in .~
integral mUltiples of 100 microseconds; the display provides for two lines
of legend typed on the keyboard together with a line of deci~al constants which
include the full-scale vertical bin value, the total number of , spikes in the
histogram, the total number of stimuli, and the full-scale t.ime value.
While our initial approach to the LINC involved programs which carried
out one specific form of ter:nporal analysis as the analog data were played into
the machine, the backlog 'of taped data awaiting processing soon brought the
real ization that it was impractical to operate with processes that require the
data to be replayed in re.l time for each analysis or change in analysis parameters. Also fundamental to our thinking was the hope to establish communication·
between the LINC and the computers. in the central university facil ity which
would not only permit use of programs already in operation in the'1604 but
would allow further development of.statistical analyses of such complexity and
size as to be awkward if not impossible on the LINC. Our first thought was to
connect the LINC directly to the CDC 160, thus obviating the need for purchase
of a digital tape system. However, plans for major changes in staff and
hardware in the central computer facil ity argued against this approach and-it
was decided to obtain a digital tape unit which could produce tapes with so-·
called IBM format which would also be compatible with CDC systems.
After investigation of the market, a unit was purchased from the DATAMEC
Corporation, the choice being based upon cost, del ivery delay, and ease of
interfacing with the LINC. The interfacing was carried out by means of· II DEC
plug-in modules installed in the LINC data terminal box. The system has been
in satisfactory operation for twelvemonths and provides read and write capabil ity
at the standard density of 200 bits per inch, all operations being carried out
under program control. The particular DATMEC unit purchased 1s also capable
of reading and writing at 556 bits per inch which will permit continuous storage
of the highest 6 bits obtained by the SAMi instruction at a 25 KC rate. The
program now in use provides the usual lateral and longitudinal parity checks as
well as detection of missing characters while reading; end.;.of-file marks can
also be written and searched for. Documentation of our use of the DATAMEC was
prepared and sent to the LINC design group.
The acquisition of the DATAMEC tape system comprised one stage in the
development of a general-purpose data processing package designed to handle
single-unit data for the entire staff of the Laboratory of Neurophysiology.
The first step in the process involves timing of the unit discharges and
stimulus events as they are introduced into .the external 1 ines'of the LINC i·n
real time, and so a general timing program was written. Timing is accomplished·
by the repeated execution of a series of instruct'ions that require 100 microseconds
for completion. During each repetition of, this basic time loop, the various
.
input lines are sampled (by means of SXL) to determine if a pulse has occurred.
If a pulse {either a unit discharge or stimulus marker} is detected, a code
word is stored in memory to register the occurrence of the event. Stimulus
markers are coded as 0; the code word for the single-unit pulse expresses ·the
12-2
time interval which has elapsed since the preceding event, the accumulator being
used as a counter which measures elapsed time in increments of 100 microseconds.
The program monitors the accumulator and, if it overflows, a code word (7777)
signifying this condition is stored in memory. Thus the end result is a table'
of words in memory which provides complete information about the timing of unit
discharges relative to stimulus events. When either the memory has been filled
or the end of the data sample has been reached, the appropriate contents of
memory are written on LINC tape. Since the number of events which can be stored
in the memory is relatively small, the program provides an option whereby the
complete analysis of long sequences of data can be accomplished by memory-sized
bites. Suitable information is recorded with the output data to permit these
bites to be recombined automatically in later processing. As presently arranged,
each experimental sequence may consist of from 1 to 21 LINC tape blocks.,
Since many of the single-unit recordings in our laboratory involve upwards
of 100 different sequences of varying experimental conditions, the abil ity to
identify and edit the'data is of great importance. An editing program has
been developed which, in addition to permitting the addition of identifying
information and comments in manuscript form, will re-arrange the experimental
sequences in any desired order, thus simpl ifying further processing and interpretation. After ptoduction of the edited LINC tape the information can either
be copied on the DATAMEC unit for processing on the CDC 1604 or can be further
processed on the LINC by programs which perform a variety of analyses which do
not require extensive arithmetic calculation. At this stage, many questions will
have been answered satisfactorily and there will be no need for further processing;
those data requiring the capabil ities of a larger computer may be selected for
1604 processing. The larger machine has proven especially useful for analyses
which require a great deal of data shuffling and/or extensive arithmetic
computatio~s.
(See Appendix 1 for details of our 1604 programs). The displayoriented LINC programs include: post-stimulus and inter-spike interval histograms
having scope displays with keyboard-controlled selection of parameters; dot
displays showing the time of occurrence of unit discharges with respect to
stimulus events; and plots of discharge rate versus time with variable count
interval s.
Permanent output records from LINC are obtained by Polaroid oscilloscope
photos, by a MOSELY X-V plotter, or by a recently arrived Teletype unit. The
basic scheme for util izing the X-V plotter is that of Woodbury and Gordon
(April 13, 1964 communication) modified to allow the automatic plotting of
waveforms in an array ranging up to 8 x 8 in size. Positioning of the waveforms
within the array is controlled by the contents of the relay register; simple
digital-analog ladder networds are connected to e'ach of the two octal digits
of this register to provide X and Y coordinate voltages which are added to
the waveform signals (see p. 50 of log book).
2. Specific Projects. Discharges from INFERIOR COLLICULUS NEURONS have
been studied in experiments which stressed binaural interaction (ref. 3). The
LINC post-stimulus histogram display of these data has been particularly useful
in depicting the time-course of inhibitory and facil itory effects from the
respective ears through the use of stimul i which partially overlapped in time.
The fine structure of the discharge trains has been examined by means of
Dr. J. Rose1s 1604 computer programs.' (See Appendix 1).
Discharges of SUPERIOR OLIVARY COMPLEX NEURONS to tone-burst stimul i have
been studied using LINC. Dr. J. Goldberg developed a special real-time
12-3
histogramming program, because he found that the- 100 microsecond resolution of
the original programs was not sufficient to resolve the time-dependent behavior
of units in the superior olivary complex. The program utilized a sequential
array of SXL instructions and could achieve 8 microsecond resolution.
CORTICAL NEURON discharges have been studied using the LINC histogram
programs. In addition, Dr. N. Dubrovsky extended Dr. Goldberg's program to
provide the mean and standard deviation of the initial latency hlstogram, both
values expressed as decimal numbers with precision equivalent to the individual
latency measurements.
COCHLEAR NUCLEUS NEURON discharges have been studied extensively (ref's
1,2). Most of the analyses were done on the CDC 1604, using the LINC-DATAMEC
system to prepare the data for the larger machine. Statistical programs
calculating mean interval, standard deviation of intervals, and correlation
between intervals were util ized (see Appendix 1 for details). In addition,
Dr. D. Greenwood developed and utilized a LINC program which displayed· the data
in a Lettvin dot-pattern display. Dr. Greenwood's program was one of the
first which operated directly on the digital data previously written by the
100 microsecond tim}ng program.
The discharges of THALAMIC NEURONS RESPONDING TO THERMAL STIMULATION OF
THE TONGUE are being studied using the LINC (ref. 4). Data prepared by the 100
microsecond timing program is handled by a program of Dr. R. Bernard. The
number of discharges in variable-width time bins is calculated and displayed
both on the oscilloscope and on the X-Y plotter.
'
A study of NEURONS OF THE AUDITORY NERVE is just beginning and so far
has utilized the standard histogramrning programs.
B.
Evoked Potential Averaging.
Although virtually all of our applications of the LINC have thus far been
directed toward the temporal analysis of single-unit discharges, Mrs. J. Hirsch,
a graduate student, has used the machine to average evoked potentials. Mrs.
Hirsch's program, which has been used on-l ine, will sample up to 8 electrodes.
After each stimulus the last response is displayed, together with the running
average, the average being divided by the proper integer to maintain a constant
vertical scale factor. The output can be displayed on the oscilloscope or
written on the X-Y plotter.
C.
Pilot Study of Automated Cl inical Procedures.
Taking advantage of time when the LINC was not being used by the members
of the Laboratory of Neurophysiology, Dr. P. Hicks of the Department of Medicine
used it to study the feasibility of automating the processing of laboratory test
results and the collecting of patient history data. Using our display subroutines,
Dr. Hicks developed several programs for the manipulation of laboratory test data
and for the display of medical history· questions to patients. On the basis
of his favorable experience with the LINC, Dr. Hicks hopes The Clinical Laboratory
will shortly obtain its own computer equipment (see Appendix 3 for Dr. Hick's
report) •
12-4
III.
FUTURE RESEARCH
A. Neurophysiology Computer Laboratory. Assuming that the,LINC is assign~d
to the Laboratory of Neurophysiology on a permanent basis, we plan to expand
our present appl ications including both the analysis of the temporal characteristi'cs
of unit discharge and the processing of evoked potential data. In, our use of
the LINC thus far it has proven impractic~l to use the machine on-l ine ,during
experiments. This 1 imitation arose because the LINC could not be installed~,
permanently in the auditory microelectrode laboratory in which most of the
experiments are conducted and which is located several hundred feet away from
the only other available site. The d'ifficulty was due not only to the severe
shortage of space in the auditory laboratory but also to the fact that this
facil ity was used frequently 'for other experiments which would have interfered
with the use of LINC for off-l ine appl ications including processing of a large
backlog of magnetically taped data. Attempts to move the machine between the
auditory laboratory and the tape processing site were 1 imited to a few trials
by the awkward character of the procedure.
From this experience arose plans for a new Neurophysiology Computer
Laboratory which will provide for the use of LINC both in on-l ine experiments
and in off-line appl ications. The Medical School administration is furnishing
construction funds and approximately 900 square feet of floor space for this
project which is in the final stage of architectural design. A folding partition
will separate the LINC- room from a sound-insulated electrophysiology laboratory
so that the two facilities can be used either together or independently.
The on-l ine use of LINC during experiments is important from two standpoints.
There is first an obvious advantage in having the results of analysis immediately
available in order to fully exploit the experimental situation. On several
occasions, off-line analysis of our taped data has revealed results which were
completely unsuspected during the experimental session. A second and potentially
more significant benefit of on-line operation 1 ies in the possibil ity of using
the computer to control the experiment. ,In our typical auditory single unit
recording procedure, much time is expended in deciding the programming of the
stimuli and in setting and checking the many knobs and dials. If stimulus
programming can be reduced to a systematic formula which may include response
contingencies, computer control can effect a great savings in time' and el iminate
human error in adjustment of stimulus parameters. Moreover, the experimental
procedure would then be rigorously defined in contrast to the subjective
approach which characterizes our present efforts. \
We are currently designing equipment in which the LINC will be able to
control the timing, frequency, and intensity of tonal ,stimul i. This system will
be used to continue our studies of stimulus coding in single units at several
levels of the auditory system. This will include a continuation of our present
investigation of binaural interaction in units of the inferior colliculus as
well as a new series of experiments on eighth nerve fibers in the squirrel
monkey. While the initial effort in the Neurophysiology Computer Laboratory
will thus be concentrated upon auditory studies, the facil ity will be designed
to handle a broad range of electrophysiological experiments.
B. Biomedical Com utin Division and LINC/ 600 Conn~ction. Early in
1963, one of us JEH was appointed chairman of an ad hoc Medical Center Computing
Services Committee which was requested by the Dean to determine means for
facilitating the application of computer science. to the research and teaching
12-5
program of the Institution. The committee's del iberations culminated in plans
to' establish a Bi()~dical Computing Division (BCD) as a part of the University
of Wis~onsin Computing ~~..,ter (UWCC); an application for support of this activity
is currently u~der t~y'e~by th~ NIH. Our fruitful exp~rience'with LINC resulted
in ~lans'for the initial 'equip~ent configuration of the BCD to be centered
aro~nd ~ second LI~C~~hic~will be installed 'in a new Medical School Data
Ac~~isition~[~borai6ri:withan on~l iMe, pri6rity interrupt connection to the
CDC 3600
computer a't the UWCC.: The ~olJfiguration
, ...
' . . will also include··a line .....
p'rinter'arid card reade'r on..;line to tHe 3600, together with.a digital plotter
and Teletype to"be driven by' the LINC (a block diagram of the equipment is
included as Appendix 4). ,\
.
,
"
,
.
.
I,
•
As an initial step in this development, the LINC now in the Laboratory of
Neurophysioldgy un~~~ 'the Evaluation Program isin the process bfbeing connected
to the 3600. Communication ~etween the LINC and 3600 will be handled by a
. pair ofshi~lde~ 3600 input/output cables, each cable 5/8 iri~hin di~meter and
~ontaining 29~wi~ted-pair transmission lines. Th~'cables will be r6uted through
an existing tunnel 'beneath the' street which separates Sterl ing Hall (site of
the UWCC) and the Medical Sciences·Building. The required length of cable
has been estimated' to be well below the 1000 foot maxi'mum distance which can
be"handled 'by the: 3600 circuits when high-powered 1 ine ~river car~s are used.
The cables include 12 bi-~irectiohal data lines plus a parity bit together
with status and ' control rines. '
The desigry of the interface between the LINC and the CDC transmission
cable is ~rim~ril~ ~he responsibility of John Keenan who recently joined the
BCD o~ganization as an engineer. Keenan has designated the interface as the
SUTURE'box~'a d~aft copy of his description is included as Appendix 5.
The.
design provides f?r th~ exc~angeof control information by ~eans of a common
flag register which i~sharedby the two'computers. Keenan has proposed a
special version of'the OPR instruction which he has named the "FUNCTIONu
instruction which will perform a variety of tasks, e.g., setting, clearilJg,.
or sen~ing of' bits' in the flag register and a word-count register. A draft
copy description of the proposed "FUNCTION" instru~tionis included as Appendix 6.
The system will permit two modes of transmission from the LINC to the 3600.
The usual·'mode of operation wi I I involve transfers of one or more blocks of data
from LINC memory via GULP 'output at a rate of one LINC word every 8 to 14 microseconds.··· Ul'lder' th~ influence of control words at the beginning'of each block,
the','3600 will 'either process each' block as ~t is receive~ or store the information
on 'tape" or drum unti I the last block of a' sequence is received. A second mode
will ~e pr6vided' for ~~ta which require relatjyely long periods of uninterrupted
analog s~mp I iiig 'at the'maxiinum' rat~ 'of'one 'word every' 24mi crose~onds . In' this
mOde t~e 'd~~a ~i 1'1. b~ ~'~ored at'the 3600 on' tape or drum f()r subsequent process ing.
Output from'the 3600' wi 11 be sent back 'to the BCD laborato'ry either to. the 1 ine
printer or' ~o 'the L'fNC ~here it may' be displayed' on the scope or'digital plotter.
'
IV.
.
'
.
TRAINING PROGRAM
,
; ".
~.
~
.,
.
Strong Points
.
'","'"
:" ... ,,~~ ..~
,.' "
"
The greatest training ben~fit that we haye ~eceiyed from. the evaluation
program is the· experience: of' having had it compute'f'in our laboratory during a
critical time in e generated,
and the LINC will not know that there is a character to be read from the keyboard. The LINC will continue to wait, no KBO instruction will be issued and
thus the keyboard will only be released if the reset button is pushed by the
operator. On the other hand if the KCC contacts close too early, data errors
can result from failure of the data 1 ines to "settle" properly before reading
occurs. In the extreme case the KCC contacts may operate so early that they
are always in the operated position, IIKSTR" may be always generated, "RLSO" will
never be generated and again the keyboa.rd will stick because the proper reading
and releasing sequence cannot take place. For example if "KSTRII operates allright but IIRLSOII does not, the key wi 11 be read but no "R~SOIl wi 11 appear to cause
resetting of the RELIP fl ip-flop·. In this case -the KAR relay will remain in the
operated position, the keys may remain down and operating the reset button by the
operator will not clear the trouble.
12-17
--ltSv
,
•
I
,--_~..J(,
""c,
~o;i,
~...iit~
N,f=:
.
..-_--=----~
___
SOL
'MA'~IS
Kc..
I
~ .
Fi'u~e
.
. -----_.
1
.
-
'c
••
·
•
-
•
.._;~~. l(;l t<~TR (Tt>L.'N(~
•
--\].
••
,
N
CD
':-Q..6) R"SD tTl) """.!t.l
1
l.
f-&
......
,
"NO
•
6NI>
'Nt)
.
8 ~.. ,-.N( J:'tt~"'E
.'
1
I
::t>
'-J
I
..
,
,
•
•
,••
•
•
•
•
•
•
•
•
•
•
•I
••
•
•
•
•~--~~--~~~~~~~m~~~tir~~
0%
~
•
=-~'.'"
