Micro_Cornucopia_#53_May90 Micro Cornucopia #53 May90

Micro_Cornucopia_#53_May90 Micro_Cornucopia_%2353_May90

User Manual: Micro_Cornucopia_#53_May90

Open the PDF directly: View PDF .
Page Count: 100

Download
Open PDF In Browser	View PDF

No. 53

May 1990

THE

$3.95

TECHNICAL

JOURNA L

ORNUCOPIA
Controllers
In this, our hysteric final issue, we cover
one of our favorite topics, micro
controllers and their cousins, the data
loggers.

Designing A Micro
Controller
Designing a controller is simple when
you start out with the right micro.

Building A Two~chip
. Terminal

page 26

The perfect accessory for your micro
controller.

Waylaid By Snakes
A low-down look at objects
(in the grass) .

Save The Floppies
Routine, complete with code,
recovering bad sectors on floppies.

And More ...
Shareware/Favorite Tools
Initializing Variables
in Turbo Pascal
Faster Neural Networks

7447019388

with 4MB RAM and Rotary Voice Coil Hard Drive
For extra speed and reliability we've included a 66MB Mitsubishi
MR535 voice coil hard drive with an access time of 28ms,
and a super fast 1: 1 interleave HD controller card.
•
•
•
•
•
•
•
•
•

25Mhz 80386 CPU, AMI Bios, Full Size MB
4MB of RAM, expands to 8MB on MB
1.2MB and 1.44MB Floppy Drives
12" Amber Monitor w/Mono Graphics Card
101-Key Enhanced Keyboard, Focus or Maxiswitch
2 Serial, 1 Parallel, 1 Game, ClOck/Calendar
200-Watt Power Supply
Socket for 80287,80387 or Weitek
Full 1 Year Warranty
w/Std AT Case ...... $2250
wI1MB ................. $250 Less
(Mega Tower) 20Mhz 386 .......... $100 Less

$2350
AT /386S SYSTEMS
Includes: 1MB RAM, 1.2MB & 1.44MB FD,
40MB Hard Drive (28ms), Fast Hard Drive
Controller, Mono Graphics Video Card,
12" Amber Monitor, 101 Key Keyboard,
Serial(2)/Parallel(1 )/Game(1) Ports,
Clock/Calendar, Full 1 Year Warranty.
FREE assembly and testing.
6/1 OMhz 286/AT.. ........................ $1249
8/12Mhz286/AT.. .......................... 1295
8/16Mhz 286/AT.. .......................... 1395
16Mhz 386-SX .............................. 1449
Video upgrades for our systems
(includes video card and monitor)
I

ATI Graphics Solution ....................... 39
CGA Color ....................................... 175
EGA Color ....................................... 349
VGA Color (Analog, 16 bit
w/Amazing monitor) ..................... 379
CGAlEGAlVGA (Multisync) ............. 450

360K

FLOPPY DRIVES

.................................................. 69
1.2MB................................................ 77
31/2" 720K ....................................... 79
31/2" 1.44MB ................................... 85

HARD DRIVES
XT 20MB Miniscribe/Kalok
8425 (65ms) ................................. 229
8425 w/controller .......................... 269
XT 30MB Miniscribe/Kalok
8438 (65ms) ................................. 229
8438 w/controller.......................... 269
AT 44MB Mitsubishi MR535,
Rotary Voice Coil (28ms) ............. 429
AT71MBMFM
Voice Coil (28ms) ......................... 595

Mini Tower

Standard Tower

$225

$269

MOTHERBOARDS

The Most Cost Effective Way to Speed Up
your AT or 386 System!

XTlTurbo 4.77/1 OMhz ...................... $75
AT 6/10Mhz ..................................... 175
AT 8/12Mhz ..................................... 195
AT 8/16Mhz ..................................... 249
386SX 16Mhz w/C& T chip set.. ....... 359
386 8/20Mhz w/Phoenix Bios,
holds up to 8MB on board ............ 679
386 16/25Mhz w/AMI Bios
holds up to 8MB on board ............ 795
386 33Mhz ...................................... Call

Upgrade your hard .drive controller. Discover the NEXT
generation of hard drive controller cards. As the chart below
demons~rates, the speed improvements are incredible!
Controller Card Type
Standard IBM AT, 3:1 Interleave, MFM
DTK WA2, 2:1 Interleave, MFM
AT/3S6 1:1 Interleave, MFM
Western Digital 1:1, RLL, w/Cache

Data Xfer Rate
167 KB/Sec.
261 KB/Sec.
522 KB/Sec.
799 Kb/Sec.

POWER SUPPLIES
150-Watt XT Power Supply ............... 49
200-Watt AT Power Supply ...............59
230 Watt AT Power Supply ............... 75

Note: Test results using 10Mhz AT, Mitsubishi MR535 Hard Drive and

SpmRH.CHSk~t;A Color ONLY $379..._

EXPANSION CARDS

When you buy a complete MicroSphere Computer System; Up~
grade includes: 16 bit VGAvideocard and 1411 Analog VGA
Monitorwitha.28 mm dot pitch.

CONTROLLER CARDS

KEYBOARDS

AT/3861:1 MFM Controller ........... $1 05
WD 1006-SR2 1:1 RLL w/Cache .... 139

Maxiswitch (AT Only) 101-Key ........ $59
Keytronic· Enhanced 101-Key .......... 55
Focus· 101-Key Tactile, Switchable
Control/Caps Lock, Dust Cover...... 59
(#1 Find by Micro C Staff)
• Keyboards are XT/AT switchable

CABINETS
XT Slide Case, Lock, LED ................. 35
AT Slide Case, Lock, LED
3 half ht., 1 full ht. drives ................ 59
Baby Tower, w/200 watt PS,
2 half ht., 2 - 31/2" Drives,
Baby Motherboard ONLY ............. 129
Mini Tower, w/200 Watt PS,
3 half ht, 2 - 3 1/2" Drives,
holds Full or Baby size MB ........... 195
5 Bay Std. Tower, 220 Watt PS,
3 half ht., 1 full ht drives ............... 269
6 Bay Mega Tower, 230WattPS .... 229

MONITORS

VIDEO CARDS

~~:~~I~~ ~~~~~~~.~~.~~::::::::::2::

Mono Graphicsw/Parallel port .......... 37
ATI Graphics Solution
Herc w/CGA Emulation .................. 79
CGAlEGAlVGA 8-Bit.. ..................... 119
CGAlEGAlVGA 16-Bit.. ................... 129

EGAlCGAAutoswitch ..................... 362
CGAlEGAlVGA
Multisync (.31 dot) ........................ 489
VGA Analog
(Amazing .28 dot) ......................... 369

SOFTWARE
MS-DOS 3.3 w/GW Basic ................. 95
DR-DOS 3.3 w/GEM ........................ .45
SpinRite II Disk Optimizer
by Gibson Research ....................... 79
386Max Memory Manager for
386 Systems .................................. 69

==ii

~M·

Clock ................................................. 18
Game (Joystick) ................................ 14
Parallel Port (LPT1, 2 or 3) ................ 18
Serial Port, 2 ports, 1 installed,
(COM1 or 2) ................................... 18
2nd Serial Port Kit.. ......................... 18
Serial Port, 4 ports installed ............... 99
Multi Drive Controller, up to
2 drives, Supports 360K,
720K, 1.2MB & 1.44MB .................. 39
3 & 4th Floppy Controller ................... 79

AT/386
AT Multi-IO, 2 Serial, 1 Parallel,
1 Game ........................................... 35
AT 3MB EMS Memory Card (OK) ...... 99

CHIPS/ETC.
Memory 64K, 256K, 1MB ............... Call
Math Coprocessors ........................ Call
XT Components & Systems ........... Call

here

,crO~D

COMPUTERSrH;RDW~:c~~~B~~ACT~~~R

Orders Only Please!
Tech Calls:

1-800-234-8086

(503) 388-1194 Hours: Mon-Fri 9:00-5:30

855 N.W. WALL· BEND, OREGON 97701

'Prices are subject to change without notice. Shipping CHARGES will be added. '1-year warranty/3D-day money back (subject to restrictions)

Reader Service Number 2

!M[EWU AT&T C++ RELEASE 2.0
SPECIFICATION
C\!J~~J'g MS WINDOWS
COMPATIBILITY

~fE~J'U EASY PORTABILITY
FROM MICROSOFT C
[K!J~~1fg C++DEBUGGER &
EXPANDED C++ TOOLS

[MgWU OS/2 UPGRADE
AVAILABLE NOW!
We listened carefully to
DEBUGGER, and now the
For many, EMS
what you wanted in a next
wait is over.
programming support, built
into the compiler is
generation MS DOS C++
You wanted expanded and
compiler. The answer is
important, so it's in there
improved documentation,
too.
Zortech C++ V2.0
----------w we both
Developer's
listened
You were happy using the
Edition.
and
18 classes provided in
You wanted
delivered.
C++ TOOLS, but we
the latest
revised and expanded it
You
anyway.
AT&TV2.0
wanted to
features
You never asked for a free
be able to
with the
upgrade
TSR
library to be included,
power
but we knew you'd love to
to an
offered by
OS/2 version compiler
use our neat little package,
multiple inheritance and
supporting Presentation
so we included it free.
type safe linkage, so here it
Manager, you did not want
is.
it to cost a fortune, so it's
You liked our FLASH
available r - - - - - - - - - - - - - - - - , GRAPHICS
You wanted compatibility
for $150.
Here is our list of highly
package
with MS WINDOWS, we
recommended C++ Dooks:
for its
added it.
You want
speed, but
to
look
at
c+
+
Language;Stroustrup
$32.25
wanted
a
You repeatedly asked for
C+ + AnsW3r Book;Hansen $26.95
easier portability from
the
C+ + for C Programmers
$29.95
CCI+as+s
Microsoft C, we got the
standard
c+ + Primer ;(.ippman
$30.25
message, and have written
library
interface,
the library functions you
SOURCE
Ask about our new c++ Video Tutorial
so we've
CODE, so
written it.
need.

~Bo'w 10 Order:
Already own Zortech
C++? Call the order
hotline for details of our
low cost upgrades.
To order Zortech C++
for the first time, just call
the order hotline. We
accept payment by,
Mastercard/Visa/COD.
Alternatively, mail the
coupon below with your
check or credit card
details.

ZORTECH INC.,
1165 MassachuseHs
Avenue, Arlington, MA
02174, USA
Voice 617-646-6703
Fax 617-643-7969
ZORTECH LTD.,
106- 108 Powis Street,
London, SE 18 6LU,
ENGLAND.
Voice (44)- 1-316-7777
Fax (44)-1-316-4138

L..-._ _ _ _ _ _ _ _ _ _- - - - '

You wanted the world's first
MS DOS C++ source level

Yes! Please rush me the following
Name
Addres-s----------City-,.--=--_ _ _ _State_Zip _ __
Visa/MC# _ _;-;-_ _ _ _ _ __
Exp.Date _ _ _Tel _ _ _ _ _ __
All MicroSoft trademarks are acknowledged.

C++ V2. a items:

o DEVELOPER'S EDITION$450 (Save $200)
o C++ COMPILER $199.95
o C++ DEBUGGER $149.95
o C++ TOOLS $149.95
o LIBRARY SOURCE CODE $149.95
o COMPILER & LIBRARY SOURCE $299.95
For US orders please add $5.05 shipping
Reader Service Number 70

o OS/2 COMPILER UPGRADE $ 749.95
o C++ VIDEO COURSE $499.95
o C++ Language /Stroustrup $32.25
o C++ Answer Book/Hansen $26.95
o C++ for C Programmers/Phol $29.95
o C++ Primer/Lippman $30.25

Overseas orders at international mail rates.

ENAl
l
Y
.
A debugging
tool tough enough to
handle the DOS
Nasties.

New Version 2.0

How Soft-ICE Works

Soft-ICE memory range break points help you
track down memory over-write problems
whether you are doing the over-writing or
another program is over-writing you.

Soft-ICE uses the power of the 80386 to surround your program in a virtual machine.
This gives you complete control of the
DOS environment, while Soft-ICE runs
safely in protected mode. Soft-ICE uses the
80386 to provide real-time break points on
memory locations, memory ranges, execution,
II 0 ports, hardware &' software interrupts.
With Soft-ICE you get all the speed and power of
a hardware-assisted debugger at a software price.

Hung program? No problem!

Don't want to switch debuggers?

Nasty over-write? No sweat!

You don't have to!

When the system hangs, you now have hope.
With Soft-ICE you can break out of hung
programs no matter how bad the system has
been trashed. And with Soft-ICE's back trace
ranges you can re-play the instructions that led
up to the crash.

Soft-ICE can run stand-alone or it can add its
powerful break points to the debugger you already
use. Use your favorite debugger until you require
Soft-ICE. Simply pop up the Soft-ICE window to set
powerful real-time break points. When a break point
is reached, your debugger will be activated automatically.

Program too large? Not with Soft-ICE!
Soft-ICE runs entirely in extended memory. This means
you can debug even the largest DOS programs. And
since your program runs at the same address whether
Soft-ICE is loaded or not you can find those subtle
bugs that change when the starting address of your
code changes.

System debugging? Soft-ICE is a natural!
Soft-ICE is ideal for full source level debugging of TSRs,
interrupt service routines, self booting programs, DOS
loadable device drivers, real-time kernels, non-DOS O/Ss
and ROMs. Soft-ICE can even debug within DOS &' BIOS.

MagicCV with Soft-ICE
Using Soft-ICE with CodeView gives you the features
necessary for professional level systems debugging.
MagicCV and Soft-ICE can work in concert with CodeView to provide the most powerful debugging platform
you will find anywhere.
"These may be the only two products I've seen in
the last two or three years that exceeded my wildest
expectations for power, compatibility and ease-of-use."
-Paul Mace
Paul Mace Software

New Soft-ICE 2.0 features
$386
$199
$199

Soft-ICE
MagicCV
MagicCV for Windows

[] Back Trace Ranges
[] Symbolic &' Source level debUgging
[] EMS 4.0 support with special EMS
debUgging commands
[] Windowed user interface

Buy Soft-ICE &' MagicCV(W)
-Save $86.
Buy MagicCV and MagicCVW
-Save $100.
-Save $186.
Buy All 3
I •• ~ •

30 day money-back guarantee
Visa. MasterCard and
AmEx accepted

™DNu-Mega

\}-TECHNOLOGIES
•
A",

T •

CALL TODAY (603) 888-2386
or FAX (603) 888-2465

0 OJ

O'\T.III(lIO ....

PO. BOX 7607 • NASHUA, NH • 03060- 7607

MICRO CORNUCOPIA, #53, May, 1990

Reader Service Number 110

RUN
IN
8KCODEVIEW~M
~ ..

MagicCV

);~,

Code View is a great integrated debugger, but
it uses over 200K of conventional memory.
MagicCV uses advanced features of the
80386to load CodeView and symbols in
extended memory. This allows MagicCV to
run CodeView in less than 8K of conventional memory on your 80386 Pc.

NEW VERSION FOR C 6.0
Attention Windows Developers!
Version available for CVW.

THE

M I C R 0

TECHNICAL

JOURNAL

MICRO CORNUCOPIA
MAY 1990-ISSUE NO. 53

H. Ward Silver

Designing A Microcontroller
Ward puts together a very simple data logger
and microcontroller based on the 68HCll.

Bruce Eckel

Karl Lunt

Waylaid By Snakes
A fun look at c++ if you don't object to serpents.

Building A Two-chip Terminal
Here's a simple, low-power terminal with a two-line by twenty-column
display.

46
54
70

86 World

Culture Comer

On Your Own

Units and Modules

Shareware

Techtips

Tidbits

Last Page

Lance Dannon Bresee

A Roundoff Roundup
Tired of round-off errors? This should help your calculations.

Dave Gwillim

Initializing Variables In Turbo Pascal
An elegant fix for a popular compiler.
D. Gilbert Lee

Fast Neural Networks
Larry Fogg

Save The Floppies
Stuart R. Bell

Build A Composite Video Adaptor

Cover Illustration by Greg Cross.
MICRO CORNUCOPIA, #53, May, 1990

THE M I C ROT E C H N'I CAL J 0 URN A. L

MICRO CORNUCOPIA
Editor and Publisher

David J. Thompson
Associate Editors

Gary Entsminger
Larry Fogg
Cary Gatton
Contributing Writers

Bruce Eckel
Michael S. Hunt
Alfred E. Newman
Karl Lunt

..... :>

Accounting

Sandy Thompson
Advertising & Distribution
Reader Services

Nancy Ellen Locke & Laura Shaw
Graphic Design & Production

Carol Steffy
MICRO CORNUCOPIA aSSN 0747-587X) was
published bi-monthly for $18 per year by Micro
Cornucopia, Inc. 155 NW Hawthorne, Bend, OR
97701. Second-class postage paid at Bend, OR and
additional mailing offices. POS1MASTER: Send
address changes to MICRO CORNUCOPIA, PO
Box 223, Bend, OR 97709.
Recycled Paper

SUBSCRIPTION RATES:
1 yr. (6 issues)
2 yr. (12 issues)
3 yr. (18 issues)
1 yr. Canada & Mexico
1 yr. Other foreign (surface)
1 yr. Foreign (airmail)

$18.00
$34.00
$48.00
$26.00
$36.00
$50.00

Make orders payable in US. funds on a U.S. bank.
CHANGE OF ADDRESS:
Please send your old label and new address to:

MICRO CORNUCOPIA
P.O. Box 223
Bend, Oregon 97709
READER SERVICES:
For orders and subscription problems call (503)
382-5060, 9 am to 5 pm, Pacific time, M-F. FAX
your VISA or MC order to us, our FAX number is
(503) 389-6833.

BBS - 24 hrs. 300-1200-2400 baud
8 Bits, No Parity, 1 Stop Bit 503-382-7643

.t}

The
Audit
Bureau

......................................................' ' .. .

.,."r~·notsureh()w·to begin. I've ~tarted this edlitolrial
:~()'inanytimes, in my daydreams, in my sleep, in the
. , shOwer.... But nothing I've come up with really feels
like I feel.
I'm closing down Micro C and I don't know what I'll
be doing next. (Wait till mom reads this .... "Dad! Our
little David's quitting his job. Quick, clean up the spare
. room.")
You may be wondering why it's closing rather than
moving on to other ownership. Good question. Three
groups have come forward, interested in taking on
'Micro C. After long discussions, none felt they could put
together both the finances and the talent to make it fly.
When a magazine folds (a strangely appropriate
word), other magazines often pick up the subscribers.
As far as I know, the only subscribers no one wanted
were the ones taking Profiles (aimed at Kaypro owners)
and CP/M Review. So we've had offers from several magazines wanting to finish your subscriptions. How did I
choose?
Some were willing to pay for your names. Some
weren't. Some were already popular with Micro C sub, scribers. Some weren't. Some were very interested in
picking up the articles already in the pipe for future is,:> sues. One even wanted to continue "Around The Bend"
~ in its entirety. (Led to an, uhm, lengthy discussion
about sanity.)
After considering all the options, I selected Computer
Language. To finish out your Micro C subscription, you'll
get your choice of Computer Language, AI Expert, or
Embedded Systems Programming. If you already subscribe (25% of you already take Computer Language),
they'll extend your subscription. Just fill out the form
you'll receive in April.
If you've got a year of Micro C coming, you'll get a
year of Computer Language-and you'll be getting it at a
really good discount (CL's normally $29 a year, Micro
C's just $18). Plus, Computer Language's sta'ndard
policy is: if you decide you don't like their magazine,
they'll send you a refund for all remaining issues. All
you have to do is let them know.
Look, I know that Computer Language isn't Micro C.
When I looked closely the past few months, I was surprised how far we'd moved away from the other com-

Continued on page 74

MICRO CORNUCOPIA, #53, May, 1990

c-:J rying to modify

a program without its
II source code is tough enough without
U having to use an inferior disassembler.
That's why you should look at DASM™
before you invest in any disassembler
product.
Most disassemblers complicate the
process with errors: giving you incorrectly
placed labels, incorrect and misplaced code
and data, data treated as code, or code
treated as data.
DASMTM makes the modification process
easy by keeping track of segment register
usage and mapping code in its execution
sequence. It determines which parts of the
file are code, which are data, and what
labels exist. You simply edit the assembly
language output, make the necessary
changes to the program, and reassemble.

Japan's #1 Seller - Made in America

TlI

Try DASM Risk-Free!

DASM's simplicity, accuracy and power
DASM™ is available through the Prohave made it the top-selling disassembler
grammer's Shop for only $225. And it'sguarin Japan for two straight years. The
anteed to outperform any disassembler on
reason - DASMTM does what other disthe market.
assemblers can't:
Order your copy now and throw your
o Handle problematic segment and instrucworst disassembly problem at it. We think
tion address register loads through interacyou'll be impressed. If not, just return
tive mode
DASMTM for a full refund.
o Track segment register usage
Why settle for anything less?
o Generate appropriate ASSUME's and
segment maps
r=-===:~::""l
[J Handle mUltiple entry points
o Track memory references
• Direct
• Via register
Software
• Implied in OS function requests
To Order Call
o Handle transfer vectors
301·752·1348
o Handle .EXE, .COM and .BIN files
up to 200K

To demonstrate DASM's superior performance, we gave the same disassembling task
to both DASMT~ and SourcerH'(Version 2.18).

bx~

C1WOH
C1N2H:
mov
mov
int

@) Since Sourcer™ generates no segment labels relative to the DS segment,
changes in the DS segment will not be
reflected.
{j) Sourcer™ can't handle jumps or calls
via register, and does not generate the,
code segment label (or reference).

proc

far

mov
mov
jmp

cl, ds:data...3e ~ @
ds, cs:data.4 ~
short loc_1 ~ ~

1)5

endp

seg-h

mov
mov
ASSUME
jmp
ENDS

'Print string message$'
'P'

ends
end

';.JJc:~,.

t:( db

"'::€>
d

21H

ends
~
segment para public
assume cs:seg-h , ds:seg

In short, if you're not using DASM™, be
prepared to spend a long time editing the
file to add and fix the missing and incorrect labels - and expect a lot of bugs!

dX,OFFSET D3BOH
ah,9

a1,D3B15H
bx, OFFSET C1B11H
bx ~ /~

OB4h, 4Ch, OCDh, 21 h.d
start:

Trademark/Owner: DASMTMIJBSoftware;
Sourcer™/V Communications.

10c_1:

€b Sourcer™ continues to have the

@ Sourcer™ does not reference the segment value symbolically, so any changes
to the program which alter its location
will not be reflected.

t"Jhile DASMH' generated the correct
assembly code, Sourcer made at least six
major errors.

start

D3SEG

SEGMENT
LABEL BYTE
'Print string message$'
db
LABEL BYTE
'P'
db
org
20H
ENDS
END
C1F15H

Reader Service Number 213

MICRO CORNUCOPIA, #53, May, 1990

Rack &Desk
PC/AT Chassis
Integrand's new Chassis/System is not
another IBM mechanical and electrical
clone. An entirely fresh packaging design
approach has been taken using modular
construction. At present, over 40 optional
stock modules allow you to customize our
standard chassis to nearly any requirement.
Integrand offers high quality, advanced
design hardware along with applications
and technical support all at prices competitive with imports. Why settle for less?

Accepts PC, XT, A T Motherboards
and Passive Backplanes
Doesn't Look Like IBM
Rugged, Modular Construction
Excellent Air Flow & Cooling
Optional Card Cage Fan
Designed to meet FCC
204 Watt Supply, UL Recognized
145W & 85Walso available
Reasonably Priced

~QW

~
-_......
.......
- . -.........
.... ..

Passive
Backplanes

ii . . .=;r~~~r=:;.

...--

~~ ~ ~ !!:=~~

.... • r=:ta

--.

!!"-e!

Call or write for descriptive brochure and prices:
8620 Roosevelt Ave. • Visalia, CA 93291

209/651-1203

TELEX 5106012830 (INTEGRAND UD)
FAX 209/651-1353

We accept Bank AmericardlVlSA and MasterCard
IBM. PC. XT, AT trademarks of International Business Machines.
Drives and computer boards not included.

Reader Service Number 22

MICRO CORNUCOPIA, #53, May, 1990

Letters

So Long
Today I heard that Micro Cornucopia
is to be no more. This greatly saddens
me even as I recognize the need for
change in all things. Your magazine
filled a niche in the journalistic world
that was very special and unique. You
represented all of us plain folks, who
did not hide behind the shirt and tie but
who loved working with computers at
every level.
You spoke to us, as one enthusiast to
another, not using technojargon to hide
a lack of real understanding so common
in other journals. You held gatherings
where we laughed and joked with the
literary mentors who taught us and
guided us on our own paths to knowledge, even while we wiped the bar-bque sauce off our lips.
You became family. You will be
missed. My best wishes and thoughts go
out to each and every one of you at
Micro C who have touched my life.
Godspeed And Thank You.
Al J. Szymanski (Big AI)
8991 Edcliff Ct. SE
Aumsville, OR 97325

Editor's note: Thanks AI. The entire staff
remembers, only too vividly, your smashing
presence during SOG volleyball. Godspeed
to you, also.
Bend Memories
I have just sent in my subscription renewal. Late. I can't believe that just because I forgot to renew, you stopped
sending me the magazine! Didn't you
know that I still wanted it?
Seriously, I think Micro C is the best
computer mag available. You can imagine my shock when I first saw it a
couple years ago. I couldn't believe such
a great mag was published in Bend, of

all places. I mean, I used to live in Bend
when I was 14, in the year 19-mumblemumble. I don't suppose Bend has
changed any since then? As I recall,
Bend was very small.
I say, Bend was very small. (Did you
say, "How small was it?") It was so
small that if you wanted a business
phone number, you looked it up on the
yellow page. It was so small that every
time I turned on my radio, the electric
trolly slowed down. It was so small we
didn't have a regular village idiot; we all
took turns. Bend had a newspaper,
though. ("Really?") Yeah, a traveling
salesman left it at the feed store.
I may be exaggerating slightly, but
the most interesting thing I did all year
there was ride my bike up the dead volcano outside town. I had just seen The
Last Days of Pompeii and, being 14, I had
hopes it might erupt again and liven
things up a bit. No such luck.
Sounds like I missed something by
not attending the Rocky Mt. SOG. But I
had been to Gunnison before, and
having recently seen Deliverance, I
Letters continued on page 68

,,~ • .e ~"'>~'

, __

"""

" "

,."

~,,,.,.

• ~

"'Y~" ,_.

'~"'_"

•

",>~,,_,., .. ,

,~. ,~,~ •..,,,,,~,

«..,...""

~'"""'"' .. ~~" .... "" ,~.~

.. ~••

,.,W~~

• v~ ","""",,~" . , " _ N M w _ , ' . "

>0>''"'' "' ..,....

~,,~

... ""' '"

~ '_".~V'

...~~,

""~<

~ ....... "

,"",

•.• ,,"

""~ ....

""''''''''''''.'V",~"M"..,."

"\(jood Lucl(f'=::·:·~·: :. ~'
/) rJJave & Sanay
Jennifer & 'Erin
II

Try the 1st truly low cost LAN
• Connect 2 or 3 PCs, XTs, ATs
• Uses serial ports and 5 wire cable
• Runs at 115 K baud
• Runs in background, totally transparent
• Share any device, any file
• Needs only 14K of ram

Skeptical? We make believers!

·~

OveJt;

15. 000 ~JJ ~

Information
Modes
P.o. Drawer F

I 0 e

I
Denton, TX 76202
...--_ _ _...., 817-387-3339 or 1-800-628-7992

Why waste money on Simple file transfer systems?

Love,
'Don & Xim
'Dona{d & Court{and
Reader Service Number 149

MICRO CORNUCOPIA, #53, May, 1990

Designing A MicroControlier
The MC68HCll In Action

Ward walks us through a microcontroller
design project based on the 6811. If you're
thinking about designing a monitor to watch
the real world or control a process this
project, the DataBear, should make great
reading. (Now that I've spilled the name of
his board, you suspect I'm going to make one
of my unbearable puns. Right?)
he best way to describe a particular
microprocessor is to show how it's
used. In this article, I'm using the
Motorola MC68HCll in a batterypowered data logger, the DataBear from
Langan Products, Inc., of San Francisco.
The 6811 is one of the more flexible
and powerful eight-bit microcontrollers.

8 MICRO CORNUCOPIA, #53, May, 1990

See Figure 1 for a list of features and Figure 2 for the way it's organized.
Not only is the 6811 well supported
by software and hardware development
tools (including C compilers and emulators), it's quite cheap. Unlike some of the
latest whiz-bang processors, a 52-pin
version with 512 bytes of EEPROM costs
only about $12 in onesies. Plus, you can
easily interface inexpensive 8-bit support
chips to the 6811.
See Figure 3 for a list of variations on
the 6811.
What Is A DataBear?
I designed the DataBear as a small,
simple unit, powered by just about anything and capable of monitoring ordinary analog data at very ordinary rates.

Of course, I wanted to include as
many features of more expensive loggers as I could while keeping down the
chip count and printed-circuit size. The
6811 makes that possible.
The DataBear packs the 6811, 8K or
32K of static RAM, 8K or 32K of
EPROM, bus logic, real-time clock, serial
interface, temperature sensor, and two
external channels of analog input onto a
board 2.6x3.0 inches. With the battery
packs, the whole unit is only 1x4x6 inches. It's portable and powerful, thanks
to the 6811.
Slow And Steady Continuous Mode
I set up the DataBear to collect data
over a long period of time, from once
per second to once every 18 hours. Each

By H. Ward Silver
RBRDesign
P.O. Box 1608
Vashon, W A 98070
(206) 463-2833/9173
Fax (206)463-9122

Figure 1-Basic 6811 Features
Internal Memory: 256 bytes static RAM
512 bytes EEPROM
o - 12K bytes mask ROM (programmed at factory)
External Memory: 64K byte address space with mux'd bus
Serial Communications Interface (SCI), to 131K baud
Serial Peripheral Interface (SPI), clock speed to 1/2 crystal freq.
Parallel I/O: up to 38 bits, some dedicated Input or Output
16-bit Timers: Four-Stage Prescalar
3 Event Capture Counters (ICF)
5 Counter Controlled Outputs (OCF)
Pulse Accumulator Input (PAl)
8 Channels of 8-bit Ratiometric A/D Conversion
2 Hardware External Interrupts: XIRQ (non-maskable) and IRQ (maskable)
Programmable Real-Time Interrupt and watchdog Timer
WAIT/STOP Low Power Modes
6801-based instruction set plus: 16x16 Integer and Fractional Divide
Bit Manipulation Instructions
Direct, Extended, and Indexed (2 16-bit Index Registers) Addressing

•••

sample is an average of eight rapid
samples,
thus
minimizing
high
frequency noise.
You'll probably use Continuous
Mode for most field, shipping, or remote
applications. When you're sampling at a
very slow rate, the processor can shut
down between samples, saving power.
Programmable Fixed Mode
A second mode uses the real-time
clock to take bursts of data at preset
times. This is called Fixed Mode and it
lets you control the time resolution of
the data.
Real Time Mode
The third logging mode samples on
request by a host computer. A command
string from the host triggers the sampling. The Bear then reports its data in
either binary or ASCII. The maximum
rate is approximately 20 samples per
second, limited mainly by the speed of
the serial link.
The
DataBear's
time-controlled
operation makes heavy use of the 6811' s
SPI port to service clock interrupts and
read the clock time and date. Because of
the SPI's high speed, these transactions
don't take long, providing for low
power consumption and a high sampling rate.
Analog Conversion Numbers
The DataBear has three analog inputs. The 6811 has an eight-channel
AID Converter, but I used two to monitor the power and left three for expansion. Of the three left, I used one channel
to monitor internal temperature. You get
the other two for data. The converter resolution is eight bits, its accuracy limited
by the voltage reference and by the performance of your sensor.
Eight bits of resolution is acceptable
resolution for the vast majority of field
data logging requirements. For example,
MICRO CORNUCOPIA, #53, May, 1990

with the built-in LM34 temperature sensor, we measure 0 to 128' degrees
Fahrenheit with a resolution of 1 degree.
Host Interface
Interaction with the DataBear happens over an RS-232 serial port at 9600
baud. Any type of computer can talk to a
DataBear. To avoid the bulky and expensive DB-25 type connector, I used a telephone handset (RJ-11) style jack.
The serial port uses only three lines:
receive data, transmit data, and ground.
The modular jack has proven to be a
wise choice because of its durability. I
chose a handset-size jack to prevent accidental connection to a 50 Volt phone
line with its 100 VAC ringing signals.
These voltages would damage the Bear.
Processor & Memory
See Figure 4 (on pages 14-15) for the
functional schematic. If you're familiar
with the Motorola 6801 family, you'll see
that I've implemented a standard multiplexed memory bus. The data lines and
the eight least-significant address lines
are multiplexed at the processor, and
demultiplexed by a 74373 octal latch.
The 6811 Address Strobe (AS) signal
controls the '373.
The 6264 8Kx8 static RAM and 27C64
8Kx8 EPROM are the only bussed devices. The RAM is mapped at $0000 and
the EPROM at $EOOO. Because of the
simplicity of the external memory-map,
you can use A15 as the master chip
select, gated with the processor's E clock
for timing. The 6811's on-chip resources
reduce or eliminate the need for separate
peripheral chips such as DARTs, parallel
I/O, timer/counters, etc.
As memory prices fall, I'll change to
32Kx8 RAM. This way the Bear will
handle more data and more channels.
Adding more EPROM space for enhanced firmware is easy also. Because of
the size of the printed-circuit board,
adding more complex memory mapping
would require a CMOS PAL.
A neat feature of the 6811 is that the
internal RAM (256 bytes) and the internal I/O control registers (64 bytes) can
be "moved" by firmware to any 4K
boundary. These can be overlaid with
external memory without bus conflicts.
Both the internal RAM and control registers can overlay each other, sacrificing
the lowest 64 bytes of internal RAM.
A less-than-neat feature of the 6801family multiplexed bus timing, often
overlooked by designers, is that data is

MICRO CORNUCOPIA, #53, May, 1990

Figure 2-0rganization of 6811

MODA

(lJ!<)

11m

MODB
(VSTBY)

(Vpp)

XIl'm"

ROM BK BYTES

EEPROM 512 BYTES

[5
U

TIMER
SYSTEM

CPU CORE

RAM 256 BYTES

SERIAL
COMMUNICATION
INTERFACE

SCI

4B-PIN VERSIONS

A-Block Diagram

I~5
1

01 7

a IS-BIT ACCUMULATORS A ANa B

I INDE~ REGISTER X

I INDEX REGISTER Y

a 1 STACK POINTER

I PROGRAM COUNTER

I CONDITI ON

OR 16-BIT DOUBLE ACCUMULATOR 0

CODE REGISTER

CARRY/BORROW FROM MSB
OVERFLOW
ZERO
NEGATIVE
I-INTERRUPT MASK
HALF CARRY (FROM BIT 3)
X-INTERRUPT MASK
STOP DISABLE

B-Programming Model

Introducing ...

The 1st Family of
Low Cost, Powerful
Text Editors

VEDIT Jr.
$ 29
VEDIT
$ 69
VEDIT PLUS $185
Finally, you can choose the best editor for your needs without
compromising performance or paying too much. And organizations that want the "same" editor for everyone can pick VEDIT®
for most users and VEDIT PLUS for their power users.
The new family of VEDIT text editors are upwards compatible,
easy to use and offer exceptional performance, flexibility and
stunning speed. (3 to 30 times faster than the competition on
large files where speed really counts.)
Call for your free evaluation copy today. See why VEDIT has
been the #1 choice of programmers, writers and engineers
since 1980.

VEDIT Jr.- Unmatched performance for only $29.
All VEDIT editors include a pull-down menu system with "hot
keys:' context sensitive on-line help, pop-up status and ASCII
table, a configurable keyboard layout and flexible, unlimited
keystroke macros. Edit files of any size and any line length.
Perform block operations bycharacter, line, file or column. Undo
up to 1000 keystrokes- keystroke by. keystroke, line by line, or
deletion by deletion. Automatic indent, block indent and parentheses matching speed program development. Word wrap,
paragraph formatting, justification, centering, adjustable margins and printing for word processing. Run DOS programs.
VEDIT-A best value at only $69.
Simultaneously edit up to 36 files and split the screen into windows. Search/replace with regular expressions. Includes the
best compiler support available-menu driven, easy selection
of compiler options, supports "Include" files and MAKE utilities.
VEDIT PLUS-Ultimate programmer's tool for only $185.
VEDIT PLUS adds the most powerful macro programming
language of any editor. It eliminates repetitive editing tasks and
permits creating your own editing functions. The macro
language includes testing, branching, looping, user prompts,
keyboard input, string and numeric variables and control over
the size, position and color of windows. Source level macro
debugging with breakpoints and tracing. Macros developed
with VEDIT PLUS also run under VEDIT.
30 day money-back guarantee. Call for pricing of XENIX, OS/2
and FlexOS versions. Very attractive quantity pricing is available for schools, hardware and software vendors.

FREE Evaluation Copy*
Call1·800·45·VEDIT
Compare Features and Speed
Pull-Down menus
Pop-Up ASCII table
Keystroke macros
Regular Expressions
"Cut and Paste" buffers
Text (book) markers
Undo keystroke by keystroke
Undo line by line
Normal/max Undo levels
Variable tab positions
Configurable keyboard
Integrated mouse support
FILE LIMITS
Edit files larger memory
Maximum line length
Maximum lines/file
COMPILER SUPPORT
Menu driven
Select Compiler options
Support "Include" files
BENCHMARKS 5DK FILE
Simple search
Save and continue
1000 replacements
BENCHMARKS 3 MEG FILE
Simple search
Save and continue
60,000 replacements
Block-column copy (40 x 200)
Insert 1 Meg file in
middle of 1 Meg file
PRICE

VEDIT and CompuView are registered trademarks of CompuView Products, Inc. BRIEF is a
trademark of UnderWare, Inc. Norton Editor is a trademark of Peter Norton Computing Inc. OEd it
is a trademark of SemWare.
"Supports IBM PC, XT, AT, PS/2 and clones with CGA, MGA, EGA, VGA, Wyse 700, Amdek 1280
and other displays. Also supports Concurrent DOS, DESOview, Microsoft Windows,
PC-MOS/386 and most networks.
"Also available for MS-DOS (CRT terminals), TI Professional and others.
"Free evaluation disk is fully functional and can edit small files.
Reader Service

VEDIT
BRIEF 2.10 Norton 1.3 OEdit 2.07
Yes
No
No
Yes
No
No
Yes.
No
1
100+
No
100+
Yes
Yes
No
No
100
36
1
1
10
10
No
No
Yes
Yes
No
No
Yes
No
No
No
500/1000
30/300
Yes
No
Yes
No
Yes
Difficult
No
Yes
Yes
No
Yes
No
Difficult
65,535
>65,535
None

Yes
>8096
8,388,607

Yes
512
65,535

Yes
Menu
Yes

No
Difficult
No

0.2 sec
1 sec
3 sec

1 sec
2 sec
19 sec

1 sec
2 sec
17 sec

0.3 sec
1 sec
2.5 sec

1:40 min
1:05 min
3:18 min
2 sec

1:36 min
3:23 min
1:44 hour
30 sec

Cannot
Cannot
Cannot
Cannot

Cannot
Cannot
Cannot
2 sec

1:11 min
$69

15:13 min
$195

Cannot
$75

Cannot
$54.95

No
512
20,000
None

CompuView
Number 7

1955 Pauline Blvd., Ann Arbor, MI 48103
(313) 996-1299, Fax (313) 996-1308
MICRO CORNUCOPIA, #53, May, 1990

Figure 3-MC68HCll Family Members
ROM

EEPROM

RAM

CONFIG3

Comments

512

256

$OF

Family Built Around this Device

MC68HCIIAI

512

256

$OD

Same Die as ' A8 but ROM Disabled

MC68HCIIAO

256

$OC

Same Die as ' A8 but ROM and EEPROM Disabled

XC68HCIIB8

512

256

$OF

Early Experimental Version

XC68HCllBl

5121

256

$OD

'B8 with ROM Disabled

XC68HCIIBO

256

$OC

'B8 with ROM and EEPROM Disabled

MC68HCI1E9

12K

512

$OF

Four Input Captures and Bigger RAM and 12K ROut

MC68HCI1El

512

$00

'E9 with ROM Disabled4

MC68HCI1E0

512

$OC

'E9 with ROM and EEPROM Disabled4

MC68HC811A2

2K2

256

$FF

No ROM Part for Expanded Systems

MC68HCIID3

192

N/A

Economy Version, No A-D and Smaller Memories4

Device Number

' MC68HCIIA8

Notes:
1. The EEPROM on B Series parts requires an external 19-volt supply for programming and is not byte erasable.
2. This 2K EEPROM is relocatable to the top of any 4K memory page. Relocation is done with four bits in the CONFIG register.
3. CONFIG register values in this table reflect the value programmed prior to shipment from Motorola.
4. Available in 1988.

only guaranteed for 10 nanoseconds following the trailing E-clock edge. Either
you carefully watch the number of gate
delays between the processor E-clock
and data receiving device or you
generate an intermediate clock edge.
6809 users have the quadrature (Q) clock
edge available to help grab data.
You can also operate the 6811 in the
Single-Chip Mode if you install the program in the internal masked ROM (great
for high-volume developers).
For the best of both worlds, Motorola
makes a companion chip, the 68HC24,
which restores lots of parallel I/O to an
external memory design .. This device
connects to the 6811's address/data bus
and restores the 16 bits of parallel 1/ 0
lost to the bus. It's as if the parallel ports
were built into the 6811! So you can have
it all .. " For a price.
SCI Software
The 6811 has an on-chip SCI for serial
communications. The SCI can generate
interrupts or set status flags and it
senses Framing and Overrun errors.

MICRO CORNUCOPIA, #53, May, 1990

Though the SCI won't automatically
generate and check parity, it will send
and receive parity bits supplied by software. Plus, you can use incoming data to
wake up the processor after you've left it
in WAIT or STOP (low-power) mode.
Since baud rates get generated by
dividing the processor's clock, you'll
want to choose the processor frequency
carefully. I selected 8.0 MHz so the Bear
can talk at all the standard baud rates.
The 6811's SCI output is logic-level,
not the ±12V standard for RS-232. I
chose not to use the familiar 1488/89 RS232 interface because of their hefty
power needs and their use of ±12V.
Instead I used the Maxim MAX232. It
has two line receivers and two line
drivers and it has a charge pump for
generating RS-232 line-output voltages.
By controlling the amount of power to
the charge pump, I control power consumption.
.
Real-Time Clock And Serial
Peripheral Interface (SP!) System
During the initial development of the

DataBear, I used the National MM58167
clock chip. This venerable device
worked well enough, but it didn't offer
absolute year clocking. And it required
four I/O lines plus control lines. On a
2.5x3.0 inch board, that's expensive!
Then I found the RCA 68HC68Tl
clock family, now second sourced by
Motorola. The' 68Tl family offers a 6811compatible SPI port, absolute year,
weekday, time to hundredths of a second, and 32 bytes of static RAM in some
family members. I selected a 32 KHz
tuning fork watch crystal as the timebase, although you can use crystals up to
8 MHz with higher power consumption.
A caveat for clock circuit designers:
tuning fork watch crystals have a poor
temperature coefficient compared to the
standard HC-6/U quartz plate crystals.
A typical watch cry~tal can be adjusted
to 32,768.0 Hz ±0.1 Hz. However, even
at that, the clock can gain or lose several
seconds per week.
It's funny; tell people that you have 3
ppm accuracy and they love it, but
several seconds per week is awful! How

do watches keep time so well? It helps to
keep the crystal at a constant 90 degrees
by strapping it to your wrist! (A warm
hand means a timely person.)
The DataBear communicates with the
clock via the 6811 SPI port. The SPI system is a high bit-rate, synchronous,
Master I Slave, serial data port using two
data lines and a clock.
The interface supports full duplex,
multiple masters, handles limited data
collision detection, and even offers interrupts and status flags.
In most simple single-processor systems, the processor gets to be bus
Master. All other SPI devices become
Slaves. You might scratch your head trying to work out clock polarity (CPOL)
and phase (CPHA), but careful scrutiny
of the data sheets usually solves the
problem. With only four possibilities,
you can always psyche it out with a
scope or debug monitor.
Once you've configured the SPI, all
you need is software which knows how
to talk with each slave. Each slave will
probably speak a slightly different protocol.
Our 68HC68T1 clock speaks data in
nybbles and (of course) in a reverse

order, but that's about par for the
course.
The Master sends data serially out the
Master Out Slave In (MaS!) pin, while
simultaneously generating a shift clock
on the SCK pin. Slaves send data to the
Master on the Master In Slave Out
(MISO) pin. Tie the 6811's SS line high
when it's the Master Device; use the pin
as a device select (Slave Select) when it's
the Slave.
Make sure that the SPI clock rate is
not too fast for any of the Slave devices,
especially CMOS parts. Each Slave will
probably have its own device select line,
so you can program a custom clock rate,
polarity and phase for each device.
I've found the SPI is probably the
most trouble-free subsystem in the entire
DataBear. More SPI parts seem to be
available each month. Other serial devices, such as AID or D I A Converters,
often work fine even though they are not
marketed as SPI compatible;
AID Converter Subsystem
The 6811 AID Converter is a ra-

tiometric converter. The output of the
converter is a digital ratio of the input
signal to the voltage references. With the

QEdl1t™

6811, these references are Vr! and Vrh.
Any voltage between Vrl and Vrh will result in an eight-bit output according to
the formula:
Output (0-255)

The Fast, Easy-To-Use, Full-Featured Text Editor. Now Also Available as a TSR Utility!

QEdit is packed with features:
o Completely configurable, including keyboard and colors
o Simple to install, requires less than
SOK disk space
o Simultaneous file editing-limited
only by available memory space
o Open up to eight windows
o "Pop-Down" menu system and
customizable Help screen
o Easy-to-use macro capabilityincluding keyboard recording

QEdif $5495
M

ADVANCED

from within QEdit
[J Wordwrap and paragraph
reformat capability
o Import files and export blocks
D Great for use with laptops-QEdit
saves battery drain by editing files
entirely in memory

System Requirements _ _ _ __
QEdit requires an IBM PS/2, POAT, PCIXT,
PC, POJr, or compatible. Minimum system
requirements are 64 KB of memory, PC-DOS
2.0 or MS-DOS 2.0 or greater, 50 KB of disk
space. QEdit runs GREAT on floppy based
slstems and laptops.
QEdit TSR also requires EMS, XMS, or a
hard disk.

$7900
OS/2

ALSO.~yAILABLE

Q Edit for

x 255

Continued on page 16

THEt;1

o Column blocks
o Recover deleted text
o Exit to DOS or DOS shell from
within QEdit
o Execute command-line compilers

Vrh _ Vr/

It's easy to tie Vrl to ground and Vrh
to Veo but digital noise can be a problem.
You can reduce the noise by adding an
RC-filter to the Vee power at Vrh. Better
still, use a zener diode or adjustable voltage reference to regulate Vrh.
Also be careful about how you route
a ground to Vr!. If digital currents flow in
ground traces used for Vr!, you'll have a
noise problem again.
I used a National LM235-2.5 as a reference diode for the DataBear. The
LM136/236/336 series of references are
adjustable, conditioned zener diodes.
The nominal tolerance is 5%, and I use a
pot to divide the reference's output
down to 1.280 Volts. Refer to the "Voltage Reference" section of Figure 4 for a
schematic of the DataBear reference circuit.
The AID Converter in the 6811 has
different modes of conversion controlled

~
ttl EDITOR
ADVANCED :-':,;;..'"____________
_

QEdit has always been the text
editor that combined price, performance and value. QEdit TSR makes
that value even better by giving
you instant access to the power of
QEdit-no matter what program
you're working in.

Input - vr/

•• QEdit is a
great piece of software. Highly
recommended. ~ ~
John C. Dvorak,
PC Magazine

September 12, 1989
•• QEdit is, without question,
the smallest, fastest, most
versatile text editor in the
DOSworld.~~

P. L. Olympia, Ph.D.,
DBMS
July, 1989

INTRODUCING QED IT TSa
Instant Access to the
World's Best Editor
SPECIAL PACKAGE PRICE OF ONLY
$®fiJ'OO INCLUDES
'i:I
QEDIT ADVANCED

To order direct call

404·641·9002
extension 12

[EB ~

.cards
. Now, whose
system would benefit from a case of slithers?
nderstand now, I set out to do
what I was supposed to-build a
voice-control system using the
Motorola VCP-200 (Radio Shack #2761308). This chip will recognize five
phrases independent of the speaker: go,
stop, turn left, turn right, and reverse. Its
data sheet (from Radio Shack) includes a
schematic for the chip's circuit. All I
needed to do was connect the chip to the
microphone and analog amplifier/filter
from Micro C #50.
But I didn't do it yet. I became distracted trying to imagine how I'd use my
apparatus. (I should have listened to
Larry, who never gets bogged down in
these kinds of details.)
So I digressed-what was I to do
with the signals from the chip? How
could I use commands like: "left turn"
and "reverse?"
I speculated-draw something on a
screen: an entity which moved under
voice control (leading perhaps to a vocal
PAC MAN). But what entity?
Then (as though in a vision on a hot,
hazy day in the high desert) it came to
me: snakes. A snake (an object) is a
bunch of linked segments (also objects).
To move the snake forward, I'd create a
new segment, call it the head, then tell it
(the lead segment) to send a message
back to the tail saying "add a new segment" or "drop a segment."
So I built snakes instead of connecting hardware. In the process, I generated

MICRO CORNUCOPIA, #53, May, 1990

some interesting C++ code and utilized
several C++ 2.0 features. SNAKES, my
code creation, is (of course) a riot.
When you run SNAKES, tell it how
many snakes you want, how long you
want each snake to be, and how quickly
you want to see action. By increasing the
delay factor, you can slow things down
as much as you like. Each has its own
ASCII character.
I compiled SNAKES using Zortech
C++ version 2.06 and their disp package
to increase display speed. In the Micro C
listing, I've #ifdefed disp out so SNAKES
will work (albeit slower) with other implementations of C++. (These include
UNIX as long as you have an ANSI terminal, or are willing to rewrite the brief
segment::drawO and segment::eraseO
routines.) You can also easily change
these routines to support graphics.
Classes
Figure 1 (SNAKE.HPP), the header
file, contains the class definitions for
point (which holds x-y information),
segment, and snake. I made point a
struct (a class with all members public)
because I wanted the convenience of
having coordinate information packaged
together without having to bother with
private data.
The constructor simply initializes the
data elements x and y using the initializer list. (That is, you can treat built-in
types as if they had constructors in the
initializer list only.)
We overload the == operator to determine whether one point is equivalent to
another. This prevents a snake from
backing over itself.
random_pointO is a static member
function which returns a random point.
random_pointO creates a temporary object by calling the point constructor.

Note: we don't need an identifier since
the object (a point) immediately returns.
A static member function addresses
the class as a whole, not a specific instance of the class. This could have been
an independent (non-object contained)
function, but I made it static so anyone
using class point can now see (and
access) it. Also the identifier random_point is local to class point, and
doesn't "pollute the global name space."
For this and other reasons, static member functions are an important addition
to C++ 2.0.
Segments
The segment class contains a pointer
to the previous segment in the list. So
any segment can send a message to the
segment behind it ("drop the tail segment, add a segment").
The point object lets the segment locate itself on the screen. It draws itself
and prevents its own redundancy-the
entire snake slithering over itself (although other snakes may slither over it).
segment displays itself on the screen
using a pattern and orients itself through
a variable called heading of the
enumerated type direction. heading determines a snake's direction and tries to
keep it going that way.

(Aside: my original design didn't include
heading. I added it after I got the system
working and saw snakes doubling back on
themselves, creating an ugly mess. Fortunately, it was a simple matter to add heading to the existing system. In general, try to
get the system up quickly. You can't know
some things while you're working on the design, only after you see the design working.
Local Enumerations & Anti-pollution
Notice that enum direction is inside
the class definition. In C++ 2.0, enumera-

By Bruce Eckel
Revolution2
. P.O. Box 760
Kennett Square, P A 19348

Figure 1- SNAKE.HPP
#include

size_t definition

struct pOint { II coordinate representation
int x, y;
point (int xi = 0, int yi = 0) : x(xi), y(yi) {}
int operator==(point rval) { II test equivalence
return x
rval.x && y == rval.y;

When you run
SNAKES, tell it how
many snakes you
want, how long you
want each snake to
be, and how quickly
you want to see
action.

static point random-point();

generate random pOint

};

class segment {
segment * previous;
point sp;
II position on screen
char pattern; II to display on the screen
enum direction heading; II "tendency of motion" of the snake
void erase () ;
void checkheap(); II make sure this is on the heap!
public:
II this enumeration is local to the class:
enum direction { UP, DOWN, RIGHT, LEFT };
void draw () ;
II re-define operator new for this class ONLY:
void * operator new(size t sz);
segment(point p, char ptrn);
segment(segment * prv, direction dir);
-segment () { erase (); }
int shed tail(); II go back to the tail and drop it off
point seq-point() { return sp; }
int cross over(point); II is this point an existing segment?
direction-path() { return heading; }
void redraw tail(); II redraw everything from here back

};

class snake
segment * head;
int length, maxlength;
public:
snake(point p, int size, char ptrn);
void crawl(direction);
void slither () ;
};

•••

tions are local to classes. The enumeration tag is global (everyone can refer to
direction). But the enumeration names
are local to the class, so they don't pollute the global namespace-you must
refer to them as segment::UP, etc.
Preventing namespace pollution is a
very important feature of object-oriented
languages. Big projects often become unmanageable simply because they create
more names than they can keep up with.
You can also use enumerations local
to a class to create "local const" values.
You can define a const inside a class, but
you can't initialize it when you define it
(unlike its definition everywhere else,
when you must initialize it). Do it instead in the constructor initializer list
(thus you can end up with different
values for the const depending on which
constructor you call).
However, you can create the equivalent of a local const and initialize it at
the same time using an untagged enum,
like this:
class x {

MICRO CORNUCOPIA, #53, May, 1990

enum {red=7, blue=4};

Iino

tag!

Figure 2 -SNAKE.CPP

};

Now you can use the identifiers red
and blue just like const values. Well, almost-at the time I wrote this, different
compilers exhibited different behaviors
when I tried to use red as a dimension in
an array definition. I'm not sure how it's
supposed to work. Note that you can
never use even a const as an array dimension in ANSI C.
Class Overloading Of "new"
Another feature of C++ 2.0 lets you
overload the operators new and delete
(which control dynamic memory allocation) on a class-by-class basis. Normally
you do this when you want to create dynamic objects more efficiently. I've done
it here because segment objects must
never be created on the stack, only on
the heap.
To signal the programmer, segment::operator newO (shown in Figure
2) puts a "signature" in the dynamically-allocated memory. The function
segment::checkheapO looks for this signature. If it doesn't find the signature, it
sends a message to the programmer. It
isn't foolproof, but will usually detect
misuse. All segments must be created on
the stack because they're only destroyed
by segment::shed_tailO when it calls delete.
"segment" Constructors
segment has two constructors. The
first-for the initial segment of the
snake-establishes the starting point and
the pattern to be used. The second
makes a segment which is linked to an
existing segment, in a specified direction. Note that both constructors call
checkheapO to insure the object was
created on the heap ..
The first constructor (Figure 2) is
straightforward, but the second must
"wrap" the coordinate of the new segment around the display screen if it goes
off the edge. Also, the drawO method
isn't called in the second constructor,
since a new snake may need a new segment which crosses over an old segment
of itself. The calling function must check
the new segment before actually drawing it.
Recursive Member Functions
The next three functions are pure fun.
They recurse back down the snake

MICRO CORNUCOPIA, #53, May, 1990

#include "snake.hpp"
#include
#include II assuming ANSI terminal or ANSI.SYS on a PC
#include
#include II memset()
#ifdef
ZTC
II use disp package wi Zortech
#includ;-
#include
#endif
const height = 23; II screen height
const width = 78; II screen width
point pOint::random-point() {
return point(rand() I (RAND MAX/width),
rand() I (RAND:MAX/height»;
void * segment::operator new(size t sz) {
void * tmp = ::new unsigned cha;[sz]; II allocate with global new()
memset(tmp, 'x', sizeof(segment»; II fill with x's
return tmp;
void segment::checkheap()
if(pattern != 'x') { II check for memset pattern
fputs("can only create segments on the heap!", stderr);
exit(l);

segment::segment(point p, char ptrn)
II capriciously choose old heading:
: sp(p), previous(NULL), heading (LEFT)
checkheap();
pattern = ptrn;
draw 0 ;
segment::segment(segment * prv, direction dir)
: previous(prv), sp(prv->sp), heading (dir) {
checkheap();
pattern
previous->pattern;
II if(pattern++
'z' + 1) pattern
'a'; II pattern test ...
II create new segment in appropriate direction & wrap:
switch (dir) {
case UP :
if(--sp.y < 0) sp.y
height;
break;
break;
case DOWN :
if(++sp.y > height) sp.y = 0;
break;
if(--sp.x < 0) sp.x = width;
case LEFT :
case RIGHT : if(++sp.x > width) sp.x = 0;
break;

II
II

Recursive function to go back to the beginning of the
sequence of segments, to find the tail and remove it:
int segment::shed tail() {
if (previous
-NULL) {
delete this; II NULL means we're at the tail
return 1; II indicates to the next call 'up' that the tail was
found
}

recursive call until tail is found:
if(previous->shed_tail(»
II 1 means the call removed the tail
previous = NULL;
return 0; II means the call didn't remove the tail

II Recursive check to see if p crosses an existing segment:
int segment::cross over(point p) {
if(previous == NULL) return p == sp;
if(p
sp) return 1;
II recursive call to go to the end, or a cross point:
return(previous->cross_over(p»;

II Recursively redraw entire snake:
void segment::redraw tail() {
draw();
if(previous == NULL) return;
Continued on page 23

towards the tail. It's as if you're traversing a linked list, with each link an
object with its own member functions.
This is inspired by the much more
sophisticated "agent" example in Ravi
Sethi's Programming Languages-Concepts
& Constructs (Addison-Wesley, 1989). If
the rest of the book is anything like this
example, it's a gem.
The function segment::shed_tailO
must do three things. If the current segment is the tail (in which case the previous pointer will be NULL), then it
deletes the tail (thus the reason for forcing segments to be made on the heap).
However, if the tail was just deleted,
then the segment just ahead of the tail is
the new tail.
Thus, when the tail is deleted, the recursive function returns a 1 to tell the
next segment up that it should set its
previous pointer to NULL to indicate it's
the new tail. Finally, if nothing happens,
it returns a a to tell the next segment to
do nothing but return ain turn.

Continued from page 22
previous->redraw tail();

void segment::draw() {
#ifdef
ZTC
disp move (sp.y, sp.x);
disp:?utc(pattern);
#else
printf("\x1b[%d;%dH", sp.y, sp.x);
putchar(pattern);
#endif

II move cursor

}

void segment::erase()
#ifdef
ZTC
disp move (sp.y, sp.x);
disp:?utc(' ');
#else
printf("\x1b[%d;%dH", sp.y, sp.x);
putchar(' ');
#endif

// move cursor

}

snake::snake(point p, int size, char ptrn)
: length(O), maxlength(size) {
head = new segment(p, ptrn);

// Just mindlessly goes where you tell 'it:
void snake::crawl(direction dir) {
head = new segment(head,dir);
if(++length >= maxlength) {
--length;
head->shed_tail();

Don't Tread On Me!
segment::cross_overO takes its argu-

// find it's own way, tending to the current path:
void snake::slither() {
const int chance = 10; // 1 in 10 chance to change directions
direction old = head->path();
1/ Choose a new direction by first throwing the dice to decide
/1 whether to actually change direction. If so, choose a new
II direction randomly:
if (rand() I (RAND MAX/chance) != chance/2)
direction dir ; old; II change only if magic number is rolled
else {
dir = (direction) (rand()/(RAND MAX/4»;
head->redraw_tail(); 1/ occasionally redraw everything
segment * new seg = new segment(head, dir);
1/ check that-the new segment doesn't cross over the existing snake:
if(head->cross_over(new_seg->seg~oint(») {
delete new seg;
return;
new direction causes cross-over

// successfully found new direction
head = new seg;
head->draw() ;
if(++length >= maxlength)
--length;
head->shed_tail();

static void pause (int rate)
for (int i = 0; i < rate; i++)

Unique
Programmer's
Books & Tools for
MS-DOS & OS/2
Professional function
libraries and advanced books
by Michael J . Young. Create
high-performance MS-DOS
programs, TSRs, OS/2
programs, and Presentation
Manager applications.
Items not available
anywhere else!
C and assembly language!

main(int argc, char * arqv[]) {
if( argc < 4 ) {
fputs("usage: snake num of snakes size factor pause rate\n"
"CTRL-BREAK to quit \n", stderr); exit(l);
}

printf("\xlb[=7l"); // turn off ANSI line wrap
#ifdef _ZTC_

Continued on page 24

Please call or write for
free catalog:
Young Software Engineering
20 Sunnyside Avenue, Suite A
Department MC01
Mill Valley, CA 94941

415/383-5354
Reader Service Number 207

MICRO CORNUCOPIA, #53, May, 1990

ment, a point, and checks to see if it's
equivalent to the segment's point. If so,
it returns 1 to indicate yes (which means
the point isn't a good choice). If not it
calls previous->cross_overO to check
back until it finds a crossover point or
the tail.
segment::redraw_tailO calls drawO,
then recursively calls itself until it encounters its tail, thus redrawing the entire snake. drawO and eraseO are the
only functions that output anything. So
if you want to adapt the program to a
new system, you only have to change
these (and a couple of minor places in
main()).
The snake" Class
The snake (Figure 1) is quite simple.
A snake consists· of a pointer to a segment called head, and numbers indicating the current length and maximum
length. A single constructor (Figure 2)
creates a new segment for the head, and
initializes length and maxlength. After
that, the snake will crawlO in any direction you say (for the voice command
system).
snake::crawlO simply assigns head to
a new segment made from the previous
head (thus chaining the snake together).
If the snake is too long, it sheds its tail.
II

Chance Slithering
snake::slitherO is a more interesting
function. It "throws dice" using the
ANSI ·C library function randO, which
generates random numbers, to decide
whether to change directions. Note that
RAND_MAX is a constant defined in
stdlib.h which is the maximum value
randO will return. Thus the statement:
if (rand() I
chance/2)

disp open () ;
disp-move(O,O);
disp- eeop () ;
#else 71 ANSI terminal
printf("\xlb[2J"); II clear screen
#endif
II seed the random number generator:
time_t tnow;
time (&tnow);
srand(tnow);
II Create a whole nest of snakes:
const num_of_snakes
atoi(argv[1]);
const size factor
atoi(argv[2]);
const pause rate
atoi(argv[3]);
snake ** snakes
new snake*[num of snakes];
for(int i
0; i < num of snakes; i++)
snakes[i]
new snake(point::random~oint(),
#if 0 II for randomly-chose sizes:
rand() I (RAND MAX/size factor), i + '0');
#else II for fixed si~es: (avoid the bell)
size_factor, «i + 1) % 254 == 7) ? 1 : (i + 1)
#endif
#ifdef
ZTC
while(Tkbhi t () )
#else
while (1) {
#endif
for(i = 0; i < num of snakes; i++) {
snakes[i]->slither();
pause(pause_rate);

=
=
=

254);

•••
Figure 3-PAUSE.C : Portable function for delays
#include

1* This is a pure ANSI C function, so it's portable: *1
void pause (int hundredths) {
clock t start time = clock(); 1* current time *1
while«(clock() - start time) * 100)/CLOCKS PER SEC < hundredths)
; II wait in this loop for the delay time
-

•••

(RAND_MAXI chance) !=

creates a random number between 0 and
chance, and tests to see if it's equivalent
to chance/2. Thus, in a perfect distribution you have a one in chance probability the statement will be true.
Notice that normally you would
think like this: randO/RAND_MAX returns a number between 0 and 1, so just
multiply that by chance to get a number
between 0 and chance. But this would be
wrong. Since the compiler will do integer division, randO /RAND_MAX will
almost always return O!
Thus you must force RAND_MAX to
be divided by chance first to insure integer division doesn't mess up the an-

Continued from page 23

MICRO CORNUCOPIA, #53, May, 1990

swer. This logic works in any situation
(there are others in SNAKE) where you
generate a random number in a particular range.
If the above "if" statement is true,
snake::slitherO will slither the snake in
the same direction by using the old
direction. (Thus you can make the
snakes do more twists and turns by reducing chance.) If it's false, generate a
random number between 0 and 4 and
cast it into a direction (to indicate one of
the directions in segment).
Note that C++ 2.0 forces you to cast
an integer into an enum explicitly; it
won't just allow you to sneak it by-better type checking!
Next, we create a new segment, but

we don't assign it to the head-yet. First,
we test to insure that the new segment
won't cross over an existing segment. If
it does, we delete' the segment, and the
function returns without doing anything. The next call to slitherO will try
again with a new direction.
When running the simulation, you'll
see that sometimes a snake will coil in
upon itself and be unable to move. (If
you run the simulation long enough, all
snakes will eventually end up this way.)
A Flexible mainO
In The Tao of Objects (a book I am
coauthoring with Gary Entsminger), and
we maintain that one of the fundamental
tenets of object-oriented programming is

"dynamic programming style." In other
words, we want to determine the conditions (Le., the number and type of objects) of a program at runtime rather than
at compile time. By delaying system decisions, we allow new discoveries and adjustments to be made after we've
"finished" the system.
Not only does this allow us to modify
a program without reprogramming, but
it also allows the user to discover things
about the system the programmer never
even dreamed of (which may work
themselves back into a future version of
the program).
Whether designers like it or not, all
programs go through cycles of design,
use, redesign, reuse, redesign, etc. This
(we maintain) is one of the reasons existing design techniques don't work so
well-they provide good structure and
documentation (at least theoretically),
but not good design flexibility.
In mainO (Figure 2), you can see that
the user specifies the number of snakes,
the size of the snakes, and the speed of
the simulation by entering parameters
on the command line.
After the ANSI C random number
generator is seeded using the current
time, the arguments are picked off the
command line and an array of pointers
to snakes is created on the heap. (Astute
readers will notice I never free this
space-sloppy, but fairly safe since the
program releases the space when it
exits.)
The #ifdef statements determine
whether the sizes of the snakes are
chosen randomly or are a fixed size
(both options use size_factor).
I create a delay function (called
pauseO defined just before main()) using
the ANSI C time functions declared in
time.h (see Figure 3).
Get Distracted
I had a lot of fun playing with this
program-got mesmerized actually. If
nothing else, it's a terrific way to waste
CPU cycles.
The source code and SNAKE.EXE are
available through the usual Micro C
channels. The source- is also part of the
Using C++ source-code disk (see the
Revolution2 ad this issue).

Eventually, I combined the articles
into a self-published book, Computer Interfacing with Pascal & c. (Academic
Press will publish a very revised and expanded version called PC Interfacing with
C & C++ this fall). Next came Using C++,
and soon you'll see The Tao of Objects cowritten with our pal Gary Entsminger.
Writing for Micro C opened the doors
to other magazines (I'm the C++ editor
at The C Gazette), and speaking at SOGs
gave me the practice and nerve to speak
at "real" conferences. Most important, I
met Larry, Gary, and the gang at Micro
C. No question: writing that first article
was one of the best moves I ever made.
I started in the "brown wrapper"
days ("Hmmm ... is this Micro C or
Naked Volleyball Quarterly?"). The magazine has vastly improved in look and
style, and it seems like we've just hit our
stride. I'm sad to see it go, but perhaps
we've finished creating it. Maybe it
would have become an Institution rather
than a platform for experimentation
(Micro C never did figure out what it was
about.. .. ).
There are all kinds of reasons that
Micro C shouldn't have happened: it
wasn't commercial enough, not enough
market research, blah, blah, blah. It did
happen though-a rare and magical
fractal-chaotic event. It happened because we wanted it to happen, and more
important, because you wanted it to happen.
Farewell, and look for me in the computermags ....
Bruce Eckel is the author of Using C++

(Osborne/McGraw-Hill, 1989) and a member of the ANSI C++ committee. He's the
owner of Revolution2, a firm specializing in
C++ training and consulting.
Reference
Sethi, Ravi; Programming LanguagesConcepts & Constructs (Addison-Wesley,
1989).

•••

Farewell
I wrote my first article for Micro C
four years ago. I expected to be
swamped with consulting offers. I
wasn't, but I wrote for every issue since.
MICRO CORNUCOPIA, #53, May, 1990

Building A Two-chip Terminal
Motorola's 68HC705 Makes It Si111ple

Karl builds the world's simplest terminal
around one of Motorola's newest controller
chips.

his battery-powered RS-232 terminal has it all; it's cheap, easy to
build, uses readily-available parts,
and boasts the newest micro controller
technology. With this project, you can
get started in 68HC705 design while you
build a versatile piece of computer gear.
The terminal's features make it ideal
for remote and low-power use. The LCD
readout and CMOS micro controller
(MCU) keep the current drain to below
40 mAo Use of a special-purpose levelshifter IC allows RS-232 operation from
a single 5 Volt supply.
The 68HC705's on-chip serial port
and timer system help the terminal run
at 9600 baud while controlling the display and monitoring the keyboard. The
design is flexible enough to allow parts
substitution, if you can't find exactly
what I used in my project.

The Display
My screen is a two-line liquid crystal
display (LCD). You can find many formats of LCDs in most of the large mailorder ads. Although this project uses a
2-line by 16-character display, feel free to
use anything else you like. Just be sure
your display is compatible with the one
shown.
How do you determine compatibility? Look for an LCD that uses the Hitachi HC44780 LCD controller chip. This
is a surface-mount technology (SMT)
chip, soldered onto the back of the display. It usually appears with one or
more Hitachi HC44100 driver chips (also
SMT).
The HC44780 chip does more than
handle character generation. It contains
an 80-byte buffer. LCDs built with this

MICRO CORNUCOPIA, #53, May, 1990

chip can therefore display only a maximum of 80 characters.
But for many applications, a 2-line by
40-character (or 4-line by 20-character)
display will do. By shopping around,
you can find several different formats of
these LCDs. Examples include the Hitachi H2750 (1 row of 16 characters), Hitachi LM032L or Optrex DMC20215 (2 by
20), or Hitachi LM044L (4 by 20).
Mail-order suppliers known to have
carried LCDs recently include Alltronics,
TimeLine, and Digi-Key. Also check
your favorite surplus house; these displays are becoming common surplus
items.
Most LCDs include some form of
backlighting. The most common style
uses a strip of electroluminescent material (ELM), mounted behind the panel.
When you apply the proper AC voltage
to the two pins on the ELM, it emits a
soft blue or green light. This makes your
terminal easy to use at night or in low
light.
I added a special, three-terminal voltage converter designed for use with
ELM. This module changes 5 Volts DC
to 200 Volts AC at about 500 Hz, giving
my display a soft blue backlight.
You might have to scrounge pretty
hard to come up with such a converter;
mine came off a discarded laser-tag arcade game. Half the size of an ice cube,
the converter's case carries the part
number NEL-D32-45, but no manufacturer's name.
I have also seen a similar part offered
in the Digi-Key catalog. The terminal
works fine without the converter, of
course, but backlighting adds a nice
touch.
The Keyboard
I chose a surplus Cherry keyboard
(very old, it uses TTL chips). This unit
provides a IS-pin connector for hooking
up to a computer. When you press a key,

the ASCII code associated with that
character appears on seven pins of the
connector, while a strobe pin goes from
ground to +5 Volts, signaling there's
valid data.
This means I only need ten wires to
hook up my keyboard; seven data bits,
one strobe line, power, and ground. I
can use a single eight-bit input port to
handle all my keyboard data. (Of course,
the TTL chips make this keyboard a real
power hog; oh well, it only cost $4.)
With this arrangement, I get all the
control codes, all upper-case letters,
special characters, and numbers; no
lower-case and no PC function-key
sequences. Still, it'll serve my purpose.
There are always other options. For
example, you could use a surplus, unencoded switch matrix. This type of keyboard consists of a grid of keys witches,
with the wires tied to each row and
column brought out to a connector.
Pressing a key shorts a specific row to a
specific column; your software just scans
each row (or column) until it detects a
closed circuit. You can then use the row
and column where the short occurred to
determine which key was pressed.
Though this sounds like extra software (it is), the scanned matrix keyboard
offers some benefits. Since the matrix
consumes no power, it makes a perfect
choice for low-power applications. The
surplus market seems glutted with a
variety of unencoded keyboards. If you
shop around, you will surely find just
the right size, shape and style. And the
price will be right; I picked up a brandnew unencoded keyboard from United
Products for just a few dollars.
You could also opt for a PC-style keyboard. The Keytronics 100-style keyboards are usually available; I recently
saw an ad from TimeLine that offered a
new XT keyboard for $15.
Unlike the other two keyboard systems, the PC units provide data for a

By Karl Lunt
th
2133186 Pl., SE
Bothell, WA 98012
(206) 483-0447

Wth this project,
you can get started
in 68HC705 design
while you build a
versatile piece of
computer gear.

pressed key via a serial line. A corresponding clock line tells the host computer exactly when each bit of the data
packet is valid. (Check out Bill Curlew's
article, "Building An IBM PC Keyboard
Translator," in the February IMarch
1990 issue of Circuit Cellar INK, for
details on PC keyboards.)
This synchronous transmission system means the keyboard can send (and,
in some cases, receive) data over just
three wires. At first glance, none of
IBM's three available formats will work
without additional logic. They just aren't
compatible with the 68HC705' s synchronous serial line (Motorola calls this
port the serial peripheral interface, or
SPI).
The problem lies with the number of
bits transmitted per keypress. Mode 1
(the oldest and simplest format) uses
eight data bits and one start bit for each
transmission. The other two modes use

additional bits and support bidirectional
transfers.
The SPI, however, can only accept a
transfer of exactly eight data bits; it does
not need nor expect a start bit. To hook a
PC keyboard to the SPI, you could add a
serial-to-parallel converter to the
68HC705. Or, you could put extra circuitry inside the keyboard to turn the
nine bits into two eight-bit packets.
Neither way was acceptable (I wanted
this project to take only two chips), so
the PC keyboard was out.
As I look that last paragraph over, I
realize there is (as always) another way.
Hooking the keyboard's clock line to an
input port line and the keyboard's data
line to the MCU's TCAP (timer capture)
line would do the trick nicely.
Following a keypress, the data line's
start bit going high-to-Iow would cause
an interrupt and transfer control to the
TCAP server routine. This routine

samples the clock line on the input port,
looking for a low-to-high transition.
Every time such a change occurs, the
routine reads the level on the TCAP line
to get the value for that bit-time. The
server could then construct the keycode,
wait for the stop bit to appear, and exit.
Okay, so you can add a PC keyboard
to this project and still use only two
chips.
The Maxim MAX232 Chip
I've built a couple of projects with
this chip and love using it. Being able to
run RS-232 from a single +5 Volt source
reduces circuit complexity, chip count,
and board size.
The Maxim chip works by generating
+8 and -8 Volts on-chip from the system's +5 Volt supply, using a chargepump technique. You hook up four 22
mfd capacitors for the charge-pump; the
chip gives you two output and two
input buffers that are RS-232 compatible.
You might run into some problems,
however. The chip does not supply a lot
of current to the RS-232 buffers. If you
try to push 9600 baud over a long cable,
you could get errors.
A new version of this chip, the
MICRO CORNUCOPIA, #53, May, 1990

MAX233, puts the capacitors on the
chip. This will further reduce the parts
count for your RS-232 interface. But this
chip suffers from the same low drive
current as its MAX232 brother. At least
one engineer I know has cautioned me
about using either chip to talk distances.
Having said this, I'll add that I have
yet to see any data errors directly due to
the Maxim devices. Plus, both enjoy a
wide following in the industry. I certainly intend to keep using them.
TheMCU
Consider what it takes to make an
RS-232 terminal. You need a bidirectional serial port good for 9600
baud and plenty of I/O lines to hook up
to your LCD and keyboard.
You also need a timer system for
properly clocking the control signals into
the LCD, an interrupt system for catching the high-speed data flow on the serial line, and enough RAM and EPROM
to hold your program.
Motorola's 68HC705 provides all this,
at a current drain of less than lOrnA.
Check out Micro C Issue #49 for an indepth look at the 68HC705. For now, I
will only discuss those aspects of the
MCV needed by the two-chip terminal
project.
You control the extensive array of
I/O and timing functions on the MCV
by writing 8-bit values into specially assigned memory locations, called registers. For example, you can use all eight
bits of I/O port A as outputs by storing
the value $FF into location $04, known
as DDRA (data direction register A).
By using a 4.0 MHz crystal, the MCV
can hit 9600 baud on the serial port. Although the project currently does not
support it, you could easily change baud
rates from the keyboard or a row of dip
switches. To alter baud rates, simply
write different timing values to two
baud-rate selector registers.
You must wait a prescribed amount
of time after each change of the LCD. I
use the MCV's 16-bit timer system to
generate 200 Ilsec time slices. The LCD
routines simply wait until the correct
number of slices (called tics) elapse
before sending the next command.
Since the terminal could receive a
stream of 9600 baud data, the MCV
must handle the incoming data immediately. The MCV's serial communication
interface (SCI) can generate an interrupt
upon receiving a character from the serial port.