"D770
..f
Jt.- No~ 7fJ S!:I'I&(! ~...,~no\.J?~,~~~!JT.\
...- -." . .- -=..~ s:1)~C fa..,
;(1t1£. 2r., S. :J
llruSfc1 0 'Vb IN Su nO
(A oru S'i3o 1b '''I cuac
•
tt
At-I. DATA t1 (ot.,r.\oN
NOR.'.\A~LV
CLDSc-D
(ONTl)e?s A~(j Cl.~S~O
IIJ feel.Olii6l> PClSITI '.: : J•
,
I
COt"r,-toN (tftc) eoftJi12t'n
III1JG t:Ol..ltll1~a tj (oN~'~7~/n•
c'o~onG or. NO~(·1tlt.t.. V oPoN
(ONYt"'fS IN VUbl.Y ()Pol1llY'ao
t~ADy .. t"')iG) C.at'JOJilo~J.
\.
Reeort of LINe use by the el inlcal Laboratory, Dept. of Medicine
Programs have been wrItten dealIng wIth two kInds of data:
laboratory test
results and patient medical history data. The following Is a summary of these
programs.
LABORATORY
Programs have been developed which allow the direct Input and manIpulatIon
of laboratory results through the keyboard.
as the decImal values
famllla~
Laboratory results can be typed In
to the technologIsts,
~.g.
% for hematocrIt, mg%
. for glucose, etc. After data Is.typed tn, depressing the "S" key causes all
.
.
data to be converted. to octal values and stored on magnetIc tape. DepressIng
the "P" key recalls th, data· and appropriate subroutInes ,to sort the results.·
Into hIstogram bIns. After sortIng, the histogram
d.l~play
wIth approprIate
legend comments Is started •.
In sumnary, three basic subroutInes have
~ctually
been"Wrltten for
laboratory data:
I. Keyboard Input for laboratory values
2. DecImal to octal conversIon
3. Sort and search of randomly entered data
We are currently trainIng one medIcal
tech~ol~glst
In our clinIcal laboratory
to program the LINC. One student who has had some programming experIence on
other systems spends part tIme helplng'wlth our laboratory computer actIvIty.
Plans for the laboratory LINe actIvity In the Immediate future Include:
I. Development of programs to store laboratory data
accordIng to date 6nonth, day, year) and to retrIeve
stored date tn any tIme IntervaJ requested.
12-20
2. A master program to scale subroutInes to handle e
varIety of laboratory data ••
3. Modtflcatlon of hIstogram routInes supplIed by
NeurophysIology to allow complete operatIon from
the keyboard, IncludIng tenmlnatlon and selectIon
of other routInes.
4. The IntegratIon of all of the programs for laboratory
data wIth questIon and answer dlspJay programs so
that the computer mIght be operated by a technologIst
not traIned In programmIng.
MEDICAL HISTORY DATA
.' ,Programs have been wrItten to dIsplay medical history questIons to patIents
wIth the oscIlloscope and accept the patIent's response through the LINC keyboard. The fIrst test program was wrItten for allergy and the questIons were
desIgned by Dr. Warner Slack.
The Q & A subroutIne of J. Cox was used for the basIc dIsplay.
A separate
control program was wrItten to present a standard response format along wIth
each questIon so that only questIons need to be stored on tape. The control
program also returns to the approprIate locatIon at the end of each question
for InstructIons about the next dIsplay dependIng on the response of the patIent.
thIs control program wIll enable Dr. Slack to wrIte medIcal hIstory routInes
In the LINC wIthout learnIng all aspects of LINC programmIng.
At the'present tIme, the program to take medIcal hIstorIes Is runnIng.
ImmedIate plans are to use the teletype output to generate medIcal hIstory
summarIes after the patIent has fInIshed the examInatIon.
FUTURE ACTIVITY
It Is hoped that In AprIl 1965 we can make flnm plans toward the purchase
or rental of computer equIpment to be based In the clInical laboratory. The
12-21
-AI L-
exact nature of these plans wi 11 depend upon our evaluations which wi II be based
upon our test programming up to April.
Since many of the programs wi II eventuaJ Iy find application In the service
functIons of the medIcal center (taking historIes, handlIng laboratory datal'l
do expect that our comp'utlng actlvJtywl II
evantu~lly
we
JustIfy a laboratory based
computer.
ACKNOWLEDGEMENT
We would lIke to acknowledge the generousrty of the Department of NeurophysIology In making LINC computer time avaIlable to us.
us
to
ThIs tIme has allowed
rapIdly evaluate many aspects of our problems whIch would be Impossrble
otherwIse.
thIs experience has had a very srgnlfIcant impact on our thInking
about computers.
We wou I d a I so J I ke to acknowl edge the many subrout I nes wh I ch have be,en, made
avaIlable to us and WhIch largely account for the progress whIch we have,been
able to make.
G. Phllirp Hicks, Ph. D.
AssIstant Professor of MedIcIne
G~~
12-22
. FACILITIES LOCATED AT THE U wee
(STERLING HALL)
BCD FACILITIES LOCATED AT THE
DATA ACQUISITION LABORATORY
(MEDICAL SCIENCES BUILDING)
REMOTE
LABORATORIES
-
OJ
0
n
'7'
PLUG
-D~
CBRCUDlf
0
-.
OJ
~
-.
M!)DUILE~
1200 CARlY'
~,u.
RUDE.R
·1.000 &J
3606
OJ
3
360G~-
c:
~
("")
nJ.
Pmoo1flElffi
~~1~ ~o61
0
-
-0
+:-
0
X
3
CD
PRINTER
::l
rt
I
. DRUM
~~iAWORDS
OJ
0
I.
I ~ __-iXI'5"O
I
rn
Z
-.
~~1
-0
("")
.........
("")
I
»
-0
CoRE STtlRAaS
65,5:;6
¥8
SIT DATA
\WORDS
I
026
CARD PUNCH
056
VERJFIER
»
N
I
-A13APPENDIX
1 .0
5
GenerClI
t • 1 Three types of Data' Transm iss i on th·st
t'J i
1 J occur betv-!cen the L H!e S,olte 1 i i ti~
and the 36,00 computer are considered ,in this paper.
1.1.1
Mode 1 is a single fixed length
~lock
of dat.a
~'1hich
wi 11 be prc(;cssed or
anal ized by a program in the 3600 with feedb·3c::k provided to the
If.~:C Pr1Id.:e,·
or to the
LINe computer Clt the Bio-t1edical COfnputing Division (B[O) or \'Jith resu!t:~ stored on.
magnetic tape at the' University of Wisconsin Computing Center (UHCC).
1.1.2 Mode 2 covers multi-fixed length blocks of data which
~Jo'uld
be stored on
magnetic tape (or Drum) at the 3600 and procassed or analyzed upon receipt of the
last block of information (as pr.iorities permitted).
to be transmitted to the
~
The results of tili!; ontillysis
printer or to the LINe computer at BCD
01-
!;tored on
magnetic tape at the UWCC.
1.1.3 'Mode 3, for \~hich programming implementation would.be delayed untIl after
Modes 1 and 2, consists of variable length 12 bit word
to the 3600 for recording
On
magnetic tape.
take place on LINe-data channel;
microseconds.
l!I
trans~issions
from the
LINe
Our-ing this mode no ot~,er .operations would
twelve bit "lord \-Iould be transmitted every 24
These tranmmissions might be for periods of 10 to 30 seconds initially
with longer periods possible at still later date.
This latter mode (moce3). imp) ies thst .an extended pe'riod of real-t.ime operation
woul d be requi red since part is 1 I-etransmi 5S i cn or de 1ays cool d not be accommodated. I?'.
In addition, these data rates imply th~t about 624,000 -48 bit ''lords ~sould be received
ea.ch minute" thus a cepab i1 it Y for chang i n9 output magnet i c tapes wou 1d b~ requ ired.
1.2 There are a number of general requirements (or assumptions) '\fhich are basic.
to this paper.
1.2.1
The 3600 will be the 'mQster" co,'llputer and the ·lUfe· a "sJsve" for all
transfers.
The L INC, however, wi 11 Mve the capab iIi ty to request 3600 sct i on, land
will have the capability of terminating a transfer that has been started should the
need sri se.
12-24
-A141.2.2 A control g~oup of three 3600 words (12 LINC words) will always be trans~
mitted as the first words of a block tramsfer.'
The control group
f(j;·m~t \'in
specified by the UWCC and \olilt allow some area for U$er information.
i be
The first word
(the fi rst four LINe words) wi 11 include the number of \"Iords in the block and an
indic8tion
8S
to whether another block fol10115.
Bit 11 'of the first LINe wOid, (bit
47 of the first 3600 word) will be Bone if'at least one more block· follows this
block.
(This is the chaining bit position in the I/o control word for the 3600).
Bits
24 through 32 indicate the number of 3600 words to be transmitted (the number of LINt
words modu10
4), bit position 24 Is the least significant.
(This word count will
occupy positions 0 -through 8 of the second LINe word transmitted
~-
the choice of
location is based upon the word count bit lc~tions in the 3600 input - output control
word.)
The value of this count can vary fran 3 to 703 8, but most transfers will be
for 7038 words (i .e. 3 control words and 448 data words)
It is 8ssunled that the
3600 will modify its I/o order to'reflect the correct number of words after the. first
,word i·s in the 36~2.0
~
Sequence of Events LINe to 3600 transfers (mode 1 &. 2)
"
2 . . 1 The 3600 Monitor program must be operating in the 3600, and the 'mask in
the "Synchronizer andUniver581 Transceiver for University Research Experiments"
~
(SUTURE) box at the LINt must be set to allow' interrupt, on "LINC 'Transmit" (status bi.t M.
"
~ CMd].
2.2 The LINt loads its A register with the word count' (value of the count in
,
011<::'
.
'
paragraph 1.2.2 times four), and sets the,proper bits inlocation'.~ to perf~rm
8
,
specific LINe operate instruction (hereafter called a LINC function instruction).
"
This specific function
'.
instructi~n
will Set status bit II in the SlTfURE box 1 'if the
3600 is running, put the A register into the count register, and
LINe whether the 3600 is running.
indi~te to the
If the 3600 is running, the LINe then loads A
wi th the start ;'ng address in L INC memory to be used for the transfer (norma J 1y 3648)
- and execute a "Gulp"instruction (to enable the Gulp mode).
The,SUTURE box w:i~
cause loading of A through B into the memory address counter (S) and then wait, if
necessary, for the 3600 read instruction.
(If no operation takes place for an
12-25
-A15-
r'.
"".'
extended period, t5Ne N~)tdd ie.uve the Gulp mt::ad and S~1t ST.~TUS bit i;.}
2.3
The 3600 would
(wh~ch
thf.: inter:'upt, determine that the lINC.desi.ed
8
The 3600 would clei'Jr Status b,t l~ in the SUTURE box"nnd ISSu~ a connect
·treed" ordero
order
s~:rvice
would be Bccepted by the
'k
SUTUF~E
box hardware) • .The 3600 \'Joulcl enable
t4
interrupt on Status bit ~ ar~ and issue a read instruction (nonmal~Y.for 7038 words)
,
~9zt
t'lhich would enab1e timing",the "MOUr'· level in the SUTURE box caus,ing an output gulp
mode to take
2.4
pla~e.
T\'/elve bit words would be tranSf€=,Tad automatically between the LINe and 3606
gover~ed by
data cha'nnel)as
the leneta" sod· '!replyl' signals)unt 7 1 an' "end" condition
appeared.
Normal End would be: (LINe <:ount =0) and (3600 count a'O), the SUTURE
(a)
'bit~ to
box sets STATUS
after the SUTURE.count
(b)
Tal
If the 3606 sends another date signal
indicate normal end.
0, t.he SU1'URE will return ;'end of rec9rd" instead ,of
"reply"~
Premature end by 3600 would be: the 3606 drops Its read'line (due to
parity error received or due to some other malfunction), the "LINe box must terminate
"
q
"
its gulpmode and set status bit ~ (error detenmined by LINe)
(c) Premature end
,"end of rocord"on the
At the 3606 the
box.
2.5
ne~t
b}I LINC~ould
.be: The SUTURE count a 0, thus liNe sends
data Signal.
LINt also gets status bit
~ount would
'f
~ setting
2.6
the SUTURE
z~ro.
If the end is normal from the LINe standpoint then status bit
will cause a 3600 interrupt.
bit
not equal
~
II in
2--
~
setting
If the end. is abnorwal·from the LINe, standpoint status
will cause the 3600 interrupt.
lhe -3600 prog.-am wi 11 check the "'lord count after the interrupt ·and check '
the status regi ster to determi ne whi ch end cond,; t ion exi sts.
~
(8)
NORMAL: (count
. ,
III
0' (status bit J
.
II:
1) (no pc:rl \:yerrcr)
q
(b) Prematurely ended by 3600: (Status bit' .. 1)
(c)
(parity error.
Error detected by 3600: (status bit
t
II'J
1) but {word count
III
0)" or
1).
,.
.'1
The 3600 must clear status bits ,S and/or. end, in addition, it must set
o
status bit i if condi tion "C" exists.
In case of condl tlon (b) or (c) the LINe Is
12 _26
-A16expected to retransmit the same block again (or seek operato~· action).
~
st~tu~
2.7 When the LINe notes status bit' is clear, it must check
0
bit. as
(. eZl(A(s
well.
If status bit
bitQ
sa
.1i)f
OJ then L INC can proceed to' next block tr~.nsfer; if status
JJLINC must retransmit the block (oc· seek operator action).
an
either case
step 2.2 is the next action required by LINCo
When 3600 has receiv~d a good block (condition 2.6 (e»
2.8
j't
must decide what
further action is required (i.e. sto,"e on mag tepe, store on drum, call in user's
progr~3.0 Sequence of Events 3600 to LINC transfers, (modes 1 :;. 2)
.
If the 3600 desires to transmit to the LINe, the 3600 sets status bit 2 in
3.1.
'-
the SUTURE box and masks on status bit ~ (LINe ready) to wait for an 'i~terrupt.
9
3.2.
The LINe checks for status bit. with a function instruction.
if bit
I
4t .. 1, then LINe loads 3414 into A, and issues a functi,on instruction t-aoic:h will put
~"-
g
A into the count ;-egister, clear status bit fD and set status bit <; (LHlt ready) LINe
then puts the starting location (364 ) into A and
8
enable the gulp mode).
e~ecutes a gulp
inst:'uction(to
Gulp mode puts the A value into the, memory address counter
(via B) and waits, if, necessary, for a write instruction from the 3600.
• ;z....
I, It c 1ears status b .~t ~
v an d ena b 1es
3 • 3 Wh en t he 36 00 .Interrupts on status b It
interrupt on status bits
'f
z..
~ and ~ (normal
or error ending) and on·end of chain (in
. -f,'&t.tfH{! -fa SetThe 3600 then issues a w."ite instruction (which enabl'~?t ',t~HNP:' in the SUTURE
the 3606)..
box and makes the LINC gulp instruction an input to the LINe type of tr~nsfer).
3.4 "Data" and 'Cfta~I{1 control the transferring of 12 bit words betwelan the 3606
and SUTURE box until and end condition:
(a)
SUTURE
Normal End would be:
box sets status bit
v when
~
~SUTURE count • 0) and (3600
count.- 0); the'
its count is zero 6nd leaves the gulp mode.
I
(b)
~
LINt prematurely ends transfer would be: '(SUTURE count
)
,z..
IS
0) but (3600
.
count?p. 0 ; the SUTURE would set status bit 9 when its count equalled zero ant leave
gulp mode.
of "repl y ".
Any future data signals from the 3606 would receive :'e,nd of recorc l ' instead
(This assumes 3600 looks for "end of record la on WRITE" an assumption which
has not been checked).
12-27
-A17-
faj'~llre
or by 3600 program if on error \tJas tbtec',:eclj-
SU1'IJRfS
san:~,;;~ tlT;;:C
line
I;
drc)pping Nith count
:z
0; SlJru:tr: sets status bit ~ (e."i"or detected i:,y L~N(;l;i End leave!'
GULP mode.
'l!'
3-5
When the 3600 interrupted by status bit §, if the ccu".".:
r.n !)
fmc nc i;rc!i1~j~'
7.!,..
mi 5S i on par i ty errors \,Je;·e detected, the block transfer \\](}s cc;rro::t
can b~~ c:leared ,'!nd the
bit
S\",i~
n·2xt
a one but the count is incorrect
~
when it notes that status bit one
mU~it
l' i':
block or requi r(~ment c·~n bs han
srl1Tv~
/-..IN ~
PO~,nON
o
3
-t;--'l--/-{)--+--~-rr'-b-' 'k-Z;~7c7f5t1T(}i?