28 MICRO CORNUCOPIA, #53, May, 1990

Figure l-LCD and Keyboard Interface to '705

68HC7~5

11
10
9
8

7
6

LCD

PA~

DB4

PAl
PA2

f[)

t::J

PA5

DB7

PA4

DBS

PA3

DBS

47pf ~ 1~

+5V

5.6M

5
Vee

~ 9

4 MHz
+5V

+5V
4

22f.1F
+

+5V

+
2 21JF
•

l~K

IRQ
2
30
29

RS-232

POl

PD0

+5V

.. 7

~~
7805 VIEW

OUT

GND

+SV

IN4~~1

l~IJF

9-3~V

-=
68HC7~5

12
13
14
15
16
17
18
19

.
---- ....
P2

PB0
PBl

PB3
PB4
PBS
PB6

P87

(15)

BIT 1

(14)

BIT 2

(13)

.
--. .•

PB2

KEYBOARD
BIT 0

BIT 3

(12)

BIT 4

(10)

BIT S

(11)

BIT 6

(9)

STROBE

(6)

7
8

•
•

+5V

~ 9 ••
.-----+5V

r---

GND

(4)
(8)

End diskette compatibility problems.
Call Emerald Microware.
CompatiCard I by Micro Solutions
This four drive universal floppy controller will let you
run up to 16 disk drives (4 per CompatiCard), including
standard 360K, 96 TPI, high density 1.2M, 8" (SSSD
or DSDD), and 720k/1.44M 31f2" drives. The CompatiCard I comes with its own MS-DOS driver, utility programs, and will let you boot on an XT (must be used
as a secondary controller on an AT or 386). Use it
with UniForm-PC for maximum versatility.
CompatiCard I Board. . . . . . . . . . . . . .. $ 119.95
CompatiCard I with UniFORM-PC. . . .. $ 179.95
8" drive adaptor board .............. $ 15.00
External drive cable set. . . . . . . . . . . .. $ 15.00

CompatiCard II by Micro Solutions
Two drive version of the CompatiCard, for the XT or
AT. Same drive support as the CompatiCard I except
no 8" or single density.
CompatiCard II . . . . . . . . . . . . . . . . . . .. $ 89.95
Special
CompatiCard II with
internal 1.2M or 1.44M drive. . . .. $ 199.95

***

CompatiCard IV by Micro Solutions
Meet the newest four drive controller in the CompatiCard family. This CompatiCard may be used as a
primary or secondary controller in almost any PC, AT,
or 386 System. Boot or use 360k, 720k, 1.2M, 1.44M,
or 2.88M, at any location in your system. The CompatiCard IV has a BIOS ROM on board so no external
driver software is required.
CompartiCard IV. . . . . . . . . . . . . . . . . .. $ 139.95

CP/M

II ~ ]
"

'MS.oOS

EJAP:- [1]
UniForm-PC

by Micro Solutions

Have you ever needed to use your CP/M diskettes
on your PC? Now you can access your CP/M files
and programs on your MS-DOS computer just as you
would a standard MS-DOS diskette. UniForm allows
you to use standard DOS commands and programs
right on your original diskette without modifying or
copying your files. UniForm-PC allows you to read,
write, format, and copy diskettes from over 275 CP/M
and MS-DOS computers on your PC, XT, AT, OR 386.
With UniForm-PC and the CompatiCard, you can use
5%" high density, 96TPI, 31f2" (720k/1.44M), and
even 8" drives.
UniForm-PC by Micro Solutions ...... $ 64.95
Also available for Kaypro, & other CP/M computers

zao Coprocessor Board
by Micro Solutions

Don't throw out all of those old, reliable CP/M programs, run them at LIGHTNING speed on your PC
or AT with the UniDOS 8MHz. Z80 coprocessor
board. And the UniDOS Z80 runs so smoothly and
transparently that you won't even be able to tell
whether you're running DOS or CP/M. UniDOS emulates most common computers and terminals such
as Kaypro, Xerox 820, Morrow, Osborne, VT100, and
many others. Supports all standard CP/M system
calls, and now works with MS-DOS version 4. Includes UniForm-PC.
UniDOS Z80 Coprocessor Card ...... $ 169.95

UniDOS
Megamate

by Micro Solutions

Apple .. MS-DOS
MatchPoint-PC by Micro Solutions
Apple II or NorthStar diskettes in your IBM? The
MatchPoint-PC board for the PC/XT/AT works with
your standard controller card to let you read and write
to NorthStar hard sector and Apple II diskettes on
your PC. INCLUDES UnIForm-PC program, as well
as utilities to format disks, copy, delete, and view files
from Apple DOS, PRODOS, and Apple CP/M disks.
MatchPoint-PC Board. . . . . . . . . . . . . .. $ 179.95

•

Copy II PC Deluxe Option Board
by Central Point Software
Have a copy protected diskette with a particularly
stubborn protection scheme? Would you like to be
able to read your Macintosh disks in your 31/2" internal drive in your PC or AT? Repair a disk that's
damaged, even between sectors? How about speeding up your hard disk backups (if you are using PC
Tools Deluxe)? The Copy II Deluxe Option Board can
help you do all of this, and more. A must for the
sophisticated user.
Copy II Deluxe Option Board ........ $ 139.95

CentralPoinl

Sof!J£~~~

CP/M .. MS-DOS
UniDOS

You don't have to be a computer expert to install this
attractive 3 1/2" external drive on your PC or AT. Just
plug the MegaMate controller board into any empty
slot, attach the drive cable, run the installation software, and you're ready to run 720k or 1.44M diskettes.
Megamate . . . . . . . . . . . . . . . . . . . . . . .. $ 329.95

Copy II PC by Central Point Software
Don't let a damaged copy protected diskette stop you
cold. Copy II PC lets you back up your master disks so
you can keep going even when your key disk can't.
Copy II PC . . . . . . . . . . . . . . . . . . . . . .. $ 24.95

by Micro Solutions

If you have a fast machine or have a V20 chip installed, you may not need to use a card slot to run
your CP /M programs. Run 8080 code directly on the
V20, or use emulation mode for Z80 programs.
UniDOS by Micro Solutions. . . . . . . . .. $ 64.95
UniDOS w/UniForm & V20-8 chip ..... $ 135.00

MatchMaker

by Micro Solutions

Now you can copy your Macintosh diskettes right on
your PC/XT/ATwith the MatchMaker. Just plug your·
external 3 1/2" Macintosh drive into the MatchMaker
board and experience EASY access to your Mac diskettes. Includes programs to read, write, initialize, and
delete files on your single or double sided Mac disks.
MatchMaker Board. . . . . . . . . . . . . . . .. $ 139.95
MatchMaker w/External Mac Drive ... $ 325.00

PC Tools Deluxe VS.S
by Central Point Software
This is one of the great bargains in MS-DOS utility
software. But with so many features built in, we think
that many people are overlooking the REAL value
in PC Tools: DISASTER RECOVERYI Sure, the shell
is great for copying, viewing, editing, and deleting
files, and the desk top environment is nice for its
appointment calendar, note pad, phone dialer, calculators, and ASCII table. The fast hard drive backups
can't be beaten by any other program, and the file
unfragmenter speeds up hard drive accesses. But
where else can you get all of that along with UNDELETE, REBUILD, and UNFORMAT? Have you
ever entered "ERASE ..... and realised when all was
said and done that you were in the wrong directory?
How about those dreaded messages from CHKDSK,
like "File allocation error ...... ? The very first time you
recover your missing files will make you a believer.
Don't wait until it happens.
PC Tools Deluxe V5.5 . . . . . . . . . . . . .. $ 99.95

EMERRLO
MICROWRRE

P.o. Box 1726
(503) 641-8088 Beaverton, OR 97075
Call or write for our complete catalog of
software, parts, accessories and
complete repair services for the
Kaypro, Xerox 820, and IBM PC/ AT.

Reader Service Number 10

VISA and Mastercard accepted. Please include $6.00
shipping and handling, $8.50 for COD, UPS-Blue or
RED Label additional according to weight. Prices
subject to change without notice. Please include your
phone number with all correspondence.

MICRO CORNUCOPIA, #53, May, 1990

I use the SCI's interrupt to grab each
character as it comes in, then stash the
data in a circular queue. Eventually the
mainline code gets around to pulling the
character from the queue and processing
it.
Data from the keyboard arrives much
more slowly. (In my somewhat advanced years, I can no longer type much
above 4800 baud.) The MCU simply
glances from time to time at the keyboard port. If the line tied to the keyboard's strobe signal shows that data is
available, the MCU grabs it.
Data from the terminal to the host
computer also uses the SCI, but the program has no need to send anything in a
steady stream. It (currently) sends a
single character to the computer each
time a key is pressed, so this function
hardly requires interrupts.
On To The Software
I've divided the assembly language
code into several major functions. The
LCD routines initialize and update the
display. The output compare (OCMP)
server handles the 200 Ilsec tic interrupts. The SCI server takes care of incoming data from the computer. The
mainline code does everything else, such
as processing keyboard data and keeping the display looking pretty.
I based my LCD library on information from Ed Nisley's article, "The True
Secrets of Working with LCDs," from
the April/May 1989 issue of Circuit Cellar INK. Ed's explanation of the LCD's
inner workings really helped.
Because he did all his software in C,
Ed didn't concern himself with the requited delays between LCD commands;
the C code ran slow enough all by itself.
Adding Delays
Writing in assembly language, however, means adding delays. These waits
can be significant. The initialization
ritual, for example, calls for delays as
long as 5 msec; the shortest delay (between sending two visible characters to
the display) still takes 120 Ilsec. (Note:
All LCD commands have a required
minimum delay before a subsequent
command will be correctly processed.
However, there is no maximum delay.)
I solved this delay problem with the
OCMP server and a global variable
called WAIT$.
The MCU's 16-bit timer continually
counts down from $FFFF, through
$0000, and starts over again at $FFFF.

MICRO CORNUCOPIA, #53, May, 1990

The output compare register (OCR) can
hold an arbitrary 16-bit value. When the
value of the free-running timer matches
the value stored in the OCR, the MCU
generates a timer interrupt.
To generate a fixed-length time slice,
or tic, simply figure out how many
timing counts correspond to the delay
you need, add that value to the current
value of the timer, and store the sum in
the OCR. After the proper number of
counts, the timer will reach the value
you stored in the OCR and trigger an interrupt. Your interrupt service routine
then calculates the sum for the next tic,
updates the OCR. ...
In my case, I needed to count off the
correct number of 200 Ilsec tics. So my
interrupt routine does more than just set
up the next interrupt count. It also
checks the value in WAIT$; if that value
is not yet zero, the routine decrements it.
Therefore, any routine that must wait
(for example) 400 JlSec simply loads a
value of 2 (to wait for two tics) into
WAIT$, then hangs around until WAIT$
becomesO.
You will notice my software only
uses the A and X registers. That's all you
get! The tiny programming model on
this chip really makes you work to get
the most out of your code.
More About LCDs
I connected my LCD to the terminal
board using a 4-bit interface. Since the
LCD interface also requires a register
select line (RS) and an enable line (E), I
could run the whole display with only
six I/O lines. Refer to the schematic for
details. Pay particular attention to the
data lines. Note that the low four bits of
the MCU's I/O port connect to the high
four data bits on the LCD.
Unfortunately, the LCD comes out of
reset in an 8-bit interface mode. You
must follow a carefully defined ritual in
setting the LCD to read the 4-bit interface. The code in LCDINIT does the job.
First off, I reset all the lines to the
LCD and load a 15 msec power-up delay
into WAIT$. I then write a sequence of
three commands, each resetting the LCD
to an 8-bit interface. (Yes, I know we
want a 4-bit interface, but you gotta do
this just like the book says, or it isn't
going to work.) Note the different delays
used following each command.
Next, I set up the display for a 4-bit
interface. The remaining commands fix
the type of display format, cursor control, font, and cursor address. If you

want to try other display characteristics,
check the Hitachi manual or Ed's article
in Circuit Cellar INK for details.
The LCD accepts two types of data.
Commands, such as those discussed
above, must be written to the display
with the RS line low. Data, such as displayed text, must be sent while RS is
high. To make. these operations easier, I
included LCDCMD and LCDCHAR.
These routines use a common section
of code, found at LCDCHARl. Each sets
up the RS line as needed, loads an appropriate delay value into the AR, and
drops into the code at LCDCHARI.
From here, the byte gets divided into
two 4-bit nybbles. Calls to LCDOUT
then transfer the nybbles (MSB first) to
the display.
LCDSTRING simply sends each character of a null-terminated string to the
display. The only tricky bit involves the
load indexed, indirect addressing mode
of the 68HC70S; the chip doesn't have
such a mode.
To provide this essential addressing
mode, I resorted to self-modifying code
(actually, a RAM-based subroutine,
which is just as bad).
The initialization code for the terminal program builds up a two-instruction
subroutine called LDAIND by writing
two opcodes into low RAM. Address
LDAIND holds a $D6 (load accumulator, indexed, with a 16-bit offset), while
address LDAIND+3 holds a $81 (return
from subroutine). The two bytes at
and LDAIND+2 get
LDAIND+l
changed by another routine before calling LDAIND.
At run-time, a section of code stores
the 16-bit address of a null-terminated
string into the two bytes at LDAIND+l.
It clears the XR (X index register), then
calls LCDSTRING. Since the LDAIND
subroutine is all set up, LCDSTRING can
simply call the routine to get the character in the string pointed to by the XR. It
then increments the XR, gets the next
character, etc., until it gets the null marking the string's end.
And Now The Hardware
I built the prototype terminal on a
Radio Shack experimenter's card (276IS8A). You can find nearly all the parts
in a Radio Shack or Active Electronics
catalog; write these companies for a catalog if you don't have one already.
The only part you might have trouble
with is the 68HC70S MCU. You will
need to contact a Motorola distributor

C CODE FOR THE PC
source code, of course

NEW!

Upgrade!
NEW!

Cheaper!

NEW!

MS-DOS File Compatibility Package (create, read & write MS-DOS file systems on non-MS-DOS computers)
CQL Query System (SQL retrievals on B-trees plus windows) . . . . . . . . . . . . . . . . . . . .
GraphiC 5.0 (high-resolution, DISSPLA-style scientific plots in color & hardcopy) . . . . . . . . . . .
PC a.mes (Aspen, Software, System V compatible, extensive documentation) . . . . . . . . . . . . . . . . . . . . . . . . .
CoData Manager (Object-oriented data management, persistent objects from runtime definitions, network and entity models) . . . . .
MEWEL (extensible window and even~ library by Magma Software; message-passing & object-oriented; SAA-compatible; dIalog editor)
ThrboTEX (Release 2.0; HP, PS, dot dnvers; CM fonts; LaTEX; MetaFont) . . . . . . . . . . . . . . . . . . . . . . . . . . .
db_File & db_Retrieve by Raima (B-tree and network database with SQL query and report writer; multi-user $475) . . . . . . . . . .
Greenleaf Communications Library (interrupt mode, modem control, XON-XOFF; specify compiler)
CDirect (multi-user hashed file manager; variable length fields, binary or ASCII data, alternate keys). . . . . . . . . .
SilverComm (complete asynchronous communications library) . . . . . . . . . . . . . . . . . . . . . . . . . .
Wendin-DOS Plus (self-bootable, multitasking, multiuser MS-DOS replacement; includes XTC editor) . . . . . . . . .
QuickGeometry Library (1~rge col1e~tion of, mathematics, graphics, display & DXF subroutines for CAD/CAM/CAE/CNC)
CBme (B+tree ISAM dnver, multIple vanable-Iength keys) . . . . . . . . . . . . . . . . . . . . . . . . . .
ThrboGeometry (library of routines for computational geometry, Version 3.0). . . . . . . . . . . .
AT BIOS Kit (roll your own BIOS with this complete set of baSIC input/output functions for ATs) . . .
WKS Library VersIon 2.01 (C program interface to Lotus 1-2-3, dBase, Supercalc 4, Quatro, & Clipper)
C Generator (generates C code to read & write file records defined with C structure syntax) . . . . . . .
OS/88 (industrial-strength U .. x-like operating system, many tools, cross-development from MS-DOS) . . .
Cephes Mathematical Library (over 100 high-~uality, double-precision scientific functions) . . . . . . . .
ME Version 2.1 (programmer's editor with C-hke macro language by Magma Software; Version 1.31 still $75)
Vmem/C (virtual memory manager; least-recently used pager; dynamic expansion of swap file) . . . . . . . . . . . . . . . . . .
Thrbo G Graphics Library (al1 popular adapters, hidden line removal) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rogue Wave Vector & Matrix Classes (inc. C++ overloadin~s for standard operators, matrix inversion & FFT; Zortech or GNU C++)
Power Search by Blaise Computing (regular-expression compIler; generates machine code on the fiy). . . . . . . . . . . . . . . .
Instal12.3 (automatic installation program; user-selected partial installation; CRC checking) . . . . . . . . . . . . . . . .
TE Editor beveloper's Kit (ful1 screen editor, undo command, multiple windows) . . . . . . . . . . . . . . . . . . . . .
Hold Eve~hing (spawn new programs; swap parent to EMS or disk; handles video, interrupts, & environment; returns error level)
B-.St,rings (dy~amlc string han~ling; c~t, copy" paste, se~rch, user input, etc.; non-fragmenting memory management) . . . . . .
MInIX Operatmg System (VersIon 1.3, U .. x-hke operatmg system, Includ~s manual) . . . . . . . . . . . . . . . . . . . .
PC/IP (CMU/MIT TCP/I}> for PCS; Ethernet, Appletalk & NETBIOS dnvers, RVD, gateways) . . . . . . . . .
B-me Library & ISAM Driver (file system utilities by Softfocus) . . . . . . . . . . . . . . . . . . . . . .
The Profiler (program execution profile tool) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
QC88 C compiler (ASM output, small model, no longs, Hoats or bit fields, 80+ function library) . . . . . . . . . . . . . .
Booter ToolkIt (HOrpy disk bootstrap routines, DOS file system, light-weight multitasking, windows, fast memory management)
Otter 1.0 (beautifu theorem-prover by Bill McCune; includes manual & two books by Wos; complete starter kit). . . . . . .
JATE Async Terminal Emulator (includes file transfer and menu subsystem) . . . . . . . . . . . . . . . . . . . . . .
PowerSTOR (upto a gigabyte of heap space on extended memory, expanded memory, and/or hard disk) . . . . . . . . . .
MultiDOS Plus ~DOS-based multitasking, intertask messaging, semaphores) . . . . . . . . .
HY -PHEN-EX a hyphenator for American English with over 4,800 rules) . . . . . . . . . .
Makx ~macros, a I languages, built-in rules) . . . . . . . . . . . . . . . . . . . . . . .
evalU C function to evaluate ASCII infix expression string; 17 built-in functions) . . . . . . .
XT BI S Kit (roll your own BIOS with this complete set of basic input/output functions for XTs) . .
Professional C Windows (lean & mean window and keyboard handler). . . . . . . . . . . . . .
Heap Expander (virtual memory manager using expanded memory, extended memory, and disk space) . . . . . . . .
Quincy (interactive C interpreter) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Symtab/Ptree (general-purpose symbol table/parse tree construction and management package; specify Symtab or Ptree)
Coder's Prolog (Version 3.0; inference engine for use with C programs) . . . . . . . . . . . . . . . . . . . . .
Async-Termio (Unix V compatible serial interface for MS-DOS; stty, iocti, SIGINT, etc.) . . . . . . . . . . . . .
Backup & Restore Utility by Blake McBride (multiple volumes, file compression & encryption) . .
Floppy TAR (TAR backup and restore on MS-DOS devices; direct access to non-standard deVIces)
SuperGrep (exception~lly fast, revolutionary tex~ searching algorit~m; also searches sub-directories)
OBJASM (convert .0bJ files to .asm files; output IS MASM compatIble) . . . . . . . . . . . .
Multi-User BBS (chat, mail, menus, sysop displays; does not include Hayes modem driver) . . . .
LaplaceB (LaPlace polynomials, real and complex) . . . . . . . . . . . . . . . . . . . . .
CLIPS (rule-based expert system generator, Version 4.3; advanced manuals available) . . . . . . . . . . . . . .
Pascal P-Code Compiler & Interpreter (full ISO standard Pascal) . . . . . . . . . . . . . . . . . . . . . . .
PCHRT (40 functions to manage multiple microsecond timers; generate precision delays; insert timers on any interrupt)
Kier DateLib (al1 kinds of date manipulation; translation, validation, formatting, & arithmetic) . . . . . . . . .
Fortran-to-C 'franslator by Polyglot (Fortran-IV -like Fortran to ugly C; plan to adapt to your own Havor of Fortran)
DES i?ncryption ~ Decryption (25,<)() bits/seco~d on 4.~7 MHz PC for on-the-Hy encryption at 2400 baud). . . . .
F1exLlst (doubly-hnked hsts of arbItrary data WIth multIple access methods) . . . . . . . . . . . . . . . . .
Virtual Memory Manager by Blake McBride (LRU pager, dynamic swap file, image save/restore) . . . . . . . . . . . . . '.
Heap I/O (treat all or part of a disk file as heap storage) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Bison & BYACC (YACC workalike parser generators; documentation; no restrictions on use of BYACC output) . . . . . . . .
PC-XINU (Comer's XINU operating system for PC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
RXC & EGREP (Regular Expression Compiler and Pattern Matching; RXC makes finite state machine from regular expression) .
REGX Plus (search and replace string mampulation routines based on regular expressions) . . . . . . . . . . . . . . . . . .
CCALC (handy extended-precision calculator; real and complex models; many built-in functions) . . . . . . . . . . . . . . .
GNU Awk & Diff for PC (both programs in one package) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6-Pack of Editors (baker's half:dozen publi~ domain editors for use, study & hacking; includes microEmacs 3.10 & Stevie, a vi clone)
Crunch Pack (14 file compressIOn & expansIon programs) . . . . . . . . . . . . . . . . . . .
PC-MAIL (UtlCP mailer by Wietse Z. Venema; send, receive, and manage UUCP mail) . . . . . .
FLEX (fast lexical analyzer generator; new, improved LEX; official BSD Version 2.1 with docs) . . .
List-Pac (C functions for lists, stacks, and queues) . . . . . . . . . . . . . . . . . . . . . .
Using C++ Library (the code from the book by Bruce Eckel and then some; Zortech 2.0 compatible)
A68 (68000 cross-assembler) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
XLTMacro Processor (general purpose text translator) . ... . . . . . . . . . . . . . . . . .

Data

$750
$325
$325
$290
$250
$250
$250
$245
$225
$210
$210
$180
$170
$165
$160
$160
$155
$150
$150
$150
$140
$140
$135
$125
$120
$120
$115
$105
$105
$105
$100
$100
$100
$90
$85
$80
$80
$80
$80
$75
$75
$75
$75
$70
$65
$60
$60
$60
$55
$50
$50
$50
$50
$50
$50
$50
$50
$45
$45
$40
$40
$40
$40
$40
$35
$35
$35
$30
$30
$30
$30
$30
$25
$25
$25
$25
$20
$20

Mo~
Mo~
Mo~

Pronunciator (150,000 words & phrases encoded with full IPA pronunciation & emphasis points; 900 distinguished by part-or-speech)
$160
Part-of-Speech (200,000 words and phrases described by prioritized part(s)-of-speech) . . . . . . . . . . . . . . . . . , . .
$120
Hyphenator (150,000 words fully hyphenated/syllabified) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
$105
Mol:1; Words (500,000 words & phrases, 9,000 stars, 15,000 names) . . . . . . . . . . . . . . . . . . . . . . . . . . .
$65
Smithsonian Astronomical Observatory Subset (right ascension, declination, & magnitude of 258,997 stars) . . . . . . . . . .
$60
U. S. Cities (~a~es & longitude/lat.itude of.32,OOO U.S. cities and 6,000 sta~e boundary points) ~ . . . . . . . . . . . . . . .
$35
The World DIgItIzed (100,000 !ongltud~/IatItude ofwo~ld country 1?oundanes) . . . . . . . . . . . . . . . . . . . . . .
$30
KST Fonts (13,200 characters In 139 mlXed fonts: SpeCIfy 'lEX or bItmap format) . . . . . . . . ; . . . . . . . . . . . .
$30
Interactive Computer Ephemerii (high-precision moon, sun, planet & star positions; USNO (no source) & Downey 4.8 (C source)) . . .
. $30

The Austin Code Works
11100 Leafwood Lane
Austin, Texas 78750-3409 USA
Free surface shippmg for cash in advance

info@acw.com
For delivery in Texas add 7%

Voice: (512) 258-0785
BBS: (512) 258-8831
FAX: (512) 258-1342
MasterCard/VISA

Reader Service Number 4

MICRO CORNUCOPIA, #53, May, 1990

directly for this chip, as I don't think it is
in the mail-order houses yet. Almac
Electronics (206-643-9992) in Bellevue,
Washington, will handle the parts mailorder in singles; price fO'1- the -S version
of this part (with the EPROM window)
is about $18.
Be sure to get a copy of Motorola's
HC705 CB Technical Data book, literature
number MC68HC705C8/D. Besides all
the technical info on using the 'HC705,
this book includes a full schematic for
building a programming board (piece 0'
cake). It also includes instructions on
using the programming board.
You can use just about any wiring
technique you want for this project. I
used point-to-point soldering with 30
AWG wire; wirewrapping would work
also.
By the time you read this, I will have
a
general-purpose
experimenter's
printed wiring board (PWB) for the
68HC705. The board will contain all the
basic circuitry needed to build a working MCU project, and could easily be the
only PWB you might need to develop an
application.
For example, this two-chip terminal
project would consist of a working ex-

STOMP OUT

EPROMMADRESS
+

'.
(

1,r

",'

' /

•

-.~,;:

•

'"
~

'Ij ,: ~ ~~ j:',;::r.-'ll,',\~l'

~ .... ~

.
"

The PROM KING emulates EPROMS, saving
both time and money during your development
cycle. Programmable in seconds via your PC
printer port or any computer RS232 port, it can
emulate most 27xxx devices.
• 8K-8M bit devices
• 8-256 bit downloads
• High speed download: • Easily expandable:
-Universal RS232
-4 EPROMS per unit
-PC printer port
-Up to 8 units
• Menu driven software • Also programs like
a real EPROM
• Battery backup
$599 for 150nS units with 256K bits
Ask for pricing of other options
- - - - - Made in USA b y : - - - - -

TI¥IXEL lABS INC.
BOX 239, RONKONKOMA,NY'11779
516·737·5147

Reader Service Number 178

MICRO CORNUCOPIA, #53, May, 1990

perimenter's PWB connected to a keyboard and LCD.
After you get your terminal wired up,
you need to assemble the source code. I
use a PC/XT clone running Motorola's
ASS Freeware cross-assembler. Borland's Sidekick handles my editing, and
I burn the finished code into the MCU
using Motorola's M68HC05PGMR programmerboard and PROG7 support
software.
You can download both PROG7 and
ASS from the Motorola Freeware BBS
(512-891-3733; 8 data, no parity, 1 stop).
Motorola uses the BBS to distribute free
cross-assemblers (available for both the
PC and Mac), working source code for
MCU designs, technical updates, and
marketing announcements. Dial up and
take a look.
Even if you don't yet have a keyboard, you can still hook up the display
and apply power (use a DC source from
+7 to +25 Volts). Your LCD should show
the power-up announcement "Working ... "
If you don't get the proper display,
use an oscilloscope or logic probe to
check pin 35 of the MCU. My software
toggles this line each tic; you should see
a 50% duty-cycle square wave that
changes state every 200 Jlsecs. If you see
this signal, you likely have an error in
your LCD connection.
If you don't see this signal, check
your board's power supply and oscillator circuitry; you may not be getting +5
Volts to the chip, or your crystal may not
be oscillating.
After you get the display working,
wire up your keyboard using the schematic as a guide. The software expects
you to supply seven data bits (on port B,
bits 0-6) and a positive-going strobe
(port B, bit 7); you can use any keyboard
that will supply these signals. Note that
the circuitry does not supply +12 Volts;
keep that in mind when you hit the surplus stores.
With your keyboard installed, temporarily short pins 2 and 3 of the RS-232
connector. When you apply power and
press a key, the display should echo that
key. If so, you have a working terminal!
Time To Tinker
Since everything about this terminal
can be found in the software, you can really customize this project. Consult the
Hitachi manual for the low-level LCD
commands if you want to change the
shape/blink of the cursor. The Maxim

chip has a pair of unused RS-232 buffers;
you could hook these up to I/O pins on
the MCU and provide hardware handshaking on the serial line.
Since a keyboard character goes
through the MCU before it gets shipped
out the serial port, you can easily imbed
function codes or signature keys into the
software.
This design leaves an entire 8-bit port
vacant; how about hooking up relays,
lights, a piezo-beeper (for a control-G
bell code), or other control devices? You
could even hook up an external A/D
converter and have a stand-alone en. vironmental control station. The station
could then download data over the serial port to a host computer, while the
keyboard lets you run programs on the
host for evaluating data.
Anyone doing field work would
probably find a use for this terminal.
How about hooking up some rechargeable batteries (or a solar panel, if you use
a switch matrix keyboard) and taking
the terminal outdoors?
If you run road rallies, you could
easily connect the terminal (complete
with backlighting) to a small rally computer stashed in the backseat, and do
some real digital navigation.
That's About It
This project was great fun. Everything I develop with the '705 makes me
want to try another design or two. If you
customize this project (as I am sure you
will), I'd like to hear what you have
done.
Active Electronics
P.O. Box 9100
Westborough, MA 01581
(800) 888-9939 (outside New England)
(800) ACTIVE6 (New England)
TimeLine Inc.
1490 W. Artesia Blvd.
Gardena, CA 90247
(800) 872-8878 (outside California)
(800) 223-9977 (California)
United Products
1123 Valley Street
Seattle, W A 98109-4425
(206) 682-5025

•••

SAYW!'!O greatT~IS.

The lightning-fast
screen generator.

The breakthrough
DBMS toolkit for Pascal
If you'd like to combine the raw power
~
and speed of Turbo Pascal or Micro~*'
soft's QuickPascal with the simplicity.? ~ 'X~S
and elegance of dBASE, Topaz is L~~~Z
just what you're looking for.
That's because Topaz was
1
specially created to let you
Y,. ~ ~
/.
enjoy the best of both worlds./.~,:¥ )
The result? You create com-OVf'/-...-'~
plete, truly dazzling applicat- : / ' _ (/-'1 ~
ions in a very short time. And no wonder.
Topaz is a comprehensive toolkit of
~~
dB~SE-like commands and functions,
"'\ \ \\
deSigned to help you produce
outstanding, polished programs, fast.
Check out these powerful features:

Whether you're a novice
programmer longing for
simplicity, or a seasoned
pro searching for higher
productivity, you owe it
to yourself to check out
(/'.")
Saywhat. You see, with
\ \
:;/";;. Saywhat, you can build
~ ~
beautiful, elaborate, colorcoded screens in minutes!
(( A ~~
/
That's right. Trulyfantastic
\ t-\
~ screens for menus, data
~
entry, data display, and
help-panels that can be
displayed with as little as one
line of code in any language.
"

c.5:.i]""'-/1
I

I '""

)/)

--j

Here's what you get:

moving bar menus!
• Easy-to-use, powerful editor lets
you create screens in a jiffy.
• Pop up your screens and menus
with one line of code in dBASE,
all the dBASE compilers, your
favorite BASIC, Pascal, or any

other language!
• Screen Library Manager.
• Generates runtime code.
• No runtime license or royalty fees.
• Comes with a 100 page manual,
plus dozens of sample programs
and free utilities.

0(C3

4L"t1

Software Science, Inc.
100 Valley Drive, Brisbane, CA 94005
_.

,., > ") ) T) ) ') ~ ) \)

_.___
I »)
J
\)

OlUf
-

• Data entry routines like SAY, GET,
PICTURE, RANGE, color selection, unlimited data validation.
• Open up to 10 DBF files, with up to
15 indexes with USE, SELECT,
SKIP, APPEND, PACK, INDEX
ON, SET INDEX TO, and FIND.
• No need to buy dBASE. CREATE,
BROWSE and REPORT utilities
included.
• Easily implement Saywh at and
Lotus-style moving bar menus.
• BROWSE or EDIT any DBF file
with just one line of code! Programmable and windowed too.

)

or Topaz, for any reason, return them wlthlO
30 days for a prompt, friendly refund.
Ct
r.
~

oJ \

..'>.Jo.,..........,--

Dealers: SAYWHAT?! and TOPAZ are
available from Kenfil Distribution.
and in Europe from

• Pick from windowed data or filenames with one line of code.
• Comprehensive Time & Date
math in 7 international formats.

____

/1-\'> '- \, I MONEY BACK GUARANTEE.
I~y/ I'I'~. Ifyouaren'tcomp1etelydelightedwithSa~h31t

• Over 220 routines all with easy-touse, dBASE-like syntax.

ORDER NOW.
YOU RISK NOTHING.
Thousands of satisfied customers
have already ordered from us. Why
not visit your dealer or call toll-free,
right now and putSaywhatand Topaz
to the test yourself? They're fully
guaranteed. You don't risk a penny.
Special limited-time offer! Save $26.
Buy Saywhat?! and Topaz together
for just $99 (plus $5 shipping and
handling, $10 Canada, $25 International, Calif. residents add 7%).
Visit your nearest dealer
or call toll-free:
800-468-9273
In California: 800-231-7849
International: 415-571-5019

• Design screens, windows, and

t:::';:;

,<:,,..!)\

•
•

~owledrfiudl~odeandreportgenerators

lnc u e .
C omes WI'th a compIete 320 page
manual, plus samp Ie programs to
get you started.

comFmiifOOd!
SOFTWARE

SCIENCE

INC.

Reader Service Number 129

MICRO CORNUCOPIA, #53, May, 1990

A Roundoff Roundup
When Rational Numbers Aren't Really Rational

Are rational numbers really rational?
Perhaps not. Does it matter? Most likely.
ary Entsminger, in his Tidbits
column "Faith in Numbers:
Chaos in Chaos" (Micro C #48,
July-August 1989), discusses the problem of roundoff error in recursive iterations of non-linear equations. Mr. Entsminger points out a problem facing
computer mathematics which, at least
for some applications, has a simple solution. This article makes several references to Mr. Entsminger's column, so
you may wish to reread it.
Mr. Entsminger defines a rational
number as a real number which is either
reducible to an integer, expressible as a
terminating decimal fraction, or expressible as a decimal fraction with a repeating
sequence of digits. This is a common and
workable definition, but not an exact one.
Take an ordered pair (x,y) of integers
and define an equivalence relation = such
that for any two ordered pairs (x,y) and
(a,b) with x= N then begin
Pr := true;
Done := true:
end
else if « N mod Divisor)
0) then
Done := true
else begin
Count := (Count + 1) mod 4;
If count = 1 then
divisor := divisor + 4
else divisor := divisor + 2;
end;
Prime := Prj
Endi {Prime}
Function GCD( A,B : longint) : longint;
Var
x,M,N : longinti
Done : boolean;
Begin
M := abs (A);
N := abs(B);
x := M;
If N <= x then x := Ni
Done := false;
While not done do
If «N mod X)=O) and «M mod X)=O) then
done:=true
else x := x - li
GCD := x;

Endi
Function LCM( A,B : longint ): longinti
Var
X,Y,N,M : longint;
Done : boolean;
Begin
M := abs(A):
N := abs(B);
Y := M;
IfN < Y then Y := Ni
X:=Y;
Done := false:
While not done do
I f «X modM)
0) and.( eX mod N) = 0) then
done := true
else x := x +Y;

LCM := X;

End;

Procedure Reduce ( Var X
Var
A
longinti
Begin

IfX[l]

0 then begin

X[l) := Oi
X[2) := 1;
, end
else begin
A := GCD(X{l),X[2)i
X[l) := X[l) div Ai
X[2] := X[2] div A;
end;

End;
Procedure Invert ( Var X Rational) i
Var
A : .longint;
Begin
If X[l) = 0 then
Writeln('DIVIDE BY ZERO ERROR!')
else if X[l) < 0 then begin
A := X[2)i
X[2] := Abs(X[l]);
X[l] := 0 - A;
end
else begin
A := X[l);
X[l] := X[2);
X[2) := Ai
end;
End;
Procedure Radd( Xl,X2: rational; Var Y: rational);
Var
Z,A : longint;
Begin
Z := LCM(Xl[2),X2[2)i
Y[2) := Zi
A := Z div Xl[2);
Y[l) := Xl [1) * Ai
A := Z divX2[2);
Y[l] := Y(l] + ( X2 [1) * A) ;
Reduce(Y):
End;
Procedure Rsub( Xl,X2: rational; Var Y: rational);
Var
Z,A
longint;
Begin
Z := LCM(Xl[2],X2[2);
Y(2) := Z;
A := Z div Xl[2];
,Y[l) := Xl(l] * Ai
A := Z div X2[2];
Y[l) := Y[l) - (X2(l) * A);
Reduce(Y)i
End;

rational) ;
Procedure Rmult(Xl,X2: Rational; Var Y: Rational);
Var
A,B : Rational;

Continued on page 38

MICRO CORNUCOPIA, #53, May, 1990

;868Z:9991 »xref=<96817>«
:868Z:9994 »xref=<8&999>«
:COaversiOD ta~le

call

.:lV

.:lV
EV

is -.aiale-

xor

si fttil

:e68Z:8897
;e68Z:8889

Su~roath.e

;868Z:989b
;Loai rqister
;e68Z:898i

Orilil
.:lV
.:lV

lax

:»»> Conversion Section
les
Us
EVZX

repz
sti

,
•
-

.':'
w

xrer

,.,

4 ;
•

;868Z:8818

address

;e68Z:9912 >}xref=<86888>«
;coaversiol ta~le
b x . - - - - - - h :8682:9916
;9682:991h get 'Yte COUlt
cx Help
s t I!:::=======!II ;9682:981£
;Store AL at £5:(11]
;968Z:ee21
;Turn,On Ilterrupts
;e6e2:ee22
;e68l:ee25 DOS:1c-te~iaate
di Yard
Address

Continued from page 36
{Subtract D*l from 2}

Begin
A[l] := X1[1]:
A[2] := X2[2];
B[l] :=X2[1]:
B[2] := X1[2];
Reduce (A) ;
Reduce (B) ;
Y[l] :=A[l] * B[l];
Y[2] := A(2) * B[2);
Reduce (Y):
End;

x := d;

Procedure Rdiv(X1,X2: Rational; Var Y: Rational):
Var
X : Rational:
Begin
X := X2;
Invert (X);
Rmult(X1,X,Y);
End;
End.

•••
Figure 2-GAUSS.PAS
Program GAUSS;
Uses RATIONALS;
Var
A,B,C,D,E,F,X,Y,Z : Rational;
Ar,Br,Cr,Dr,Er,Fr,Xr,Yr,Zr : Real;
Begin
A[l] := 131;
A[2] := 1;
Ar := 131;
B[l] := 23:
B[2] := 1;
Br := 23;
C[l] := 31;
C[2] := 1;
Cr := 31;
D[l] := 7;
D[2] := 1;
Dr := 7:
E[l] := 27;
E[2] := 1:
Er :=.21;
F[l] := 67;
F[2] := 1;
Fr := 67:
{Using Rationals}{First divide 1 by a}
X :=·A;
Rdiv(a;x,y) :
a := y:
Rdiv(b,x,y);
b

:=y;

Rdiv(c,x,y);
c :=y;

MICRO CORNUCOPIA, #53, May, 1990

Rmult(x,a,y);
. Rsub(d,y,z):
d := z;
Rmult(x,b,y):
Rsub(e,y,z);
e := z;
Rmult(x,c,y);
Rsub(f,y, z):
f := z;
{Divide 2 bye}
x := e;
Rdiv(e,x,y) ;
e := y;
Rdiv(f,x,y) ;
f

:= y;

{Subtract b*2 from 1}
x := b;
Rmult(e,x,y):
Rsub(b,y,z);
b := z;
Rmult(f,x,y);
Rsub(c,y,z);
c := z;
Writeln('Using rationals we get ... ');
Writeln(a[11,'/',a[2],'X
',c[11,'/',c[2]):
Writeln(e[1],'/',e[2],,'Y = ',f[1],'/',f[2]);
writeln;
{Using reals}{Divide 1 by a}
xr := art
ar := ar / xr:
br := br / xr;
cr := cr / xr:
{Subtract D*1 from 2}
xr := ar * dr;
yr := br * dr;
zr := cr * dr;
dr := dr - xr:
er := er - yr;
fr := fr - zr;
{Divide 2 bye}
xr .- er:
dr .- dr / xr;
er .- er / xr:
fr
fr / xr;
{Subtract B*2 from 1}
xr := br * dr;
yr := br * er;
zr := br * fr;
ar := ar - xr:
br := br - yr;
cr := cr - zr;
Writeln(' Using REALS we get');
Writeln;Writeln(ar:S:9,' *x = ',cr:S:9);
Writeln(er:S:9,' *y = ',fr:5:9):
End.

•••

cant properties of the rational numbers is
that given any two rationals a and b,
there must be a rational z such that a< z<
b, I can never represent a complete range
of the rationals. In modeling natural systems, I do not need to.
The set of all numbers which exist in
nature is discrete. In making such a bold
claim, I feel compelled to defend it.
Using the most accurate measuring
instruments available, and the best techniques of extrapolation, you can measure
a quantity to only finite tolerance. You
will then be able to get two measures for
which no intermediate measure can be
found.
Consider Zeno's paradox. Zeno
asserts that if a turtle and a rabbit were
to run a race, and the turtle were given a
head start, the rabbit could never pass
the turtle. His reasoning is that when the
rabbit achieved the turtle's starting position, the turtle would have moved some
distance ahead. The distance between the
two, though less, would still be positive.
Continuing in this manner, Zeno says
that the turtle must always be some positive distance ahead.
This is true only if distance is infinitely divisible. Try this experiment
yourself with two objects moving at constant velocity, one faster than the other.
Give the slow one a head start and you
will see that the fast one does pass the
slow one. The distance between them reduces to a point where no smaller distance exists.
Mr. Entsminger's example of the
equation Nextx + RX(1 - X) yields values
between a and 1, representing some portion of the maximum population of
wolves. If the maximum number of
wolves is some number z and the population after any iteration is given by
z*Nextx, we see that values of Nextx of
minute enough significance are not useful. Suppose Nextx = (K(z + 1) + l)/(z +
1) for some rational K. We then have a
fraction of a wolf, which is not likely to
reproduce.
For many applications, all discrete
representations of the rational numbers
are equally flawed. Trigonometric equations, for example, rely on the irrational
number pi. The results obtained using
ordered pairs will be at least as inaccurate as those obtained using floating
point notation. But for many applications, representing rational numbers as
ordered pairs of integers will greatly increase accuracy.

•••

CITIZEN PRINTERS
Citizen
Citizen
Citizen
Citizen
Citizen

120D 120/25 CPS 80 COL
180D 180/35 CPS 80 COL
HSP500 300/66 CPS 80
HSP550 300/66 CPS 120
GSX140 200 CPS 24 PIN

$165.00
$180.00
$350.00
$475.00
$345.00

GOLDSTAR MONITORS
MBM-1210A 12" TTL AMBER
MBM-1401A 14" Flar Screen TTL
MCH-1420 14" EGA 640x350
MCH-1430 14" VGA 640x480

$ 90.00
$135.00
$345.00
$375.00

VIDEO CARDS
VIP Card VGA,CGA,EGA,MGA,
ANALOG AND TTL 800x560
ATI VGA Wonder 256 1024x768
ATI VGA Wonder 256 16 bit
Paradise Basic EGA 640x350
Paradise Basic VGA 640x480
Paradise VGA Plus 800x600
Paridise VGA Plus 16 Bit

$175.00
$225.00
$375.00
$ 99.00
$175.00
$199.00
$275.00

ATI

HARD DRIVE CONTROLLERS DOR XT & AT
WD WX1 Auto Config
PC/XT
WD WX2 Auto ConfigPC/XT
LONGSHINE LCS-6210D PC/XT
WD WAH AT 16 Bit 2 Hard Drive
WD 1003SM1 AT 2:1 Int.
WD 1003SM2 Hard/Floppy 2:1
WD 1006SM1 AT 1:1 8K Cashe
WD 1006MS2 Hard/Floppy 1:1
WD 1003SR1 RLL 2:1 Int
WD 1003SR2 RRL Hard/Floppy
WD 1006SR1 RLL 1:1 8K Cashe
WD 1003SR2 RLL Hard/Floppy

$ 65.00
$ 55.00
$ 45.00
$ 75.00
$ 99.00
$125.00
$125.00
$135.00
$125.00
$135.00
$135.00
$145.00

SEAGATE DRIVES AND CONTROLLER IN STOCK
Please call for current price.

**********

2400 BAUD MODEM **************
by Computer Peripherals
2400 Baud Internal and Software $85.00
2400 Baud External and Software $140.00
CASCADE ELECTRONICS, INC.
ROUTE 1· BOX 8
RANDOLPH, MN 55065
507-645-7997
Please ADD shipping on all Orders
COD Add $3.00
Credit Card ADD 5%
MN Add 6% Sales Tax Subject to change
Reader Service Number 15

MICRO CORNUCOPIA, #53, May, 1990

Initializing Variables In Turbo Pascal
Handling Initialization During The Link Rather Than At Run Time

This is hacking at its best. Dave obviously understands the limitations of Turbo
and he's sent along some fixes. Fun stuff.
he most recent versions of Turbo
Pascal (5.0 and 5.5), have a great
deal to offer in terms of flexibility,
code readability, and development
speed. However, because of the way the
compiler and its internal linker work,
and because of Borland's support for
only one memory model, it has some
real limitations.

Procedure Tables To Go
One of Turbo Pascal's annoying limitations is the compiler's inability to initialize arrays of procedures (procedure
tables) at compile time.
Procedure tables are easy in C. The external linker automatically initializes the
address values in an initialized array of
procedures (at link time after they have
been resolved). See Figure 1 for a simple
example. (I'm not a C programmer by
trade, so the example may not be elegant.)
If you were to translate this directly
into Pascal (see Figure 2) and then try to
compile it, Turbo complains about the
line that initializes the array of procedures.
So what do you do if you want to create a procedure table in Turbo Pascal?
Normally you have to waste some runtime code to assign each of the procedures to the array members (see Figure
3).

For an array of four procedures, this
doesn't add much to the run-time. But
how about 500 procedures? Then it adds
up. The .EXE file for the small interpreter
you just knocked together using procedure tables would be fatter than it should
be. Plus, I think it's kludgy to do all those

MICRO CORNUCOPIA, #53, May, 1990

By Dave Gwillim
159 Woodbury Road
Hicksville, NY 11801-3030
(516) 942-8697

One

begin
writeln('Procedure 3');
end;
{$F-}
const
Proc : array[O .. 3] of procedure = (ProcOiProcl,Proc2,Proc3)i
var
n : word;
begin
for n := 0 to 3 do
Proc[n);
end.

•••
Figure 3-PROCTABL1.PAS
var
Proc : array[O .. 31 of procedure;
n: word;

of Turbo
Pascal's annoying
limitations is the
compiler's inability
to initialize arrays
of procedures
(procedure tables)
at compile time.
assignments, especially when the linker
knows what the addresses are before it
creates the .EXE file.
Is there a way to make the compiler
handle this at compile time? Using only
Turbo Pascal, no. The compiler handles
the initialization of typed constants
before linking. At that time it isn't sure
where the procedures will be located.
However, by using a dash of assembly language, you can force the compiler'to handle most of the initialization
during the link, leaving only the initialization of a single pointer variable for the
run-time code. Figures 4 and 5 show one
way you can do this.
The fake procedure Procs is never
called (and never should be, unless you
feel like reaching for the red button). It's
simply used to export a pointer from an
external assembly language program.
Since you cannot declare any data PUBLIC in an external procedure and still
have Turbo Pascal see it, you have to
cheat.
Fortunately, cheating will make your
.EXE file smaller, especially when you
MICRO CORNUCOPIA, #53, May, 1990

have lots of procedures in your table. The
trade off is that you have to maintain a
PROCS.ASM file in addition to the Pascal
source code.
Typed Constants Without Limit,
Anyone?
We all know how valuable that 64K of
data segment for variables in Turbo Pascal is, right? Every typed constant (initialized variable) gets its bite (or two). Sometimes 64K seems downright tiny!
Of course, you can allocate all your
variables on the heap. But then you have
to initialize them with assignments to
constants embedded in your code or load
in a separate data file. Both approaches
add considerably to the size of your .EXE

me.

You don't have this problem in any
implementation of C that supports multiple data segments. So wouldn't it be nice
if Turbo Pascal could manage it? If only
you had some way to assign initialized
variables to their own segment. Perhaps
you wouldn't have to switch to C!
Using a little assembly language, you
can do this in Turbo Pascal. Using only 4
bytes of that valuable Turbo Pascal variable space you get access to as much as
64K of initialized data. This trick turns
code space into initialized data space. See
Figures 6 and 7 for a way to do this.
Note that to get 64K of initialized data
space using this technique, you would
have to make the external procedure containing the data the sole member of a
separately compiled unit. And do it so
that virtually none of the code space
would be taken up by executable code.
When unitizing, you should also include the pointer variable in the unit and
use the run-time initialization capability
(placing an assignment statement between a BEGIN and END in the implementation part) to set up the pointer
variable. You only use the unit to have all
the initialized data available. Using the
unit approach is also a good way to hide
the fake procedure _DataSegl.
See Figures 8 and 9 for a sample using
the unitized approach.
Putting the initialized data into the
.ASM me also means you can use the
power of TASM or MASM operators and
macros to generate complex patterns of
initialized data that Turbo Pascal can't
match. Just try initializing a typed constant with the equivalent of 1000
dup('TEXT ') in Turbo!

MICRO CORNUCOPIA, #53, May, 1990

DefaultOUtputFile db 12,'DATA_OUT,DTA',
InitFileOffset
dd 1623447
BigAnay
db 1000 dup (' TEXT ')

: string (79)
; longint
; array(1 •• 5000] of char;

Confusing Code?
'Include
text count(){lnt c,nllnes,nwords,nchars,lnword;
Inword.tlO; nllnes-nwords-nchars-O;wh Ile( (c
·getchar(» I·EOF) {.. nchars: If (c •• ' \n' )
.. nllnes: If «c·.' '):: (c •• ' \n' » Inword·tlO;e Ise
1ft Inword··NO){ I nword.YES: .. nwords:} }prlntf (
"%d %d %d\n", nllnes,nwords,nchars):}

_OataSeql endp
code ends
end

•••

CIt Your Way!

Figure 7-INITDATA.PAS

.Include
~ext_count

{$F+}
{$L OATASEG1.0BJ}
procedure _DataSeql; external:
{$F-}

type
DataSeqlType
record
OefaultlnputFile : strinq[79];
DefaultOUtputFile : strinq(79];
InitFileOffset : lonqint:
BiqArray : array(1 .• 5000] of char;

()

Int c, nlines, nwords, nchars, Inword:

'DATA IN.DTA' }
, DATAOUT. DTA' }
1623447 }
1000 dup('TEXT ')

end;

Inword • NO:
nllnes • nwords • nchars • 0:
while «c· getchar() I· EOF) {
.. nchars:
If (c •• '\n')
.. nllnes:
if « c •• ' ') :: (c •• '\n'»
Inword • NO:
else if (Inword •• NO) {
Inword • YES:
Hnwords:
}

prlntf ("%d \d \d\n", nllnes, nwords, nchars):

var
DataSeg1
n : word:

ADataSeqlType:

begin
DataSeql := @ DataSeg1; {point to initialized data in DATASEG1.0BJ}
with DataSeqlAdo
beqin
writeln(DefaultlnputFile);
writeln(OefaultOUtputFile)i
writeln(InitFileOffset):
for n := 1 to 30 do '
write(BiqArray(n])i
writeln;

with C·ClearlytM, format C source
NOW
code exactly the way you want it.
C-Clearly's context sensitive analysis will
format any C program in your own personal
or corporate style.
to use, C-Clearly's style templates
EASY
are a snap to modify, since they
resemble C source code you edit into your
preferred format. Templates are included for
several common styles as well as standard
K & R.
can a!so be created with
LISTINGS
function names and
comments highlighted for improved readability.
Listing options include line numbers, headers
and/or footers and flow lines.
for making obtuse code clear.
Allows all of your source code
to be presented in a consistent format of your
chOOSing. Also great for code walkthroughs and
final documentation listings.

IDEAL

with all IBM PC, XT, AT, PS/2
WO
RKS
and compatibles, with 512K
RAM. Automatically processes all include files
and pre-processor statements. ANSI-C compatible. Not copy protected.

C-C/ear/y $129.95
Shipping & Handling USA $5; Canada/Mexico $10; Other Countries
$15; CA Residents add sales tax, Visa/MasterCard/COD accepted,

For orders or information call:

1-800-662-8266
V COMMUNICATIONS, INC.
4320 Stevens Creek Blvd., Ste 275. Dept. MC6
San Jose. CA 95129 (408) 296-4224

Reader Service Number 62

MICRO CORNUCOPIA, #53, May, 1990 43

Initializing Large Amounts Of Data
For large amounts of initialized data,
a program that reads the data structure
definition and then creates the appropriate db, dw, dd directives, etc., would
make creating the .ASM file easier. You
might write an include file that has the
initialized data appear after each of the
members in the record structure definition. Then feed this include file to your
program as a data file and have it create
the .ASM file.
See Figure 10 for an example.
Your program could parse each line to
determine the equivalent assembly language data type and number of elements
and then initialize it with the data in the
comments. As the famous cop-out goes,
"Such a program is left as an exercise for
the student."
Editor's note: Yes, yes. I enjoy writing in
Turbo Pascal but I, too, am running up
against limitations in the language, especially
with the 64K limit on the data segment.

•••

MultiPurpose Lab Interface for Your IBW
Tum your IBM-compatible computer into a powerful laboratory
instrument with our MultiPurpose Lab Interface (MPLI) hardware
and software. MPLI allows you to use a wide variety of sensors to
make measurements and to analyze and graph the results. Data
can be collected at a rate of thousands of samples per second or
slowly over periods as long as several months. Sensors available
include assembled. ready-te-use units and parts kits. The MPLI
hardware-software combination consists of:

• A 12-blt AID Interface Board: A 12-bit. 8-input analog-to-digital
converter with built-in sample and hold. Conversion time of the A-to-D is 9 J!Sec. There are three bipolar
and three unipolar software-selectable voltage ranges. The board also includes one 12-bit analog output
and access to 16 digital 110 lines. Includes manual and sample programs disk. (Order Code AI B-PC. $240)

• MultiPurpose Lab Interface Box: Includes three 8-pin DIN sockets which allow quick connection to 3 of the analog inputs. voltage out and
power leads. A prototyping area for building your own circuits is included on the circuit board inside the box. A variety of sensors and probes
are available which plug into the MPLI box: pH. thermocouple. temperature (AD590). force. microphone. etc. (Order Code BOX. $45)

it . .'."_'..(

• MultiPurpose ~b Interf~ce Software:

~e Mu.ltiPurpose Lab Interface Program allows three input ~;~
....._....
t:
]
channels to be calibrated to display any type of Input signal; for example. channel A can read temperature.
{>:. .....
channel Bcan read pH. and channel Ccan read pressure. Data from each ofthe channels can becollected... .. ·,·· .. ·tt·t,~· .. ·· .. ·:·:r·,. ..... ·;:I ..~.~.' .... '....
graphed. and saved on disk. The output voltage can be controlled by the program. The Oscilloscope mode
V' '.......
!i
lets the computer act as a triple-trace. storage oscilloscope with a sampling rate of at least 40.000 samplesl
sec (on an 4.77 MHz. IBM PC). much faster on other computers. The program even does X vs. Yplotting.
ICY Plottl_:-';;_~~=--("AA-::"'=--=-B~)· so you can display Lissajous figures on the monitor. (Order Code MPP-IBM. $49.95)
~:~:~~: ....·i~_~

f'"- . ./

~Q~:u.ulu.lJWuili.Luhw.L.t of time goes into
incrementing the index and testing
when to exit the loop. To save time you
can "open the loop" by enlarging the

/* Initialize Sum */

}

•••

/* Use normal approach */
/* for the last few
*/
/* Multiply and Accumulates */

code within the loop to perform several
multiply / add calculations (see Figure 4).
This approach will improve performance (to varying degrees) no matter
which processor you use. In fact, the
fastest code (assuming we're using a
non-caching processor) would require
no looping. Of course this is unrealistic
for all but the smallest neural network
because of the memory required for the
instructions.
How much should you open the
loop? You can determine this by taking
advantage of how the transputer works
with constants. The fastest executing instructions are a byte long and use the
first nybble to indicate the instruction
and the second nybble for a constant.
Thus, constants between 0 and 15 are
easily handled in one-byte instructions.
If it needs larger numbers, the transputer builds them a nybble at a time
(also using a byte-long instruction). The
instruction associated with the last
nybble necessary to form a constant will
also operate on the constant (e.g., load,
store, etc.). For example, if you need a
number between 256 and 4095, you need
three nybbles (with an additional nybble
instruction for each, i.e., a total of 3
bytes). Of course each additional nybble
requires an additional clock cycle.
Therefore a good amount to open the
loop up by is 16 (for constants between
0-15). This requires only one nybble (and
therefore one instruction) to access each
operand. To make the code usable for a
neural network of any size, add regular
loop code after the opened loop code to
handle the situation when the number of
nodes is not divisible by 16.
RAM
The second key feature of the transputer is its internal RAM. Most compilers use this RAM for stacks and
workspace pointers, but we can tell the
compiler to set aside space within internal RAM for a small routine or two. Instructions and data found in this area
(4K) can be retrieved in a single clock
cycle.
Since the less expensive TRAMs use
external memory requiring four clock
cycles for access, we can speed up the
matrix vector multiply by using the internal RAM. It's too bad we can't get the
arrays in there also.
Now what are we up to? Using the
hand-optimized assembly language to
maximize the floating point operation
and opening the loop by 16 to minimize

OK, piece a cake!
Berry Computers presents The Tele Operating System
A MULTITASKING MS-DOS COMPATIBLE
Computer software exists in layers.
Tele contains the layers between
your application and the hardware.
Tele executes binary programs
intended for MS-DOS in a multitasking, windowing environment.
You can use Tele to run several programs at once thereby.eliminating
boring and unproductive delays.
For instance, in a software development environment, you can set a
compiler running and Simultaneously
edit an()ther source file.
Tele uses a preemptive task scheduler.
Its file system is fully compatible
with MS-DOS. It runs on any processor compatible with the Intel
8086 family.
Most programs intended for MSDOS will multi-task under Tele.
Microsoft could find no way to make
MS-DOS multi-tasking and still support all its existing applications.
Tele simply does the best possible.
Tele services the documented interface to MS-DOS, fits in less than
lOOK of memory, and proVides the
important features of OS/2., including installable file systems.
Tele improves display performance
by a factor of 2, whether you use
multiple windows or not! On processors slower than 20 MHz the
improvement is even greater.

Source code in C and assembly is
included for the interface to MSDOS. If you have to have MS-DOS
idiosyncracies, you can easily put
them in yourself.
Treat yourself to an operating
system where the marketing and
financial folks had no say at all in
the design. The result is an efficient,
structured program that either provides or specifically anticipates all
features of modern mainframe and
desktop operating systems.
Tele Operating System
Tele Tool Kit
Multitasking (MT)*
Windows (Wl)*
File System (FS)
Tele Operating System
and Tool Kit
Demonstration Disk

$100
$ 50

$ 40
$ 40

$200
$ 5

*MT formally ,ailed SK. WI formally ,ailed CD

Telephone support is freely available.
Tele is available from:
Crosby Associates
P.o. Box 248
Sutter Creek, California
95642

CALL NOW TO ORDER:
(209) 267-0362
FAX (209) 267-9246*
* Note new Fax

Visa, Mastercard. American Express &. Discover Card accepted.

MS· DOS and OS/2 are trademarks of Mkrosoft Corporation

Reader Service Number 147

MICRO CORNUCOPIA, #53, May, 1990

the loop instruction overhead, I got
around 675,000 MAC/sec. When I put
the code in internal RAM, the speed improved by 50,000 MAC/sec to over
725,000 MAC/sec. Now we're talking!

Note: You can download the code for the
matrix vector multiply and the code for performing the transposed vector matrix multiply (needed during the error correction
phase of the neural network simulation) from
the Micro C BBS. The transposed code isn't
quite as efficient as the first but still does
over 700,000 MAC/sec.
Whew! I think we've eked every bit
of speed out of the transputer. You can
use this code with both online and epoch
training. Take your PC code and modify

it to insure that the data for the weight
matrix is arranged correctly. Otherwise,
code written for the PC using Turbo C
(from Borland International) or Microsoft C compiles under the Logical Systems' C compiler with little or no
modification.
Using Transputers In Parallel
So far, I've used a single transputer to
simulate neural networks. If you want to
process more data and/or process data
as quickly as possible, then a parallel approach to transputing is the ticket.
The underlying architecture of the
transputer, with its multitasking and
synchronized communications capabilities, almost eliminates the programming
hassles associated with controlling
several tasks concurrently.
Processor Farms
To implement a problem on several
transputers, you need to distribute the
workload as evenly as possible.
Some algorithms fit more naturally
into a parallel implementation than
others. So, you look for algorithms
which (when applied to parallel proces-

MICRO CORNUCOPIA, #53, May, 1990

sors) increase linearly in performance
with an increase in processors.
One such approach is a "processor
farm.,,4 In a processor farm one processor serves as the "farmer" and the remaining processors serve as "workers."
The worker processors execute identical
code. The farmer processor sends packets of data to each worker. When a
worker finishes with its work (its
packet), the farmer retrieves the results
and sends a new packet. Farming continues until the work is done.
This approach is probably the most
appropriate for using multiple transputers to simulate neural networks. Each
worker runs its own neural network.

weights and the weight corrections
through the pipeline. Each pipe processor has a complete copy of the neural
network's nodes as well as the input and
output values for each exemplar vector
that it's responsible for.
The weights are stored and modified
in the administrator processor· at one
end of the pipeline (typically another
transputer). The administrator processor
is connected to the host PC and to the
first and last processors in the pipeline.
It gets the topology and exemplar and
test vectors from the PC (typically in
files) and then uses this information to
configure the pipe processors.
When the pipe processors have their

Pipelining
While a processor farm makes sense
once you know the neural network's
weights, it's not good for training. An algorithm called pipelining is a better approach for training.
Pomerleau et. al. 5 describes the algorithm for use with the Carnegie Mellon
University Warp machine (a very expensive systolic array processor). Chong
and Fallside6 describe it for the transputer.
This algorithm uses a group of processors arranged in a pipeline (see Figure 5). In a pipeline, each processor
receives intermediate results and/or
data from its upstream neighbor,
processes the data, and then sends its results/ data to its downstream neighbor.
For best results, try to communicate
(send/receive data) and compute
(process data) in parallel. Also, each processor's work load should be relatively
equal. So no processor waits for another.
For neural network simulations on an
array of transputers, you typically send
a subset of the exemplar vectors to each
pipe processor. Then you pipe the

exemplar vectors and a copy of the
neural network nodes, the administrator
processor sends a copy of all the weights
down the pipeline. As a pipe processor
receives the weights, it immediately
sends them to the next processor
downstream. Then (in parallel) it uses
the weights to make a pass through the
neural network for each of its exemplar
vectors.
The pipe processor compares the calculated output values with the exemplar
output values, then calculates an error
value which it sends back to the administrator processor. The administrator
node receives the error values from all
the pipe processors and determines if
the total error for all the exemplar vectors meets the desired criteria.
If the error is low enough, the administrator processor transfers the weights
back to the PC and runs through any test
vectors that have been supplied.
If the error is too high, the administrator sends a message to each pipe processor instructing it to calculate a change
of weights based on its exemplar vectors. The administrator processor then
sends an empty array down the pipeline.

When a pipe processor receives the
array, it adds the change in weights that
it's calculated to the array, then sends it
downstream to the next pipe processor.
When the administrator processor
gets the array back, it will have the sum
of all the changes for each weight based
on all the exemplar vectors. The array is
then multiplied by the lea~ning rate constant and added to the previous changed
weight vector (which has been multiplied by the desired momentum constant) to produce a new weight vector.
The administrator creates a new set of
weights by adding the present weight
vector to the new weight vector. Then
the whole process begins again with the
administrator node sending the new
weights down the pipeline.
Programming Transputers
To implement the pipelined neural
network simulation on the transputers
requires only three programs, no matter
how many processors are involved. The
first program (ADMIN) is the administrator processor program.
This program interfaces to the host

PC and to the pipeline. It's responsible
for:
(1) Retrieving the topology of the
neural network;
(2) Transmitting the information
down the pipeline so that each pipe processor can configure itself;
(3) Retrieving all the exemplar vectors from the host;
(4) Transmitting evenly divided subsets of the exemplar vectors to each pipe
processor;
(5) Transmitting the weights into the
pipeline;
(6) Getting error measurements from
each processor in the pipe;
(7) Determining whether the total
system error is below a user specified
value;
(8) If necessary, instructing the pipe
processors to calculate the weight
changes;
(9) Sending an empty weight change
array downstream;
(10) Collecting the filled weight
change array from the pipe processors;
(11) Using the host (Le., user) supplied momentum and learning rate con-

stants along with the weight change
array to calculate a new weight vector;
(12) Looping back to step 5;
(13) When training is complete,
saving the weights;
(14) And, if necessary, running test
vectors on the trained neural network
and collecting statistics.
The program (PIPE) is run on all the
pipe processors. It's responsible for:
(1) Receiving and retransmitting the
topology of the neural network;
(2) Setting aside enough space for the
weights and its own exemplar vectors;
(3) Reading and storing its own exemplar vectors;
(4) Receiving and retransmitting any
exemplar vectors destined for processors
downstream;
(5) Receiving and retransmitting the
weights;
(6) Using the weights to calculate the
output for the neural network for each of
its exemplar input vectors;
(7) Comparing the neural network
calculated outputs with the exemplar
output vectors and calculating an error
value;

Save "Mao-Years of Effort~ with Thrbo 5.5
Don't Start from Scratch
with Object-Oriented Pascal

Object Professional includes
clear, comprehensive documeiltation,
on-line help, full source code,
technical support, and hot demo
programs. Pay NO royalties. You'll
get up to speed fast with OOP!

Object Professional is a huge library
of over 200 object types and 2000
methods that
will multiply
iliThe range of objects is
your productanttm;ttc. Object Professional
tivity. Window
could literally save you
object types let "';;;=;;;~II ~ ~I'!!!!~a.
man-years of effort. JJ
you use overJeff Duntemann
lapping and
resizeable
~iiiiiiiiiaii• •l',',j(",:!~
Object Professional 1.0,
windows. The ...
only $150.
windows include _ scrolling data
entry screens _ pick lists _ menus
_ file selection _ printed forms
AMulti-User B-Tree Toolkit
_ help capability and more.
Write powerful network
Build your programs using
compatible databases faster and
proven data object types like stacks,
easier using B-Tree Filer 5.0.
linked lists, virtual arrays, and more.
You'll have the fastest, safest, most
System-oriented routines provide
flexible databases - no rigid strucswappable TSRs in only 6K of RAM,
ture, no TSR hassles, no running out
EMS management, and much more.
Satisfaction guaranteed or your money back within 30 days.
Add $5 per order for shipping in U.S. and Canada. Inquire
about other shipping charges. OPro requires Turbo 5.5.
BTF requires Turbo 4.0. 5.0. 5.5. or QuickPascal.

of files. And they're compatible with
Novell, 3Com, MS-NET, and others.
You get _ Fixed and variable
length records _ Two billion
records per database _ Up to 100
indexes per index file _ Fail-safe
mode with joumaling _ Units for
sorting, browsing, reindexing, and
network control.
B-Tree Filer includes full source
code, documentation, technical
support, and you pay NO royalties.
iii B-Tree

Filer ... a well rounded,
feature-rich approach to
B-Tree databases. JJ
Computer Language, 1/90

B-Tree Filer 5.0, only
$125. (single user)
With network support, $175.

Call toll-free to order.
1-800-333-4160

8AM - 5PM PST Monday through Friday. USA & Canada.
For more information call (408) 438-8608. Fax: (408) 438~86\O.
TurboPower Software PO Box 66747 Scotts Valley. CA 95067-0747

Reader Service Number 194

MICRO CORNUCOPIA, #53, May, 1990

(8) Transmitting its error value
downstream;
(9) Receiving and retransmitting any
error values produced by upstream processors;
(10) Receiving and retransmitting any
instructions regarding a backward pass
through the network;
(11) If necessary, using the output errors from all the exemplar vectors to calculate the changes in the weights;
(12) Receiving the weight change
array from upstream, adding the locally
calculated changes, and transmitting the
array downstream;
(13) And looping back to 5.
The third program (TAIL) runs on the
last transputer in the pipeline in parallel
with a copy of the PIPE program. Unlike
the other processors that used the hardware DMA channels Oinks) to transfer
data from one processor to the next, the
PIPE and TAIL programs on the last
processor communicate with each other
over a memory channel.
From a programmer's perspective,
memory channels operate identically to
hardware channels. This means that the
last PIPE program is identical to the
PIPE programs on the other processors.
The TAIL program primarily transmits information coming out of the pipeline back to the administrator processor
and collects garbage. When the neural
network topology information reaches
the last pipe processor, there's nowhere
to send the data.
Since all the PIPE programs are identical, the last PIPE program will try to
send the data on anyway. TAIL receives
the topology data and discards it. The
weights transmitted down the pipeline
are likewise garbage collected. The error
values and .the weight change array,
however, are sent back to the administrator.
Each of the processors has a harness
program. The harness programs specify
whether specific channels are hardware
or memory channels and which programs (ADMIN, PIPE, or TAIL) are to be
run together. The TAIL program can run
on the same processor as the ADMIN
program instead of with the last PIPE
program. Or, if space allows, a copy of
the PIPE program can run on the administrator processor. To maintain load
balance, it might be necessary to reduce
the number of exemplar vectors that this
first PIPE program processes.
Finally we must write a short configuration file (used by the network

MICRO CORNUCOPIA, #53, May, 1990

loader program) to indicate which program is to run on which transputer. The
network loader is responsible for resetting all the transputers and loading each
one with its specific program. From the
file, the network loader can tell which
transputers and which links a program
must be transmitted through before it arrives at the transputer on which it's to
(ultimately) run.
Unfortunately, programming transputers isn't always as simple as I've
made it sound. Fortunately, vendors are
improving their operating environments.
For example, some vendors support
message handling protocols. With some
of the protocols, a communications
process runs continuously in parallel
with the application program. Any time
data must be sent to another transputer,
the application program uses a vendor
supplied function or a memory channel
to pass the data to the communications
process. This process adds a header to
the data and sends the message out the
appropriate link.
Another transputer receiving the
message uses the header to determine
whether the message is intended for
another transputer or for itself. If necessary, the process forwards the message
off to another processor. Otherwise the
process strips the header off the message
and sends the data through a memory
channel to the appropriate application
process. While this scheme slows down
applications, it can make communications simpler.
More important, these communications processes give you a means for debugging remote transputers. Although
the application program may have
stopped on a remote processor because
needed data hasn't communicated to it,
the communications process can still
send and receive messages. One of those
messages can be an instruction to read a
particular memory location or register
and send the data back over a link.
Discussion
The topology of the neural network,
the number of test vectors, and transputer memory resources will determine
how you manage the transmission of
weights. In most situations the administrator transputer transmits the weights,
storing them in the pipeline processors.
During the error correction phase of
the training cycle, each processor generates changes for the weights. These

changes are then sent back to the administrator where momentum and learning
rate adjustments get made. This approach requires few (if any) parallel
processes.
If, however, the network is very large
and there are memory limitations at the
nodes, you might have to break the
weights into packets (say all the inbound
weights for a hidden unit) and send
them sequentially down the pipeline.
The PIPE program would then calculate all the partial sums for each exemplar vector. With all the weights for one
layer received, the activation function
would be calculated for each unit.
There are, of course, other approaches to dividing a neural network
simulation among transputers. For example, reference 7 describes how each
transputer can process a subset of the
nodes. Other researchers have used
transputers with non-back-propagation
neural networks. 8,9,lO
In Sum
The back-propagation simulation of
neural networks maps very nicely into a
parallel processing environment. While
other parallel processing platforms are
available, transputers provide a cheap
way to optimize neural network simulations. Plus, we can use these transputers
for other applications as well.
References
1 Pelczarski, M., "System Review:
Microsoft Softcard," BYTE, vol. 6, no. 11,
November, 1981, pp 152-162.
2 Inmos Limited, The Transputer Reference Manual, Prentice Hall, London,
1988.