J
-
.~ilA:'l,t18t.~
/'Of{
8vr NOT CONUlc7l:0
IttrJw1ur.lT1J~e-
36 tn> P1~
:
~~r; ~
C'1 tJ
6~{- CowtlT¥o
,(J.J
rl 'zE
-
b
?
9
-
/.INt:.
-
"IN(
R(.)AllVflfJ~
~1d:::::'f ~~r:
(/.left SIC 1'111 f.. ~IA V 13~
Re.4dJ
'tl.rAl(J:~~€ ~~~
6~~ for;,,- 3 C:,(J"'l) w~
1~~--I---_--+-/..-I.-'lJ/-c.--w-~-·--::::-----··---~ ~~~~
/I
-
A INC
f>~7i ~
(.I./~Jof.f) (-h~eI 0-#)
'AJVN~.'t~"
,~1' ·~c.oG. --lorIS M.o. ~tJotJN
6-i1!A~~ 3'~ ~
11.10 ,PVh'/~
Mer}' ,:.
L INC
II
I
»
ex>
I
APPENDIX
referred to ES the FUNCTION instruction.
perfo~m
the
This
in~truct;on
c5~terna1
under program ccntrci cartsin functions
st~tus
location
-A19-
6
of
e;;~tern~l counters
one~ hf~resfter callec~
or
flip~rlops,
lo~..,ciing
.;~-nd
the contrel word,
has been
to
tD the L!NC such DS: raoding
of e;(ternal cOtmters¥ ;;md/oi"
the ]2 bits of inf(u·mr.:tion in
the l\ register are used to give detailed logre::"l instf"Jctions to thu
instruction.
d8s~9nsd
FUNC1im~
Yhe e;'tecution time is 2ft- U sec un1e~;s hit 3 1n the contro~ \fJord i5.
zero when .s 3600 interface type
FUNCT[m~
which ind!cates that an ext:e&-n1l11 count
instruct!on executed.
in the ietter case;-
h:: to be t~eud into the A register,
~egister
requires 32 u sec for execution.
1.2
On all function instructions: the condi~;ion {ona or-zero} of the appropriate
exte rna 1 status 1- i nas pr i or to cxecut i on of the ins t ruct ion will cause comp 1ement i n9
of the correspondi n9 hi 9h order' bits in the control
condition is a·one.
It is expected that this 'reading·' of external status
conditions wi 11 always immediately followtfiY a
2.0.
2.2
if the external status
This compiementing will change the conf:guration of the
control word in memory.
2.1
1J10rd
'gulp' mode .type operate instruction.
Procedure
The general procedure for performing a FUNCTION INSTRUCTION IS:
t.oad core memory 1ocat! on one
(b)
toad the A regi sterwE th
(e)
execute the function instruction
(d)
examine(or ~aAthe A register if the count register has been read
(~.)
examine control ·"'lord to detennine tha condition of external status
lines, if required.
B
\,/i
th
th:~
(a)
proper control word
count, if r'equi red
move ~ro~ IOCA~ -/we»
Figure one shows the format. for a contrel \1m'"d to work wi th the LINC/3600
interface ($UTU:~E box).
If thel~~est order· ~zf3rc,) bit is set, the instruction will
12-30
,!
-A20-
cause
cle~ring
of designated (see bits 6 through 11).
lCo'.iest order bit is zero no ciear-ing will
bit (the one bit) is a one (i
designated
t~ke
piace.
st~tus
fl ip-fiopsi if tha
~f
ne~t
the
.e. is set);t the instruction will
most signifi'i;ant
~use ·seti.:ing of
rf brt one is a zero, the flrp-flops will not be set.
f\~p-flops;
If
"the next higher bit, bit 2, is set then the value in the A register will b~::
transferred to the external count register. if it iz zero the 'count
I
will be read
into the A register and into core memory location tt...Jt).
2.3
Bits
3, 4, and 5 in the control word designate the external area
operated on by the FUNCTION instruction.
SUTURE
b~x
(3600/LINC
interf~ce)
When bits 3,4, and 5 are all zero, the
is specified.
The high order bits 6 through 11
indicate the. condition of status 1 ines and the status 1 ines
depending upon the settir." of bits 0 and 1.
~,::
~lhich .a~a i~J
•
h",' ;. .tered
The status lines \ihtch are a one always
comp 1ament the i r corresr~l'id i ng bit in the control
6 through 11 are all zero:
t~ be
wo,-d
\ '111 be'.'reed"
(thus if bits 0, 1, and
intO' the control word.
none of the status flip-flops can be both set. and reset by the
Since
LINe, both setting
and resulting (clearing) can take place with the same instruction.
Bits 6 and 7
can only be reset, bits 8 and 9 can only be set, and bits 10 and 11 can neitherb
be set nor reset, only 'read".
Since the re,!lding of status.is a compl imenting
act on, certain bits sn locations 6 through 11 could be 'masked off" by setting
them as a
'one·' in the control word if it was expacted that
"one" state when read.
A setting
Qf
zero when the status 1 ine
be a zero would also achieve the same effect.
~re
~hey
would be in the
,,r~
0,0
~
~)t..';'E~
:.,' .::' ~ I tso.'
)AIh(~
Ie
(it C"
:J,o
~.I
;Z,2.,
~
.J;
~
~
~
'"
..
~~r \:e,,~
tl{l('
('12-
Jj1"
. j,l,
,.
':! , .~
-
iP)
CiS')
btJl.P~ ~~
'\., ...',£: ....
...
4'
.......
......
- ..
,
(,)
... -- --
-
-
,
rD
- --
XI.. ~
-
Mot),..
-- -
~,AlP
B~G,i(r.~
- .-
--
bUt... P
+
l~) (op,:z~): ) :!f. P;:"'
CrF2 ) ("itl')
:
0
v':'
In.:: t.
-
-
\
.
-I
--I
.},,)'J('
R~~~'
~ F"F'
(F~I)(~t/) (~:
'!:'
-
-.-
w~U4!~
\ ~rFL.
I
(~, I) i-~')(8B2.f)+tfr-I)(BS~0'il(FF'k):
:t>
0
N
-=P F F' 2-
Fi6URE
FUN C1
I0
~
W
I
I WI)
't.Js" ~OC71
()IJ
FINAL REPORT
LINe Evaluation Program
J. Lederberg
L. Hundley
Department of Genetics
Stanford University
March 1965
13-1'
CONTENTS
I.
II.
III.
IV.
I ntroduct ion
General Use
l~O
Equipment and Programs
Util i ty Programs.
Experiment-Related Programs and Hardware
I
V.
VI.
VII.
The LINC EvaluatIon Program as a Training Technique
Computer Performance
Conclusion
Appendix A:
Selected Programs
i.
I.
INTRODUCTION
The instrumentation Research Laboratory wIthin the Department of
Genetics has as· its' purpose the design of special purpose instruments for
biological research. ThIs includes electrical, mechanIcal and optical
design. The LINC in our lab~ratory has been used as a system element in
a number of experimental. situations and its use has p.roven to be both
education to us and experimentally rewarding.
Headed by Dr. Joshua Lederberg and under the direction of Dr. EllIot
Levinthal, the laboratory has as its primary mission the development of
1 ife detection systems on a microbial level for remote MartIan exploration. In order to accomplish this end, a number of different types of
physical measurements have been investigated in great detaIl. We believe
that these studies, a number of which involve LINC, will also result in
new instrumentation and techniques of general laboratory util ity.
We wish to request that the LINC be permanently assigned to our laboratory.
.
ii.
,
13-3
· II.
GENERAL USAGE 1-0 EQUIPMENT AND PROGRAMS
Our LINC has been e~uipped ~ith a number of peripheral devices. These
include a Datamec IBM compatible tape recorder, a Calcomp plotter, and a
teletype. In the process of being installed is a 4096 word external memory.
The Datamec is equipped for two speed (45 and 4.5 ips) two density
(200 and 566 bpi) operation, with both read and write capability. These
speeds and densities' give us a wide range of data rates. The upper limit
is 25,000 six bit characters per second. The interface is very simple and
required only two cards. One of these would be eliminated if the gated
accumulator lines were being used for nothing else.
Programs for the Datamec 'include those to read and write IBM compatible
forma~
generate data tapes from continuous on-line input, and to regroup
the input data on LINC tape blocks and then, if desired, to rewrite these
blocks into IBM format. All of these combinations form a highly flexible
system. Use of the Datamec has completely superseded the IBM 026 keypunch.
The. Calcomp plotter has been in operation for some time now and has
proven, to be extremely useful. Programs for plotting all forms of data
have been written. These i'nclude both ordinate and abscissa scaling and
linear, interpolation. A program has also been written for character generation which includes character size scaling and positioning.
Th~ teletype has proven to be a very good means of getting both program
and data into LINC and getting hard copy of both out. Its major drawbacks
are its low speed, lack of tabs and that it is somewhat noisy; however, we
know of no 9heaper means of getting printed output., Input and output routines ,have been written whIch calculate teletype code from LAP code and
viseversa which take about twenty locations each, so memory usage is not
excess i ve~
The 4096 word memory, which should beln operation within the next few
weeks, will be used both for program and data handling- There wIll be
three modes of operation which are: 256 word Input and output gulps at
1•
13-4
eight microseconds per word and single word input which indexes the memory address ,register with each input.
This later mode Is designed mainly
for data handling.
III.
UTILITY PROGRAMS
These programs include those for program input, assembly, and debugging, for keyboard data input and computation and for data display.
Most
of the programs to be mentioned are more completely described in Appendix
A.
The LINCT system is our teletype program text input-output system which
has a number of useful features.
It is tied into a modified LAP which
will assemble for the 2K memory.
We have operating on the IBM 7090 a compiler for liNe which uses a
modified Balgol language.
This system, called "BLINC", and a program op-.
erating system which was written in BllNC are described in some detail in
the appendix.
Debugging routines include an octal to Mnemonic converter and printout prpgram, and a program which follows another program through all of
its br~nching to determine which locations contain instructions and which
contain constants.
This is used with
th~
converter program to get a prop-
er print-out. A print-out of LAP III was obtained in this way.
A Floating point package with two word mantissa has been written.
" Copies of this program and a usage explanation will be avai.lable shortly.
This program has been incorporated into a desk calculator with storage
routine. This routine has the usual arithmetic operations as well as
square root, eX, log X, sin XJ cos X , and 2 X 2 Chi square. It is
e
. arranged fo~ the easy addition of other arithmetIc s~broutines. Teletype
input
ed.
A
~nd
output and certain manipulations of the stored data are includ-
n.~mber
of simple algebraic pr.')grams have been written, such as those
for mean and standard deviation,
C~
i square and other statistical opera-
t ions'.
Display programs include those 'for point and bar graph display wlth X
and A scaling keyboard calling
2.
of~jata
sets.
These data sets may be
13-5
manipulated in a number of ways including inversion, addition, multipl ication and rptation.
These are the major programs of a general usage nature now in operation.
The only major programing effort now
bei~g
considered in this class
is a simple arithmetic compiler based on the two word floating point system.
A more complete symbolic compiler is a possibil ity, but due to the
large amount of effort involved will probably not be undertaken for some
time.
IV.
EXPER I MEf\4' RELATED PROGRAMS AND HARD\·/ARE
Most of the research in our department is involved with
experimen~a
tion either on a bacterial or molecular level; therefore all 6f the online LiNC experiments that have been done have involved physical methods
such as mass spectroscopy, radioactive and fluorescent tagging, fluorescent decay times and particle counting.
An anticipated
~xperimentlin
volves the interpretation of Raman, spectra.
The LINC has been directly connected to the output of the Bendix timeo~-flight
mass spectrometer~ Output from the mass spectrometer is reduced as it comes into LINC into mass amplitude ana time of occurrence.
The direct determination of mass number is difficult due to instabil tty
in the Bendix's scanning ramp.
One means of overcoming this, which will
be tried, is to allow LINC to generate the scanning ramp by the use of a
mechan,cal D-A converter which has been built in our shop.
This consists
of a 290 step per revolution stepping motor driving a ten turn pot.
is a very simple system and has proven"most useful.
This
This'use of LINC ties
in with a much larger system which is a computer'program for the direct
determination of compound composition from mass spectra.
This
worl'~
is be-
ing done under a separate grant and the initial program is being run on
the IBM 7090 at the Stanford
Comp~tation
Center.
: The LINC has been used in a number of ways in experiments with fluore?cent compounds.
tOG~hase
oremeter.
The first experiment of this type used LINe as modula-
locked detector, and integrator in an extremely sensitive fluWith integration times of ten minutes .. the detection of 10_ 13
13-6·
3·
molar solutions of fluorescein with a signal to noise of 15 to 1 wereobtained using,a 400 milliwatt 1 ight source.
This experiment was performed
to determine parameters for a sensitive fluoremeter as part of our effort
to design. apparatus for the detection of I ife on Mars.
A program is now
being written which will determine the best fluorescent system transfer
function for a given material by generating ail possible combinations of
filters, light sources, and phototubes.
The data for the components of
this system will be stored as sets on LINe tape.
A system has been built for the determination of fluorescent decay
times in the low nanosecond region.
This consists of a fast flash lamp,
photomultipl ier tube, sampl ing scope and LINC as a 512 channel integrator.
Calculation shows that we will get about two quanta per channel per flash.
Our design goal is to investigate materials with decay times on the order
of five nanoseconds.
second region.
improv~d
To date, our best results have been in the 10 nano-
The 1 imiting factor is the lamp decay time.
by the use of a different type of lamp.
most admirably in .this appl ication.
This will be
The LINC has performed.
No external hardware was required ex-
cept the mechanical D-A convertor for driving the sampl ing ?cope sweep.
This experiment is being conducted in cooperation with Dr. Lubert Stryer
of the .. Stanford Biochemistry Department.
A program will be written to get
a best.·exponential fit to the experimental data so that direct time constant output will be available.
Programs have been written for the keyboard input of data from nuclear
counters which determine man and standard deviations as well as sorting
data sets according to size distributions and normalizing the data.
These programs have been in routine use by a group in the Genetics Department under the direction of
D~.
Leonard Herzenberg.
This group 'is study-
ing antibody 'reactions in mice.
These programs have, by the rapid presentation of, results, allowed the
/
e~peri~enter$
b~ w~th
to determine what. the next step in their procedure should
very I ittle delay, and has therefore increased the number of ex-
periments 'which they are able to perform by a factor of two to three.
4.
13-7
V.
THE LINC EVALUATION PROGRAM AS A'TRAINING'TECHNIQUE. ,'" ,
In general,:the experience gained with digital techniques has been of
great value to all of us here. The instruction initially received on,
LINe was quite adequate with one exception. It would have been very desirable to spend mo're time on use and misuse of the various 1.-0 'functions.
I·t has been In this area that most of our nonproduct~ve time has been
spent. From 'the overall point of view, LINC has been a most demanding
teacher in its own right. It has changed and simplified our approach to
many problems. It has also made' possible experiments which ,would otherwise have been too time consuming to perform.
Several undergraduate and medical students have gained proficiency in
systems programing on LINC. It is an excellent machine from the standpoint of man-machine i."teraction but higher level languages would gl,ve a,'
more realistic interaction to sophisticated systems.
VI.
COMPUTER PERFORMANCE
The 'performance of LI NC 'i n respect to ma i nten'ance has far exceeded'
reasonable ~xpectation. After approxim~tely ;200 hours of operation,
the only failures have been one bad cable connection and two ~utput tral'lsistors whose failure can be traced to ext~rnal misuse.
The general perf~rmance'of LINC in the laboratory has been entirely
adequate and most rewardi,rag. Most of the recommendations that come to
~ind must be admitted to be generated by our own special requirements;
However, there are three recommendations which it is felt are of-general
interest to most users.
"
/
.
The first area is that of multiple word arithmetic. Any instruction
changes which would reduce program length ,and running time would be a
great', he 1p. These mi ght i ncl ude cJ ear i ng the accumul ater on a LAM t n~1
struc,ion' and recovery of both halves of ~ multiply.
'; Th,~ second suggestion is to make all' of the 2K memory programable.
Tris ~ould be very useful when performing complex computations and
13-8
would reduce the running time of a number of programs which we now oper'ate by minimizing the number of tape transfers involved. A suggested
means of achieving this 1s:being transmitted ,under separate cover to S.~
~rinsten at the Computor Research Laboratory.
Our third point is that a problem-oriented compiler (e.g. artran or
Atyol) would be extremely useful. Even if the compilation were somewhat
5.1 ow" , the reduction in' programing time should still ~e very large.
Mnemonic print-outs of the compiled program can allow the programer to
see' exactly what is happening and give him a framework in'which to get
machine code zonatIons.
VII.
CONCLUSIONS
.The concept of what an ideal laboratory computer should be will vary
greatry among various investigators. From our point of view" LINC has
proven to be a very useful system. The careful attention of the designclrs to those points which are most important for the on~l ine use of a
domputer is obvious and most gratifying.
It:has become apparent that in the fuiure we,will warit to have on-line
cbmpui~r capability even greate~ than that provided by LINC.
Greater
word l'bngth" higher A-D resolution" larger memory" greater speed and smaller physical size will be the types of improvements that we will be looking for in new machines. A system such as IBM's 1800 is a step in the
right direction. This desire for a larger,capability has certainly been
the result of the use of'LINC itself. We feel that future developments
must proceed in this direction if full advantage is to be taken of the
experience gained from the LINC program.
,
,
6~
13-9
Appendix:
A
. Selected Program Discriptions
13-10
Double Precision Floating Point
Winter 1965 '
t. coburn
13-11
General Informa~ion
1. A double precision floating point word consls~8 of three 12-bit
words in the following sequence: exponent, high order word, low
order word. The last two of these are collectively called the
"mantissa".
2. Exponent and mantissa each conts,in a sign in the leftmost bit,
i.e. the 11 bit of the exponent or 23 bit of the mantissa.
3. The ,mantissa is a fraction between +1 and -li that is, the
decimal point is assumed to be at the left of bit,22.
4.' The mantissa is left adjusted. This means-that exoept for zero
words, all positiv~ mantissas will contain a 1 in bit a2, and
all negative words will contain a zero in'bit 22.
5. Integer? c~n and indeed must be used for some of the (routines
,available.~These are automatically flqated before ~hey are used.'
6. A floating . point accumulator(:B'AC') is maintained in locations
1120, ~12l~ and 1122. It is use~ in the same way that the
regula~ accumulator is used.
...
. -, 7. The' otl;ler r-alf of any operat-i-on,'is"cal,led the operand or argument, " "
8. The ad~res~ of a double precision floating point word is the
location
of.... the exponent.' Inte3~rs a.re addressed as 'usual.
,
9. The fl~ati~3 point routines uselndex regisiers'12-l7,' These
registrrs ~re' not restored on lEfa,ving _the floating point paoltsge,.
,10. Entrance to the floating point package is accomplis~ed by
jumpi~g tg a three instruot,ion routinelooated aome plaoe in
- core (.seeanext page).
'
11., After"j~he :'last operation code ~he pr~gram exi ts and oontiuea
exeouting regular ,Linc instructions. ',
12. There is no rounding off within the floating point paokage,
,
~
,
I
"
"
_
"
~.
,
(
"
13-12
Instructions for using the oackage
The following sequence of instructions will serve as an example
of the necessary format.
176
177
200
201
202
203
204
205
.206
Jmp 375 -----------------------------------375
0400
(operand address)
376
4001
(Qperation ccide)
377
·0403
(operand address)
(operation code)
4002
(operand address)
0400
0023(operation co~e)
\
Lda
o
Jmp 1000
.
207
210
Operand Address
This may be a'direct add~ess: 400
or a~ inqirect address:
4002
or it ma~ be zero.
1. In0a dire~t address the loc8tion, 400, contai,ns the' exponent
of the flo~ting point word, or an integer as ~he case maybe.
2. In an ind~rect address the index ~e3ister, 2,1 refers to the
correspon~1n3 address. Bit 12, the 4000, bit s1gn1f10s that the
'address ia indirect.,
3. A zerb operand refers to the floating point acoumulator. Hence, to
square a number in the FAC, one executes a multiply specifying a
,zero operand.
Operations
1. Ope~ations av~ilable are listed in the f?llowin~ table.
2. The 4000 ,bit in the operation code is: used to indicate whether
this ope~ation is the last in a series. In the example above,
if th,e na,rt location follo\'I1"ng the code 0023 contained 0400,
this woul. d be interpreted as "sxll!. If, the last code had been
.
4023, th~n the next location would be the a.ddress of an operand.',
3. Some :iope~"ations, ,fix and sign, are meaningless unless they are
the ~ast ~in a series since the result is left in the regular
a 0 cum \.11 a ti 0 r •
~.
,
-.
~...
'
.
'
"
13~13
Table bf codes and operations
Code
1
2
3
4·
5
6
7
10
11
12
13
14
15
16
, 17
20
CIa
Add
Com
l~ul
FAC/OP
OF/FAC
I+FAC
IxFAC
FAC/I
' I/FAC
Fix
Flt
Clr
Max
Min
SGn .
0
21
22
Sub
23
Sto
' SSP
'SSM
Operation
Clear and add operand to FAC.
Add a floating point w9rd to FAC.
Complement operand; leave in FAC.
Multiply FAC times floating point word.
Divide Fac by floating point word.
Divide operand by FAC, result in FAC.
Add an integer to FAC
Multiply FAC by an integer.
Divide FAC by an integer.
Divide int~ger by FAC, result in FAC.
Convert a floating point word to an integer. Result
is left in regular accumulator.
,
Float an integer, result in FAC.
Zero put in operand and FAC.
Compare size, of, operand with FAC. Larger left in FAC.'
'n
"
T'
Smaller"
"
If operand is less ,than zero, -1 is left in regular ace.
If operand equals zero, 0 is left ,in regular acc.
If operand is greater than zero, ,+1 is left in reg. acc.
,Increment operand by FAC, +eave in' ,Fac-p.,s ,wel1. This is'
e'qui valent to an, add to memOry. :, I:',
. ', "._.,', '" ,..:.. .. .....
1
~SUbtract
••
__
...
operand from FAC. Result ileiJt ,in
"
'
~JAc..:
.;
J
~
, "
--;--",---:,---'
'\
'\
'24
25
Set sign of operand minus; "
"
"
"
positive.
"I' 3::·~··;l'.I}J...
.
I
,
"
; .
•• '
1.
I
AnD. l'
1 (,~ (/\ 1 1776
1 (~(.I.~
1 (Mtly
STr.
17
1 til 01 ,1 SF.Tr 1 :3
-
1 (I; (,11.J
1JQ 1
1 (~(ilS
SFT~
1 17, (1\/)
1 (I, (l, 7
1 I??
SPTT
J (t\ J
1 1Q5
J til
(7j
11
]4
16
SETT 15
...
1 (I) 1 ~ I lP.4
1 rt\ 1 ~~ LJ)tlr17
, 1 (11 1 LJ tl7.F.
1 (I) J 5 •.IM, ? 1 0? ~
1"1,' , i (;j 3
1 (11] 6 \,.'
Ifill7
ll7i? (1\
ll/1~l
-'.'
I
j ~jJ-=t
j
~oA-:"
i I £J. CJ
p(c. ~ \..\.~
~P()T
JrviP
J (/'?:3
6("(7,(11
STC
HLT
1 (Ii?
')
1 til? 5
1 ("?A STpr
1 W?7 (1
1 ~.1i/1 STC 4Q'
J.,
1 ()\~ 1 1_ f) ,6 12
1 r/l~? STC } 12 ~1
10~,l
f e. \",,,,\1\. J.
c)..v\~ O~e. \o...\CO\l\.
1(7J26
1 (1,2? SCQr
I (IJ?4
0
f
I
f
Ln~iJ?
I
1 (;1~, L. STC 1 I? 4
1 (135 L.nAr I?
1 fi"~h siC 1 l? 5
.. _-
......
...-... , - ..
1 (I, 'J 7 CLl~
1 0. /1 rl) I..DAi 1 -,
1 (,/I Li 1 ~(;L r
I (I,/J? LdA (I. Ci'l
j (I. L: ~, Ail A r
1 ('/iLl L.J 1 (/,72
1 ~,,/, ~ STt 12
J (I, ~6 LDA IP.
1 (1,1.J7 ST(;,
1 C/·5(;) HLT
1 ,,, ~ J
STC
i v.':)? L. j) A,
IV/53 A Pc.
1 [;>:14 ,];V. tJ
1 (I, S ') L f)A i'
1 Oi C) 6 6r~1
1 (I. ~ 7 ~\;l 0
1 (/. (, e. SiC
1 fi~ r-, 1 LnAT
1 (/,~?
"
107 1 .J'IjP
1 r;j -, () , ) ~/J tJ
-; ~j
j
-, 4
c~
17
1 0',13
17
1 (1,63
(ij
1(.'063 HLT
1 (,~6 4 J:vjp
1 (ilh S Jiv'IP
1 (i', (, () J:V, P
1 (/·67 .. 1."~ I~,
1 (1:701 "J,Yj P
1 (,I,
1062
,J:V; ~
•.H'i P
1 (,~ 'IS ,,1:v,P
1 ~76 ,J ~·i fJ
1 (1,77 JMP
1LJS01
1 L.4 1 .
j (7, ') 1
}
j
).650
1~?1
j LJ
4 1
j (A
51
107 1
1 177 iP
l~)
~.;.
1 51,.;
J(I."'7
J,''iP 1 1 35
1 l(l)tS J,'viP 1 (ijAL,
1 107 J;'VlP lh53
1 1 1 (tl ... /i\1? I~~t
J J j I ,j,'''IP j 5"}6
1 1 J? JMP j555
1 1 13 ,,i.\1 F' 1 1 ~l
I ! 1 /., ,J;'Ij? 177/1
1 1 l~ ,j,\I, P 1137
1 1 It) ,1,vI P ]571
1 1 17 ,.1 t''lP 157 L/
J I :::> r~ til
'".'
J 1 P. J (1
1 1 ('-.4
1 j(/)5
(11
1?1
V'I
j
'. - ----.---.-::t: --'" ..
,:>
, j .\~
J J:::>~
,.---....".--...-...-- .... -.:.
~/)
:j
.j
,.--.--.----.--~.
J
"}.
Il?Li ("
(I,
] I?
1P.6
1
7
1 J :1 li1
r" ] . '.
r"
I 1~ 1
1 ] .l~
J 1 .l.l
1 1 ~~L'
J 11')
J J.lf,
',..
j'
1 1:::> S
(11
fl,
(,)
i
J
.i~ p 1 171
,;,v;,J 1 1 ~7
.1 ....,;-'
1 LJ4J
. JV.:=-
147 L!
1 ],:.,7
1 ll.;(i,
s;;-~·
l?
llL;j
L[).0.
•
1 (/;'? 7
1 ] I,? 11 Pli1
1 ) ~,l ST (.l ,1:::> ,:.
I jilL! !-O-" 13
I I LIS STt':Ir 1?]1l.(" LDA·)L!
liL,7 ST~~12
1] srtj ,)'v;P. 1 (/151
ljSl'SFT.)~
1 j 5?\ 0
i lS:1 LD(.\'
1 15LI 1120
1 ]':-)5
COM
1 I 5 tS ;:-, n f) ) 12 8·
1 j 57 AZFi
1 1 h Cil , ) ~., P 1 1 66
1 1 () i A PO T
11 {)P, I..'-;K r 12
1 1 ():-1 ,-; ~V; ;.:1 12
ll()~ '-)FT
12
1 1 () ~ (I;
1 j~() Li.lf:. 13
1 ]t;7 cn,V)
] j 70 f:.i)~
1'5
1 17 1 ~:/~
I
117? . . jV,.~ 1161.
1 17 ~1 l. j) f. 1 4
li"i'1j cnl'Vl
1 1 I S An A 16
1
i·' n
• j
J Ivl P ,1 16 1
77 LDA'
]
J,
\ 't'\c. ~e~eV\.-t
J F'"t/< .
,
~=-'=-'
·
__ .- -."-
...
.... -..
- .. -.
n J
'::;TC
) ',? ,.;~?
• Lv,i';'
J :.)
1?17.1 ,1,v,P
1 ,,? 01.1
. --_....... -. _....
i?l.ll
1 /J? Lj.
3.
160111
STC 1? 3 1
1?
(~5 ADO
1 P(7'n r.OM
1 ':) (.~'7 Ar)D
1 120
1 123
1 ? 1 (1
1211
J? l?
1?11
l?ILJ
JIVlP 12~7
?IPO 'f
.IMP 1?21
)9)')
.JMP
~
-( E r
STC
12
16Ci'1LJ
1? 1 6
~) M P
] n 13
1?17
1 ?? 0
•.JMP
~J:V: P
1622
]?? 3
1P?1 COM
1??P. SiC
12
]';:13
1P~:1
.ji'liP.
1 ?? LJ
1 ?? 5
1P? 6
12?7
X.') KT 1? ~
.J:V:;:' 1 ? ?' 3
1\1 (i
i-',
.,I,v,p
1 2:-Hi',
~ROT
12:11
1 2 :1?
.. I (Vi P
12~8
,J;v)P
1231.1
1':)3')
1? ~ 6
1 ? :17
l,?LJ(tl
l?LJl
1 ':) LJP
1 ? /J~'
-
1666
(J.
.AJJ
_
1 2 ..& 2
1600
p?nr
1?i:l
J tlJ P ] 6 :3:1
J i'l: P 1 7 3,5 .
~.lY:~
N()r:;
~LT
,):Vi P,
1 7 (1(-l
.j ,'F, P
,W,}J
1? L., 6
175 1
.j,V,?
,ji,/: P
1?,Lj1
1 7 l{P
..F.t!?
12L!1
i t;l I.J L,
1?L<5
1 ? IJ f..
19,1;7
1 p ~ (/j
1 P '=) 1
1 ? 5?
~
.J iVi P
.J~t1 P
17 ? 1
]975
1276
1?77
.~FTT
I?
776LJ
CLR
,-1:V;
1 2 I.! 6
1 60(''i
\?S~ ,~Tr.
1316
1 ':) 5 LJ A j) D. :. 1 1 ~·3
1255 ~DD 1120
1256 STC ] 120
IPS? ~J;v,p 1756
1 ? 6 G1 ... E'I: P I 4 32
1261 .IMP 1751
1 ?f..'?. SFT} 12
1?fl~~ 7763
]?6Li CI.P
1 ?6S S,'-\Ci
1266 11301
1?67 J:V1P' ]1)6'6
I ? 7&1 L. Ii A 1 3 ~
1271 ~('\Kj'l
1P7? .. lY:P Ih41
1?7,~ XS~T 12
1 Q 7 L; ..J (Vi r> I? 6, L!
,
"
i ~(" (f,
l:-::n 1
:::: i·~
n
1 :V~~.
1 ~ (~ ~
! 1'?7
,F'i? 1 IS 6 6
!. n A i ~
~~ O!< r 1
1 1 (1\ 5
,j:vl P
J ~ ...~ Q
1 (J. ()
1 64 1
1 ....~1(1
1.111
X S K r I?
·\1 P j '? 7 7 .
,cOD 11?6
STC 11?(i?
j
1 .1 r;l7
.r
1 .11?
JiVl?
17 ; 2
131.3
,J~.~?
1 ::s 1 LJ
.. 1'Y)
1315
17 P.. 1
P
l1i5 SRC)r
1 ....1 1 () Vl
1 11 7 JM P 1 7 L,2
J.'3?r" ,J"'iP 1(A51
1 3? 1 ,j['.1? 16C-I(1,
1:l? ~
S Tel
.~?
1
1:3?3 J;vjp
1756
1 :-~? 4 ,J IV) P
i J? 5 ,.1 !Vi 'p
1 7 (Id})
1 1 ~~ 17.
8? 6 L; .''-: ';..:>
1? '5 1
11?7 .j:-/,P
J 1 ~ r~ J ~ ?
1(1:51'
1 16 .4
13 l 1
1 ~~ ~ (1
J
.y., P
11.1? ,J>'iP
jLo~?
13.11 J:v,.':'-
lLd.jl
334 ,J:.tJp j 633
8 1 5 f\. r) ;:J'
111 n ,j,Vj r=>. 1 6 .1 3
.1 1."3 7 ..i ,v, P 11111
j
j
i ....~ LA'?, ,F., P
13/lj
,6!)D
1432
1126
J 31.J?
C()M
.,
l1/j~
,t.fji.)
1 ]?,:3
O\v\de
.~
Jn~5
l.'71L:S STC 1 1 ~6
l1LJt; ~TG 1 j 01
11/17 t:;TC 113P..
11,:>Oi .SFT,. I?,
1 :1 '\,1 1 i:-~ 1
.
1 :1 /.1 /; ?\ i') I')
13S?