3 Atkin, Phil, "Performance Maximisation," Technical Note 17, InmosSGS Thomson, Bristol, UK, March 1987.
4 "Some Issues in Scientific Language
Application Porting and Farming Using
Transputers," The Transputer Development and iq Systems Databook, Inmos Ltd.,
1989.

5 Pomerleau, D. A., Gusciora, G. L.,
Touretzky, D. S., Kung, H. T., "Neural
Network Simulation at Warp Speed:
How We Got 17 Million Connections per
Second," Proceedings of IEEE International Conference on Neural Networks, San Diego, USA, July 1988, pp
11-143-150.

6 Chang, M. W. H., and Fallside, F.,
"Implementation of Neural Networks
for Speech Recognition on a Transputer
Array," Technical Report CUED/F-IN-

FENG/TR8, Cambridge University, Department of Engineering, Cambridge,
UK, March, 1988.
7 Beynon, T., "A Parallel Implementation of the Back-Propagation Algorithm on a Network of Transputers,"
Research Initiative in Pattern Recognition, Royal Signals and Radar Establishment, Malvern, UK, Poster Paper,
1988 IEEE International Conference on
Neural Networks.
8 Di Zitti, E., Caviglia, D. D., Bisio, G.
M., and Parodi, G., "Neural Networks
on a Transputer Array," Proceedings of
the 1989 International Conference on
Acoustics, Speech, and Signal Processing, pp 2513-2516.
9 Abbruzzese, F., "A Transputer Implementation of a McCulloch & Pitts
Network," Parallel Processing and Applications, E. Chiricozzi and A. D'Amico
eds., North-Holland, 1988, pp 135-140.
10 Board, J. A., Jr., and Lu, J. S. J.,
"Performance of Parallel Neural Network Simulations," Proceedings of the
Second Conference of the North American Transputer Users Group, J. A. Board,
Jr., ed., Durham, North Carolina, USA,
North American Transputer Users
Group, 1989, pp 185-200.
11 Lee, D. G., Jr., "Preliminary Results of Applying Neural Networks to
Ship Image Recognition," Proceedings
of the 1989 International Joint Conference on Neural Networks, June 18-22,
1989, Washington DC, IEEE, San Diego,
II-576.
12 Rumelhart, D. E., Hinton, G. E.,
and Williams, R. J., "Learning Internal
Representations by Error Propagation,"
in Parallel Distributed Processing Volume
1: Foundations, Rumelhart and McClelland ed., Cambridge, Massachusetts,
USA: MIT Press, 1986, Ch. 8, pp 318-362.
13 Leung, H. c., and Zue, V. W.,
"Applications of Error Back-Propagation
to Phonetic Classification," Proceedings
of the Neural Information Processing
Systems - Natural and Synthetic Conference, November 28 - December I, 1988,
D. S. Touretzky, ed., Morgan-Kaufman,
1989.
Vendors
Companies advertising Transputer
TRAMs and/or language compilers for
use with the Transputer:
Inmos-SGS Thomson
1000 Aztec West
Almondsbury Bristol BS12 4SQ
UK

Levco
6181 Cornerstone Ct. East, Ste. 101
San Diego, CA 92121

Transtech Devices Ltd.
Unit 17, Wye Industrial Estate
London Road
High Wycombe
Buckinghamshire HPlllLH
UK

Multis Corporation
99 Willie Street
Lowell, MA 01854

Companies advertising neural network software using coprocessor cards
and IBM PC compatible computers:

MicroWay, Inc.
P.O. Box 79
Kingston, MA 02364

SAIC
10260 Campus Point Dr.
San Diego, CA 92121

Computer Systems Architects
950 North University Avenue
Provo, UT 84604

HNC
5501 Oberlin Drive
San Diego, CA 92121

Logical Systems
P.O. Box 1702
Corvallis, OR 97339

Company advertising neural network
software for Transputers and Apple
Macintosh II computers:

3L Limited
Peel House
Ladywell Livingston EH54 6AG
Scotland

Neurix
One Kendall Square, Suite 2200
Cambridge, MA 02139

Sension
Denton Drive
Northwich Cheshire CW9 7LU
UK

•••

You're in charge
Don't be limited by someone
else's idea of the perfect menu. SuperMenu's unique graphical interface
adds a level of creativity, individuality, and ease-of-use not found anywhere else. Why shoehorn your
software into a stuffy, preconceived
mold when SuperMenu gives you
complete control over the structure,
format, and appearance of your
menu? Most menu systems give you
a modest choice of screen colors and
then claim to be customizable. SuperMenu gives you color plus a palette
of drawing tools, multiple fonts, fill
patterns, line styles, pop-up notes,
and more. You can even edit the
background pattern. In fact, SuperMenu has many features you'd only
expect to see in an expensive drawing
program, much less a menu generator.
• Graphical mouse-driven
interface
• Instant access to your softw::re :md DOS options
• Worksw/allpopularprograms
• Easy-to-use drawing
tools, fonts, fill patterns,
line styles, and colors

e
u t c Opy

...

r1"
•
.

.,,'

F ron t B k G roup

MASTER

ISN..ESPEOPlEII

mITJ
L
Too'.'

PIAlLISHING

ACCOUNTING

Pr.v .. N.xt

III

Scale

'h
•

rSus.n'sClienu

~ ~

W _~

Backups I FORMAT

I
I
G'~--I
I
I
SACKUP - TUESOA'!'

BACKUP -

WEONESDf'IY

I BACKUP - HlJRS(lAV J
I BACKUP - FRIDAY I

1~Il-KlRDPRQC[SSlNGI

Ho•• l { End

--w

DOS Cornlrlands

.. Ga. . .

flIiJ"-IOIOI0IAI

SuperMenu's easy-to-use
mouse driven interface makes
running your favorite software a
snap. Other computer tasks,
such as daily backups, can be
easily automated and accessed
with a simple mouse click.

• Multiple menus, each
with up to 99 pages
• Hypertext-like links
between pages
• Pop-up notes
• Passwords prevent unauthorized modifications
to your menus

AI,gn

~!zDp.

1Inli'1I jIlilUtD

1lARRY'SNOIJEI.."

Unogroup

Because SuperMenu has a
wide range of powerful tools,
menus can be as complex or as
simple as you desire. Like we
said, you're in charge.

~lID~~~lID

Add$S.OOS&H
• Date and time display
Please specify 5W' or
• NOT memory resident
3
Vi' when ordering
• Enter DOS commands or
-----------invoke a DOS Shell
• Supports eGA, EGA,
VGA, and Hercules
graphics
1293 NW Wall Street,· Suite 71

49 95

~~[ZJ
Bend, Oregon 97701

Requlfes: mMpc,·.XT, 286;386; psiiOrcompatible.··rios io· or.·greatei,256K RAM?

. .•. rvtic1?so.f~ .~.it~h?!.c9r?~~ti"I~r?9\1s~: SgA,§gA,ygA.~ ~~~\lI~~of~ompati"ledisplaX'

(503) 389-5489

Reader Service Number 216

MICRO CORNUCOPIA, #53, May, 1990

Save The Floppies!
Rx For Slipped Disks

This is probably the easiest to use disk
recovery program I've ever seen. You just
insert the flaky disk and fire up the program.
When it finishes, every sector on the disk
will be readable, probably.
ata error reading drive B
Abort, Retry, Ignore?
Grrrrr ...
Back in the old days, when CP1M
had trouble reading a floppy, I could
break out the dependable and userhostile program DU (Disk Utility) for a
little rescue work. DU didn't care about
bad sectors; it would read with reckless
abandon. Often, the "bad sector" had
just a single bad byte. An easy repair job
would restore the file's readability.
I thought it would be useful to have
this ability in an MS-DOS system. Forget
for the moment that MS-DOS already
lets you repair disks with its RECOVER
program. At the beginning of this project, I hadn't made the acquaintance of
RECOVER.

A Little Background
Each sector on a disk consists of sync
bytes, ID address marks, ID fields (cylinder, head, sector, bytes per sector), a
CRC value for the ID fields, a gap, more
sync bytes, data address marks, data
(whew, at last), a CRC for the data, and
yet another gap. That's at least 105 bytes
of overhead for each 512 bytes of data.
Add another 146 (or so) bytes of header
info for each track and you'll get an idea
of what the Floppy Disk Controller
(FDC) has to wade through.
At the lowest level, the FDC, acting
under orders from the CPU, selects a
drive, fires up the drive motor, seeks to
the track, performs the requested function, and reports the results to the CPU.
During a read, the FDC looks at the

54 MICRO CORNUCOPIA, #53, May, 1990

Assuming that a
CRC error means a
data error, and
barring physical
damage to the disk,
simply rewriting the
questionable data
to the same sector
should (and has, on
numerous
occasions) restore
the file.

data have changed (or that the CRC has
been corrupted). Why'd they change?
Beats me: a stray magnetic field, a glob
of peanut butter, cosmic rays ....
It would be nice to isolate the data
CRC error from the ID field CRC error.
But after a wade through the XT BIOS
listing, I can't see that it differentiates between the two. So we can either make
the rash assumption that any CRC error
will come from the data, or we can dive
into the FDC to get a closer look at the
error.
Leave your trunks at home (leave
your rash, too). It doesn't make sense to
try to fix the sector ID since it gets written only during formatting; reformatting
a track would wipe out the information
we want to recover. We probably
couldn't read the sector anyway, since
the corrupted ID fields wouldn't match
those requested by the CPU.
Assuming that a CRC error means a
data error, and barring physical damage
to the disk, simply rewriting the
questionable data to the same sector
should (and has, on numerous occasions) restore the file.

track until it either finds ID fields matching those requested, or sees the disk's
index hole pass twice without finding
the matching fields (sector not found). If
it finds the sector, the FDC sends the 512
bytes of data to the host-in MS-DOS
systems, to the Disk Transfer Address
(DTA).
As the FDC scans sector ID fields, it
generates a CRC value for the IDs and
compares this with the CRC recorded on
the disk. If the two don't match, the controller returns a CRC error.
A second type of CRC error comes
when the CRC calculated during the
data read doesn't match the data CRC
recorded on disk. This is the error we're
after. It means that one or more bytes of

Levels Of Disk Access
Most programs use DOS interrupt
21h to read a floppy. DOS deals with
files, so application programmers like it.
No fuss, no muss. But we'd like to get
into some real trouble here. We can get
at the whole disk by dropping one level
closer to the hardware and making use
of the BIOS.
Nestled in ROM, the BIOS doesn't
know or care about directories, files, or
any of the other high level nonsense that
DOS requires. With the BIOS you can
access any sector on any disk, whether
it's allocated to a file or not (including
the directory and FATs). And, like DOS
INT 21h, you can choose to ignore error
conditions.

Larry Fogg
187 NW Jefferson PI.
Bend, OR 97701

Figure I-FLOPFIX.C
1* FLOPFIX.C - Reads all sectors of a floppy. Any sectors with bad
CRCs are rewritten. *1
#include
#include
int crc failures;
int sectors recovered;
int non_crc:failures;

1* reset the floppy disk controller *1

void reset FDC ()
{
union REGS in, out;
in.h.ah = 0;
int86 (Ox13, 'in, 'out);
1* reset_FOC () *1

1* service 0 = reset *1
1* INT 13h *1

unsigned char read_sector (unsigned char track, unsigned char side,
unsigned char sector)
union REGS in, out;
in.h.ah = Ox2f;
intdos (&in, 'out);

1* get OTA *1
1* do the INT, ES:BX holds address on return *1

in.h.ah = 2;
in.h.al = 1;
in.h.ch = track;
in.h.cl
sector;
in.h.dh = side;
in.h.dl = 1;
int86 (Ox13, 'in, &out);
return (out.h.ah);
1* read_sector () *1

1* read diskette *1
1* # sectors to read *1

1* drive B: *1
1* do the read *1
1* return read status *1

unsigned char write_sector (unsigned char track, unsigned char side,
unsigned char sector)
union REGS in, out;
in.h.ah = Ox2f;
intdos ('in, &out);

1* get OTA *1
1* do the INT ,ES:BX holds address on return *1

in.h.ah = 3:
in.h.al
1;
in.h.ch = track:
in.h.cl = sector:
in.h.dh
side;
in.h.dl = 1;
int86 (Ox13, &in, &out);
return (out.h.ah);
1* write_sector () *1

=
=

void fix_disk ()

1* write diskette *1
1* # sectors to write *1

1* drive B: *1
1* do the write *1
1* return write status *1

continued on page 56

The key here is that, even though an
error condition exists, the data has been
read. It's alive and well at the DTA, and
all you have to do is rewrite it to disk.
Code Notes
You won't find anything very startling in FLOPFIX.C (see Figure 1). fixdiskO makes three attempts to read each
sector. It takes a while for the drive
motor to spin up, so a well behaved program should do these retries to ensure
that any errors are real. (Reads from a
half-fast drive don't have much chance
of success.) A call to reset_FDCO follows
any read failure-again, a required programming practice.
Once we're sure an error has occurred, fix_diskO checks the status returned by read_sectorO. If the status
indicates a CRC error, write_sectorO attempts to rewrite the sector. Global variables for the number of bad sectors and
recovered sectors get updated and we
move onto the next sector.
You'll notice that each sector access
(in write_sectorO and read_sector())
starts with a call to determine the DTA.
This could have been done only once
during program initialization, since the
DTA doesn't change during execution
(unless the program changes it explicitly). But the call to DOS function 2Fh
provides a convenient way to load
ES:BX with the DTA.
But Why Bother?
Sure, sure. Once again, I've reinvented the wheel. But in some situations
it's a better wheel, a bit more round. RECOVER has not been as successful as
FLOP FIX in restoring damaged sectors
to readability (at least, in my humble
opinion). On one occasion a RECOVERed file was not TYPEable. Besides, diving into the guts of the
machine keeps me out of trouble. It's alMICRO CORNUCOPIA, #53, May, 1990

ways instructional to try to understand
the inner workings of a system.
If I had more gumption, I'd disassemble RECOVER to see what makes it
tick. I could rip off the good stuff and
add it to FLOPFIX. But wisps of cloud
over the nearby Cascades promise a fine
day of early spring back country skiing.
Some other time for the disassembly.

Continued from page 55

unsigned char track, Side, sector, status;
char tries;
reset FDC 0;
for (track=O; track<40; track++)
for (side=O; side<2; side++)
for (sector=l; sector
'include
linclude
'include
#include
extern int okhigbuf = 0;
extern unsigned _stack = 1024;

1* these

reduce resident program size

1*~----~------------77------~---------~-------------------------7---*1
void khon (void)
(
/* enable input from the keyboard */
outp (Ox64, OxAE) ;
1* enable keyboard command
,ifdef DEBUG

c++ Tools
The C++ Tools include classes for
queues, singly and doubly linked lists,
stacks, bit vectors, virtual arrays, hashed
search tables, windows, and all kinds of
other things. Included is a book which
explains each class, gives examples of its
use, and gives the complete source
listing for the class (the source is on disk
as well). If nothing else, the C++ Tools
are good examples for those just starting
with C++.
I had wanted to play with these
classes some and get a feel for how useful they were. But, as with the TSR package, there just wasn't enough time. Rain
check.
Documentation
The old version of Zortech had a
single manual for the compiler and library and another manual for the debugger. This was more or less adequate,
but I still found myself reaching for my
old Turbo C 1.0 manual now and then
(especially in the old days, when I was
weaning myself from Pascal-it has a
great little section showing equivalent
operations in C and Pascal side by side).
The manual set for 2.0 has a separate
manual for the compiler and utilities, the
libraries, the debugger, and the C++
Tools. I've done 'qUite a bit of thumbing
through them in the last two days and
have found most of the information I
needed. There is even a thorough explanation of symbol construction in C++,
and many other implementation details
that almost nobody would need. But
when they did ....
There are sections both on AT&T C++
compatibility and MS C compatibility. It
also has a section titled "Incompatibilities with ANSI C," which deals not with
incompatibilities between Zortech and
ANSI, but with incompatibilities between any C++ and ANSI C. This is very
useful reading for a new convert from C.
Price
Unfortunately, the $29.95 JRTs and
$49.95 Turbos are a thing of the past.
Well, not really. I've been seeing JRT ads
again recently but, not able to believe,
MICRO CORNUCOPIA, #53, May, 1990

thought that I had entered a strange time
warp like one of those episodes from
"Twilight Zone." Seriously, I thought
the magazine had misplaced some of the
copy from their April issue.
The problem is that the amount of
development (and the amount of advertising dollars) that has gone into this
kind of software has skyrocketed since
the days of JRT and Turbo 1.0.
Even Borland realized long ago that
$49.95 per copy wasn't even keeping the
lights on, much less paying for the
$10,000 BYTE ads and financing all those
hot tub parties and hotel suite bashes so
essential in modern (marketing) warfare.
The price of Turbo Pascal has risen with
each new release (it seems like the update fee for the new edition is about the
same as the retail price of the previous).
I stopped at 3.0.
The retail price of the Zortech
Developer's Package 2.0 is $450. The
C++ compiler by itself is $199. Prices
from discount houses will probably run
considerably less.
On the other hand, the functionality
of JRT Pascal, or Turbo Pascal isn't even
close to the functionality of Zortech C++.
Although considered breakthroughs in
their time, they are now mere toys.
Also, the retail price of Microsoft C
5.1 is about the same. And it's just a C
compiler, no C++ (it does include OS/2
support, however, which is a $150 upgrade from Zortech).
Still, it isn't even fair to compare Zortech C++ with MS C. Or any C, for that
matter. Because it isn't C. It's C++. And
it has a debugger that is light years
ahead of the last version I've seen of
CodeView.
Even if the price was a bit higher
(which it isn't), it would be worth it in
all the development time saved. After
all, how many days does it take you to
make $450? Don't you think it's about
time you worked a few nights so you
could afford to have some fun?
Zortech Inc.
1165 Massachusetts Ave.
Arlington, MA 02174
Voice: (617) 646-6703
Fax: (617) 643-7969

Batch File Prophylactics
Before closing this thing up and tossing the smelly cat out for the night (you
thought I was just kidding about smelly
cats), I want to pass along a little "cuteC" utility I wrote. I compiled it with

MICRO CORNUCOPIA, #53, May, 1990

Zortech, but it will work with MS or
Turbo with just a few changes in places I
~ave pointed out in the source. See Figurel.
KBD
KBD is a program that can enable
and disable the keyboard on a PC compatible. This can be useful when writing
a BAT file which you don't want the
stupid user to screw up, or running a
program which you don't want the ....
You get the idea.
The syntax of KBD is:
kbd lei I d [conunand]

/ e enables the keyboard, / d disables
the keyboard, and command is an optional parameter, which is a DOS command to execute while the keyboard is
enabled/ disabled. If command is given,
the keyboard is turned back off! on
when the command finishes. For example:
kbd Id

turns the keyboard off. If you type this at
the DOS prompt, the only way you can
get the machine going again is the reset
button or the Big Red Switch! Usually
you would do it at the beginning of a
sensitive batch file. Another example:
kbd I d chkdsk

password with the command LOGIN
before continuing. For example's sake,
LOGIN returns 0 if the login was
successful, or 1 otherwise. You want to
put LOGIN in a BAT loop that looks at
errorlevel. The danger is that the user
can simply type ctl-C and break out of
the batch file. This is where KBD comes
in. For example, you might use this BAT
file:
echo off
break off
kbd Id
: loop
kbd Ie login
if errorlevel 1 goto loop
kbd Ie
break on

Notice that, although ''break off"
minimizes the number of times DOS
checks for ctl-C, it doesn't eliminate it.
That is what makes KBD essential. KBD
also traps the DOS ctl-C interrupt (23h)
while it is executing the command,
making sure nobody can ctl-C out of that
program.
Of course, you have the problem that,
while the first two lines of the ~AT file
are executing, the keyboard is still
enabled. A good way to solve this is to
make a device driver that turns off the
keyboard. This way the keyboard is dis-'
abled before the user has any chance to
screw things up. I'll leave that one as an
exercise for the reader.

or
kbd Id brot

(I thought Larry might appreciate
that second one.) These commands turn
the keyboard off while executing chkdsk
(or brot), then turn the keyboard back
on. As I said, KBD is mostly useful in
BAT files (especially AUTOEXEC).
As an example, say that you have a
BAT file that gets a user's name and

How?
KBD turns the keyboard on and off
by outputting a command to the keyboard controller port (64h). OADh turns
off the keyboard, and OAEh turns it on.
This physically disables anything from
the keyboard. Even ctl-alt-Del falls on
deaf ears.
Interrupt Package
Another interesting part of KBD is

the use of the int_intercept function from
the Zortech interrupt package. The interrupt package handles all the details of
saving the original interrupt vector, putting the new one in place, setting up a
stack for the service routine, chaining to
the old service routine ~f requested, and
restoring the old vector when finished.
Note that breaktrapO (my replacement for the ctl-C interrupt) returns a 1.
This means "don't chain to original interrupt." If it returned a 0 it would mean
"chain to the original interrupt now,
then return." Since our entire reason for
trapping the interrupt was to keep the
original from being called, we don't
want to return O.
DOS Exit Values
In the example above, we used errorlevel to check the exit value of the
LOGIN program. But the program being
executed (by the BAT file) is KBD, not
LOGIN! How can I do that? With the
clever trick of using one of the spawnO
functions (spawnvpO in this case), which
returns the program's exit code. Then I
return that value from mainO in KBD

(the return value of mainO is sent to DOS
as the exit value of the program).
spawnvpO searches the DOS path for
the command and can use the arguments
in argv[] without modification, but it
cannot call a BAT file. If I wanted the
ability to execute BAT files, I could use
systemO instead, but then I wouldn't get
exit values. (systemO uses command.com, which eats the exit value.)

XT SCHEMATIC
Micro Cornucopia
Phone Orders:
(503) 382-5060. M-F, 9 AM-5 PM PST
Mall Orders:
P.O. Box 223. Bend. Oregon. 97709

IBM PC-XT Schematic

_okbigbuf and _stack
The declarations of the variables okbigbuf and _stack at the top of KBD.C
are included to reduce the resident size
of the program. This leaves more
memory for the program called with
spawnvpO.
In Zortech C, when _okbigbuf is 1
(the default), the maximum 64K of data
space (for small data models) is allocated
on program startup. If _okbigbuf is 0,
data space is only allocated from DOS as
needed. _stack controls the size of the
program's stack. By setting these two
variables, I was able to trim the resident
size of the program from 72K to 10K. A
big improvement for two lines!

••• .$15.00

As always, you can do anything you
like with KBD. Even swear at it if you
want. I simply don't care.
Pseudoepidermatologistical
Meanderings
Sorry, but it's nearly 2 a.m., and you
can't expect everything I say at 2 a.m. to
be coherent. The column is finished now.
But it's cold out, and the smelly cat looks
so pitiful. (If she just wasn't about to explode with babies.) I guess she can stay
inside tonight. But just this once. (Anybody want a Turkish kitten?)

• ••

DEVELOP YOUR OWN

ICs

PROMPT DELIVERY!!!

SAME DAY SHIPPING (USUALLY)

QUANTITY ONE PRICES SHOWN for FEB. 25,1990

BOOTABlE
OPERATING SYSTEM
WITH BOOTER TOOLKIT'M

DYNAMIC RAM

1MB
SIMM
SIMM
SIMM
1Mbit

41256
--41256
BB41256
41256
4464
41264*

MEM DeskPro 386/20
AST Prem386/33Mhz
1Mx9
80 ns
256Kx9 100 ns
1Mx1
80 ns
256Kx1
60 ns
256Kx1
80 ns
256Kx1 1 00 ns
256Kx1 120 ns
64Kx4 120 ns
64Kx4 100 ns

EPROM

$295.00

225.00
92.00
35.00
8.75
4.15
2.50
2.20
1.95
2.50
7.50

200
200
150
250

ns $18.00
ns
7.80
ns
6.50
ns
3.75
STATIC RAM
62256p-10 32Kx8 100 ns $10.00
128Kx8
64Kx8
32Kx8
16Kx8

6264p-12
6116AP-12

8Kx8
2Kx8

120 ns
120 ns

4.50
4.25

•
•
•
•
•
•
•

Real-time scheduler
DOS-compatible file system
Window system
Memory manager
Bootstrap loader
100+ page user manual
ONLY $99 WITH SOURCE CODE

For immediate shipment, send your
check or money order for $99 to:

GENERAL

____ SOFIWARETM
SAT DELIVERY
INCLUDED ON

MICROPROCESSORS UNLIMITED, INC.

MasterCardNiSA or UPS CASH COD

1h: $.2 . $6.25/41b
Fr: "'1
S18.50/11b

No minimum order. Please note: prices subject to changel
Shipping, insurance extra, up to $1 for packing materials.

F~~~V~D:::S ~~~~s~·cJ:.o;t4:~e., (918)

Reader Service Number 37

267·4961

General Software

P.O. Box 2571, Redmond, WA 98073
Telephone ordars shipped UPS BLUE COD only. To order by crad~ card. call Programmer's Connection or Austin Codawo",s.
authorized General So~ar~ de~lers. Secon.d-day air s~lpping provK:ted tree to destlnati~n only within the United States. Foreign
ordars add $20.00 lor a,rmail delivery. Wash'ngton ResidenlS add 8.1 % sale. lax. COPYright (e) 1990 General Software. All rights
reserved. Booter Toolkit and General Software are trademartls of General Software

Reader Service Number 212

MICRO CORNUCOPIA, #53, May, 1990

"":"'::".::''::''''';:'

GUlliJ"URE:
C.®RNER

MICRO CORNUCOPIA, #53, May, 1990

You're On Your Own

By David Thompson
Laura Shaw &
Nancy Ellen Locke
Micro C Staff

Well, here it is: probably for the first time in
the history of computers, an article in a computer magazine on advertising in other computer magazines. So many advertisers have
asked for suggestions of where to advertise that
we've put together this information.
Who's Offering The Best Deal?
The standard way of judging what kind of
deal you're getting on advertising is to divide
the one-time B&W full-page price by the circulation (in thousands). That's the cost/thousand.
That way you can compare rates from smaller
magazines with rates for the big guys.
Once you figure the cost/thousand readers,
you can factor in the less precise variables. What
percentage of the audience would be interested
in your product? How much training would
they require? How much will you have to explain in the ad? (You can say ICE to designers;
you have to explain what it means to managers.)
Would they pay what you're asking?
Be aware that corporations look for image
and support, hackers look for performance and
price.
But when you're all done, the real question
is: Will advertising your product in Computer
Language or Programmer's Journal or PC Techniques be profitable? If you make money in all
three, then advertise in all three. If your product
doesn't make money anywhere, maybe you
need a new product.
As for what to expect, check out "On Your
Own" in Issue #51. Don and Kim jindra's experience advertising their $25 network should
give you a good idea what happens.
My advice for prospective advertisers is:
(1) Figure out how much space you need to
describe your product. If you're selling something that's easy to describe, you don't need
much space. If it's a totally new idea, or a substantial improvement on an old idea and requires a lot of explanation, plan to buy a lot of
space.
(2) Figure out how much image you need to
buy. (This is very different from #1.) Let me give

you an example: if you're selling a $1,500 CAD
package, be prepared to buy full-page four-color
ads and find the best ad agency to produce
them. Your ads must create a very solid image
for your product. If you're selling a $25 network, a simple micro ad may draw more orders
than any other size.
(3) Make sure your ad gets seen. (This
sounds a bit like #2.) If you only need a tiny ad
(a cheap product that's easy to describe), you
may find that a micro-sized ad grouped with
other micro ads may be better read than a quarter page hidden among the articles. If you need
at least a half page to tell your story and you're
afraid of getting lost, then go for a full page. Full
pages are very visible. Four-color full pages are
most visible.
Also, if you're buying a half page or more,
you might as well act like an experienced advertiser. Ask for a right-hand page, outside corner,
and ask them to place it with a particular
column or article. (If it's a very large magazine,
ask them to place it in the front third). They'll be
glad to charge you extra to guarantee you a particular position, but ad people can often finagle
you a visible spot, no extra charge, if you simply
mention you'd like it.
Plus, you may as well tell them you're acting
as your own agency. That way you'll get the
standard 15% agency discount, off the top.
(4) Does the ad have to make the sale? Not
always. You can send out a lot of information in
a I-ounce letter. Just make sure your ad is as
good at weeding out weak prospects as it is at
attracting strong ones.
(5) Something other than ad space? There's
direct mail. Figure on spending anywhere from
$.75 to $1.50 per piece. (That counts label, envelope, return envelope, literature, postage, design.... ) Also figure on a 2% to 5% response, so
you'll spend from $15 to $75 on each order.
Generally, magazine ads are a faster,
cheaper, easier way to reach a lot of people.
Mail works best as a follow up to enquiries
generated by your ads and for selling additional
products to recent purchasers.
MICRO CORNUCOPIA, #53, May, 1990

Now That You're Going To Advertise
(1) Order media kits. The kits contain
the latest ad rates (rates can change from
one month to the next), ad deadlines, ad
sizes, how they should be produced ....
(2) Watch the deadlines. If you ask for
space, you'll probably have to pay for it
even if you don't get your materials to
them on time. Stay on top of your ad. It
may be obvious to you and me that ads
need to change as products change. But
you'd be surprised how many small
companies run outdated ads for six
months simply because they keep
missing deadlines.
(3) Make sure your ad fits. A lot of
people send in incorrectly sized ads.
(4) Make sure it looks good. You want
solid black type, clean lines, white paper,
no smudges, no grey, no dirt.
(5) Keep a second copy of your ad (a
high-quality, printable copy). Remember, Murphy works for the U.S. Post Office and his brother runs Federal
Express.
(6) Check everything for accuracy
(particularly the phone number). I know
more than one old lady who's been inundated with orders for hard drive utilities.
(7) If your product runs under $100
and your ad's doing the selling, you may
not want a reader service number. I
know several companies who have been
crippled by the costs of replying to
reader service numbers. (They assumed
most of the recipients would respond;
turned out few, sometimes none,
ordered.)
(8) Talk to the magazine's editor. Ask
him if he thinks you'd do well advertising there. Also, ask him if there's someone you should send a copy to. (Don't,

absolutely don't, mention that your decision to advertise depends on his decision
to review the product. You won't feel
good about it and he won't either.)
Once The Ad Has Run
(1) Ask everyone who calls where
they heard about your product.
(2) Make sure you can respond immediately to orders. We started Micro C out
of our house and we lived on a dead-end
road. We'd meet the mailman as he
stopped at our box and many times we
had reply cards addressed and ready to
hand back to him by the time he'd gotten
back to our house. We had 10-minute
turnaround and subscribers loved it.
(3) Be prepared to lose money and
patience on some orders. Occasionally
you'll run into a real bother. He'll want a
custom product, he'll want you to install
it, he'll call you day and night instead of
reading the manual, he'll erase his disks;
and after you finally return his money,
he'll write a nasty letter to your state's
attorney general suggesting that the
death penalty might be too lenient. (The
guy was from New York City, of course.)
Note: Circulation figures are, at best,
optimistic. When you're on the phone
with a new magazine, ask for a breakdown of their circulation. (Ask the questions in the following order.)
(1) Is your circulation audited by ABC
or BPI? If so, great; have them send or
fax you the latest publisher's statement
and skip the following questions.
(2) How many paid subscribers in the
most recent issue? (Not next issue, or the
end of next year.)
(3) How many copies of the most recent issue were sent to the newsstands?
(4) What percentage of the newsstand

copies normally sell? (40% to 50% is normal, 60% is high, 70% is almost unheard
of)

(5) What was their total print run for
the most recent issue?
Listen for pregnant pauses, hemming
and hawing, fudging, explaining, etc. If
the totals of #2 and #3 add up to #5, then
they're probably fudging. Either way,
take printed circulation figures with a
giant grain of salt.
Also, don't be afraid to call a magazine's advertisers. They'll tell you a
lot. Some magazines are comfortable to
work with, others aren't. Some will
make deals, others won't. Sometimes
companies find the best response in
brand new magazines, other times they
do best in the old timers. In some publications the response is immediate, in
others it takes three ads. Sometimes ad
placement is very important, other times
it isn't.
You'll be amazed how much information you can glean in half a dozen calls.
The Carrot
Of course, there's no feeling like
opening your post office box and seeing
it full. Really full. Tends to make all the
hassles, all the preparation, worthwhile.
Reminds me of a little known saying
that's been handed down from hacker to
hacker over the centuries.
"May your bugs be gentle, may your
CRT be warm upon your face, and may
your mailbox be ever filled with checks.
Good checks."
Writers Market
Speaking of checks in the mailbox,
there's more than one way to squeeze
cash out of a computer. If writing software isn't your bag, then how about articles? The most effective writers:
(1) Have a specialty: processor design, graphics algorithms, database diagnostics, entrepreneurial experience ....
(2) Have a conversational style: An
open, easy style rather than a stiff, aca-:
demic, jargonizing (agonizing) presentation.
(3) Write for the magazines they keep
in their bathrooms.
(4) Don't take themselves too seriously. I'm talking about both their dealings with editors, and the way they
write.
I wish you well, my friends.

•••
Reader Service Number 2lO

MICRO CORNUCOPIA, #53, May, 1990

MARKETING ALTERNATIVES
Phone

Cire.