.1·'·iP
1tS?2
1 :1 5 .1 .. 1"; P 1 7 4?
1 3 5 Ij J.v. .J 1 3 6
V' ......
11'55 ,): ..
16?.?
.
rv~
p 1 f,(1.6
11StS
12-S7 ,j~', F- 1666
5
1 3 tS 1
136?
,jV;P
161.3
Lf')~
15
~f-'O T
1 2> (S~:. ~jJP 1366
i364 ,ji'~ ,~. 1 7 (?) (/,
13('S ,j:\I) P. 137. 6
1.166 LDA 12 .11
f)
7
1 17 ell
R C (l.r
I; (M~ (i).
j~71
~T.0
J87?
~L (I)
j L: (V:' r I.. hi
1 I, r.n ~\nn j 1.12
I/j!(,1 STC
li??
j LI 1! AD j)
1 13 1
1/JI,/j
I .I: I?
i I, 1 ~
1 /; 1 4
lLJi5
1 L: 1.5
1 .t~ 17
Nn p
1~?f7;
SRCi
1 L;? 1
1 L; ~::•.~
i /.23
1 L:? Ll
j .6~5
1 42 h
J LJ?,7
) L
)
J
j,',
Lj~:
S T (:l 1 3
P. PO r
J ~I, P j L;? [1
JiV1? 1633
\.J M P 1 7 ;3 5
~-:
,
,j .\1:, r..J
17
..J• "'il. f=-
1 r::'i~ 1
i 7 ~i(/)
J:v.?
j:v,~"
l',/,:J
,jV:~.
'. i:<
,-'
~2
1 L? 7
1 (/7 1
1 (,,~ 4
1 f. ) 3
~n
• jV:;';"
16
1 Ll:-~?
1 L::-~3
i .. ()('
STC
1J
1 1? 7
1 L:::1.6
l~
i
i"J .C:.
Lj
j LJ~i7
C:TC
; 1 :-~0
1 J 2 Vi
1 126
j LI LJ I'll
.i :1': ,.,)
(ij
iL!L.:l
1 LL.?
U)~
,
1 .6~tS
;'::';")j)
Ln.:~
15
j 12]
i6
11,LIL,
S'i'C
1 102
1 L:' L,:-)
:.:",:) j)
S ril
.J;"'}?
L.O.6
.~J'!'~'
C:Tr,
}
f" A~_.~~\() . .. ~
_---
.c:. 't"C
,.1 L5 i
LI
r;,
STC
1
L~?
"
LJS~
i
L.' :) ..:';
..1'/ r)
PonD
~
STA
- L::)6
S c~:
j
1 L: l)
1.,-,
i LI '57 •.<::,--0
1 IJ t) I;' .6Jn
c;:.;- .....
'-.I
1 L! ,~; 1
e
J /,
~/;>
1 /1:;
~
1 .U;LJ
•
-:.{..)
.. !
,
.J .',; r.~Jivi
,t'
1467
:432
1 i~3
j :3
J J .
1 1~2
16
i I? (il
17 (.;j(i'!
1 .!J 7 ell
0?
17 ~ i.
oj
,w! }:->
1613
J /; fJ..
•;:v', p
..IL. T
16~?'
1,67
1 1.;7 (.j
i
LIO
j
,J.\~
P ·171,?
,J,v,P 1465
1 47? ..i(/. ;:> 1 7? 1
1 LJ7 '-1 .Iiv') P 1470
1 ~7 /1 LD ..6
1 LJ7 5 1 12'1
1 .676 p.z.r-.:
1 477 p., P'CJ
J
L, -,
j
1
. \0
.
1 :) (~ LJ ,J:'oj.::' 1 5 1 7
1 ~)~~ cn;'v1
1 S(l,il P-DAr
1 5(/)7 ~L,(,l
1 S 1 (Jj S Tel S 1 ~
i S 1; p i) il 1 j 2 1
1'5]2 ~L~(
1 ~ J ~~ ..J ~/: !.) 1 p·5 1
lSj/.J'LilD. 13
1515 SCf~
1 5 j 6 ,]iV1 P
13
1 ':'15 1
lSi"? LDA 13
1 ~? C~
IS?!
IS:? ?
SCR
p.~o
13
(ii)
15~i'~
P. 1) iJ .17 2
1664
JM?' 1 r" 5 1
.J~1 p' ] 604 )
.]:-'1 P
,J,\/ P 1613
if, ??
,j IV! ,i. ] 3? 1
'
1')31
,w·p
IS?3
] 524
1 52 5
1 52 h
IS?7
~Dn
J'
] 45(1
J
1 5,l~ ,J;'''iP
1177
l~l~
.i'\':P
1 ("'~ I
1~~L.
,J:v, P 1.:.50 }
15:1S
IS3'1
I
s~n
j? ~?
,j,,~ ~
\J~
':1
~ ~~~
J~'!, ~
16 };8
1 t) LJ 1 .1~1~
~i
j 5L.? .... ) '1
' ..
1 ~ 4~3 ,J:vJ ;-J
1622
1 ':.I.,fll
1544 J~'tP'
1545 ..Ji'r.}-'
1546 ..J:'-';P,
i 5LO ,-WI?
155fi1 . .ltv; FJ,
1 S,) 1 "LV.?
1S5? ,j.v,?
J 5 5::~ .. ii'lt ;.=.
155~
l
J iV,
•..J
J5S~
.JMP
1SSf,
\..),\1 ~
1 ':JS7
,j,V;P
1 ~f){.'.
CLR
156
.) ~
j
]~;:,p
j
J
?
LilA
5 ()~: SCt-~
5,;4 A ?rJ
1 LILl 1
174:?
1 (Ii ~ 1
1 iSl
15SLJ
1 L: 41
1 (tl~ 1
14~1
17 910
1562
!
J
1
1 (1'1 S, 1
13 ,
i3 ~
i 57(/)
17?,PI
1 C-15' 1
1 57 r,
] 577
..-.--- .... ~ ....
II? 1 }
11, Li 1
1 I/i 5. j
pDf)
15-/(1,
1':,/' j
1 57?
1 c)7 3
j 57 4
j ':,7 S
_- .._-
1321
156 r, pf)fj
15(,7 ,J,''iP
J~65
'}
...........
~~
777.'5
17:',6 ')
...i ~-I; fl 1 L; Ll 1 ~
JI'vj~
,III,i.)
,):,':,:--
,liviP
\..lv, p
,..iMP'
1 (i:5.1
j 7 5,6
144~1
1 -; L;?
10~'1
.
..)1
13-20,
.
"
.'
,.....
1.
)
: :: ," :<. 1..;) ;:.
R (~(,
j
S(;:~
,-1 .\~ r~
(I,
LDA 1 ()
1 13(t1
1 124
c;-,...
1 127
" I ll'S ) rlj ;oOf) '1 123
1/) 1 j
SiC J 126
1 () 12 J l~·; tJ (I,
1('1;-~ LDA
1'::'
~TC
1 /.,(1, f.
1 -::,(~"/
/lijD
,\3
:.4 ST e l l ? ~
If) 15 /lDD 11?1
1h
16j{) STC
1124
16j7';'DD liP.0
j 6? (~ S Tel 12 3
1 tS2 l .li'~ P .V:
1 f)?2 eLF<,
J (:.? J ,cl j) (L 1 1 3 VI
1 fi? Ll Sjr, 11?2
16?!1 /If")u 11?7
STC,1121
1 62"' t-oUIJ. 1 126
1 () :-iCl, STC 1 120
]'~ ..:.J
(i)
163 j
16l? 13
....
16~J
I
tl
CL i~:
i r;
1 () Ld"
'i 72~
i 1~0
,. v 1 12 VI
_.J.
Li)A 13
S r: ~ T 1
1(4)
STC
1 (1LI?
A,")D
1f)3L;
j ()
~Df)
.15 AD:,
~r'"
3-'
1 i21
i 128
J h L~/. STC 1 l??
16;.5 J;V,? til
If>I.Jr; y/.;:. 1 6221 6/,7 ,J·v'
.. (..
j 7 v.r~(
J 6 :~ (,j ...1::' p i? (ij3
Ih:11 "W, tJ lLi~ i;
1 r:.~? ,j:vi? 12Li 1.
I
1 (-) S l
1 h~/.;
lASS
) () 5 ()
i~~r.
jf>1)3
1
.,.
\"J"'/I.j
J7 ~ 1
.i,"I/:"
1137
}
1
c.. "'e.c. \2-
~
'\
,1
I
Not
~seJ
t7. I
i ()()L; 3777
'1 hh ~ "r1"16
i66{) CLi'~
1667 L:) P1 () 7 «. L ;.'1,
1 ()"I i L.t"', A
16"/ ? L ~:Vi
1 () -;;~ S ';' c;
1r; 7 L; Lllv
..... I
1 () 7~) s-rc
1':' 7 (. l_A:\I;
1,{, 'I"i JIl,?
16
1 .u
15
~ 3
i (;~62
1 L.
1 (/162
1.'3
;'i
II
I
+'O~
~e,,"o
\Y\
' AJJ.
=-=':'C~'-:"-'-'I_I
j -f(~.,~
{:.
i ;.:. ...
'r~'~-""""-"'-'-
• } 7 (,'·2 ..J;/j rJ 0
17(/;'J seRr 1
17(1,4 LD;' 14
1 7 {I,5 R() K j" 1
17rJJ(, A7E'.
17 (I; 7 J~tj f-J (IJ
17 1 (1 X Sf( .,
] 7 1]
,ff.
7. t.
,Pit P
0 () e.s
F A~ =
0
~
(i1
171? LOA 13
] 7 1:1 pnL 1
Ben .13
1i 1
'I
17 1 S APOr
1716 XSf< f
}717 JMP OJ
]7R0
1
17? 1 CLR •
17P.2 POD 166.5
17?3 ADO: 1 1~·0
17'?.11 STC: 1120
] 725 LDt{ 14
r1726 ROLlI
1727 STA. j 4
173~ LDfJ.~ 13
] 731 f~nL'! 1
173'?. STC 1)21
173::-1 LM'w1 1-4
1 734 ,J ,'vi P C'l
1735 LDA 13
)
178f> BCOr
1737 4('100.
','
17L;0 STC~1121
I 7 II 1 .J !VI P
17L;2 LDA
j 7 /.j 3 CO;'-;
17/.;Lj STC
'(i)
13
1 1r..
'j.
Co
17Lj5 fJ.DD· 112r.
17/16
17 4 7
cn,Vj-~
~ -r c;'
.G
1 122
I
"
c.t)~~ec,f ~V\. 'O~i: •.
J
.
~~ ~~ ~:_
:;
: J?0 ' } ..
1758 S1C',l121
1 ~I 5 Ii S T C • 1 122
1 755 . Ji''i P (I,
j
1756 SI~T 12
1757 (1)
17('(J, L.OA 13
176 1 . A~O
176? JI"IPI7Lj~ .
1 7 6 3 L D A '1 5
'
17(,.4 APOf
I 7 (, 5 .,W, P 12
]7()6 COM'
]767 SiC
] 124 '
177Vl ADn
11?5
en:'!;
::
1 7 71
1772 STC' 1125
1 7 ~I 3 •..1 MP 12
.... ,."
.13
. ··t·2~2:·.
~
.
.
•...
.
.
A Prel iminary Description of an
Operating System
~6r
the LINe Computer
by
Richard K. Moore
Department of Genetics
Stanford University
./
March 1965
13-23
The
LINe
The LING is a binary, 12-bit, 2048-word, digital computer.
The core
is divided i:1to eight "quarters", each quarter being 256 words long.
Quarter 0 contains cells 0 through 377 * , quarter 1
through 777, etc.
con~ains
cells 400
LINC tapes are divided into 1000 blocks, each block
being 400 words long.
Thus, for example, the tape read instruction replaces
the contents of one quarter of LINC memory with the contents of one'tape
There are two tape units on the LINC, units 0 and 1.
block.
The LIXC·
O~erating
Systeill:
LOSS
LOS? is based on a highly
as
structure~i
quarters of memory_
t~e v~rious
in w.hich (co,ru:mnica tion among
comp~ex opera~iuns,
use of both tape units as well
LOSS .strives to
LI~C; programs is
such as compiling, to: be accomplished through the
c9~uunication
OVEI}LAY, " (2)
a,framework
very simp le, thus allowing
sucq.essiye efforts of relatively simple programs.
simple
provi~e
is based on three artifacts:
the BUFFER, and (3) TEXTS ..
This framework of
(1)
the RECURSIVE
"-
,
The RECURSIVE OVERLAY
Und~r
LOSS, tape unit 0 is reserved for a program. stack and an
/
overlay
~tack.
avai~able
stack is
,
*
Fr-pm
t~
be
The program stack consists of those programs which are
r~n
~identified
on the LINC under LOS5.
by a number.
Each of the programs in the
Program 1 occupies blocks 11 through 15,
~
n~w
on, '-'nless otherwise
no.~ed,
numbers are in the octal system •.
13-24
2. '
program 2 occupies blocks 21 through 25, etc.
following the top of the program stack.
occupies only quarters I through 5.
which are used by LINe programs.
The overlay stack begins
During execution, a program
Quarter 0 contains a number of routines
One of these routines is an overlay pro-
cedure which allows any program in the program stack to be called like a
subroutine.
As an example, suppose that there are four programs on the
proeram s tack and that program 3,' currently under execution" wishes to
call program 2.
In this case program 3 merely
pla~es
the
accumulator and transfers control to the overlay routine.
rou~ine
num~er
2 in the
The overlay
(1) saves the address frqrn which, it was called, (2) writes
1 throug.h 5
01)<
qu~rters
tape (at the top of the overlay stack, say blocks 61' through ,
65), (3);; :reads program 2 (blocks 21
executiqn at cell 400.
~hro~gh
25) into
cor~'
and (4) begins
When program'2 has completed its execution (which
may~ hav~ included overlays of otf\er 'progJ:'ams, in which case program ,2 would
ha~~ be~n
ov~rlay
written on blocks 71
th~ough
75) it merely returns control to the
routine which reads back in program 3 from the. overlay stack and
ret~rns
~
j:ontrol to the cell saved at step (1) above.
"
Thel power of this system is ~hown by the fact that in the preceding
'\
exa~le
program 2 was able to
per~orm
its function without knowing what
progjram rad called it nor in what depth of recursion the overlay process
was ,curr,encly involved.
The BUFFER
It
,
or
\{'~i~e
~rr.:lys
the elements of
mathem~tical ~F
(single cells or blocks ,of
sucl1 ope:r;ations as input-output 017 of
logical operations are variables
con~ecutive
~,nte~
cells), the elements of
"program communications are
13-25
3.
LISTS, where the elements of a LIST are either variables or alphanumeric
STRINGS.
In order to allow for the manipulation of such LISTS, LOSS includes
a general purpose nUFFER (beginning at cell 3000) together with two procedures,
PUT and GET, located in quarter O.
of storage is a RECORD.
The BUFFER is a pushdown stack whose unit
The arguments to the PUT procedure are one or more
LISTS; these LISTS are combined by PUT to form a RECORD which is placed on
top of
~he
BUFFER.
Similarly GET($$1)
causes the variables of the LIST L
to assume the values found in the top RECORD
is then erased from the
nUFFER.
1 qhrough 5, the canonical way
is. by
ot~r
m~ans
fo~
Since an
of the
BUFFER. which
OVERLAY
RECORD
affects only quarters
programs to transmit parameters to
~ach
of the BUFFER.
TEXTS
E
LOS:? rese;ves tape unit 1 for the storage of TEXTS.
A TEXT is a group
of c. .onsef,utive tape blocks preceded by a .few special code 'words and a five
character name.
100
.tape~blocks
TEXTS are
long.
groupe~
together to form BOOKS, each BOOK being
Thus BOOK 0 comprises blocks
comprises blocks 100 through 177, etc.
index
":.
wh~ch
~
a through 77, BOOK 1
lhe O'th block of each BOOK is an
contains the names of 1all the· TEXTS (in alphabetical order) in
itsi!BOOK together with their size and initial blocks.
not
mean~
This formalism is
to restrict the kind 6f.information which can be stored on tape,
but rather maKes it possible for allocation of tape storage .. space to become
autOIpatiq and somewhat "resistant to destructive over-writes.
time;,'3
b~ock
At the same
number together with .the first t\.,o letters of its name become
a concis~, as well as a securely redundan~, way to refer to TEXTS.
13-26
4.
The LINC MONITOR
The LING MONITOR is, by convention) program 1 on the program stack.
Its only capabilities are to accept instructions from the typewriter, to
perform simple operations upon the buffer, and to overlay any of the programs
on the program stack.
DISPLAY n
Its specific operations are:
n
an octal integer; the nlth
BUFFER
RECORD
of the
is displayed on the scope.
TYPE n
The nlth
EXECUTE n
The nlth program on the stack is overlayed.
" ERASE n
n
RECORD
is typed ..
RECORDS are erased from the top of the DUFFER •
.
The
LIST L
is placed on the BUFFER.
L
,
of octal integer; and alphanumeric STRINGS.
.
consists
A
STRING, in this sense, begins with the character
"
and is terminated by
%
J
Since LOSS itself includes a method for referring to and loading programs,
i.e.) the OVERLAY) the
MONITO~
does not require a "loader" or a list of
1
available "systems".
Richard ~oore
March 8, 1965
13-27
BLINK
General Description
BLI~~
is a version of Subalgol designed for use with the LINC computer.
Programs very similar to Subalgol programs are translated on the 7090 by the
BLINK compiler (which is written in Subalgol), 'into relocatable LINC code.
Reserved Word Changes with Semantics
BLINK has no "library procedures", though it retains all of Subalgol's
"intrinsic fupc{:ions".
.
The following Subalgol reserved words are without
.:
,
special meaniBg in BLIIDZ:
,',!to'
STOP, SHLT, SHRT, EXTR, STATEMENT,
~
INP"UT, OUTPUT" TRACE, DPRECISION, LIBRARY, CARDREAD, PRINTOUT,
1
: : ;
COHPLEX, RE, 1M, HRITE, READ, 5QRT, LOG,
s
ENTIRE, SINH, COSH,
-
The
o.
SEGMENT, MONITOR. STEP,
t~lILE,
READM J
\~RITEM,
F~GN,
etc.
fol~owing
I.
TA~11)
1
~XP,
SIN, COS, TAN,
Ii
ARCTAN, ROMXX,
ARC~IN.
ARCCOS, RCARD,
CHECKM, MOVEM, MOVEFILE,. ENDFILE, REWIND, UNLOAD,
reserved words are lintroduced or redefined with BLINK:
5
/1
A.
ROTL, ROTR, SCLR
B.
;
..
H•
INCR.
BTCLR, STCOM, STSET
I.
OVERLAY
c.
'LDA
J.
STRING
D.
STA
K.
LIST
E.
DO
L.
PUT, GET
F.
REPEAT
M.;
RESTART
G.
RDC, ReG, MTB, WRC,
R.
*GETCOR
'olCG, \.JRI, CHK, RDE,
0 •.: EXTERt.'lAL
•
0
13-28
2.
II.
....
A.
II, I2,
B.
M, MF, Mll
c.
Ill, 12I,
D.
POINTER
I7
....
I7I
Semantics
Co~resp?nding
I.
A.
Intrins ic func tio'n; argu11lents type integer;
ROTL(N , OPERAND)
result type integer; corresponds to ROL instruction; ai in later
~"
t~trinsic fun~tions,
9,
~ ~< ~
B.
the effect of the
,I
i-bi~":is'
obtained 'by using
value of N > 17.
BTCLR(MASK,OPE~D)
Intrinsic function; types integer;
corresponds to BeL, etc.; as with ROTL class function, a constant
firs t argument naturally reduces 'length 'of resul ting code.
( C.
LDA( $
effect to: '
DMYlc$
! ,:l
;~xcept
FOR DMY2=(1;, 1 ,DHY~ $ $
that the REPEAT loop is more
efficien~
and does not
change' the val~e' 'of loany var.iable in its indexing.
G.
RDC(i,u,QNMBR,BNMBR~
are represented by
'H.
Identical to LAP, expept that i'and u
~,O
or'
l~
INCR«expression>, <~ariable»
Identical in effect to:
~+
except that if the variable is subscripted, INCR calculates
I.
the subscript only
~nce
value of
~s
and;INCR is a function having the new
its yalue.
OVERLAY
The OVERLAY* routine is entered
",wi th the integer exp:ress ion ,: in the accumulator.
J.
STRING(~fnteger~)=«alpha
declaration is
only a single
allowed.
ident~cal
dimens~on,
The STRING
to the ARRAY declaration, except that
anq no irregular subscript ranges, are
The effect"G{c the
~STRING
declaration is different in
.
\
~
'/'f
string»
The reader should be familiar \tfi.\b LOSS at this point.
I
13-30
4.
that the zero'th position of the STRING (even though not requested
in the declaration) is reserved and filled with the size of the
STRING, this information being necessary to the PUT and GET
routines.
STRINGS may be manipulated word by word, as are ARRAYS,
through subscription.
Thus S(l) refers to the first and second
.
characters of the STRING S.
,. K.
LIST
: ' The LIST declaration is identical to the Subalgol OUTPUT
declaration except, that a STRING name "(followed by empty parenthesis)
I
is allowed 'as a LIST element, and fulfills the role served by the
alphanumeric
~9p1e'-lhat
insert~on
phr¥e in Subalgol.
{\
~IST,
however, is
more elegant than c!1 Subalgol INPUT or OUTPUT, list since
4:;LIST can be used ~r eit¥r input or output (i. e., as argume,nt
·10,
"
'jO.,;I
of either PuT'··~r -G"E';) ~··~nd ~ncludes the types of its elements,
therefore needing
n~ accomp~nying
FORMAT (which concept therefore
fails to exist ,in BIi;INK).
'L.
PUT ,GET
These
a~
simply procedures (always in core) which
can have any number pf LISTp as program reference parameters.
M.
RESTART
\~len se~ral BL~NK
the same 7090 run,
~re
non-tast programs use
compilation rather than
FINISH.
.
I:
instead of
N.
returning~contro~
~GETCOR
.'
by BLINK.
*GETCOR
i~
I.
programs are to be compiled during
n
RESTAR~
to . terminate
RESTART reloads the BLINK compiler
to the monitor.
This is 'a control ;card recognized
similqr to RESTART, except that after
•
completing the compiter output, the indicated disc file (rather
than the compiler)
i~ loade~.
5.
O.
EXTE&~AL
PROCEDURE, EXTERNAL SUBROUTINE
These declarations,
identical to those in Subalgol, allow linkages to be created on
the LINC between BLINK subprograms and subprograms created by
means other than the BLINK compiler.'
II.
A.
Unlike Subalgol,
BL~NK
has
~eserved
variables.
II through 17
(index registers), are simple variables of type integer with
absolute address 1 through 7, respectively.
These variables
are GLOBAL and their values are not restored after an overlay.
B.
M",MF, and MH are GLOBAL arttays with absolul:e base address of
",{ ,
I:"
.R: . Their
types are :integeq, floating, and
q~lf-wOrd,
,
respectively •.
...
;-.'
These are used .to__ g~ea~ advilntage together with II through I7:
= M(I~)$
M(Il)
results in the elegant
~ode.
LDA 2, STA 1
C.
Ill,
...
, 171
are psed in conjunction' with the M() and MH() arrays
in order to referenc~ consecutive words, or half-wQrds, of' core.
As an example, the f91lowing statements replace the contents of
quarter 5 by
II
CI
of quarter 4:
thecon~ents
"3777" .,· 12
REPEAT
CI
"2377" ;
"
"400" ; M(I2I)
COMMENT
= M(IlI)
LDAil
, STAi2
;
"
'I
Thus the value of the«: indicated
index register is incremented
•
before ,it is used as a subscript.
completed, II ,.,ill
(It is a quirk of
c~ntain
l:h~
LINC
When the above program is
"2377" and I2 will contain "2777" •
~at
the core is logically divided
into halves;' thus ~77.7 is the:" predecessor of 2000 and 1777 is
the predecessor of:O.)
13-32
6.
In the case
half-words, index registers are
o~
4000 rather than by 1.
increment~d
by
Thus if an index register were stepping
thr,ough. the charac ters of quarter 4, it would assume successively'
the values
2000, 6000, 2001, 6001, 2002, etc.
indicates the
right~half o~
The 4000-bit
the word;
An index register can be made to point at a variable by a statement
of the form
InI=.
The code generated is:
SET i n
The following program places the characters of the· STRING S() into
quarter 7, putting one
charfoct~r
(right justified)
in~o
each word.
STRING 5(20),=( alp¥ string);
17I~"3377";·(
CONMENT:
............ )
There is~anonical correspondnece between
reg is ters and quar~ers •.
;,
l
121=5(0); 12=12+"4000";
Ii
REPEAT 40;
~
M(I7I)=~I(I21);
COMMENT:LDHi2 , STAi7 $
"
D. P.OINTERis that cell ("155") in QUARTER 0 which points to the top
of the BUFFER.
The statement ' P01NTER=M(POINTER)
one record from the BUFFER.
would erase
If we assume. that the top record of
the BUFFER begins wi'th/an alphabetic item, then the statement:
/'
would
12=M(POINTER)+"4001 a;
a1lo\~
""I'
MN(I2I) ·to reference successive characters of that
first item.
13-33
-=::=..-----
~
-
~---~
-~.
-- -- --
~
.,.-- _. - ------- .--
indicates data 1OCo t ion
0P. 010 STC 210
STC 16
02(111 ADD 17
0101 LDAt16
(12f'l2 STA 1 1
, 0] C":? set-< 6
LIJAt13
02r~3
0.1(1)3 STC 15
(1)204 STAt 11
0. 1(i'I/l LDA 1 6
0205 x SK t 1 7
01015 Belt
02(1)6 JMP 203
0106 7'7 (It (IJ
0.2 CIl7 L O{,H
01017 AOt) 1 17
02
1V, (~
(i) 110 ,tlOU 17
JM}J
'""2 1 1 STAt 11
1 13
01 1 1
0212 JMP 160
o 1 12 :L 0 AT 1 7
0213 JMP 231
. 0113 STAt 16
0214 LOA
GET
01 14 XSK t 1 5 .
0215 0
0?115'JIVJ}J 1 12
'02
16 s~rc 176
LOAt
01 16
0217 JMP 2 (i)
01 17 6 (A0 1
/0220 STC 301
(i) 12 (il
ADD 16
(i)2~ 1 JMP
154
0121 STC 122
~,30(/\0
j'
0222 AZEt
'012? JMP (iJr(
PI(i'12~ ~OA'
EMPTY· C1~~3 HL T'
0I1?3 LOA
fi10!~1
1776
REPEAT
(
0224 STC 301
(il01P.L! POD 0
vn22! 01
0~25 JMP. 175
01125 STC 21 0
f'ilfil~5 ST AtIS '
LST. OP "OI?~ 6 STC 243 .
(71'126 JMP 2 PI
(iI(/\?6 STC' 16
(71 (il? 7 LD,tlt16
0227 ADD 111
?/ (11127 SAM t 1 5
0230 STC 271
fi'I(i'l30 ROLf l'
/ Cil 130 HLT 1'"
0231 JMP. 20
~(i'!31 ,SCR
/0131 STC 136
1
/0232 ROL 3
013~ POD
0.03P. STC 114
/0233 STC 276
(/1(1133 LDA 14
(111 :13 JMP
(A2 3 LJ J\v,p 271
/01134 J,v,p 1902
0!113L! ST,tl t 15 \
7
(N~35 ADD (~
GET;
CL
HLT
/0135
(.'(.'35 L7.~t
0P.36
STC 243 .RETURN'
/0i]36 HlT
(/\036 JMP '?.7
c
i
(1)137 LOAtl~
0237 JIV1P 15.4
OHi137 LDP
J
0.OILJ(7I ~ 1
024Cil STC 155
01140 A7..E
0)1LJ1
Pt-'O o ~ LJ 1 . ST P tIS'
LST. CL 01? LJ 1 JMP L!6
.
..
0!(i)1.J?, I~f) A
0142 JI"I~ 1 5W··' ,. "-'.. . (1)242 CLR
~
01243 JMP (i) OVERLAY'
01LJ3 CCHJ)
001L!3 1 5 r
(?lO!LlL! -STC ?~ 1
01LJ4 STC 1 ~ STRING (~2 il4 ADD (i) :
())2LJ5 JMP 14
(I)]
45 JMP'135
(iHi'LJ 5 ~J~P ~ 1 16'
T /0246 JMP 1,601
LDA
1
01146
X
SK
t
10
STEP
RESTORE~/Jf)
(i1(il/.17
/01247 HLT
01
0147 JMP 135
J
EXIT
01 (il 501 STC t(,73
(l,150l LOA
,. 0250 LDAtf7
/
"
'j
01251 STC 13
(il!i151 SET 1 5
0151 136
(/1(1,52 2 1
(11252 ADD 247
0152 STC 243
/
0253 STC 27 1
(il053 LDA ~ 15
01153 JMP 24:1
(i:12~4 JMP 46
'/
0I015LJ ,tlf' F' TJ
0l15L! SET t 1 1~
(11255 LOA 13
PO I NT ER Ci' 1 5 5 314(1 i'
. EMPTY ~(,)5S HL T'
0256 Co IV)
(A (.I':J b STP
0156 LOA' 1 1;:
0257 SRO
01157 JMP (1
0057 2 1
(i1~6(i'1
(')26(11 276
fi1160. LOA
STC 1 5
-,
(})161 1 1
0261 ADD 264
0!il61 tDAt15
0262
JIYlP 27 iI '
01162
STC
15:5
14
006? ~TC
,.
, LST. ~L 02-2,;L AOAt
006:1 I~ D,tl t. 14
0163 JMP {1
PUT
.'
01 ] h LI L.DA
0264 7776
RO!f.;LI fH)L t 1
:'
(i)265 STC 13
0165 0
(:
til f/lA 5 ~CR ~1
~0()6
(i)J66 ~TC 17-'6
.0266 ADD 0
'
STC ) 6
(i1liln7 LD,tlr!'15.
fI1167 J'VJ~ 1 ~4
"1267 JMP 74
)
f"JJ 07 (~ ~T,tl ,1 6
- /0270 JMP 16{1Jl
Cill70 LOA
.J
(1,(':7 1 ~7.~t
, /(1)27 1 HL T
017 1 1 1
/ClJ272 Ji"lP L!6
(i) 172 ~ T P t 1 11
0~72 ..IMP ;-63
01'73 .. ',YIP 16;0
0273 LOAt 17
0107:1 ,-JMP (7,
PARAM r~ r,-, 7 /.. ADO 2.3
fil174 COM
0274 STC '17,
(M;'75 c;TC ~1 7
01175 .ciTe ?7~6
0275 SROt
0,
01]76 ,:MP Vi I
0276 :~
("til76 PDD
APPEND I X 1:
O_UARTER 0 .
"/"
~1001
16
C? 0. r" 1 CLR
0?l!ilP. c)T,tl
0~~~
31L!0.
('HiH/14 ST,tl
(,.,005 30fi1o\
~(M~6
RCG
(JtC"fJl7 4(~ 1 1
(71010 .J~P 40(/)
(/\(?\ 11
HLT
OIR]? HLT
. 0013 HLT
(iHi) 14 HLT
(i1(/115 HLT
~0116 HLT
filOl]7 HL T
SAVE (.i!(.lI~R 8fT t 15
010091
I
(
;1~'
"
I
'~'
,
i,.,~,
.
/'
,"
.~
(/13",0; S ~':T
(/)3 (II
f
11
1 1/1
03 (it? lOC\f 11
tA ~J, /,1
(1)3
~~
H~I'~ f
v. LJ 6 "11" (tj
("31"~
CUr-1
11
(;, 1
"'3v 16
t-\t)J)
[1j3 (Ill
~TC
v.)31
LUt-\'l'11
(11
031 1
0312
0313
031L!
(1)315
("316
0317
(1)32 (I)
(;l321
0322
(.1323
(li324
Vi :'12 :>
~3
S" P T 13-
XSK f 17
JMP 310
JMP 2 ~~ 1
HLT
HLT
liLT
HLT
HLT
HLT
HLT
HLT
HLT
(1,3 ?6
HLT
(t1~3?7
HLT
3 t~ HLT
HLT
('13]P. HLT
(1)333 rlLT
(1133/'1 HLT
Vl335 HLT
0336 J~I}P /./6
033-' STC 341
0340 RCG
("341
(iI~~ /,2
JMt-',tt.l
(il~~ L!3 AZEf
(1344 JMP 336
0345 PPO
03LJ6 Ji"rP 3630}3 LJ7 STC 17
(1~~
0.331
'"
03~fi)
POD 0
03:>1 STC ~3 .62
0352 JMt-' 20
"'353 SCI~ 2
0354 STC 356
0355 'J) CG t
(ij356 4 ] VII
(~~~57 AOD
130
(tj;3 6 ~ ADD 3:>6
0:36 1 STC 356
c,,362 LOA
0363 17
"'364 ADAt
(~365
777CIJ
0366 APO t
-;t
(,1367 eLk
.~
(1137(11 PDAt
ClJ371 7
I
~372 ROL 3.
t<\
0373 AOAt
t-f
(137 L! 4'""0'1
(/)375 STC 377(i)376 I~CG
LOSS Character Codes.