Andrew Binstock

213-473-7414

6,000

111 E. Drake
Ft. Collins, CO
80525

Bill Gates

303-491-9092

15,000
projected

500
250
175

Keith Weiskamp

202 E. Greenway
Phoenix,AZ
80532

Jeff Duntemann

602-493-3070

20,000
projected

1,270
775
420

Tech Specialist

Robert Ward

2601 IowaSt.
Lawrence, KS
66047

Donna Ward

913-841-1631

10,000

1,150
575
290

TheCUsers
Journal

Robert Ward

2601 Iowa St.
Lawrence, KS
66047

Donna Ward

913-841-1631

27,000

1,150
575
290

Circuit Cellar

Daniel Rodrigues

12 DepotSt.
Peterborough, NH
03458

Rose Mansella

203-875-2199

20,000

995
600
350

Programmer's
Journal

Liz Oakley

PO Box 31060
Eugene,OR
97403

Jay Moore

503-747-0800

30,000

1,895
1,195
645

Computer
Language

Regina Ridley

500 Howard St.
San Fransicso, CA
94105

Regina Ridley

415-397-1881

70,000

3,270
1,940
1,145

Magazine

Publisher

The C Gazette

Andrew Binstock

1341 Ocean Ave.
Santa Monica, CA
90401

Midnight
Engineering

Bill Gates

PC Techniques

Contact

Address

The C Gazette is a code-intensive C and C++ quarterly for MS-DOS
users. It specializes in in-depth articles aimed at the experienced C programmer. Code supports the major compilers and is accompanied by
clear, cogent prose. The C Gazette is committed to the exploration of
serious programming topics. Bruce Eckel of Micro Cornucopia has been
the Gazette's C++ editor for the last year.
Midnight Engineering, the journal of personal product development, is the magazine for hardware and software developers who want
answers about developing and marketing their own products. Coverage includes just-in-time development, product pricing/marketing, and
tips/techniques for working "on your own." Published bi-monthly, a
one year ME subscription costs $19.95 for the introductory period.
PC Techniques is designed for a market of proven buyers of programming languages, tools, application software, hardware, and more.
Readers are interested in the practical applications of programming related software products. Every issue will present hands-on software reviews, programming tips, and practical advice from leading software
developers and authors.
Tech Specialist is written for advanced PC developers. This powerful group of professionals works in both software and hardware
development, using the stand-alone PC as a tool. Tech Specialist gives
PC developers a "real world" tool - integrating information about hardware manipulation, software design, and the internals of the Pc. Each
issue provides an in-depth analysis of specific development problems.
The C Users Journal provides an interactive forum for C programmers. The Journal publishes practical information for C programmers,

Ad Rates
Full page $450
1/2 page 230
1/4 page 130

including advanced and intermediate programming hints and techniques, tutorials, and information about software tools. Each issue features major "how-to" articles written by experienced C programmers, C
product reviews, book reviews, correspondence from members, and information on new products.
Circuit Cellar INK, The Computer Applications Journal, is the
premier source of practical technical information for designers and
builders of computer hardware and software applications. Circuit Cellar INK enhances the electronics design skills of its readers by offering
creative solutions and unique applications through complete projects,
practical tutorials, and useful design construction techniques.
Programmer's Journal publishes articles written by the industry's
top professional talent - working programmers who are willing to share
their valuable technical expertise for the benefit of fellow colleagues.
Every other month, professional developers worldwide go to PJ to find
language articles with useful source code on Ada, ASM, BASIC, C,
C++, COBAL, FORTRAN, FORTH, Modula-2, and Pascal - plus the
latest information on OS/2, the 80386/486, the EGA and VGA, new
product releases, industry news, and consulting tips.
Computer Language is the productivity magazine for professional
software developers. Its mission is to enhance the programmer's ability
to generate working applications on schedule. Each issue of Computer
Langague includes columns devoted to application design and analysis, productivity tools, user interface design, object-oriented programming, and "how to" articles written by the finest professional
programmers in the field.
MICRO CORNUCOPIA, #53, May, 1990

Letters continued from page 6
figured it might not be the place for an
urban keyboard potato like me. I may
have been wrong, but about the rafting:
would I actually have to go outdoors
for that, or could you digitize the experience and send it to me on a floppy?
(By the way, does SOG stand for Sages,
Oracles, & Gurus or Silly Old Geezers?)
Anyway, thanks for putting out the
best computer mag around. And don't
give a second thought to that volcano
when you're trying to get to sleep. The
chances of it suddenly exploding into a
violent torrent of boiling lava and poisonous gases and burying the entire
town under glowing cinders are really
small. Less than 1 in 10, probably.
M. N. Macleod
3043 S. Laredo Circle
Aurora, CO 80013

Editor's note: SOG stands for Anything
You Like-no, that's not it.. .. I suppose you
know that Bend almost closed when you
left. Fortunately, the other resident stayed.
As for the dead volcano, the humane society
picked it up last weekend. (The smell was
getting awful.)
Correction
As if the discussion isn't confusing
enough already, we managed to slip an
error into James Martin's letter on the
radar equation controversy (see Letters,
Issue #52).
At the bottom of the right hand
column on page 78, we refer to the total
power at the detector as proportional to
the inverse cube of R. Wrong. The correct term (as stated in James's original
letter) is the inverse fourth power of R.
How can this 25% power loss have
happened? Perhaps we're seeing proof
of the First Law of Preservation of Postal Power.
The PPP theory states that in all
Postal interactions, power is conserved.
In other words, the Postal Service's proposed 25% increase in rates will be
balanced by a corresponding decrease in
all mail contents. Good news for all bill
payers, but be sure to tell your employer to stop sending your checks
through the mail.
We hope this will serve to clear up
the confusion and save James' sullied
reputation.
Micro C Staff

68 MICRO CORNUCOPIA, #53, May, 1990

More UNIX
Bob Morein's "UNIX Packages For
The PC" in Issue #50 was exactly the
type of information I've been looking
for. I turned up several other articles as
well as two useful books. (See my list at
the end of this letter.)
One of the things I'm discovering is
that UNIX is quite particular about the
hardware. For example, at least three
flavors of UNIX (and XENIX) do not
support RLL hard disk controllers, nor
do they support many video display
boards.
I haven't decided which UNIX I'll
purchase. That decision will have to
wait until I find out more about which
flavors support which hardware. Meanwhile, I hope you will have enough responses to Bob's article to encourage
more articles in Micro Cornucopia.
Fiedler, D., "Future Imperfect,"
BYTE, May 1989, p. 113.
Fiedler, D., "Calm Approach to
UNIX," BYTE, Aug. 1989, p. 113.
Combau, G., "The UNIX Shell,"
BYTE, Sept. 1989, p. 315.
Unger, J., "One Man's Experience,"
BYTE, May 1989.
Morein, B., "UNIX Packages for the
PC," Micro C, Nov.-Dec 1989, p. 40.
Fiedler, D., "UNIX on Personal Computers: Why and How," BYTE, Sept.
1989.
Christian, K., The UNIX Operating
System, 2nd edition, John Wiley and
Sons, 1988.
Fiedler, D. & Hunter, H., UNIX System Administration, Hyden Books, 1986.
Larry Hoffman
10152 Oso Ave.
Chatsworth, CA 91311

Editor's note: Wonderful list, but you
missed the original tome on the subject: The
Hardware Requirements Of Eunuchs,
Xinu, ML VX Be.
Op-amp Expose
Dear Mr. (smug) Editor-Regarding
your comments to my letter in Issue
#52: I do not appreciate your condescending and insulting response to what
was intended as a constructive correction. Turnabout being fair play, let the
games begin!
Obviously my letter isn't the only
thing you've ignored; OpAmps-l0l
must have occurred coincident with

your nap time. For your sake (as well as
others who have yet to tame the beast),
I shall endeavor to put into words the
secret of life, op-amps, and everything.
The op-amp is a device with two
(count them) inputs: one inverting (-)
and one non-inverting (+). It also has
one output (isn't that wonderful?).
Now, whatever happens on the non-inverting (+) input causes the output to
move in the same direction. Amazingly,
whatever happens on the inverting (-)
input causes the output to move in the
opposite direction. I guess that's why
they call it the inverting input.
The real secret (I know you've been
waiting for this) is: the op-amp will
swing its output, hoping against all
hope, waxing brave in this cold, cruel
universe, that somehow, somewhere,
someway there is a feedback path that
will bring its inputs into balance by
virtue of its efforts. This is usually done
by providing feedback to the inverting
(-) input. Yes friends, indeed, that little
booger won't rest until its inputs are at
equilibrium. The same. Identical. Twins.
You get the picture.
Knowing what we now know, let's
walk through the circuit as published
with a test case. Suppose we let
VI=1.01O volts and V2=1.000 volts. By
the equation stated (Vou t=VI-V2) we
should expect Vout to equal 0.010 volts.
The (-) input will be at V2/2 (0.500
volts). The (+) input has ... well, it depends where the output was to start
with. (VI and V2 in the original article
should be swapped for the equation to
apply. Change the designations in either
the schematic or the equation, but not
both.)
Let's say the output was at 0.000
volts just a microsecond or two before
we applied the input. Since 0.000 volts
is as good as ground, the (+) input is at
Vt/2 (0.505) volts. The inputs aren't
balanced so the output will simply rise
until they are. (Ooh, we've got me
now!) There, see! The (-) input is at
0.500 volts, and the (+) input is uhm, er,
even bigger than it was before! (Just a
minute, let me grab my calculator.)
Nope, the inputs still aren't balanced.
"Hmmph" says Mr. op-amp, "I'll fix
that. I'll just raise my output even more,
and hope like crazy that somehow my
(-) input will rise to equilibrium with
my (+) input." Things are even worse;
the (+) input rises higher still.

"Now my inputs are really getting
out of balance. I'll show them!" Mr. opamp zings his output as high as he can
go. "Surely my inputs will be balanced
now."
Sadly, Mr. op-amp has pegged his
output, his (+) input higher than ever,
and his (-) input still with only 0.500
volts on it. Mr. op-amp does the only
thing he can do. He keeps on holding
his output high, for surely somewhere,
somehow, someway his (-) input will
rise, and he waits. He will be waiting a
long time.
Yes, op-amps are a bit more complicated than that, but you won't stray far
from what's really happening if you
keep "the secret" in mind. Maybe if
anyone (else) cared, mine would not
have been the only letter pointing out
the mistake. After all, this is a magazine
about micros; who cares if you lie about
anything else?
What you presented in Issue #49 was
little more than a comparator (and a
poor one at that-too much hysteresis).
If you still don't believe me, build it.
Perhaps in your haughtiness, you, not I,
were thrown by the inverting (-) input
not being on top. In closing, may I suggest you either stand corrected or sit in
shame.
Dave 5tojan
10310 Lybert Rd.
Houston, TX 77041

Editor's cursory response: At first I was
kind of plus-minus about your letter, but I
like the way you amplified the subject.

do, any computer can do. Of course, any
theorem is based on assumptions; and
unfortunately no theorem can make my
BBII do color graphics. Still, for composing this letter, my BBII is just as good as
the latest 32-bit screamer. A "Turing
theorem" for aircraft, on the other hand,
would have little practical value. There
simply aren't many tasks at which a B2
bomber and a Stinson are equally effective.
2. Unless your Stinson spent the 30
years in a museum, I'll wager the fabric
and most of the engine aren't original.
Someone has shelled out hard cash just
to keep that Stinson from becoming
worse than obsolete. My BBII has consumed negligible cash.
3. If 3 BB years equal 30 Stinson
years, then 10 BB years is a century.
That spells "antique" where I come
from. So if I hang on a bit longer, I can
contest the high ground now held by
the Imsai.
Of course you'd have to dezonk your
Stinson analogies. To atone for past
excesses, you might even start a semiregular old folk's department. "The
Slow Lane," or "80s in the 90s," or "Is
There Life After Obsolescence," or "The
Trailing Edge." Something like that.
I guess I'm not your only reader who
wants to know where to load up on 8"
disks before antique dealers drive the
price up. Maybe there are others who
believe that good software won't appear
until the hotshots have moved on.

textflle.new

textflle.old

hello

',j.Suti'
good
~

with

DELTA
TIlE. COMPARISON TOOL

• Side-by-side scrollable
presentations
• Compare individual text
files or entire directories
• Menu-driven interface
• Built-in editor allows file
change while viewing the
comparison
• Target files can be chosen
from directory comparisons
• Ideal for programmers
Requires an IBM compatible with at
least 384K. A hard disk is recommended.
DELTA runs under DOS 2.0 or higher.

R. W. Hartung
408 Orchard

East Lansing, MI 48823
Oh, Go Take A Flying @#%*
You very nearly blasted me off your
list of subscribers-back when you remarked that your Stinson had taken 30
years to become as obsolete as my BigBoard II had become in three. I never
owned a Stinson, but I used to park my
7AC in a hangar next to one, so your
remark hit me with a zonk of 10 on the
Richter scale. However, many months
went by before I needed to renew Micro
C, and time is a notorious healer. I'm
still here.
I won't attempt to refute your remark. But I'll soften the zonk ("Obsolescence, where is thy sting?") by recalling
the following:
1. Mr. Alan Turing proved a theorem
which says that what one computer can

See the difference
at a glance ...

Editor's Hand-Typed Response: It
sounds like you're really up in the air over
my analogy, but I won't let that bring me
down. Would you accept a takeoff on flying
in place of an apology?

•••

Order Now. $79.
VISA or MasterCard accepted.

Also from OPENetwork ..

The Berkeley Utilities
Enhance your DOS envirorunent with
this set of powerful UNIX tools

TextLib
A screen & keyboard management
library for C developers

Write or call our RBS for information

POWER TOOLS FOR POWER USERS
21 5 Berkeley Place, Brooklyn, NewYork 1121 7
BBS: 718-638-2239
Voice: 718-638-2240
1-800-542-0938

Reader Service Number 201

MICRO CORNUCOPIA, #53, May, 1990

Build A TTL To Composite
Video Adapter
Have some old composite B&W monitors
you'd like to use? Here's a simple project
that'll give many of them new life.
hen you built your clone, did
you wish that you could find a
use for your old composite
video monitor? I had a TTL monitor for
my clone system, but that left me with
nothing to do with myoId composite
monitor. I wanted to add the monitor to
a second, 80188-based clone that I had
designed, but it just wouldn't operate at
the IBM frequencies. It wouldn't, that is,
until I built the circuit described here.

Background
Most CP1M systems use a monitor
that operates at a horizontal sweep
frequency of 15,750 Hz and a vertical
sweep frequency of 60 Hz. This, conveniently, is the same as standard American television. For a standard monitor,
the sync, blanking, and video all transmit
on a single coaxial cable.
The IBM PC and clones generate a
horizontal frequency of 18 KHz and a
vertical frequency of 50 Hz on a 9-pin
"D" type connector. These frequencies
allow more lines and better resolution,
but they're a problem for standard monitors.
A circuit to convert the TTL-level signals from a PC or clone is easy. Unfortunately, most composite monitors cannot display the entire picture; they lose
the first few characters. The problem is
quite simple: the monitors internally
generate a horizontal sync pulse that effectively blanks the screen for a few microseconds after the external horizontal
sync pulse ends.
The PC begins sending the first characters almost immediately after the sync
pulse. The monitor ignores these characters, and the result is a partial display.

70 MICRO CORNUCOPIA, #53, May, 1990

By Stuart R. Bell
741 Okie Ridge

Yukon, OK 73099

So how do you fix this problem? First,
let's delve a bit deeper into how the
video signal works. A composite video
signal consists of three basic components:
video, vertical sync, and horizontal sync.
The video signal is made up of the
"dots" that the screen displays. The dark
areas of the screen are at the blanking
level. The dots you see are at the white
level.
.
The vertical sync occurs once per
frame and synchronizes the vertical
sweeping of the screen. The horizontal
sweep occurs once per line and synchronizes the horizontal sweeping of the
screen. We're interested in the horizontal
sync.
On each horizontal line, the PC generates a horizontal sync pulse that's a few
microseconds wide. As mentioned, the
video information begins immediately
after that. The composite monitor cannot
display anything for a few microseconds.
If we could delay the video until the
monitor had recovered from the sync
pulse, everything would be all right. The

circuit to
convert the
TTL-level signals
from a PC or clone
.
IS easy.
first character on the line would appear
at the left side, and everything else on the
line would shift to the right by the same
amount. Unfortunately, this approach
uses a lot of parts to delay the video signal.
A second approach would be to anticipate the horizontal sync pulse and
generate a new one that would occur a
few microseconds early. This would have
the same effect as delaying the video in-

Figure 2;-Parts List ,for TTL to Composite Video Adapter
Quan

Description

Comments

74LS123
U1
U2
74LS04
7406
03
;Oluf cap
Use mylar or silvered mica
.001 uf cap
Use mylar or silvered mica
20k,10ttirnpot,
Position adjust
10 turn pot Brightness adjust
100 ohm, 1/4 W resistor
lSOohm, 1/4'wresistor
390 ohm,l/4w resistor
470 ohm, 1/4w resistor
6~8kl/4·wresistor

lN914'diode
. Olto .. l uf bypass capacitors
DB9P connector

formation. This circuit works that way,
sort of.
I say sort of because I haven't yet discovered the psychic logic that anticipates
a pulse. Instead, this circuit takes into account that all the horizontal sync pulses
are identical.
Instead of anticipating a sync pulse
and substituting an earlier one, this circuit delays the horizontal sync pulse
until it's time to start the next line. Each
horizontal sync is delayed by almost-butnot-quite a full line. The horizontal
sweep frequency generated by the PC is
about 55 microseconds.
If the horizontal sync pulse for any
given line is delayed by, say, 53 microseconds, and then used to generate the
sync pulse for the next line, the second
line will apparently shift right by 2 microseconds.
About The Circuit
The circuit (see the schematic in Figure 1) is straightforward. It uses only
three common TTL ICs. The DB9P connector at the left connects to the circuit
board with a short cable and mates with
the DB95 connector on your video board.
The composite video connector at the
right is a standard phono jack, which
connects to the composite monitor with a
short piece of coaxial cable.
The 500 ohm potentiometer adjusts
the brightness. The 20K potentiometer
adjusts the horizontal position of the picture by varying the amount of horizontal
sync delay. The circuit generates standard video levels of a v for sync, 0.5 V for
black, and about 2 V for white level.
Building The Circuit
I constructed the prototype on a small
piece of perfboard. The potentiometers
are mounted for easy access with a
screwdriver. I stole 5 V from my power
supply, which is open frame. You could
use an external supply, as long as it
MICRO CORNUCOPIA, #53, May, 1990

shared a ground with the circuit board
and the computer.
Be sure to use .01 to .1 JlF capacitors
to bypass the circuit. Don't use ceramic
capacitors for Cl and C2. If you do, the
picture will probably have the jitters and
will drift. The diode shown in the schematic is a IN914, but almost any fast
switching diode will work.
Once you've built and connected the
circuit, adjustment is easy. First, adjust
the 500 ohm pot about halfway between
both ends. Then adjust the 20K pot to get
all the characters on the left side of the
screen. If you set the value of the 20K pot
too large, it will not receive sync pulses,
and the picture will be scrambled.
After adjusting the horizontal position, set the brightness using the 500 ohm
pot. If you can't get the brightness adjusted to your satisfaction, or if the picture tends to "tear" at the top, increase
the 150 ohm resistor at V3 pin 6 to 220
ohms, and the 470 ohm resistor to 680
ohms.
Adjusting The Monitor
You'll need to adjust most monitors
before you can use them. They usually

have internal oscillators that run at the
horizontal and vertical frequencies of 15
KHz and 60 Hz, and they must be adjusted.
The easiest way to adjust the monitor
is to build the circuit, but leave VI out of
its socket. Jumper VI pin 1 to VI pin 5.
This will pass the horizontal sync pulse
through to the monitor. Then connect the
adapter to the video board on the PC and
to the monitor, and set the 500 ohm pot
about halfway.
The vertical adjustment on most
monitors is a pot, and will go down to 50
Hz without any trouble. The horizontal
adjustment may be a pot, and may go up
to 18 KHz.
If the pot doesn't make it (or there
isn't a pot), there's usually a coil with a
ferrite core somewhere near the horizontal pot. There may be another coil that
adjusts the width. Tinker with the horizontal coil and the vertical pot until you
get a stable picture on the screen. Then
install VI, remove the jumper, and adjust
the adapter.

monitor. The horizontal section of the
monitor generates several thousand volts
for the anode on the CRT. The CRT acts
as a large capacitor and it can hold a
charge for days.

(Editor's note: You'll see a well insulated
wire running from a transformer on the
board to an insulated cap on the side of the
CRT. That's the wire carrying thousands of
volts to the anode, and 1'd be very careful
around that entire part of the circuit. Use an
insulated tool when adjusting the core of the
horizontal oscillator, and keep your hands out
of there (even when the monitor's turned off).
Second, keep in mind that some monitors just cannot display the entire picture
of a Pc. Most can, but some will lose a
character or two off one end or the other
no matter what you do.
Finally, a (very) few monitors will
overheat if you run them at 18 KHz.
After adjusting the adapter, let the monitor run for a few minutes, then tum it off
and see if anything is excessively hot. If it
hurts to touch it, it's probably too hot.

•••

Cautions
First, be careful working inside the

Program FASTER with •.•

BUILDING BLOCKS
FOR THE ASSEMBLY
LANGUAGE PROGRAMMER

CC-RIDER
THE C PROGRAMMER'S COMPANION

. .

PROFESSIONAL EDITION

~~
~~

QUANTAS., POWER UB
Price $99.95
With Source Code
$299 •95

StH $5.00

Outside U.S. ~15.00
CA ReSidence
Add 7% Sales Tax
MCand VISA

CROSS-REF Editing! Function Prototypes! QulckHelp, too!
Instantly recall your C
symbol defin~ions by hming
a hot-key inside your ed~orl
Along w~h standard
CCRIDER features like
Source View, Ed~ and
Defin~ion Paste, the new
Professional Edition has a
Symbol Browse mode and
even Jets you walk through
all usages of a symbol. ~'s
on-line cross-referencing
ss you editl

Build a database of useful
information for your C code
w~h CCSYM, a powerful
SOUTee code analyzer
which can generate fully
commented function
prototype finclude files
for your program's static and
global function definitions.
Imaginel- No more manual
maintenance of function
declarations - and you
know they are consistentl

CCSYM can also create a
help database for your
program which is
col11>atible w~h Microsoft's
help system, QuickHep. All
symbols in your application
are accessble from
QuickHelp's menus,
together with
documentation extracted
automatically from your
original SOUTee code.

Reader Service Number 169

Standard

Professional

~
~
~
~
~
~
~
~

Over 256 Assembly Language Routines
No Need to Re-Invent the Wheel Every Time
You Write a New Application
Permits Faster Application Development
Overlapping Windowing System in Less Than 3K
Fast and Compact
Ideal For Programming TSR's
Cand Pascal Interfaces Available
MASM 5.X Compatible

ALSO AVAILABLE:
ASMFLOW $99.95
An Assembly Language
19855 Srevens Creek Blvd, Suite 154 Flow Charting and Source
Cupertino, CA 95014
Code Analysis Tool
(408) 244-6826
IUJA~iIf Silli
r2UiJR,9mR~~lmil\'

Reader Service Number 139

MICRO CORNUCOPIA, #53, May, 1990

PLANASOG VACATION FOR 19901
Longhorn So

;:::.: :.... .... . ,' ...... ::::. .:...>::,,0,::,,:. :'.'

. '.

Announcing SOGEast '90
The Eastest SOG of them all returns for an all new show on
August 17 and 18, 1990, in York, Pennsylvania.
This SOG starts with a picnic on the evening of August 16,
around 4 or 5 o'clock. The picnic includes delicious fare fixed
on site for your dining pleasure. There will also be frisbees,
softball, volleyball, and more. Come in early and spend time
meeting fellow SOGees.
th
The 17th and 18 (Friday and Saturday) will be chock full of
seminars, ranging from Hardware Control Using C++, to DOS
Internals, ONline Secrets for Modem Lovers, and More. Friday
afternoon will sport a live panel discussing (calmly, I'm sure)
favorite programming editors, including QEdit, MultiEdit, Brief,
and others. If you want to stand up for your favorite, let us
know so we can save you a chair.
Friday evening is the First Annual SOGEast Dinner Party.
This is a sit-down, stand-Up, have a ball Ball. Don't change
your clothes from the SOG sessions, just come for the good
time. A special guest speaker will be introduced ....
Friday night the infamous Jolt SIG (with real Jolt this year!)
will convene for a night of fun and merriment. A special hardware project especially for software people is scheduled, as
well as a contest for Programmers vying for the title of "The
One Day Wizard of SOGEast." We'll even be awarding prizes!
If you are interested in the project or the contest, let us know.
We are looking for speakers on a large range of topics. If
you'd like to share some knowledge, pass along some secrets,
or amuse us for an hour, please call soon,
Current suggested retail prices are as follows:
Advance SOGEast '90 Registration (before June 1)
$25
Advance SaG East '90 Registration (June 1 - July 15) $35
SOGEast At The Door (Anything after July 15)
$50
Picnic
$10
Official SOGEast '90 Tee shirts
$10
Friday Evening Dinner
$15
Tables for Exhibitions
$25
(All prices are Per Person)
There is more, ... For the latest information, contact:
John Ribar
The CDS Group
P.O. Box 7549
York, PA 17404
(717) 792-5108 Evenings
Compuserve 735n, 1652
bix jribar
We're proud to announce that SOGEast '91 is already in
planning, This will be an annual event. This year, we plan to
hold the SOG at one of two local colleges. If you have suggestions for next year, pass them on.

AnYone
G '90
fami!
Who ha
~ reUnion's attended
WOUld be
(Without th
knows th
M'
Sad n
e aUnt
at a So
Icro Cornu
' ot to haVe
sand unde
G is like
!exas, June ~O:'a Will be a fo~~nghom SOGS);9Do~ and I feel i:
tlVe, fun-filled
30 (Friday_s
memory. S
0 JUst becau
The' , dayS,
aturday) 11
0 come to D
Se
May 5 mVltation has '
' Or two excitin ' enton,
5
So we W:II
Just one
,,
g, tnforma_
o attende
I haVe a he Condltlon-PI
We Will
es not COUnt'
ad COunt ~, ease hOl/er
, aSSUme th ' mg kids (k' ' ere 100kj
at Us by
fi
/ Ve reserved ere s not enou
I~S are We/co ng for at least
,eady giVing
COnference fi gh mterest.
me), otherwis
like to s
presentati
ooms and
e
Peak.,
ons, Be SUfi
Several 0
We'll kiCk thi
e to ca/{ soo ,f You are alOn Friday
ngs off With
n If YOU'd I
Jolt SIGs and Saturday
a HOWdy Party
a So
D'
, We 1/ hOld
Thursa
/ng,
have be '
rOund table
Interested' a nice COffege t aUt/ful Scenery
and
In COnt'
oWn '
Or ~h'
mUing th
jUst nOrth
Ite Wat
e SOG trad't'
of Dallas' s ,er raft".
lion r
' 0 If \,
,
~/m JiOd,
, Please c
J Ou re
In~
ra
ontact_
ormation
ap.O. 0 rawerFMOdes
eoton 1)(
(817) 38~ 76202
l

b~;~oZ do~sn't

diScu:~i:~:ning,

'-3339

RObo"SOG '9
Set as'd
0
I e Aug
OUr first So
ust 9-12 t "
S
G,
0 VISit th
, 'Ponsored
e Seattle
WJ/f Offer
by the S
area and b
mUch
eattle D
ea
Us know if
,Of What 0
Ifobotics,
Part
of course, th~U 'd like to S~~kexpect from aS~clety, AObo-S
'
O
We'll I
Jolt SIGs
), barbeqU
OG: qUality
G 90
reclined I a So haVe m~~
e, banqUet, freSh t~/ks (let
/, and
<.<.e.e-runn'
aIr
,oPen to all
a Junque d
mg rObots
' and,
mg YOUr
(rObots). If Yo e Jour tour 1',' camping (for
AI metal monst
u WOUld like ' he rObotics
thOse So
so, this S
Sr to d b
more '
maze
/f You 'fie '
OG Wi/{ ~
0 aWe dlio
Information
-run is
I '
mto e
leatu,'
if) Us
on P
mend), be
mbecJded
re m,crOCOnt
a POstcard
repar_
lA',
SUre to att
sYstems '0
rOl/ers a"'"
'
vve1/ b
end
I' r You'
"1..1 dat
Ii
commOd ,9 sending out' Or even bette Ve embeddP'"l a oggers.
atlons
compl
r, SPeak
"'1..1 Your b
maze speCifl' ',and SChed, tete details
: (WOOf!)
est
ICat/ons and
u es after AA on pnces tOe '
Conte
lv/arch 15
'
atlon
Mike Th
st gUidelines
. We'll al
' ac11036 ~hng
at that time So have
40 N
'
Seattle lA, .E.
II
' viA 98
VOice: (206
125
88S: (206 ~ 362"5373
Of

~362"5267

MICRO CORNUc
aPIA, #53, May, 1990 73

-I

Continuing

AROUND THE BEND
MicroC Garage Sale

· e on\'l\

''''leA \\~
\)~\\: u

Special I
,
c os eout!
While Quantities Last!
No rainchecks!
Genuine Antiques!
No Cliches!
All Sales Final!

130
1
10
12
10

70
6
900

Visual 50 Terminal (VT-52 compatible) like new
$60
Boxes of new 8" disks, 10 disks per box, SSSD
$5/bx
$100
Commodore 128 w1571 floppy drive, used 4 hrs.
81/2" reals of 1/2" mag tape, unopened
$10 ea
8" SS SOlDO Floppy Drives, Mostly Shugart 801 s
$10 ea
Cabinets, new, can hold 2 8" drives, includes PS
$10 ea
Complete Cromemco S-100 sys. w/boot disks, manuals $75
Kaypro 11-83, very used, working
$100
Kaypro 4-84, very used, working
$100
Big Board I with cabinet, two 8" drives & PS
$75
Big Board I in keyboard case with PS
$50
Baby Blue Z80 cards, run in 4.77 MHz XTs, w/utilities $10
Constant voltage transformers, Sola, 250 V.A. 11 OV $35 ea
2764s, Intel 200 ns, new (in tubes of 14)
$21/tube

Call for other goodies: For instance we have back issues of Lite:
J.i.nas (Lifelines?) and Computer Journal (CP/M anyone?). Plus, we
have numerous computer books.
All items sold as is. No refunds. All worked the last time we used
them (even the books). Call for availability as well as shipping
and handling charges. Shipping will be significant on some items
so Jf you' plan to be in the Bend area, definitely give us a shout.
(Shipping and handling are $7.50 (in the U.S.) for the Baby Blue
Z80 cards and $3.00 per order on the ROMs.)

To Order, Call Today, 503-382-5060
(9 a.m. to 1 p.m. Pacific Time)

MICRO CORNUCOPIA, #53, May, 1990

puter magazines (or how far they'd moved away from us).
What's important about CL's subscription guarantee isn't that
the magazine might replace Micro C, but that you can, without
risk, try it. You can see if it, too, doesn't become a valuable
resource.
I must also mention a couple of other magazines:
Programmers Journal (503-747-0800) is a smaller, less formal,
version of CL. It wouldn't surprise me to see PJ really grow,
both editorially and in circulation over the next few years.
Midnight Engineering (303-225-0856) could become another
Micro C. Bill Gates is aiming ME at the folks who are working
graveyard "On Their Own." Several Micro C regulars had articles in his first issue.
Meanwhile, if you want CL to look more like Micro C,
you'll just have to write articles for them. Contact Michelle
Williams at (415) 397-1881 for writers' guidelines.
We've already given the articles we've been working on to
J.D. at Computer Language and referred several embedded
authors to Tyler Sperry (a very old friend) at Embedded Sys-

tems.
Bruce Eckel's already writing for J.D. Karl Lunt, Bob Nansel, John Ribar, and Michael Hunt have already connected
with Tyler. Who knows, you might even see the Fogg in one
or both.
The Office Will Be Open
Starting April 2, we'll have a skeleton crew (Nancy) manning (womanning?) the office from 9 a.m. to 1 p.m. Pacific
Time. Nancy'll keep the office open at least through April,
probably through May. Sandy and I will no doubt be hanging
around the office, too.
We'll still take orders for back issues, XT and Kaypro schematics, as well as issue disks. Speaking of back issues, we've
hired a professional indexer to thrash through issues 1
through 52. It'll be on disk (on paper, too, we hope). We'll
send you the index for only $6. Or you get the index disk free
with an order for 10 or more back issues. It's best to order
soon; we're getting low on quite a few issues.
If you're into bargains (I'm talking historical treasures
here), check out the Micro C garage sale ad on this page. We
have lots of new 8" SSSD disks, 10 to a box (make great bases
for little league). We have 8" single-sided drives (take a rope,
tie a drive on one end, tie a boat on the other.... ).
We have (shudder) Big Boards for sale, a terminal, constant
voltage transformers for those of you with dirty power, and
something very, very special. We've got about 70 280 plug-in
cards. I was hoping to announce them at the end of the
LIMBO series (but we ended before the series did). You could
use them to develop ROMs for the 64180 (or run any other
CP 1M software on your PC/XT I clone). One problem, they
don't seem to work on any system running faster than 4.77
MHz. However, they're only $10 each (including software)
plus $7.50 each, shipping and handling.
It's best to order by phone (503-382-5060) so you can be
sure we've got what you want. (Or that you want what we've
got.)

AROUND THE BEND
Where Am I Headed?
Nowhere.
I want to have some time that isn't already spoken for. I've
spent my whole life in student mode. You know, if I'm awake
and not working, then I should be.
I can't think of anything more enticing right now than
spending two weeks or two months on a sailboat anchored off
a deserted island. By the end of two months, I'd be totally stir
crazy. I hope.
Finally, I've been feeling at odds with everything lately. I've
been tense, uncomfortable, absent-minded, jumpy, irritableall reminding me that Micro C and I are very close. But this
issue's edit-struggling line by line through every article-has
been nearly impossible. Each evening I take home the articles
which need my touch. Each morning they return untouched.
My heart's'not in it.
Sorry.
My Secret Wish
People who've called after hearing rumors of the end have
asked how I'm feeling ("sad" is the best I've come up with)
and what I'm doing next (nothing). No one has asked if something has been left out-if there's been something I've hoped
would happen, but hasn't.
Yes, there is something. For most of these nine years, I've
secretly wished that just once, for a day or an hour, every subscriber, every newsstand reader, every library browser who
regularly thumbs through Micro C, every one of you who has
followed and shared this journey-could get together. All of
us in one place at one time. Twenty-five thousand incredible
people.
That was my wish.
Solid Expectations
As of January 1, Miniscribe Corporation is under court protection from its creditors. The company threw in the brick after
a year of shipping half-baked clay blocks from warehouse to
warehouse, pretending they were hard drives.
I'm sure the bankruptcy surprised folks from the building
trades; after all, the demand for good, new bricks has never
been higher. However, these small red clay thingies aren't
commonly used to hold data.
That may change shortly. We've received late word (very
late word, actually) that Taiwanese computer manufacturers
are scrambling to retool their 256K production lines. Apparently there's a rumor running rampant around the island
that bricks are the latest in solid storage technology.
Short Note
The biggest Taiwanese exporter is IBM. (Wait until they
hear about bricks.)

Develop Powerful
Image-based Applications
Quickly and Easily

Introducing VICTOR, video capture
and image processing library
Create Image-based applications that display live video on
the VGA and offer powerful Image processing features
With Victor your applications can show live video on any portion ofthe VOA
screen and combine live video with text and graphics. Use sophisticated image processing tools, create lineart, TIFF/PCX files,
more. Victor handles all the low level functions so
you can concentrate on your application.

Rapidly test and prototype
Victor comes with ZIP Image Processing software to get you up and running quickly -- test and
prototype functions before you write any code.