APPEN DI X II:
CHARACTER
~
CHARACTER
00
01
02
03
04
05
06
07
10
11
12
13
14
15
A
B
C
D
E
rr
if"
$
to-1
.&
~
(
)
r~"
.
*
+
) t
X
.
J
"'0
.!.
.5
6
7
8
9
<
=
>
::1
@
~
.!
!
4
:':
; 17
20
i". 21
'. ,~ 22
23
24 .
25
26
27
30
31,
32
33
34
35
36
37
40
41
42
43
44
.45
46
47
.50
.51
52
53
54
.55
.56
.57
p
60
61
Q
R
62
·6'··
s·
3.
T
64
U
6.5.
66
V
w
67
.70 .
X
y.
71·
Z
72
(carr. ret.)
73
(end of text) '. 74
1q
1
2
3
\
F
G
H
I
J
K.
L
M.
N
0
..
QQP!
.'
code derivations
;.
LDA'
(teletype code)
130M
ADA 1
0277
. SeR 1 .
"-
13-35
Anpendix III:
A Detailed Description of LOSS, Especially QUARTER O.
SAVE AND RESTORE:
These routines govern a push-down stack whose presence allows the other
routines of QUARTER 0 to be recursive.
JMP SAVE
LOCATION
LOCATION~
LOCATION +4000
n
ca~ses ~he
n
~ocations
together
of the push-4own stack.
of,
loca~ions
-
JMP RESTORE
causes 'the, topmost list
on the stack to be restored to their former contents." This
,
.....
'pppo~ed
-
The call
push-doVn stack occupies cells"
as
thqir contents to be, saved' on the top
~ith
-.
3~00
to 3),40 and is called the SAVE-BUFFER,
..-....
to the PUT'-BUFFER whiqh begins at 3140.
PARAHS:
Consider the BALGOL statement:
P(3,Y$Z$Ll,L2)
wh:\::ch
w~uld
be equivalent to
.' . .
(a procedure call)
thehfollowi~g
LINe code:
LDA
Y
STC*+3
JNP P
0003
0000
Z
'
JMP Ll
JHP L2
Val~e
nama
parameters are thus represented by their values in the 'calling sequence,
pa~meters
by their
address~p,
and program reference parameters are
preceded by the JMP prefix.
13-36
2.
The heading of procedure
might appear as follows:
P()
LDA
0000
JHP PARAMS
7405
0000
0000
0000
LDAiI7
P:
R:
LDAiI7
etc.
}
STA list for Z
}
STA list for L1
Hhere 7405 derives from the formula 100(76 - no. of. value params) + (rio.
of params), R will be assigned the return address for each call of P() ,
1
r;
1
and R+l and R+2 will ,be assigned the values of the two
.:
Index
r~giste;.
,
.~
parameters.•
.
,
17 is left by the PARAMS routine so that tlie non-valu.e para-
:~
,'i
v~tue
,
•
meters can be:c··convenientIY .. :q~~.~~ned and stored where required in the bo~y of.
PO.
REPEAT:
The program
.1
"
JMP PRPEAT
JMP PAST
·0005
l
} . code
,(
JMP STEP
(
t
~AST:
causes "code"
etc.
to
be executed 5 times.
REPEAT is completely recursive (i.e.,
many REPEAT loops may be nested), and is the sale user of index register 10.
A R~PEAT loop' can be terminated ohly by ~om?leting the full number of iterations,
or
alte~natively,
.
by executing the instruction JMP EX!T within the loop.
the:: coun,t parameter (5 in the
abo~e
example) is zero or negative, the loop
If
3.
is not .cxecuted at all.
PUT and GET:
These can best be explained' through an example.
The following program
will replace each item in the LIST L2 by the corresponding item in
A,D,X,Y, are variables and Sl() and S2() are strings.
th~
LIST Ll.
First, the Balgol
statements:
STRING Sl(5)=('ALPHAS'),S2(5);
LIST L1(A,B,Sl(»,L2(X,Y,S2(»;
PUT(;;Ll);GET(;;L2);
etc.
(
Next, the
~
co~~esponding
LINC
cod~:
PUT
JNP L1
JMP GET
JMP L2
JMP GET.eL
etc.
Ll :ADD a
JHP LST.OP
LDAi
JMP
A
JMP LST.EL
3776 • • •
LDAi
(controi1 word; lthe first digit is type: 3 fs
integer", 7 a1pqabetic, and 1 floating. The
B
next 3 ,digits contain the complement of the
JNP LST.EL. size of the' item.) .'
3776
JMP STRING
Sl
JHP LST.CL
L2:ADD a
etc ••••
JMP LST.CL
Sl: 0005
(string· size)
4154
6050
4163
i
7474
(end code)
0000
7474
('extra margin' end code)
S2:0005
0000
0000
0000
0000
0000
7474
13-38
..
4.
The BUFFER:
The BUFFER has a linked-list structure, the top of which is POINTER'
(cell155).
If in the previous :example we assume that A and n have the
values 7 and 24, respectively, then after the statement
BUFFER
PUTe;;Ll), the
\vould have the following appearance:
LOCATION[CONTENTS
0155
3140
3141
3142
3143
.3144
3145
3146
3147
3150
3151
3152
3153
3154
3154
0000 • • • null contents denote nUFFERbottom
3776
0007
3776
0024
7771
4154
6050
4163
7474'·0000
7474
3140
If GET is ever entered when the BUFFER is empty, i.e., when. location 155
contains 3140, then a halt occurs at location 223; if RESTORE is called when
the SAVE-DUFFER is empty, a halt ,occurs
jit
location 55.
OVERLAY;
'fu~n
4
it is desired to
OVER~Y
•
a program from the stack, the program
I
number
is
loaded into the accumulator, and JMP OVERLAY is executed.
,
I.
a
~rogram
has c9mpleted its
JHP. RETURN is executed.
func~~on
OVERLAY
P is
When
and.:wishes to return to its caller,.
eq}fivalent to RETURJ.'l.
The
seq~ence
. (case 1) LDAi
0005
JMP OVERLA1
JMP RETURN
is much more efficiently accomplished by:
13--39
5/
(case II)
LDAi
7772
JMP
(i.e., -5)
OVERLAY
for in the second case, the present contents of core are neither written on
tape nor read back in when program 5 is
~finished;
rather program 5's RETURN
iS,placed on the same level with'the RETURN appearing in case 1.
If an argument of OVERLAY is greater than the size of the program stack.
then the last program on the stack is loaded; thus a copy of the MONITOR
is 'usuaily in both the first and last positions. If RETURN'is called ,.,hen the
OVERLAY stack is empty, a halt occurs at location 55 (since RESTORE will be
spuriously called).
TEXTS:
t
~'th
Ther first two words of the
I.
block no.
4253
;;
bloC7k (the index) of each B,OOK are
.. --"."
(0,100,- 'or 2QO, et~.)
(' BK')
,
1
Each succeeding group of four words are TEXT entries
1
charI
/
"char 2,
char
3
/
char
charS
/
size
~
...
4
,
(i.e. , length in blocks)
initial block
~nd ~f
The
f:
i,
the index is denoted
b~
"
a zerQ, where the first two characters
of the
next
name would be •
•'
r
Each~ text is headed by the fQllowin81 four word code:
/
char
3
/
char 4
charS
/
max. size
charI
.1
2i
,j
.char
current size/ type
J
;.
~.:
'1.
-
..
The purpose of having types is for file protectionj when some program is
"
written which will· create a special kind of TEXTS • it can.
p
~
y
"
'\.
6.
usa any of the 100 available types.
Thus far type a denotes a standard
alphabetic TEXT (using the half word codes in Appendix II), and type 41 ('A')
,
I
denotes an I'absolute I TEXT. i. e •• a TEXT whose first block consist of a
header alone and whose next 5 blocks are an absolute program ready to be
placed on the program stack.
13-41
Appendix IV:
HOt." to Run a BLINK Program.
The ilLINK3 compiler is stored in the disc files of Stanford's 7090
computer.
In order to use this compiler, it is ·necessary to prepare
&1
card deck as follows:
No. 1 Card:
SYSTEM
F-INFO
Mount on A3, at 1m." density,
a tape which can be removed
from the Computation Center.
.. TAPES
No. 2 Card:
SYSTEM
Control Card:
(Cols'.
F-INFO
BLIM<3 file number (changes
periodically), right justified.•
~-6
Cols. 7-11:
'}
:
1
There should follow a BLINK source deck.
1
"
This deck should be terminated by
)
~
a RESTART, FINISH, or GETCOR card.
,.,'"
BLINK>
-
i
If 'a RESTART terminator is used, it should
~
be followed by another BLINK program.
The output produced by
unit A3.
BLIW~3
will be on the tape which was mounted on
This output is quite similar to that produced by the SUBALG~L
compiler, i.e"
listings' of the source decks, diagnostic messages, symbol
tables of the compiled programs (these being especially useful for console
debuggi~g).
In addition,
howeve~,
the Jape will contain the actual LINC code
produced by the compiler.
. 7
~ If ~no compiler el;'ror messag~s are p,r-oduced, the tape is brought . over to·
-
....
..
:'I
the~' LINg,
program~
1
moun5ed on the LINC' s 1;.ape uni;,t, and read by -an appropriate LINC
One such program merely searches the tape, ignoring all that it
~
sees, uqtil it comes to
th~
n
compiled~code.
LING tape, unit,l, in the
On~'
of the principle
in~orma~ion
fo~
drawba~s
is transfered from
t~
of a
That code is then transferred to
~EXT.
of tpe BLINK3 system is the ,means by which
7090
~othe
LINC.
Not only is it
.
~ncon-
,
13-42
2.
venient to have to physically travel to the Computation Center, but jobs
requiring special tape handling are not, given top scheduling priority on
the 7090.
The BLINK4 system will employ an electrical connection between the
LINC and the Computation Center's PDP.
pr~paring
LINC.
The BLINK programmer, instead of
a,card deck at tpe Computation Center, will prepare a TEXT on the
tvhen completed, this TEXT will be sent to 'the 7090 via the PDP.
first line of the TEXT
'monito~.
In~~ead
~ill
The
actually be the No. 2 card expected by the 7090
of using tape A3, the BLINK4 compiler
~~ll
send its output
r,t
di~ctl~
wr~te t~e
to the PDP, which will
output on IBM tape.
~elay
it to the LINC.
The LINC, in, turn· will
The tape can be examined on the LINC's, scope
anq nev,er need be 'lis ted ...... If - the ,tape contains error messages J then it is
only
ne~essary
th~re a~e
BL~NK3.,~
to alter the original
no error messages,
the~
TE~T
and re-send it to the,' 7090.
If
the procedure ,becomes the same as under
THE LING TELETYPE HONITOR SYSTEM
r
LINGT
Tne System consists of a monitor which accepts Macro-instructions
and associated octal parameters from the teletype and separately coded
)
programs 't'Thich are executed to achei ve the desired result and return to
the monitor.
The requirements to use LINeT are a standard LING and
a Teletype Corporation series )) teletype attached.to relay #0 and
External Line
tr~.
Tape requirements are Blocks 200 and forward on
unit 0, as the system i:? follolied by an indefinite scratch area a
.
---!
practical upper limit of 277 is satisfactory. ___
~~
.
.
,r:::::-'
The monitor is started by an 0700 O?OO in the switches and a
START 20.
A return, line feed.is the signal that the monitor is
ready to accept input.
The operator then types a 2.letter Macro code
followed by the appropriat~ octal parameters and
unit numbers (binary only).
.
,
Commas separate fields and blanks are ignored.
All parameters need .
not be explicitly specified as they are initially defined as zero;
hm'lever, as unit 1 is desireable as a library tape, unit numbers are
assumed as 1 unless a 0 is typed in the field (in some cases a , is
"
necessary to "open" the ~ie1d, but once a field is opened 1 is assumedJ
unless zerois typed,
, " means 1).
(e.g. illegal macro code,
An error in a calling sequence
something besides 0 or 1 in a unit field,
a non-octal digi-t in a octal field, or too many parameters in some cases) .,',
't'1ill result in a NO being typed back follo'toJ'ed by a carriage return, line feed
signifying ready status for a new line.
The RUB OUT key is interpreted.
as an illegal charachter resulting in the NO and may be used to delete the' ,
line.
1"Ile RETURi"-l key effects execution of the
~Iacro.
The follOlnng pages contain write ups of the Macros with description's
and calling sequenoes.
Also a page of
ac~ull tel~type~')oper~:t1on.
·1"-44
1.' iliput
Type:
IN n,u,x
This program receives alphanumeric text froin the teletype and
record it on tape in succesive blocks beginning at Block n,
Unit u
(initially asstU..'led 1).
L~put
Description: All alphabetio, numorio, and speoial oharachters are
valid except ? l'mich is ignored.
The RUB OUT key will delete only the'.
line currently being typed (multiple depressions have no effect on the
teAt).
7
XIS
Upon depressing it the program will do a car±igge return,· type
over the juru< that is deleted and proceed to a new line.
To end a line, press RETURN.
The program gives the line feed'when
ready -to accept a ne't-T line (usually immediate, but delayed when writing'
tape.)
To terminate input, press EaT (OTRL & D keys) immediately
follm'Jing a RET0fu\J t at any other place a NO is typed back and the EOT is .
ignored. The program after an EOT will "'¢.te out the remaining text
and type the message:
Control is
~'1en
Line NUl~bering:
tor tape logging purposes.
LAST BLOCK USEDIS 111
returned to the monitor.
If xjo numbering is suppressed.
If' x=O (normal) an
octal line n'lL"Ilber for the preceeding line vTill be typed every eighth line
begm"1ing after l:i.ne zero.
The number is
pr~~eededdby
~
to avoid confusion
ufth the numerous 4digit numbers that appear.
Output Format: T'.ae first tvlO words of every block are
75758 , the third
liard is negative if the block is last in the text (never looked at in, the
system but lI'.ight/ be of value).
The text begins in the left haIf of the
fourth "lord and continues by UNO half' vIOrd indexing through the entire block.
The end code is signaled by a ~38 i'ollo.,1iIigba 128 ·(EOL Code)..
places ~~e restriction that the charachter \
This
cannot be first in the line.
13.-45
2. Type (list)
Type l'Till list the text beginning at Block n , Unit u (initally
assumed 1) under control of the remaining parameters.
lJorL"1al Format:
If Ll=L2=Nc=X=O
the printed outpu'c of the entire.
text~
has the same format as the input listing .except for deleted lines.
Unusual Fonnats, Control Parameters: .
~
Begins printing at line number
L2
Stops
~qual
to L1
printing at line number equal to L2 ;however it Lz=O
the entire text.' aft~r:~~~ is printed.
Either L may be grea.ter than
the last line number meaning equality to it.
Nc
Prints the first 1'1 charachters per line.
listings to
./: ., xlo
Notes:
get1in~
Useful for quick and
di~y
numbers after alterations •
suppre~ses line numbering.
If a ,b~ock read does not have the
immediate return to the monitor is made.
7575 text code a NO is
~yped
and,
If ~he line is longer than 6~O
charachters, the program ...1ill start a neir line an the teletype and
90ntinue printing the same line of text.
13-46
3.
Group
.Type:
GR Nl,Ul,N2,U2,N,3, U,3, ••••••• Nn, Un
Group 1vill 'group the n te:...'"ts at Block Ni , Unit Ui into one text
and store' the result at the System scratch area (aroUnd 240 depending
on the edition being used) on unit O. T.ne main purpose of this program,'
is to prepare lmlltiple texts for assembly.
called for.
Group~~g
From 0 to
1748 texts may
~e,
in equivalent to catenation, i.e. the first of
,
,
text I £01101'1s the last line of text 3:-1 and the last line of text I
'is toilot·red by the ',first line of teXt I+l.
Text 1 ,follows nothing and
tc)..-t, n is follow'ed by the text. end code.
Operating Notes:
A rIG is typed if any block' read lacks the text code.