Video frame grabber support
Victor supports EOA and VOA up to 800 x 600,
and video digitizers by: IDEC, CCI, Micromint,
DAK. HRT, Electrim, Willow, VO, and others.
Victor supports Microsoft C, QuickC, and
Turbo C, includes demonstration and prototyp- I~~~I~II
ing software, full documentation, and source code ~
for device control routines. " all for only $195.
Victor with video frame grabber $349
VISNMC/COD

Catenary Systems 470 Belleview St Louis MO 63119 (314) 962-7833
Reader Service Number 217

tl~ ON NEW PLATFORMS
~OREGONC++
•
•
•
•

Atrue compiler - not atranSlator
Three compilers in one: C++, ANSI Cor K& R C
Source-level debugger
Conforms to AT&T V2.0 C++ ~

2
PIASCAL
fi
tt#

•
•
•
•

~~~\;

f~l'~~~~I\~S
'~

Full ISO Level 1standard
Comprehensive error checking
Access to command-line arguments
Extensions include full string type, separate
compilations, extemal variables &more

OREGONMODULA~2

• Conforms to emerging ISO standard
• Excellent support of coroutine and interrupts
• Ideal for embedded systems

680XO CROSS-COMPILERS ALSO AVAILABLE

CALL 1-800-874-8501
(503) 245-2202

More Bricks
I was talking to Allyn Franklin, my favorite hard drive person (who knows, he may also repair bricks), about his "On
Your Own" article when he offered me a few tips on dealing
with a recalcitrant hard drive.

Your applications will hal'e
powerful image processing and
handling capabilities.

Call (314) 962·7833 to order

FAX: (503) 245-8449

OREGON
6915 SW Macadam, Suite 200,

SOFlWARE
Portland, OR 97219 USA

The following are trademarks: Oregon Software, Oregon Ctt, Oregon Modula-2, Pascal·2, Oregon
Software, Inc., Xenix, Sun-3 ,386i, VAX, VMS, Motorola, UNIX, AT&T, ULTRIX, NCR Tower, HP 9000

Reader Service Number 193

MICRO CORNUCOPIA, #53, May, 1990

AROUND THE BEND

BIOS

~gg~CE

The AT BiosKil gives you a complete Bios with source
code you can modify for your own applications! The
BiosKit includes a Bios on diskette ready for programming an Eprom, and includes the utilities you need to
Rom the source code. The Bios also has a Rom
Monitor/Debug and Setup. At last you have conlrol over
the core of your system. Over 380 pages, with diskette,
$199. The Xl BlasKiI is only $99, or get both for $279.
lbelnlel Wildcard Slppl.enl for the XT BiosKit is $49.
We'll include a free copy of the
pocket-sized XT-AT Handbook by
Cholsser and Foster with each BiosKit if you mention this ad when you order. Of course, this $9.95
value is also available by itself. Or buy five or more for
only $5.00 each.

FREE

~
Annabooks

800-462-1042
In California 819-271-9528

12145 Alta Carmel Ct Suite 250-262
San Diego, California 92128
Money-back guarantee
Reader Service Number 160

"It's Glorious!"
"No color PCompatible is complete without a copy of this;
get one and see what I mean."
- Jerry Pournelle, BITE
"Infinite•.. Neither of us (the chief engineer of the Galileo
Jupiter probe and I) may ever be heard from again."
- Arthur C. Clarke
"I could watch an animated image for hours."
- Larry Fogg, Micro C (MIA 89)

FractalMagic 5.0
America's Premier Fractal Program
mM PC, PS/2 (3.5" or 5.25")
CGA, EGA, & VGA: $35.00
Macintosh: Plus, 512KE, SE,
SE/30, and Mac IT's: $35.00
mM Expansion Modules 1-4:
$15.00 each. or all four for $45.00 .

Sintar Software
"Software for the Mind"
l0014th Avenue, Suite 3200
Seattle, WA 98154
(206) 625 - 1213
Visa - MC - AmEx

MICRO CORNUCOPIA, #53, May, 1990

"If something seems weird, don't turn the drive off. You
may have a damaged boot sector. Just back up everything you
hold near and dear.
"Often, however, you don't know there's been a problem
until you power up the next day. And, what often happens is
the system doesn't realize it has a hard drive. That's when I
run FDISK. Try all the versions of FDISK you've got. FDISK
might see the drive." (I just tried it. It works!)
"If you run FDISK with the same parameters as you did
during the original format, it won't eat any data. Normally
you have to do a high level format after FDISK, but in this
case you want to avoid that. With any luck, after you've run
FDISK, the system will again see your drive."
"If FDISK doesn't help, try reformatting just track O. Mace's
FORMAT-H is the tool I use a lot for this.
"You can reach Western Digital at (800) 832-4778. Key in
the type of drive and WD controller number and the system
will tell you how to jumper the card, and will give you head
and cylinder counts for the drive.
"There's an even longer list inside the newest Speedstor,
and Tandon has recently come out with DRIVES.COM.
DRIVES.COM queries your AT and your hard drive, telling
you which drive number to enter, etc.
"Both Mace and Norton have books out on data recovery,
but neither goes far enough. They are really aiming at the
average user."
Drives That Should Have Been Mortared

My brother grabbed me the other day, desperately looking
for a letter we'd run. He wanted the issue and page number of
the bit about 15 Seagate 238s that died. He wanted the copy
for a customer who wouldn't accept an RLL drive (and didn't
much care for MFM either).
Don had asked the guy if he was running 238s and 225s.
Turned out he was. And they'd failed.
If I were buying a hard drive right now, I'd get: A Sea gate
full height like the 8096, 8048 or 8038; a Miniscribe full-height;
or, even better, a Mitsubishi. The Mitsubishi I'm thinking of
can run either 40 meg MFM or 60 meg RLL. If I were concerned about speed and space, I'd run the Mitsu RLL. On the
other hand, there's nothing safer than an RLL certified platter
plodding along with MFM data.

Micro Cornucopia

Final Issue! Stock

Ue on Back Issues Now!

ISSUE #1 (8/81)
Power Supply
112 PFM.PRN
16 pages

ISSUE #17 (4184)
Voice Synthesizer
68000-Based System Review
56 pages

ISSUE #2 (10/81)
Parallel Print Driver
Drive Motor Control
16 pages

ISSUE #18 (6184)
Kaypro EPROM Programmer
1/0 Byte: A Primer
Serial To Parallel Interface
Business COBOL
60 pages

ISSUE #3 (12181)
Configuring Modem 7
Reverse Video Cursor
FORTHwords Begins
16 pages
ISSUE #4 (2182)
Keyboard Translation
Modems, Lync, and S10s
Undoing CPIM ERASE
20 pages

ISSUE #19 (8184)
Adding Winchester To BBII
6 MHz On The BBI
Bulletin Boards
Track Buffering On Slicer
4 MHz For The 820-1
64 pages

ISSUE #5 (4/82)
Two Text Editors
Double Density Review
20 pages

ISSUE #20 (10/84)
HSC 68000 Co-Processor
DynaDisk For The BBII
Serial Printer On BBI Sans S10
Cheap & Dirty Talker For Kaypro
72 pages

ISSUE #6 (6/82)
BBI EPROM Programmer
Customize Your Chars
Double Density Update
24 pages

ISSUE #21 (12184)
Analog To Digital Interface
Installing Turbo Pascal
Low Intensity BBI Video
80 pages

ISSUE #7 (8182)
6 Reviews Of C
Adding 6K Of RAM
On Your Own Begins
24 pages

ISSUE #22 (2185)
Xerox 820-11 To A Kaypro-8
Sound Generator For the
STD Bus
Reviews Of 256K
RAM Expansion
88 pages

ISSUE #8 (10/82)
SOLD OUT
ISSUE #9 (12182)
BBII EPROM Program
Relocating Your CP/M
Serial Print Driver
Big Board I Fixes
32 pages
ISSUE #10 (2183)
ISSUE #11 (4/83)
SOLD OUT
ISSUE #12 (6/83)
Bringing Up BBII
Double Sided Drives for BBI
Packet Radio
5 MHz for Kaypro
40 pages
ISSUE #13 (8/83)
CPIM Disk Directory
More 256K for BBI
Cheap Fast Modem
BBI Printer Interface
44 pages
ISSUE #14 (10/83)
BBlllnstallation
The Perfect Terminal
BBI Video Size
Video Jitter Fix
48 pages

ISSUE #23 (4185)
Automatic Disk Relogging
Interrupt Drive Serial Printer
Smart Video Controller
Review: MicroSphere RAM Disk
86 pages
ISSUE #24 (6185)
C'ing Into Turbo Pascal
8" Drives On The Kaypro
68000 Versus 80x86
Soldering: The First Steps
88 pages
ISSUE #25 (8185)
Why I Wrote A Debugger
The 32-Bit Super Chips
Program The 32032 Modula II
RS-232C: The Interface
104 pages
ISSUE #26 (10/85)
Inside ZCPR3
Two Megabytes On DSI-32
SOGIV
The Future Of Computing
Graphics In Turbo Pascal
104 pages
ISSUE #27 (12185)
SOLD OUT

ISSUE #15 (12183)
Screen Dump Listing
Fixing Serial Ports
Playing Adventure
48 pages

ISSUE #28 (2186)
Rescuing Lost Text From
Memory
Introduction To Modula-2
Inside The PC
104 pages

ISSUE #16 (2184)
Xerox 820 Column Restarts
BBI Double Density
BBII5"/8" Interface Fix
Recovering Text From Memory
52 pages

ISSUE #29 (4186)
Speeding Up Your Xl
Prototyping In C
C Interpreters Reviewed
Benchmarking The PCs
104 pages

ISSUE #30 (6186)
PROLOG On The PC
Expert Systems
Logic Programming
Building Your Own Logic
Analyzer
PC-DOS For Non-Clones
104 pages

Accessing dBase III Plus
Records from Turbo Pascal
96 pages

ISSUE #31 (8186)
RAM Resident PC Speedup
Practical Programming In
Modula-2
Game Theory In PROLOG, C
104 pages

ISSUE #45 (1/89)
Computer Aided DeSign
CAD In A Consulting Business
Choosing PCB Layout Systems
Building Circuits With Your
Computer
Secrets of Optimization
Finding Bargains in the
Surplus Market
MASM5.1
96 pages

ISSUE #32 (10/86)
Public Domain 32000:
Hardware And Software
Writing A Printer Driver for
MS-DOS
Recover A Directory By
Reading & Writing Disk
Sectors
96 pages
ISSUE #33 (12/86)
ISSUE #34 (2187)
ISSUE #35 (4187)
ISSUE #36 (6187)
ISSUE #37 (9/87)
SOLD OUT

ISSUE #46 (3/89)
Software Tools
The Art of Disassembly
Handling Interrupts With Any C
Hacking Sprint: Creating
Display Drivers
Greatest C Compilers
Turning A PC into An
Embedded Control System
96 pages

ISSUE #38 (11/87)
Parallel Processing
Laser Printers, Typesetters
Build A Graphics Scanner
For $6, Part 2
Writing A Resident Program
Extractor In C
96 pages
ISSUE #39 (1/88)
PC Graphics
Drwing Mandelbrot I Julia Sets
Desktop Graphics
Designing A PC Workstation Board
Around the TMS-34010
96 pages
ISSUE #40 (3/88)
The Great C Issue
11 C Compilers
Writing A Simple Parser In C
C++, An Object Oriented C
Source Level Debugger For
TurboC
96 pages
ISSUE #41 (5188)
Artificial Intelligence
3-D Graphics
Neural Networks
Logic Of Programming
Languages
Applying Information Theory
96 pages

ISSUE # 42 (7/88)
Maintaining PCs
Keeping Your Hard Drives
Running
Troubleshooting PCs
XT Theory of Operation
Simulating A Bus
96 pages
ISSUE #43 (9/88)
Building Databases
Build a C Database
Selecting a dBase 111
Compatible Compiler
Working with Paradox
Designing Custom PC Cards

ISSUE#44 (11/88)
Object-Oriented Programming
A Taste of Smalltalk
Actor
Thinking Objectively
Building MicroCad
Peripheral TechnologyPT68K-2
96 pages

ISSUE #47 (5/89)
Robotics
The LIMBO Project
Starting A Robotics Company
How To Write and Use A
SystemProfiler
Problem Solving and Creativity
Turn Your XT Into A Controller
Writing Code For Two
Operating Systems
96 pages
ISSUE #48 (7/89)
Tools For The Physically
Impaired
The Adventure Begins
Selecting A Talking Computer
For A Blind Friend
Writing Software For The Blind
File Transfer Via Parallel Port
The LIMBO Project-Part Two
PCX Compatibility
A 68000-Based Multitasking

Back Issues

Kernel
The Very Early Days of
Computing
96 pages
ISSUE #49 (9/89)
110,110...
Build A Computer The Easy
Way
PostScriptals
Prog. Logic Controllers
Driving Stepper Motors
Writing TSR Programs
Low Cost I/O For The PC
Interfacing 16-Bit Devices
96 pages
ISSUE # 50 (11/89)
3-D Graphics
3D Surface Generation
PC Video Frame Grabber
LIMBO, Part Three
PostScript, Part Two
UNIX For The PC
Capturing & Graphing A Voice
In Real Time: Part 1
96 pages
ISSUE #51 (1/90)
Embedded Systems
Embedding An Xl Motherboard
Writing A Neural Network in C
LIMBO, Part 4
Getting Started in Hardware
Capture & Graph a Voice,
3D Surface Generation, Pt 2
Marketing Your Own Software
96 pages
ISSUE #52 (3/90)
C++ 2.0
Training A Neural Network
Debugging C Pointers Using
MEM
The AT Keyboard Interface
Filling In The Holes On Your Xl
LogiC Families
EPROM Programming
Controlling Runaway C With
CC-Rider
96 pages

•••

To Order:
Phone:
Mail:

1-503-382-5060
PO Box 223
Bend,Oregon 97709

United States,
Issues #1-34
Issues #35-current

$3.00 each ppd.
$3.95 each ppd.

Canada & Mexico
All issues

$5.00 each ppd.

Foreign (air mail)
All Issues

$7.00 each ppd.

~----------------------------------------~
MICRO CORNUCOPIA, #53, May, 1990

AROUND THE BEND

eLI is an intelligent front & back end processor
that enhances the MS-DOS system kernel. eLI
provides many features of advanced OS's, like
VMS and UNIX, to MS-DOS without any re-training.
The core of the eLI kernel is based on recursive
definitions and references to Alias, Logicals,
Symbols, and Lexical-Functions from the command
line, batch programs, and existing application
programs (without any modifications).
Powerful command programs can be quickly
written for a wide range of purposes such as:
• flexible control of code & data management
• code development & code support tools
• automated file/device maintanence utilities
• creation of generic command procedures
• installations & system re-configurations
For the first time, designers, developers, and
managers can coordinate software development and
production efforts without the major limitations
and restrictions in the MS-DOS environment.
• VHS is a trademark of D.B.e.
• UNIX is a trademark of AT&T
• HS-DOS is a tradema~k of HicroSoft

Price: $ 149
S/H: $ 6
215-543-2658
Reader Service Number 187

16MHZ

68000
SINGLE BOARD COMPUTER

The new PT68K4 computer board has all the features of the K2 series
plus extra benefits. It is available in 2speeds -12MHZ and asuper fast
16MHZ - with increased memory capacity to a maximum of 12 MB!
* 4 RS232 Ports
* 7 XT Expansion Slots
* Floppy Disk Controller
* 1 Memory Expansion Slot
* Real-Time Clock
* 2 Parallel Ports
* Optional Operating Systems
BASIC KIT (12MHZ) - 4 layer board, Processor, Humbug Monitor,
4K SRAM, 2 Serial Ports
$220
K4 COMPLETE KIT (16MHZ) - with 512K of DRAM
$545
ASSEMBLED BOARD (16MHZ) - with 1MB DRAM
$649
PROFESSIONAL OS9 with CCOMPilER
$299
only valid if purchased with board
SK *DOS OPERATING SYSTEM:
$ 80
. COMPLETE INFORMATION AVAILABLE UPON REQUEST

PERIPHERAL TECHNOLOGY
1710 CUMBERLAND POINT DR. #8
MARIETTA, GA 30067
404/984-0742 -FAX 404/984-8248
COD /MASTERCARD /VI SA/CHECK
SK*DOS IS ATRADEMARK OF STAR-K SOFTWARE SYSTEMS
OS9 IS ATRADEMARK OF MICROWARE AND MOTOROLA
Reader Service Number 119

MICRO CORNUCOPIA, #53, May, 1990

NEC Can't Defuse Printer Problem
It's 3:57 p.m. It's quiet in Carol's office-her NEC 890 Silent
Writer is unusually silent.
"Dave, my printer doesn't work"
The power, which had been off for an hour, had just come
back on and all the computers were doing their deep knee
bends.
"Dave, my printer doesn't work"
"So give it a chance, already."
Unfortunately, it needed more than a chance. Nothing,
dead, no fan, no lights, no smoke, no whimper, no nothing. I
looked for a fuse. None in sight. None underneath, on top, by
the plug, anywhere. I opened the back and pulled out the
power supply. No fuse.
I called NEC. After being referred to three different sites, I
finally reached someone who would talk to me ....
"Where's the fuse on the 890 Silentwriter?"
"Before we talk about that, what are the symptoms?"
"It's dead."
"It could be something other than a fuse."
"Right, but for now I'd really like to know where you put
the silly fuse."
"I'm not sure. Have you tried to reboot it?"
"Reboot it? REBOOT IT?"
Of course, what was I thinking. Of course, I'll need an
operating system for a dedicated 68000. Let's see, DOS 3.1
doesn't run very well on an 8088.
Now, without a floppy drive, I'm not sure where to put the
3.1 boot disk (and I'm certainly not open to suggestions). And
with a totally dead system, how will I know when the silly
thing's been booted sufficiently? (Don't say it, I'd just hurt my
foot.)
800 Frustrations
I can name two significant events in the financial life of
Micro C. First, we got okayed as a VISA/Mastercard merchant, then we got an 800 number.
We had a dickens of a time becoming a VISA merchant. I
was talking to the last bank in town when I finally got the
okay. Banks seem awfully gun shy about small mail order
businesses like Micro C. Once we'd been okayed by one bank,
however, everyone wanted to step in. (No way.)
Working with VISA has been interesting. Wives cancel
husbands' charges. Card numbers that are okay one day are
not okay the next. People use other people's charge accounts.
But by and large, VISA has been great.
Getting an 800 number was a lot easier but a lot more expensive. We pay a monthly service charge plus a per-minute
fee during calls. The fee is often higher than the normal direct
dial rate, and like standard long distance, we pay more for
east coast calls than for west coast. But, an 800 number makes
it possible for people to order subscriptions or disks from
work, from a friend's house, or anywhere.
But I was worried that "orders" might become half-hour
tech calls. What I didn't worry about was misdialed calls. I
should have.
None of the 800 services tell you they assume no responsibility for wrong numbers. Initially that wasn't a problem. We
received maybe two or three erroneous calls a week Lately,
however, we've gotten as many as 15 wrong numbers per day

AROUND THE BEND
(besides our usual half-dozen calls for subscriptions).
A travel club called Excellence in Exercise is advertising
tours to east coasters. Their number is 800-888-TOUR. Great.
That's 800-888-8687. But if people dial zero instead of the 0 in
TOUR, they get 800-888-8087. That's Micro C.
Calling the company hasn't helped, after all they're doing
nothing illegal. So don't be surprised if you dial our 800 number after April 1 and you hear it's been disconnected. You can
still use our regular order number: 503-382-5060.
(But maybe we're missing a much better solution. We could
just put out the word there's a new toll-free Lotus 3.0 support
number-something ending with 8687.)

Who's Selling Computers?
I've been watching The Wall Street Journal again. IBM, Apple,
and Radio Shack are all reporting that computer demand is
drying up. Meanwhile, MicroSphere is scrambling to keep up
with orders.
I called my brother.
"Hello, is Don there?"
"I'm sorry, he's tied up with a customer."
"But I'm his brother."
"Then he's especially tied up."
Busy is one thing, but being tied up with a customer is
pretty kinky.
Later, after things unraveled a bit, I asked him what was
happening in the industry.
"You know, Intel's profits are way up. It's all those 386s
they're shipping. People are buying 286 and 386 systems like
they're going out of style.
"Meanwhile, Radio Shack is now suffering from its reputation for incompatibility. They set up their systems so that you
have to buy their drives, their memory cards, their everything.
"And, Apple has finally dropped its low-performance products. The variations on the Apple II, for instance, were cash
cows for Apple. Now they're gone. Plus, the older MACs are
selling just as slowly as they run.
"Even some of the big clone houses have set up their house
brands to require custom memory boards. Really helps their
bottom lines-for a while."
So it appears that the demise of the computer industry is
really more of the decline of its figureheads.

PostScript
There's nothing more worthless than a PostScript printer
when you want to output regular stuff. So you connect a silly
dot matrix printer to your PostScript machine. Nothing wrong
with that except dot matrix is slow, noisy, and its output isn't
very clean looking.
When I heard that Legend had come up with a simple TSR
package that automatically grabs anything sent to, say, LPT2:
and sends it out, say, LPT1: wrapped inside PostScript, I called
them up.
I haven't tried all the graphics or font options available on
an Epson FX-85 or IBM graphics printer, but what I have tried
has worked. Painlessly. You can run it as a TSR or as a filter.
Either way it's a godsend for outputting the ordinary stuff on a
super printer, super fast.

ODD
DO

ROBOtiC
SYSTEMS

LIMB 0 Parts Kit Series
Issue #50
Issue #51

Mobility Base
Superstructure
& Controllers

$219
$289

Please allow 3-4 weeks for delivery. Prices
subject to change without notice. International
orders must be paid in advance in US funds
drawn on a US bank. Visa, Mastercard, personal
check, and COD accepted.
Shipping &
Handling, $10. Wisconsin residents please add
sales tax. Call or write for more information.

(414) 769-9070
3620 East Layton Ave, Suite 25 • Cudahy, WI 53110-1462
Reader Service Number 166

Turbo Pascal Programmers ... are scraps
of paper keeping you awake nights?
Tired of digging through
old notes? Can't ramember whether a function returns a real or an
Integer? Want to get
procedure calls right the
first time?
Then you need MyFLIN.
•
It's a TSR. Hit the hot key to display procedure Information
right beside the code you're editing.
•
It's easy to find Information. Type part or all ofa procedure
or function name, and press the hot key MyRln will pop
upwlth the nearest match(es).
•
To save new procedure Informatlon,just place your cursor
on the same line as the procedure, pop up MyFLIN, and
tell It to add the procedure to the database. Requires
PC/YJ/AT Computer, M5-PC -DOS 2/3.xx. That's It.
Throw away the paper, get MyFLIN and get going.

OpaIFire _ _ _ _ _ _ _ _ __

Software Inc.
329 North Stata Staat,

OREM, UTAH 84057
Phone 1-800-336-6644

Price $30.00 PPD
Visa/Mastercard/American Express

Reader Service Number 161

MICRO CORNUCOPIA, #53, May, 1990

AROUND THE BEND
PSFX $85
Legend Communications
54 Rosedale Avenue West
Brampton, Ontario
Canada L6X lKl
(800) 668-7077

YOU WANT THE SOURCE1!
VELL NOV YOU CAN RAVE IT! The IIASTERPUL
iiiSiSsmoLER (1ID86) vii'i"Create HASH coapatible
source code froa program files (EXE or COH).
And the files are labeled and coaaented so
they become USEABLE. HD86 is an interactive
disassembler with an easy to use" word
processor like interface (this is crucial for
the REAL prograas you want to disassemble).
'lith its built-in help screens you
won't have to constantly refer to
the aanual either (although
there are valuable diSCUS-~
sions on the ios and outs
of disassembling which
you won't vant to ..iss).
1
~

-n. c-. .. rere'·

HD86 is a professionally
supported product and yet costs
no more than "shareware". Aod of course, it"'s
not copy protected.
VElSIOI 2 aov AVAILABLEI
IID86 Y2 fa anY $67 • .50 ($1.50 an) + tax

C.C. Software, 1907 Alvarado Ave., Walnut
Creek, CA 94596, (415) 939·8153
Reader Service Number 31

e ot ~ou covered ...
Wg

ort tor popular pC
b dded sv stem supp
. processors,
with complete em eS6 and NEC v·S~r"Ies. mlcr~th your choice
'c~~pilers on Intel ~oxan embedded apphc~tlO~:ebuggers and in'
NoW Vo~ can d~v: f~1 support for source eve
of compiler-wit
& Modula·2
circuit emulators.
0 ular C, Pascal, Bf\SIC
aradigm lOCf\T~ sUPpofrttsB~rrand and others.
P . f mMlcro so ,
compilers ro
th coprocessor
,
• ma.
ort
prehensive user s
emulatIOn supp OS
• ~:ual
& . extensible MS·O
• compiler startup code
emulato~ OMF output
examples
t
• full Inte
• free technical suppor
roblem with our Turbo
ou down? Not a P
ur target
I, of an emulator go~ V N w V
can debug VO
lac"
. f e option. 0 ou
Oebugger mter ac comfort of vour PC.
hardware trom the
~E $295.00

Paradigm LOClX Interface $195.00
Turbo Debugger
. Guaranteed
SatisfactiOn
37-5043

orders: (800 )5 rt. (508)478-0 49
Technical Sup~o .
SIX: ioin paradl~

Accepted

Thanks
I can't finish without thanking a whole lot of you.
Thanks to all the folks who attended our local SaGs (especially to Andy Bakkers from Holland, who attended every
one). You made those SaGs very special for me.
Thanks to all who organized, staffed, and attended the regional SaGs: including a special thanks to Dave, Randy, Mike,
Karl, Bob, Kim, Don, and John-those of you who did it, and
you who are crazy enough to do it again.
Also, thanks to all who subscribe. Many publications spend
most of their money and time selling-going door to door to
find new advertisers and mailing zillions of promotional
pieces to find new subscribers. Though we haven't done any
circulation promotion for 21;2 years, we're very close to our
peak circulation. That doesn't happen very often in this industry.
The reason we've done so well is you. You've stayed with
us. You've encouraged others to subscribe. In addition, you've
been willing to read the world's longest editorials year after
year after year without complaint. (Okay, you caught me. But
three of you haven't complained.)
I'd like to thank my mother and father, without whom I
wouldn't have parents. And, I'd like to thank my father-inlaw, who was one of Micro C's biggest supporters and a
wonderful friend.
I'd like to thank Jennifer and Erin. (They're like family to
me.) It's one thing to have a job, it's another to conceive and
raise a magazine. Over the past nine years, they've played second fiddle (and second piano) to 53 editorials, 53 deadlines,
and 53 hassles with the printer. Now I'm going to have some
time, and they're going to have a Dad. (They may insist I start
another magazine, real soon now.)
Plus there's Sandy. If everyone had a Sandy, there'd be a lot
more successful businesses. I'm not sure which of us has
worked harder to make Micro C go, and I'm not sure which of
us is more burned out.
Finally, there's the staff. What can I say about this collection
of miscreants who-put up with, cover for, tolerate, even occasionally get a laugh out of-what's left of me. They've produced magazines, harassed writers, filled pages with advertising, and still found time to party. A truly great group.
If You're Bummed

I'm sure many of you won't appreciate this change. I understand. It's part of the sadness I feel. If you feel strongly about
it, write to me, call me, or leave a message on the BBS. I'll be
here.

\lisa/Mastercard/C. . .

Paradigm SVS~~~;setts 01757
P.O. BoX

152 Milford. M

r
is a registered trademar'" of 6o \a

'TurbO OebU9ger

nell",.,n."o""l.

Da::::::son

Ex-Editor & Publisher
Reader Service Number 113

80 MICRO CORNUCOPIA, #53, May, 1990

By Michael Hunt
2313 N. 20 th
Boise, ID 83702
(208) 336-7413

Shortest Distance
MODULES

Michael concludes his column with a look
at connections: tying together network
nodes, for instance, with a minimum amount
of cable. An interesting problem.
veryday we use our reasoning
ability to find the shortest distance
between two points. The trip to
work, to the store-you know how
routine it is. But how good are we at
finding the shortest distance between
many points? If you've ever set up a
computer network, you were probably
concerned with keeping cable cost down.
The nodes in the network are the many
points that we need to connect.
For a small network, the solution can
be simple and straightforward. As the
number of nodes increases or offsite locations are introduced, the problem of
efficient connectivity grows quickly. We
need an algorithm to find the shortest or
lowest cost method of connecting all the
nodes in the network.

Abstraction
You can solve the network problem
by using a mathematical structure
known as a graph. A graph consists of a
set of points or vertices, and the lines
connecting the vertices, called edges. The
nodes of the network can be viewed as
the points or vertices of the graph. The
edges represent the connections between
the nodes. You assign a cost to the edge
or connection between two nodes, representing the distance between the two
nodes.
The graph is a collection of edges,
each with a cost, between vertices. The
set of edges is called a spanning tree. The
- sum of the costs of the edges in the tree
is the cost of the spanning tree. Our goal

is to find a minimum cost spanning tree.
Two algorithms address this problem.
The first is known as Prim's algorithm.
Prim's begins with one vertex and builds
the spanning tree one edge at a time. It
begins with a set of one vertex.
For each iteration, it finds the shortest
edge to a new vertex and adds the vertex
to the initial set until all vertices have
been included. Prim's uses an adjacency
matrix to store the edges and their costs.
The other algorithm is Kruskal's. I've
chosen to implement it because it uses a
linked list of the vertices and edge sets.
The linked list is dynamic, not restricted
by some upper array limit like the adjacency matrix.
If there are many fewer edges than
the square of the number of vertices,
then Kruskal's algorithm is best. But if
the number of edges is about the same as
the number of vertices squared, Prim's
performs better.
Kruskal's begins with a graph consisting of all the vertices and none of the
edges. Each vertex is in a connected component by itself. The algorithm builds a
collection of connected components, each
component consisting of a set of vertices
and the edges that form its spanning
tree.
The list of edges is sorted by cost and
then examined in order of increasing
cost. If the edge connects two vertices in
two different components, the edge is
added to the spanning tree and the two
components are combined. If the edge
connects two vertices in the same component, it is discarded. When all the vertices are in one component, the spanning
tree is of minimal cost.
'HC705 Simulator
The simulator is complete, and by
press time Karl Lunt and John Ribar will

We need an
algorithm to find the
shortest or lowest
cost method of
connecting all the

nodes in the
network.

be testing it. John is finishing up the
back end of the compiler. Look for it in
Embedded Systems Programming. For you
Amiga fans, a friend of mine is porting
the simulator to the Amiga using Benchmark Modula-2.
Last Column
Well, after nine issues, this is the last
installment of the Units & Modules
column. After talking with Karl Lunt at
Rocky Mountain SaG, I have been bitten
by the hardware bug. Recently I've been
gearing up for the change. Let's see,
multimeter, breadboard, power supply,
'HC705 programmers board, soldering
pencil, solder, and back issues of Micro
C.

The simulator for the MC6SHC705CS'
micro controller that Karl writes so much
about will be the first of many articles
about micro controllers and embedded
systems.

•••
MICRO CORNUCOPIA, #53, May, 1990

Commenting Disassembler!

• SEE HOW PROGRAMS WORK
• EASILY MODIFY PROGRAMS
SOURCER ,. creates detailed commented
source code and listings from memory and
executable files directly suitable for reassembly. Built in data analyzer and simulator resolves data across multiple segments and
provides detailed comments on interrupts and
subfunctions, 110 ports and much more. Includes a definition file facility to include your
own remarks and descriptive labels, force
data types, and more. Complete support for
8088 through 80486, 8087 to 80387, and V20/
V30 instruction sets.
We welcome comparisons with any other product, because no product comes close to the
ease of use and output clarity of SOURCEA.
"Sourcer is the best disassembler we've ever
seen." PC Magazine 1/17/89 page 101.

BIOS SOURCE
I

for PS/2, AT, XT, PC and Clones

• CHANGE AND ADD FEATURES
• CLARIFY INTERFACES
The BIOS Pre-Processor'· with SOURCER
provides the first means to obtain accurate
legal source listings for any BIOS! Identifies
entry points with full explanations. Resolves
PS/2's multiple jumps for improved clarity.
Provides highly descriptive labels such as
"video_mode" and much more. Fullyautomatic.

ASM PrOPakTM
• ASSEMBLY TOOLS PACKAGE
Save $50 with the ASM ProPak ,. assembly
tools package. It includes Sourcer with the
BIOS Pre-Processor; Unpacker,'· for unpacking packed EXE files; ASMtool,'· to automatically generate flowcharts and tree diagrams
from assembly source code; and View-It,'· to
quickly view files of any size and see up to 8
times more information on a Single screen
using EGAlVGA equipment.
Sourcer-Disassembler
$ 99.95
BIOS Pre-Processor
49.95
Unpacker-Unpacks packed files
39.95
ASMtool-Source code analyzer
89.95
View-It-Zoomable file viewer
69.95
ASM ProPak-AII Above (save $50) 299.75
Sourcer w/BIOS-(save $10)
139.95
Shipping & Handling: USA $5; Canada/Mexico $10; Other Countries
$15; CA Residents add sales tax. ViSa/MasterCard/COD accepted.

NOT COpy PROTECTED
30-DAY MONEY-BACK GUARANTEE
If within 30 days of purchase you find our product does
not perform in accordance with our claims, call our
customer service department and we will gladly
arrange a refund.

For orders or information, call:

1-800-662-8266

v COMMUNICATIONS, INC.
4320 Stevens Creek Blvd., Suite 275, Dept. MC
San Jose, CA 95129 (408) 296-4224
Reader Service Number 62

82 MICRO CORNUCOPIA, #53, May, 1990

"
,I, i
'/ ..

toost : REAL;
tkay1,tkay2 : kaystr;
I, J, oount : INTEGER;
a,. our, prev : edgeptr:
(* start at beginning of list
e := mainedge;
oount := 0;
(* then implement bubble sort
(* bubble sort is 0(n**2) so
WHILE e". fptr <> NIL DO BEGIN
(* this implementation of Kruska1's
oount := oount + 1:
(* would have to have a 0(n1ogn)
(* sort to be 0(n1ogn)
FOR I := 1 TOoount~l DO BEGIN
our := e;
prey := eA.bptr;
FOR J : = oount DOWNTO I DO BEGIN
IF ourA~oost < prev".oost 'l'HEN BEGIN
toost :=our".o.ost:
(* swap oode
our".oost := prev~.oost:
prav".oost := toost;
tkey1 :=our".key1;
ourA.key1 := prev".keyl:
prev".key1 := tkeyl;
tkey2 := our".key2;
our A.key2 := prev".key2:
. prev". key2 : = tkey2

*)
*)
*)
*)
*)
*)

END:

our := prey;
prey := prev".bptr
END
END

proo to build minimal spanning *)
(* tree if it exist *)

Get organized

our: edgeptr;
oamp1, o.omp2 : oompptr;

instantly!

PROCEDURE find(key : keystr:
(* proo to find the oomponent that *)
VARoomp : o.ampptr);
(* a key of an edge is in *)
found: BOOLEAN;
adge : edgeptr;
BEGIN
found :=FALSE;
(* set found flag to false *)
camp :=maincomp;
(* start a first oomponent *)
WHILE (oomp <> NIL) AND (NO'l' found) DO
(* while not empty *)
BEGIN
(* oomponent & key not found *)
edge :=.oomp".edge;
(* start at first edge *)
WHILE (edge<> NIL) AND (NO'l' found) DO
(* while not empty *)
BEGIN
(* edge & key not found *)
IF (edgeA.keyl=key) OR (edge".key2=key) 'l'HEN (*if key found*)
found := TRUE
(* set flag *)
ELSE edge :=edge".fptr
(* else inorement to next edge *)
END (*while *):
nNO'l' found.'l'HEN o.amp:=oomp",fptr(*if not found-next compon *)
END (* while *);
IF.NOTfound.'l'HENoomp:=NIL (* if still not found";nooomponent *)

END;

Create, Find, View, & Edit
notes, addresses,
documents, data.
*No setup.
* No keywords to define.
* Configurable,
Full-screen editor.
* Turbo fast & little

PROCEDUREjoino.omp (VAR olf 0.2 : oompptr); (*proo to joinoompons *)

. VAR

1: .edgeptr;
BEGIN
l:=ol".edga:
WHI;E.l:".fptr<> . NIL DO· BEGIN

1. .•••.. .:•.:. .
END;

1. . .::":. . :. . .•.·•.:.f. . P.:.:t
. . . . r.:::.:.

set temp to. first .. edge :l.ist
while not last edge inl.ist
(~ino. to> next edge in list

:=:
..•.••.•:..:.....•.·••...:..

WIZARD

1 A~ fptr:=o~" .e<.ilile(C*l....stin 1st: list pointsto~stiri2nd list*)
02~ .edgeA.bptr:~l;(*1st

027.;edge:=0~,:.edge;
c2~~~dge:"~bPt:r:::= . NIL:

ol";edge:=

NIL~

in 2nd list. pointsbaok\l\IARE

Helpful Shareware

In this, his first shareware Micro C
column, Nelson covers his favorite utilities.
Great stuff.
rior Shareware columnist Anthony
Barcellos is going to be a tough act
to follow. For the last two and a
half years, Tony has done an admirable
job of coming up with the real gems.
Over the past eight years, I've tested
and reviewed thousands of public
domain and shareware programs for the
Public (software) Library.
I've written a dozen shareware programs and am one of the founding members of the Association of Shareware
Professionals.

Terms
First, let me clarify a few terms.
When I say "pd/ shareware," I include "freeware," too. Public Domain is
software which the author has donated
to the public domain. (The absence of a
copyright notice is not sufficient to make
a program pd.)
Freeware used to be the same as
shareware. Andrew Fluegelman trademarked the name freeware, so other programmers invented the name shareware.
After Fluegelman's death, the term
freeware fell into generic use, meaning
software that is free, but copyrighted, as
opposed to free and publicly owned.
Since most of you spend your time on
DOS machines, I'll start out with a discussion of...
The Ten Greatest DOS Tools In The
History Of Man
I would hate to think of having to
work my 14+ hours a day on a computer
without the following:

The Public (software) Library
P.O. Box 35705
Houston, TX 77235-5705
(713) 665-7017
CIS: 71355,470
FAX: (713) 524-6398

MICRO CORNUCOPIA, #53, May, 1990

(1) LIST, by Vernon Buerg. You probably already know about this file viewing utility, but you may not know that
Buerg keeps making it better. If you
haven't seen version 7, it's time to dial
up your favorite BBS.
(2) FGREP, by Christopher J. Dunford, is the world's fastest text search.
Sure, you can get a fancy TSR like Gofer,
but you can't beat FGREP for speed and
convenience.
(3) SD-whatever, anyone's sorted
directory utility is better than what DOS
provides. I prefer HotDIR, or one of the
others that color-codes file names by extension. It's easy to spot all the OBJ files
in a crowded directory.
(4) MOVE FILES 4.2, by Bryan Higgins, helps you move files between subdirectories or disks. After years of
struggling with deficient file movers, it's
great to find one that's very close to perfect. The only question is why it took so
long to show up.
(5) CED, by Christopher Dunford, is
to the DOS command line what a fullscreen editor is to Edlin. It lets you cursor up" previous DOS command lines
and edit them using all the editing. keys
(e.g.: home, end, word left/right). CED
is freeware, but there's a commercial
version that's even better.'
(6) HindSight, by Christopher Dunford. Anyone who works in DOS needs
to see what's just scrolled off the screen.
Though there are several video backscroll utilities, HS works with Dunford's
CED, so I use it. Also check out FANSIConsole, which includes a wealth of
video related features along with backscroll.
(7) LHarc, a freeware file archiver
from Japan. LHarc is the best file compression I've seen. Some files will compress to a fifth their original size.
/I

(7a) I must sneak a related cleanup
utility in here. LHdel, by Duane Hendricks, examines the contents of an
LHarc archive file and deletes matching
files in the current directory.
(8) Newkey, by Frank Bell. If you
don't have a keyboard macro program,
you need Newkey.
(9) DESQview. Oops; okay, this multitasking program isn't shareware.
Worse yet, I lied in (8): I don't use Newkey anymore, since DV can automatically load a different set of macros each
time you open a window. Since I only
use DV for task swapping, you might
think that Software Carousel could do
just as well. But DV has too many nice
extras to give up.
(10) SitBack, by SitBack Technologies.
This TSR auto-backup utility isn't
shareware either, but I am the world's