At:lthe end of the grouping the message :n m BLOCKS GROUPED AT ???
is Hritten before return to the monitor • .flu is the number of blocks
'1,rritten at the beginning of ,the scratch area which is given in the 111.
Caution: Only a one block buffer is used so nothing' may be inserted in '~ront of
the scratch area text
but may be appended.
4. Copy
T'l1e program 'Vrill copy
m blocks from block nl t unit u l to
If ltt>l, successive blocks are copied.
ini'on~~tion
The
may be of anytyp6 and is not limited to texts. ,
Caution: A three block bufter is used so it moving more than three blocks
for:vard on the same unit care must be taken to avoid clobbering blooks
1-Thich have yet to be read.
The range of m is 0 to 1000.
/
13~47
5. llter
TiJPe: .AL n,u,x
This program will perforill a group of
inse~ions
and deletions
to the text at block n, unit, u., It makes use of the scratch area
a..'1d programs Input and Copy.
is in order.
A brief description of its operation
First the macro is j;yped, then the monitor instruots
Input to place the alteration text at the beginning of the soratch area.
me Alteration Te:h.-t is then t,yped in '(Format 'described, below.) • ended with,
a'll
EOT (no last block, message is typed).
~lter
Input then enter Alter.
reads the Alteration t~~ and the' text to be altered (n,u).
It 1'Trites the altered text in the scratoh .area but ilmnediately
, f.Ol101ving the Alteration Text (Note: The altered text is never at
the beginning of the scratch area).
:C9~1ditionally
It then
,
types a message and'
,
i:'enters Copy .to return the altered text to bloclt n.u.
In any case the monitor is re-entered.
,Alteration Text: JUteration instruction lines and liries to be inserted
in -ehe te)..-t make up the Alteration Text.
.
The alteration instruotions
'
refer to line numbers in the text to be altered and referenoes must be "
in sequence from line 0
is;
1m, nreturn
n~
fo~vard.
The format of alteration instructions'
blanks are permitted on the line.
rne program will remove lines
fl~m
the beginning of line m to the
beginning of line n and will insert any lines of'text that follow it
until another alteration instruction is encountered. (or an end of text):
Note: A slash cannot'be the first charachter in the Alteration text
/'
unless the line is an alteration instruction (no r:striction on ~riginal text~.
Tl1e message
ILLEGAL PROCEDURE' is
not contain the 7575 code,
t~ed
if: a block 1.s read which does
an alteration instruction of an illegal format'. :
or an alteration instruction of legal format but where m:>n., 'm<{ the previ~~s .
instructions n), or m>(last line number o£ the original text).
' .
As the process is a merge, line numbering 01 theOld·t~ is preserved
"
..:
\.
13
_"4 8 ".
throughout the one pass, after completion however, the line numbering
is dependent on the al terations made and a partial print may be used'
to determine line numbering in places of interest.
If x=O the text will be returned in place of the original· text 11'
the length
oIj~the
altered text is less than or equal to the original,
. ·0.
othert'lise, it -vlill be, left in the scratch area and the message,
n BLOCKS RIDIAIN AT
x
will be typed, where n is the number of blocks and '
x is' the location of the first block.' If the altered text is returned'
the message
n BLOCKS
RETUP~ED
,
,,
'
will appear. The purpose of this
criterion 'is to prevent clobbering a block of text immediately following "
the original text. 'If x=l, the text is unconditionally
retu~ed.
arid,
if x=2 the text is not returned.
}n
example is in order:
}..L3.59
/1,1
means alter block 3.56, unit 1, x=0
says II remove nothing and in,sett the .following lines
?-n
~ro.nt
of line 1"
F..LPHA
BETA
,Gl!l,Il·IA
these three lines are inserted
says "remove lines 5,6,7, and insert" but there is nothing to insert.
/12,'1:.3 says "remove lines 12 and insert before 13 11
this line is substituted for 12
/17,100 presul;d,ng a text of say 62 lines, this will remove ihine 17 through
ErSILO£l the end of the text and append whatever follows it to an end of text
Z15'rA
(a..l1othar alteration in'struction is illegal as ITl must be lhass than
or equal to 62 and greater than or equal to 100)
ETA
THEl'A
(eot)
lrnereupon the alteration is performed the text is fOWld to be smaller
and blocks are returned, the message is then typed.
/5,10
1 BLOCKS RETURNED.
A note on Grouping: Grouping is an easier vlay to insert
......
'-:.~
~ormation
before
...... ~~
an eJdsting text or appending to it.
the beginning
~f
~s
a scratch block it is
an altered text, is never, left at
necessa~
to group it to place
it at the beginning (one block of text may be grouped in tront of it
s.?i'ely) • '
13~49
6.
Type: AS
Assemble.
LINCT has been ·tied into the LAP Convert Metacommand, which
works quite satisfactorily, through a program which transforms
LINCT text at the beginning of the system scratch area into LAP
text and places the result in blocks 336 forward (LAP input area)
and enters the LAP converter which assembles to blocks 330 333
(270 to 't!77 for 2K version). The rules of line structure given in the
LAP I II Manual must be adhered to, obviously a new special charachter
set is necessary as LAP uses a somewhat unusual set. ·The changed
charachters were chosen for typing convenience and resemblance was
a secondary situation. The following is the list of changes
LAP
LINeT
(semi colon)
p
* (asterisk)
u
(exclamation mark)
/ (slash)
Origin
$ (dollar sign)
Tag
(colon.)
OPERATIONAL SAMPLE FROM TELETYPE
IN5 23,l
$220
LOA;
1777
ROL 3
:5H JMP 78
\-IRC; 10 (note: unit l; if ·from cards)
2/240
ROC ;J
\ 4A
4A=230
JMP *-5
LAST BLOCK USED IS 523
GR5 2 3
AS
/
13-50
)
7.
~JPe:
Execute
XE n,n,n, •• oo,n
This is not a program but a means of loading and executing a program.
The monitor vnll place the first parameter in location 1375 and
successive locations thereafter, l-lhen RErURN is pressed the monitor
ju.']lps to 1375 and the octal cOl1nnands typed in v1ill be Executed.
The. reason 1375 Has chosea was that one may do an RDO and a
~l'1P
and
it lfill leave parameters in 1400 fort-1ard or i f a program is to be read
into quarter 2, three
16 1 s (NOP) may be given and instructions .are
"', I ~ ~ Co
.
quarter 3.
t\
in
Overlaying is a hit and miss proposition ·as o/\tape check
on the first attempt to read
l~
cause error.
·.13~51·
A Floating Point Subroutine Package for the LINC
Jeremy Pool
This package was written for programs compiled by "Blink", an IBM 7090
Ba·1 go] comp i I er for the Lj nc, wr j tten by Richard Moore; however, it is
completelycompatib]e with any machine language program.
The arithmetic routines - add,mu]tip]y, divide - and the float subroutine
are, with minor alterations, those written by J.C. Dill, W. M. Stauffer, and
R.
w.
Stacy of the University of North Carolina.
In this package the format for floating point numbers is a one word exponent
followed by a one word mantissa.
Both words are signed, one1s complement.
numbers (standard form for the Linc).
and a zero mantissa.
Zero is designated by a zero exponent
Floating point numbers must be in standard form, so that
the mantissa has an absolute value between
01~ ~~0 0~~ ~0~
The decimal point is understood to be between bits II and
In addressing it is always the first word, the
e~ponent,
and
I~
~]1
III III 111.
of the mantissa.
which is
specifi~d.
The call ing sequence is as fo] lows:
JMP
400
A]
01
A2
02
An
On
Next instruction
A] is the address of the first operand.
are possible:
Al
Al
> 0
=0
AI < 0
Three possible formats for this address
AI = absolute address of operard
The operand is the floating accumulator
AI = indirect address of -~erand
For indirect addressing, the address is not
be set to].
Thus with AI
operand, not location
37]4.
= 4~63J
compj~rnented;
location .6)
c~nta;ns
only the 11 bit must
the address of the
'.
13~52
2
01
is the desired operation. Two forms are possible:
Execute the specified subroutine, and then continue to
01 < 0
execute the next specified subroutine.
01
>
0
Execute the specified subroutine, which is the las~ in
the series of subroutines, and return and execute the next
instruction in location p + 1.
Here again, when 01 < 0, this is specified by setting to one the 11 bit, not
by complementing the entire number. Thus 4~02 means add and continue to execute floating point instructions while ~002 means add and return from the
floating point package.
Some of the rout i nes are Iii nteger ll subrout i nes and assume one of the numbers
involved to be an integer. In this case the actual address of the integer is
specified by the operand, directly or indirectly.
The subroutine codes, their. mnemonics, and their explanations are as follows:
(op = operand; FAC = floating accumulator; ac::& Linc's accumulator)
1•
2.
3.
4.
5.
6.
7.
CLA Cl ear and add
ADD Add
COM Comp 1ement
MUL Multiply
DFA Divide (a)
OAF Divide (b)
lAD Integer Add
c(op)----> c(FAC)
c(op)··+ c(FAC) - > c(FAC)
complement of c(op) ~> c(FAC)
c(op) x c(FAC) ----> c(FAC)
c(FAC) / c(op) ---->c(FAC)
c(op) / C(FAC~-> c(FAC)
; c(op)
c(op) + c(FAC ----> c(FAC)
::&
r
In the previous subroutine and in some of the following, the operand is
assumed to be an integer
10.
11 •
12.
13.
14.
15.
16.
17.
20.
(I).
IML
DFI
OIF
FIX
Integer multiply
c.(op) x c(FAC) ----> c(FAC)
c(op) : & 1
Integer divide (a) c(FAC) / c(op) ----> c(FAC)
c(op) = I
Integer divide (b) c(op) / c(FAC) ----> c(FAC)
c(op) = I
Fix c(op) is convert4d to a fixed point number (an integer), and
is stored in the regular, Linc, accumulator. Numbers are not
roynded; all fractional parts are lost. Any number less than one
is stored as zero. Any number greater than 3777(8) or less than
-3777(8) is converted to 3777 or -3777 respectiv~T~.
FIT Float
c(op) is assumed to be an integer. It is converted to a
floating point number and replaces c(FAC).
CLR Clear Storage
~ ----> C(OR)
MAX Maximum The c(op) is compared with c(FAC). The larger value
replaces c(FAC).
MIN Minimum The C(op) is compared with c(FAC). The smaller value
replaces c(FAC).
SGN Sign
If c(op) < ~, then -1 ----> c~ac~
If c(op) ::& ~, then ~ ~ c ac
If c(op) >~, then 1 ----> c ac •
13-53
3
21.
INC
Increment
c(op) + c(FAC) ----> c(FAC) and ---->c(op). This is
the floating point counterpart of Linc'~s add to memory instruction.
22.
liN
Integer Increment c(op) + c(FAC) ----> c(FAC) and c(op); c(FAC) = I
Note that in this instruction it is the FAC, not the operand, which
is assumed to be an integer.
23.
24.
25.
STO
SSP
26.
SQT Square root; ./IOPT - > FAC
IPT Input; the number inputted on the keyboard ----> op
The number is inputted in decimal and is terminated by a space.
The number may be preceded by a minus sign. Any of the following
inputs are allowable:
27.345 6..
-.0001 ~
996 ~
-101..6..
-62. ~
~
(=0)
There is nO limit to the number of digits inputted. Pressing
"del" at any time during an input deletes what has been entered and
the entire number must be retyped.
27.
30.
Store
c(FAC) ----> c(op)
Set Sign Plus Ic(op)1 ----> c(FAC)
SS~ Set Sign Minus -lc(op)1 - > c(FAC)
OPT Output; the operand is outputted on the teletype i.n the following
format:
x.xxx,
XXX
return and line feed
The first four digits are the decimal mantissa and the last three
the characteristic as a power of ten.
A1so PKG
= JMP400.:
The mnemon i cs are used in anassemb 1y program to be descr i bed.
If locations 1472 and 3742 are altered so that they both hold "4276", the teletype
does not return after it has outputted a number; it spaces once. (Normally
these locations hold "657011 )
The actual teletype output routine is included dS a subroutine within the package,
so that it"can be Jumped to from outside the subroutine package. To type a
character, load. the accumulator with that charp~ter'~'s teletype code and jump
to location 1742,. Control wi11 automatrcally be returned to p + 1,. Index
registers 12 and 15 are used by this subrout~ne and are not restored if one
jumps to 1742. A modification is included for scope output of the same format.
The package occupies all of quarters one, two, ~nd three. Quarter 7 is used
from location 3700 to 3756. All index regi~ters are restored to their previous
4
value except;n the case mentioned above.
The floating accumulator is locations
112~
and 1121.
No error detection is provided. Overflow of exponents in arithmetic subroutines
will yield incorrect answers, not error messages. The same is true of
invalid operation codes" etc.
A sample program which calculates
23
, which stores the result in the
x -3x
floating accumulator., and which leaves +1, -1, or
depending upon the value of the result, follows:
x is in lG
3 (integer) is in lT
locations 24 and following contain
LOA
~
JMP
~
in"Linc's c:'c'cumulator,
l~~~
24
JMP
IT
Loads -3, floating point, into the' FAC
4~14
I;
41;1;3
)
x-3::: c(FAC)
}
x(x-3) ::: x -3x
J
x -3x
)
Determines sign of answer Exit because the 11 bit = 0
lG
4iJiJ2
lG
4~~
IT
4iJ12
iJ
iJiJ21J
2
2 3
= c(FAC)
::: c(FAC)
Next instruction
"
13-55
Program follower
To use:
Read in progr~m to be tested and execute it once.
Then read it out temporarily on tape and read it b~ck into memory~
in executed form" into blocks of upper core corresponding to those
blocks of lower core where the program normally operates" i.e."
quarter 0 into quarter 4"
into 5" etc.
On sense switches set the quarters which the program uses (actual lower
core quarters).
On the right switches set the address of the first executed instruction.
Read in the Program follower and start 20.
Hhen the program "halts locations 20 and fo.l:lowing will contain the
locations of your progrzm which are instructions.
example of the final output:
~
loc.
20
21
22
23
24
contents.
20
rhe
~ollowing
is an
This means instructio~s were
contained in locations
20 through 176 and 405
through 760 of your
program.
176
405
160
0
0
25
26
0
The program fo 11 O\oJer is not perfect.
I t wi 11 not catch returns from
subroutines where the return address is manipulated to be anything
~esides
p+l or a constant return address.
It will not catch jumps
executed by pull~ng addresses out of; a jump table •.
all XSK instructions can proceed to both p+l and
not always true.
Therefore~
It will assume that
p+2~
while this is,
the results may contain a few locations·
which are date and" may omit locations which contain instructions.
The program follower is just that; it does not tell you what parts of
your program were meant to be instructions~ it tells you which
locations can be reached~ as instructions~ by the various jumps and
branches of your program. Thus it· provides a good method for
troubleshooting a program by showing you where your program actually
can go.
13-57
Mnemon i c dump .
To use:
Read program to be typed into upper core in quarters corresponding
to the lower core location of the program, i.e., a program which
runs in quarters 0 and 2 should be read into quarters 4 and 6.
Have tape J? on'unit 1.
Read B1 310 into quarter 0 and start 20.
Type on the kbd one-digit numbers corresponding to the quarters
used by the program. For the case mentioned above, type 0 space 2.
Separate these digits by spaces.
Then type in the locations which are instructions in the form specified
below.
If the program postulated above ran from 20 to ;60 and from
]000 to 1377 the entire input should be as follows:
~ ~360Q)~OOO~ 1377~
EOL
Quarters
Instruction Locations
Location ~ may not be specified as an instruction location
Sense switches control the output format as follows:,
~
All set to /) =
SW 1 at 1
=
SW 1" 3 at 1 =
SW 1" 2, 3 up =
single column output
2-column output, numbering spaced by 200
2-column output, numbering spaced by 100
4-column output
13-58
Source Exif Data:
File Type : PDF File Type Extension : pdf MIME Type : application/pdf PDF Version : 1.3 Linearized : No XMP Toolkit : Adobe XMP Core 4.2.1-c043 52.372728, 2009/01/18-15:56:37 Create Date : 2011:04:04 13:52:18-08:00 Modify Date : 2011:04:04 14:50:49-07:00 Metadata Date : 2011:04:04 14:50:49-07:00 Producer : Adobe Acrobat 9.42 Paper Capture Plug-in Format : application/pdf Document ID : uuid:737307b1-f2ae-4cec-930c-ea8dae522b30 Instance ID : uuid:6383b324-1075-4e22-933c-5e78b41e12cf Page Layout : SinglePage Page Mode : UseOutlines Page Count : 517EXIF Metadata provided by EXIF.tools