worst at backing up my work and
there's no shareware solution. SitBack
will backup preselected files in the background at specified times. This TSR uses
a modest 16K.
The closest shareware equivalent is
AutoSave, by Biologic Corp. It records
your keystrokes to disk, plus it saves
your work for you as often as you wish.
How To Write Shareware And Make A
Million Dollars
Step One: Write a shareware program. Step Two: Win a lottery.
A half-dozen shareware authors have
been able to combine the two. Tom
Smith and his partners turned a
shareware communications program,
Procomm, into $15 million a year. Jim
Button (PC-File) and Bob Wallace (PCWrite) make over $2 million a year.
Marshall Magee (Automenu) and Dave
Berdan (File Express) are grossing over
$1 million a year.

These big bucks have put stars (and
dollar signs) in the eyes of thousands of
programmers. That's why thousands of
people make little from shareware while
a handful prosper. If you have ideas
about making real money from
shareware, consider these:
(1) Writing the program is the easy
part. Sending out disks to vendors,
uploading to BBSs, responding to users,
fixing bugs, adding requested features,
and improving the documentation are
not only more work, they aren't as much
fun. They take up time and cost you
money.
Two classic examples of proper
marketing are Datastorm's Procomm
and Magee's Automenu. Procomm was
not the first shareware communications
program, nor was it significantly better
than Qmodem. But Datastorm's people
knew how to market the program.
Marshall Magee took a program in
one of the most crowded arenas of PC
software-DOS menu programs-and
turned it into a winner. In shareware
circles, Magee is known as Mr. Image.
He knows how to market, and he
learned it on his own.
(2) You don't have to choose between
shareware and nonshareware marketing. There is nothing sacred about
shareware; it's just another way of
marketing. As the prices of retail programs have approached shareware
prices, programmers have had to be
aware of both arenas.
If you market through both retail and
shareware channels, don't cripple the
shareware version. It's better to sell retail only than to pawn off a crippled
shareware program.
(3) Write an outstanding program
that businesses can use. Businesses are
still the top shareware supporters.
Games and DOS utilities may bring in a
few bucks, but that's about all. Yet
Another Text Editor or Yet Another Flat
File Database Manager is not likely to
generate much cash, either.
(4) Don't reinvent the wheel. If you
want to rewrite the same old utilities
over, fine. If you want to make money,
know what's already there and be prepared to either out-program it, outmarket it, or forget it.
(5) Hang in there. Many people write
a program, get all excited while they
send the program out and upload it to
BBSs, and then get discouraged and quit
when a month or two goes by without
hearing anything. The shareware mill

grinds exceedingly slowly.
Eventually, your program will become ingrained in The System of
pdf shareware distribution. If you have
a good product and you continue to improve it and promote it, you can profit
from it for years. I still get weekly registrations from a 1984 version of Diskcat at
a P.O. Box that I have not publicized in
over three years.
None of the top moneymaking
shareware authors of today made much
money their first six months. It takes that
long for programs to start getting passed
around. If you need quick income,
advertise. Unfortunately, advertising's
not likely to be worthwhile.
(6) Shareware is not a way to make
money from inferior software. Some
folks try marketing their program as
shareware because it's not slick enough
to sell on the commercial market. They
shouldn't bother.
In theory, shareware has to be better
than retail-only software because people
get to try shareware before paying for it.
I could go on for pages and pages. In
fact, I already have in the Programmers
Guide, an on-disk guide for shareware
programmers. It contains tips on how to
polish a program and documentation,
how to protect your software legally,
how to find necessary supplies and services at the best prices, how almost anyone
can get a MC/VISA merchant account,
and more. The Programmers Guide is on
CompuServe
(GO
IBMPRO)
as
GUIDE1.ARC and GUIDE2.ARC. You
can also get it by sending $5 to my
address and asking for the Programmers
Guide disk.
The Tax Man Cometh
By the time you read this issue, you
should be starting to think about your
tax return. My longtime favorite tax program, commercial or shareware, is AMTax.
PC-Tax is another veteran of the
shareware world. It usually supports
more forms than AM-Tax, but it's lineoriented, an approach which is more appropriate to a teletype terminal than a
full-screen computer.
AM-Tax uses a full-screen, fill-in-theform approach in which you scroll
through replicas of the tax forms, fill in:
the blanks, and watch AM-Tax do calculations and carry totals forward for you.
AM-Tax works so intuitively you'll
probably never read the manual. Any
time you need a subschedule, all you

have to do is press F6. If no IRS subschedule exists, it creates a worksheet.
After you use the worksheet or subschedule, you can return to the original
schedule bringing along the balance.
In contrast, the last time I tried Turbo
Tax, a popular retail-only tax program, I
had to work through a confusing form
menu to create a subschedule. Arbitrary
worksheets were not an option. I got so
bogged down in Turbo Tax that I never
could complete the tryout.
The shareware version of AM-Tax
supports form 1040 and schedules A, B,
C, D, E, F, R, and SEe All printouts can
be filed directly to the IRS. If you need
more schedules, corporate or state
forms, etc., they have more advanced
versions.
Finding These Programs
All pd/shareware programs· mentioned in this column are available from:
public BBSs, CompuServe, user groups,
and disk distributors such as PsL. In particular, PsL coordinates quite a bit with
IBMNET on CompuServe.
IBMNET is divided into many different Forums and each Forum has many
different Data Libraries. The quickest
way to find a program on IBMNET is
through the File Finder (GO IBMFF).
Once there, you can search by program
name or Key Word. For example, to find
AM-Tax, you could search for the name
or the Key Word "TAX."
CompuServe limits file names to six
characters plus a three-character extension, so longer names may be abbreviated in ways you could never imagine.
When this happens, the actual program
name will often be recorded as one of
the Key Words, so search for the program name in the Key Word field.
References
Chip Merchant
9285 Chesapeake Dr., Ste. L9285
San Diego, CA 92123
(800) 426-6375
(619) 268-4774

SitBack Technologies
9290 Bond, Ste. 210

Overland Park, KS 66214
(913) 894-0808

•••
MICRO CORNUCOPIA, #53, May, 1990

8088 Hardware Development

Aiding The Ailing 8088
I discovered a way to diagnose a recalcitrant
8088 recently while working on controller cards.
(It turned out that the steps outlined by Ed
Ninsley for the 8031 in a back issue of Circuit
Cellar INK can be modified for use with 8088s.)
My situation was pretty down and dirty:
building an 8088 board and trying to get it
working. I couldn't even get the 8088 to execute
a single instruction.
Ed's steps boil down to the following.
Disconnect the data bus from all memory
devices and ground all the data lines. This
forces the instruction ADD AL,[BX] onto
the bus.
Ground both of the interrupt pins (NMI
and INT) and also the HOLD request pin
(used by DMA and things like that).
This way the CPU will free-run, incrementing addresses through segment FFFF.
Remember that most of these accesses will wrap
around to 00000; only the first 16 bytes come
from the high end of memory.
Basically this gives you the ability to make
sure that address latches and CPU clock are
working before you add more things (such as
memory decoding and memory).
Nathan Engle
6465 Piping Rock Lane
Indianapolis, IN 46254
High Memory Update
Since the original incarnation of my high
memory card ("Filling In The Holes On Your
XT," Micro C, Issue #52), I've rearranged and
added things to my computer. Specifically, I
added an expanded memory board.
Expanded memory is one of the greatest
things since sliced bread, but it does require 64K
of open memory space. For me, the simplest
place to map it was in block D (DOOO:OOOO to
DOOO:FFFF). With the commercial memory cards
referenced in the article, this would mean separate cards if you wanted to fill in the remaining
A and E blocks.

MICRO CORNUCOPIA, #53, May, 1990

The SRAM card, however, may be modified
as follows:
Interchange address line A18 and A17 at
the II 0 bus interface;
Move the chip select lines from pins 12 and
13 of the '138 to pins 14 and 15. (See Figure
1.)

This mod will give 128K, split between the A
and E blocks. You might also want to rethink
just which blocks, if any, you want to battery
back-up. The change to E-DISK to make it a 64K
RAMdisk is relatively trivial.
Also, in the eternal search for extra memory,
a special note to Hercules (and clone) card
owners. When operating in text mode, it is
simply not possible to make the video access
anything beyond the first 4K.
Let's call this an overly slavish emulation of
the IBM mono adapter. When running one of
the CGA simulation programs, only the top 32K
is accessed.
Notice the hole between 4K and 32K
(BOOO:1000 to BOOO:7FFF). This memory gets
used only when doing native Hercules graphics.
For ordinary text applications, this 28K sits
there idle. Scandalous! Quickly, load something
up there and use it.
While on the subject of the HI LOADing software mentioned in the article, I must say it's one
of the more useful utilities I've ever written. But
you should use some common sense in its application.
Some TSRs are design~d to speed up various
PC functions, such as video BIOS accelerators.
Since they would be executing out of I/O bus
memory, which is typically slower than motherboard memory, it doesn't make sense to load
them up there, even though it's possible.
And, some TSRs (cache programs come to
mind) are smart enough to "look around"
where they are. From their point of view there
isn't much memory left, so it wouldn't- make
much sense to install themselves. If they refuse
to load, it may very well be something like that.
Finally, in the never-ending saga of living
with SideKick, I've gleaned a few more tidbits

Searching for that IMSAI?

Need to sell your 286 to buy 386?

Micro Ads

Privata Listings is a natiooal publication far computer and
computer related items far sale by private owners. Savings
galcxc m all types of R:ccllt and vintage itemsl Home of tho
$5 two line one year adl
Call or write for a ficc subscriptionl

Gateway Technologies, Inc.
ISO Church Road
King of Prussia, PA 19406
(215) 354·0330
Reader Service Number 216

Learn a Modern Language

Experience Embedded
Power

with the new Control-R II 8031 I1Controller Module
..J 5 volt single supply operation (even with RS232)
..J Data/Address/Control Bus Headers for expansion
..J Direct Access to 8031 Ports 1 and 3
..J 8K on board EPROM space
..J Socket for 8K 01 on-board Static RAM (6264)
..J MAX232 socket to provide RS232 compatible port
..J Memory decoded to allow use 01 8052-BASIC I1P
..J Assembled and Tested, I1J1l.1LlsI1
..J Compact size: 3.5" x 4.5"

Our tutorials use a unique method to teach
you to program in a modern language. Pascal,
C, C++, Modula-2, beginning Ada, and
advanced Ada tutorials are available for
MS-DOS. Each tutorial is $39.95 + $3.00 P&H
(in US). MIC or Visa accepted.

Coronado Enterprises
12501 Coronado Ave NE, Albuquerque,
NM 87122

Control-R II, Enhanced 8031 Module
(SRAM & MAX232 optional)

64.95

(505) 293-5464

Control-R II, Bare PC Board w/docs

25.00

Reader Service Number 206

Control-R I, 8031 Module
(Port 1 & 3 access and MAX232 option only)

39.95

Technology for the Arts

Control-R I, Bare PC Board w/docs

14.00

8K Static RAM I.C.
MAX232 lor serial VO

10.00
6.95

PseudoSam 51, 8031 family cross-asserrbler 50.00
for MSDOS computers (5.25 inch disk w/docs)
Shipping $3.00 US I $5.00 Canada
illinois Residents Add 6.25% Sales Tax

Cottage Resources Corporation

Vision-16 Video Digitzer displays
32,768 colors, w software $1495
Wtllow VGA-'IV w NTSC Output $495
Publishers' VGA/Frame Grabber $595
Panasonic CC1V B&W Camera $225
MIDI (music) Interface $119
MIDI Programmers Toolkit $39
Cakewalk. 3.0 Sequencer $150

• I

•

P.O. Box 550363
Dallas, 'IX 75355

Subjects are organized into sections to give you
the most benefit in the least time. Uses Filters to
customize output; Dynamic mode to stimulate
creativity.
It's never too early to get a head start.
Introductory Offer: $79.00 for all 5 disks.
30 day money-back guarantee.
Requires 256K, DOS 2.0 or later & 2 floppies or hard
disk. Add $2.00 S/H. CA residents add 7%. Outside
U.S. add $10.

MeR Agency, Inc.
6116 Merced Ave. #81MC
Oakland, CA 94611
1·800·767·6797
FAX 415·444·6561

(214) 328-2730
BBS 328-6909

Reader Service Number 176

NEW!
N()w, '" Micro ,Cornucopia , ' ," has .. , "microform
availabiUfy;16mm microfilm, 35mm microfilm,
105m,m '. microfiChe .,coplesof.ls$ues, and article~
pr9vld$d thf(;)ug1}.~.
; University Microfl,lmslnterriaH()"ClI.

300NorthZeeb Road>Ann ArborI M148106-'1346'
313176h4700

I I I I I I I I I I I I

HEXED?
Get HEXED1~qhe HEXadecimal EDitor
Display and edit ALL characters in ANY file.
• Full-featured, menu-driven binary file editing
• Simplicity and power of a text editor
• Split·screen, simultaneous Hex value\ASCIl symbol editing
• Easy access to file data formats, headers, etc.
• For all PCs & compatibles; 256K; DOS 2.0+
• $49.95 USA, Canada. $59.95 elsewhere.

Ideal for engineering, program development, hacking.

T~[Ji)[@ID

HE,J('lnc.

Call for FREE Demo Disk.
141 Peyton Rd. Sterling, Va. 22170

703·444· TREX

Reader Service Number 218

DSP32C COPROCESSOR BOARD
XSPAWN replaces the standard C spawn functions
with functions that can transparently swap most of
the parent process to expanded memory or disk
before executing the child process. The XSPAWN
functions are as easy to use as the standard C
spawn functions. Indudes all source code and
libraries for Microsoft C and Turbo C.
Price $79.95 plus $4 shiping and handling. CA
residents add sales tax.
Whitney Software, Inc.
P.O. Box 4999
Walnut Creek, CA 94596
415-933-9019
Reader Service Number 164

Select a subject. Use the vivid parade of timeless
knowledge to act, write & speak better; earn
more.

Reader Service Number 158

•

First electronic book of quotes, sayings & ideas
brings over 1,000 of the world's greatest minds
to PC screens & printers.

joel Sampson Engineering

Suite 3·672C, 1405 Stevenson Drive
Springfield, lllinois 62703
(217) 529-7679

WISDOM OF THE AGES

~
~
~
~
~
~

25 MFlOP Floating Point DSP
High speed NUMERICS and GRAPHICS
32 BIT on board data bus
16 bit PC/AT interface
All ON BOARD MEMORY DUAL PORTED
Parallel and codec serial 10
15 ms 1024 point FFT
Assembler, monitor, and libraries
Base board and All software .......... $950
640K 85ns memory ......................... $300

SYMMETRIC RESEARCH
15 Central Way, Suite #9, Kirkland, WA 98033

(206) 828-6560
Reader Service Number 182

Reader Service Number 181

8748 EMULATOR
Simulate your 8748 programs in software before
burning the EPROM. See al\ internal registers
and i/o pins. Script files can simulate comp'lex
external events. Includes source. .............. $15

Other Products
DOS Source

Fully commented 1.1, 2.1 ................ $15/$45

Hercules Tools

Graphics stuff w/source ..................... $15

Math Library

ascii Floating pl. w/source ........... , ........ $15

I.~ ~ ml'nformat'on
P.O. Drawer F Mode.

I 0• e

Denton, TX 76202
817-387-3339

1-8O().628-7992

Reader Service Number 149

EGA FRACTAL MASTER
Whether fractal novice or fractal hacker, this easy to use
yet powerful package lets you explore:
The Mandlebrot set
Dragonland
Biomorphs
Peano curves

JUlia sets

~:~~~:r~~~:9ons

Fractal landscapes

Interrupt, save and resume plots at any time. Easily
change colors, set boundaries. animate colors, create
mirror images. Supports fast integer computations (even
WITHOUT an 80386) and 80x87 math co-processors.
Includes a slideshow program. Specify 5114 or 31/2 disk.

Only $25 (postage included)
Paul W. Carlson
602 North Avenue, #23
Burlington, Vermont 05401

Reader Service Number 185

MICRO CORNUCOPIA, #53, May, 1990

//Editor's comment: Larry discovered
c++ comments in FRACTINT (see the
Shareware column in Issue #52) when converting the code to compile under Turbo C.
FRACTINT was written in MS C 5.1, so
there's your answer, Steven.

of its inner workings. Yes, it does use
undocumented INT 21h calls (function
34h, at least: the INDOS flag) but many
TSR writers do that. While surprising at
first, I now use it myself!
Second, it seems that on every clock
tick SideKick scans the interrupt vector
table. If it sees someone else grabbing an
interrupt it wants, it rearranges things
its way. Every clock tick! Talk about
chutzpah!
Oh yes, SideKick is unloadable. Don't
try it, though, if you've HILOADed it. A
guaranteed lockup.
Larry Shannon
5615 Truscott Terrace
Lakeview, NY 14085
Comment Comments
I agonized on whether to submit this
as a technical tip, as it is merely an undocumented feature of a popular compiler. I decided to go ahead on the basis
that it is a potentially useful feature that
may increase programming productivity
(but I may be overstating its usefulness).
I've always found it somewhat awk-

MICRO CORNUCOPIA, #53, May, 1990

ward to enter C-style comments with
their slash-star, star-slash pairs (/* *I).
So I was quite surprised to find out that
Microsoft QuickC 2.00 accepts C++ style
comments! This feature, as far as I can
tell, is completely undocumented in the
QuickC manual. I haven't been able to
verify this feature in MS C 5.1. Have any
MS C owners out there seen this feature?
I stumbled upon the feature while
looking through the on-line help system.
It's only used in one place: the sample
program LOCK.C (to comment out two
lines of code). There are no other references to C++ comments in the help system. (Note: C++ comments consist of a
single slash-slash pair (/ /) with the
comment text to the right, similar to assembly comments.) I've tested this feature with flawless results.
Things like this make you wonder if
the programmers at Microsoft don't
have a few other things they haven't told
us about. Enjoy!
Steven Byrnes
10510 Emnora
Houston, TX 77043

Friday The 13th !
I spoke to you Micro C folk a few
weeks ago about some odd happenings
on a Haupage 386 mammy board: that
was, it appeared to kick into low gear
approximately 30 minutes after boot.
Well, I've tracked that dog down, and
had my suspicions confirmed by one of
the Houston area gurus.
Lo and behold, the poor thing had a
virus. The infamous Friday the 13 th
virus, according to the guru. It's difficult
to chronicle the sequence of events as
nothing made any sense until it was
over. I ignored the first law of troubleshooting (well, all right, the third law; it
was (1) plugged in and (2) the power
supply was fine). I should have known
better when I saw that interrupts 08h
and 21h were hooked. Unless you're
running a legitimate program, there's no
excuse for them to be hooked.
Anyway, to make a long story short,
here are some of the clues for sniffing
out the beast:
1) Interrupts 08h and 21h get hooked
from the get-go. No need to run anything first. It gets into the boot code.
2) It expands uninfected COM and
EXE files as soon as they are run. Note
that it's sneaky about this as it restores
the original time and date of the file, so
no clues via time stamp.
3) In every case, the string "sUMsDos" was embedded in the infected files.
Using the Norton Utility disk search or
grep are two ways to find victims.
There may be easier ways to rid the
beast, but I found DOS's FORMAT effective, saving the agony of real low level
formatting (I can't always remember
G=C800:5, or something like that). Anyway, there ye have it. May the fleas of a
thousand camels invade the armpits of
the worms that invent these demons.
Disinfectedly yours,
Dave Stojan
10310 Lybert Rd.
Houston, TX 77041

Determining Required Stack Size
I used to find it difficult to determine
the required stack size for my C programs. Then it occurred to me that the

PSFX TSR prints any Epson FX-85 or IBM
Graphics Printer compatible output as POSTSCRWI',
with graphics and IBM screen characters.
$85.00
Also available for Novell's NetWare.

Micro
Ads

FIGHT PIRACY

Since 1986. companies worldwide have been choosing Az-Tech
security products. If you demand the strongest protection available.
why not choose one of these 'proven leaders':
• EVERLOCK Copy Protection
• EVERKEY Hardware 'Key"
• EVERTRAK Software Security
Software Security

PSPlot Convert HP-GL to POSTSCRWI' or EPS,
any line width or color. POSTSCRWI' oriented editor
with test downloader and error display. ASCII file
print, any font or size. AutoCAD to POSTSCRWI'.
Batc~ download of PC and MAC fonts.
$175.00

For IBM and Compatibles.
30 day money back guarantee.
Free info and demo disk available.

EPScreen Capture PC screens as tiny EPS files,
frames optional. Includes IBM screen font.

PROTECT YOUR
PROGRAM $$$'5!

AZ
@

$99.00

Az-Tech Software. Inc.
305 East Franklin
Richmond. MO 640SS
FAX (816) 776-8398

(800) 227-0644

(816) 776-2700

Reader Service Number 197

XenoCopy-PC $79.9S+S/H
PC-DOS program lets your PC
Read/Write/Format
over 350 formats

XENOFONT $49.95 +S/H
high quality text screen printouts
ideal for use in software documentation
Bold face and reverse video supported.
2210 SIXTH STREET
BERKELEY. CA 94710
415-644-9366

Reader Service Number 39

Please see

UPPER DECK FORTH $49
Based on Forth-83 Standard Fully segmented
architecture Fast compile, no link required
Uses standard DOS files Integrated multi-file full
screen
editor
Assembler,
decompiler,
source-level debugger Turnkey application
support For IBM PC/XT/AT and compatibles
with 256K, hard disk or floppy, DOS 2.0 or later
Add $3 for shipping and handling (outside
USA $15). CA residents add sales tax.
Upper Deck Systems
P.O. Box 263342,
Escondido, CA 92026 (619) 741-1075
Reader Service Number 211

us at Electro Booth # 1121

IN CIRCUIT EMULATORS

8051 FAMILY
68HC11
CALL OR WRITE FOR FREE DEMO DISK AND
BROCHURE!
51 East Campbell Avenue

nOHaU
CORPORATION

(408) 866-1820
FAX (408) 378-7869
Campbell.CA95008

Reader Service Number 186

3112 or 5114 Disks
Available
$6 ppd (U.S.) or
$8 ppd (Foreign)

LOW-BAND PC LOGIC ANALYZER

• Designed for general applications (below 1 MHz)
where hlgh-cost analysis Is not needed
• Works with 5V digital logic
• 8 color-coded data lines & external clock Input
• 6 range Intemal clock (1 Hz-100KHz) for periodic sampling
• 8-blt logic probe mode
• Card plugs Into IBM PC!>IT/AT/386

'"a
_

0.11.

.... I ...

t ......

:!:
O·Ot

.iM
•

1::''''

II •

.,:~•••~l!..~ ...OO

•

36" cable with micro-clips
Full·featured software
Mlcrosoft·mouse compatible
Fast assembler data acquisition
Help system & manual

IPHDIltONICS I $99.95
109 Cam,lIe Str ••t
Am,t •• LA 70422

MC,VISA,
Company PO,

504-748-7090

COD

Reader Service Number 189

FLOPPY DISK FILE
A utility that sends print fllas to dlsk.fW
Lets you work and print simultaneously
• Uses!U:l than 6k' Uses any LPTICOM port
• Provides printer selection' Supports Xonlo"
• Auto-releases disk space when file Is printed
• You select amount of disk space to be used
• Doesn't Interfere with computer usage
~0(
Only $49D5 1-800-274-3007
~0'O-s rjlr:.\ Send check or money order to:
0\ ~0
Gribbs & Associates
331 Glencrest Dr. Solana Beach CA 92075

• Stackable - For Convenient Storage
• Adjustable Metal Follower Keeps Disks Upright
and Neat
• Sturdy Fiberboard
Construction
• Easy Pullout Drawer
With Metal Brasstone
Handle and Card Holder
• 10 Dividers Included
• Available for 5.25" and
3.5" Floppy Disks

Reader Service Number 214

sPORTt

Get
All the
Software
in Issue 51.
on Disk!

serial PORT tester
sPORTt is a serial port tester for people who don't want to
be data communications experts, but use or write software
for serial devices, such as plotters. mice, digitizers,
modems, serial printers, or computer to computer links.

• displays all data going in and out
• displays ASCII, decimal, hex, bina~

Disk Storage
Model Size Capacity
33-1
3.50"
100
55-1
5.25"
150
Price Size (HxWxL) In.
$15.70 5.25x4.25x15
$17.20
7x6.5x15

• lets you choose handshaking lines
1-800-888-8087
(orders only)
or (503) 382-8048

Plus shipping $3.20 Per Unit -In Florida, Add 6% Sales Tax
.. No shipping charges with orders of 12 or more boxes"

• sends/receives disk files
• automatically runs port diagnostics
Order today only $80

916-885-2480

LOALCO Inc.
P.O. Box 7117, Delray Beach, Florida 33484
Telephone: 800-869-9241 • (407) 243-6222

Reader Service Number 192

Reader Service Number 215

WHS
3037 Grass Valley Hwy #8201 Auburn, CA 95603

Hercules
Console Driver
Fully emulated text output in graphics mode on
your Hercules Graphics Card. Great for programmers debugging & working with graphics
programs. Write directly to the screen or run
DEBUG without leaving HGC graphics mode.

Just $24 (Shipping & Handling Included)
Winter City Software
8723 - 162 St
Edmonton, AB Canada T5R 2M2
( 403)484-3187
Reader Service Number 203

WE ACCEPT VISA AND MASTERCARD

16 Megabytes

EMS andlor
Extended Memory

-Works on 8 or 16 bit bus
-16 bit transfer on AT bus
-Single board design
-Includes RAM disk and
extensive diagnostics
-Quantity/OEM discounts

~:~~;::;ured,
Sold and Serviced by

XT and AT
Compatible

m~ ~~

~LJ~~lfU

807 North 6th 8t. l.ak8 City, lIN 55041 (812)345-t555
Reader Service Number 3

Why you want BAlCOM!
BATCOM is a batch file compiler that compiles
your· .bat- files to -.exe- files to make them faster,
more professional, and more capable. BATCOM extends DOS with new commands so you
can read keyboard input. perform arithmetic, use
subroutines, and much more. In addition, BAT·
COM protects your source code, and you can
distrbute your compiled programs without royalties. For IBM PC. Only $59.95. Order today I
/-'-,. Wenham Software Company
t • ! 5 Burley St.
' _ . / Wenham, Ma. 01984
(508) 774-7036
Reader Service Number 124

MICRO CORNUCOPIA, #53, May, 1990

NEW COMPUTER BOOKS
Select from thousands of new
titles in C++, C, ADA, Pascal,
Modula, Assembler, Basic, Lisp,
Fortran for PC's, MAC, Unix and
VAX. CAD, Desktop Publishing,
Dbase,
Word
Processing,
Business & more. Not a book
club, no membership fee or
required purchases. SeNice for
businesses & professionals.
Quality at low prices.
up to 40% OFF LIST
SMALL MINIMUM ORDER
send for FREE information to

Figure 2-STKSIZ.ASM
. MODEL

SMALL

EXTRN
EXTRN

end: WORD
: ;...amsg_exit:PROC

. DATA
S'l'KHQQ

__stksiz

PUBLIC
dw
PUBLIC
dw

__chkstk
IF

PUBLIC
LABEL

-1

larger than possible

__chkstk
PROC

pop
cx
@codesize
pop
dx
mov
sub
jc
cmp

jb
mov
sub
cmp

jae
mov
return:
IF

Coming Up In Issue #54

You won't find this issue at your local
newsstand, nor will it be available by
subscription. Issue #54, coming soon to a place
very close to you. Watch for it. (Thanks, Dave Stojan.)

stack. bottom + slop

return offset
return segment

ENDIF

PO 7076
Sunnyvale, CA
94086-0746

The meaning of life, the universe and
everything. This series will take a quantitative
look at this engrossing subject rather than
mucking about in silly philosophical mud
puddles.

S'l'KHQQ

offset DGROUP:_end+256
__stksiz

. CODE

Computer Book Co.

Reader Service Number 167

stack bottom
write message and exit

dummy

bx,sp
bx,ax
overflow
bx, [S'l'KHQQ]
overflow
sp,bx
bx, [S'l'KHQQ]
bx, C _stksiz]
return
C _stksiz] ,bx

subtract local requirement
stack overflow
stack overflow
new stack pointer
get free stack space
replace with smaller value

@codesize
push
dx
push
cx

return segment
return offset

PROC
ret
ENDP

FAR

ELSE
ENDIF

jmp

return to CX

overflow:

xor
jmp

ax,ax
__amsg_exit

write message and exit

dummy

return to DX:CX

END

•••
SHAREWARE AUTHORS: Give your
programs the chance they deserve! With
Megapost, your program is uploaded to
Compuserve, GEnie, Delphi, SIX, and
EXEC-PC. The low fee is probably less than
you would pay just to open accounts with all
these services. Included are one half-price
update and notification of the downloading
procedure and 3-month download count for
each service. 15% discount for ASP
members. Write for a free brochure.
Andrew M. Saucci, Jr.
.641 Koelbel Ct.

Baldwin, NY

11510-3915

Reader Service Number 200

Micro
Ads
94

MICRO CORNUCOPIA, #53, May, 1990

"stack probe" routine (called at every
function entry point to verify that there
is enough remaining stack space to allocate local variables required by the function) could be modified to also maintain
a global variable that would contain the
smallest amount of free stack space up to
that point.
The routine presented here (see Figure 2) was written for Microsoft C and is
functionally equivalent to the Microsoft
stack probe routine (called _chkstk) except that a global variable called _stksiz
is maintained. My routine is only a little
larger and slower than its counterpart.
To use my routine you must first delete the chkstk module from the Microsoft run-time library you are using. This
is because __ chkstk is not a procedure
label.
When testing a program, I print

_stksiz just before exiting. Assuming
stack checking is enabled (see your compiler documentation), _stksiz will contain the smallest amount of free stack
space for the portion of the program that
was actually executed.
The routine presented here is for the
small memory model and was written
for MASM 5.0. To change memory models you need only change the .MODEL
directive.
Jack Whitney
Whitney Software, Inc.
P.O. Box 4999
Walnut Creek, CA 94596
(415) 933-9019

•••

ADVERTISER'S INDEX
ISSUE 53
72 Acquired Intelligence . . . . . . . . .
187 Alphax Systems, Inc. . . . . . . . . .
160 Annabooks...............
4
Austin Codeworks . . . . . . . . . .
197 Az-Tech Software, Inc. . . . . . . . .

83
78
76
31
93

181
37
2

MCR . . . . . . . . . . . . . . . . . . 91
MicroCornucopia . . . . . . . . . 74,77
Microprocessors Unlimited . . . . . . 63
Microsphere, Inc. . .. 94, Inside Front
Midnight Engineering . . . . . . . . . 87

147
214

Berry Computer . . . . . . . . . . . . 49
Bluebird Systems . . . . . . . . . . . 93

186 NOHAU . . . . . . . . . . . . . . . . 93
110 Nu-Mega Tech . . . . . . . . . . . . . . 2

31
185
15
217
167
7
206
158

CCSoftware . . . . . . . . . . . . . .
Carlson . . . . . . . . . . . . . . . . .
Cascade Electronics . . . . . . . . . .
Catenary Systems . . . . . . . . . . .
Computer Book Co.
. . . . . . . .
Compuview Products, Inc. . . . . . .
Coronado Enterprises . . . . . . . .
Cottage Resources . . . . . . . . . . .

201
193
216
161

OPENetwork . . . . . . . . . . . . . .
Oregon Software . . . . . . . . . . . .
OSCS . . . . . . . . . . . . . . . . . .
Opal Fire Software . . . . . . . . . .

3
113
119
189

10
93

Emerald Microware . . . . . . . . . . 29
Erac Co . . . . . . . . . . . . . . . . 45

PC Tech . . . . . . . . . . 93, Back Cover
Paradigm Systems . . . . . . . . . . . 80
Peripheral Technology . . . . . . . . 78
Photronics . . . . . . . . . . . . . . . 93
Programmer's Journal ... Inside Back

80
91
39
75
94
11
91
91

69
75
53
79

139 Quantasm Corp . . . . . . . . . . . . 72
204 Gateway Technologies, Inc.. . . . . . 91
212 General Software . . . . . . . . . . . 63
Genus Microprogramming . . . . . . 19
149
22

Information Modes . . . . . . . . . 7,91
Integrand................. 6

213 JB Software . . . . . . . . . . . . . . . . 5
198
215

LegendCommunications . . . . . . . 93
Loalco . . . . . . . . . . . . . . . . . 93

RJSwantek Inc . . . . . . . . . . . . .
Regan Consulting & Programming .
Revolution 2 . . . . . . . . . . . . . .
166 Robotic Systems . . . . . . . . . . . .
171 Ryle Design . . . . . . . . . . . . . .

37
66
25
79
84

176
200
127
162

91
94
13
84

210

Sampson Engineering . . . . . . . . .
Saucci . . . . . . . . . . . . . . . . . .
SemWare . . . . . . . . . . . . . . . .
Semi-Disk Systems . . . . . . . . . .

....

Sintar Software . . . . . . . . . . . .
Software Blacksmiths, Inc. . . . . . .
Software Science, Inc. . . . . . . . . .
Star-K Software Systems . . . . . . .
Symbologic . . . . . . . . . . . . . . .
Symmetric Research . . . . . . . . . .

76
56
33
18
59
91

Turbo Users Group . . . . . . . . . .
218 Technica Rex . . . . . . . . . . . . . .
178 Traxel Labs . . . . . . . . . . . . . . .
194 Turbo Power . . . . . . . . . . . . . .

87
91
32
51

219
129
40
202
182

211

Upper Deck Systems

. . . . . . . . . 93

62
208

V Communications . . . . . . . . 43,82
Vernier Software . . . . . . . . . . . . 44

124
192
169
164
203

WenhamSoftwareCo . . . . . . . . .
Western Hydrologic Systems . . . . .
Western Wares . . . . . . . . . . . . .
Whitney Software . . . . . . . . . . .
Winter City Software . . . . . . . . .

Xenosoft................. 93

207

Young Software Engineering . . . . . 23

Zortech.................. 1

93
93
72
91
93

**
Contact Advertiser Directly.
When you write for information, please tell these folks you read about their products in Micro Cornucopia.

MICRO CORNUCOPIA, #53, May, 1990

By Gary Entsminger
P.O. Box 2091
Davis, CA 95617

THE

Man-Machine
InteractionsThe Edges of Reality

Can you control your own reality-even
a little bit? Here's a way to find out.
hanks in part to the "strange"
point of view cast by quantum
physics in this century, philosophers, psychologists, other-ologists, and
just plain folk have readdressed the classic question: What role does consciousness
play in the creation of reality?
In particular, the Princeton Engineering Anomalies Research Program
(PEAR), has been challenging the common sense view of reality through carefully controlled experiments.
Robert Jahn and Brenda Dunne have
collected ten years' worth of PEAR results in a fascinating book, Margins of Reality, the Role of Consciousness in the
Physical World. Margins attempts to:
(1) produce a history of scientific attempts to credit and discredit paranormal events;
(2) define parapsychology;
(3) review the design, operation, and
results of an ensemble of engineering experiments. These experiments address
the interaction of consciousness with
both inanimate devices and systems
embodying random processes;
(4) create a theoretical framework to
explain the PEAR results.
The story Margins relates is, to say the
least, ver-y in-ter-est-ing.

Man-Machine Connections
The PEAR Program's experiments fall
into two broad categories:
(1) man-machine interactions;
(2) precognitive remote perception.
PEAR's interest in man-machine interactions stems from concern about the
sensitivity of technology. One possible

MICRO CORNUCOPIA, #53, May, 1990

subtle influence is human consciousness.
I found two of the PEAR experiments
especially intriguing. Both utilize random event generators (REGs) to create
random patterns. One uses microelectronic circuitry; the other uses a random
cascade apparatus (a macroscopic contraption, 10 feet wide and 6 feet high,
which lets 3/4" balls funnel down
through a quincunx array of 330 pegs).
In both experiments, PEAR establishes a control by running the REGs unattended. These unattended runs
establish the likelihood of random output. Then an operator positions herself
several feet from the REGs and tries to
affect the output by thinking about it.
In the microelectronic experiment, the
experiment simulates a coin toss (the
outcome is binary: 1 or 0). The operator
tries to increase the output of either Is or
Os (in the controlled runs the output of
Is and Os is roughly equal).
In the macroscopic experiment, the
controlled runs produce a cascade of
balls that's bell-shaped. In this scenario
the operator now tries to shift the cascading of balls either left or right.
The results, as you might expect, vary
among operators. Some had an extraordinary ability to affect the output, particularly as the number of experimental
runs increased.
Typically, an operator's effect wasn't
significant in the first few hundred or so
runs. But as the number of runs increased, the operators got better.
My Test/Your Test
During the past 15 years, I've encountered many paranormal claims,
most of them seriously flawed one way
or another. Most often, experimenters
claim more than their experiments show,
fail to completely isolate variables, or are

unable to replicate their results.
PEAR avoids the common pitfalls of
paranormal research, doesn't claim too
much, and seems well on track to discovering important clues to the role consciousness plays in creating reality.
Just for fun, try an experiment. Set up
an REG using the randomize and random functions in Turbo Pascal or C.
Make sure you call randomize before
each call to random, otherwise each run
is really only one random event and seriously flawed as a control for the experiment. Calling randomize reshuffles the
deck by calling the system clock to
reseed the random number generator.
. ,For exampleuses CRT;
var
I, R, Tot1, TotO: integer; C:char;
begin
Totl := 0; TotO := 0;
for I := 1 to 10000 do begin
randomize;

R:= random(2);

if R = 1 then Tot1 := Tot1 + 1
else TotO := TotO + 1;
end;
writeln('Total1: ',Tot1);
writeln('TotalO: ',TotO);
end.

Generate ten groups of ten thousand
samples. Call it the control. Then repeat
the experiment, but this time concentrate
throughout the process (an event at a
time) on either a 1 or 0 outcome. Who
knows-maybe the consciousness of
Micro C affects reality?
References
Jahn, Robert G. & Brenda J. Dunne;
Margins of Reality; Harcourt, Brace,
Jovanovich; 1987; $17.95 (paperback).

•••

The complete Programmer's Journal "On Graphics" series-assembled in one volume for the very first time!

Maximize your EGA/VGA programming
graphics with Que's Power Graphics
Programming! Que has teamed up with
Programmer's Journal to bring you
this nuts-and-bolts graphics programming guide. Now, for the first
time, you can have the expertise
of Michael Abrash consolidated
in a single volume. His insights
and approach to efficient:
effective programming have
earned him high acclaim for
his "On Graphics" series.
If you are seeking that elusive graphic edge, you
can't be without Power
Graphics Programming!
Order yours
directly from
Programmer's Journal
To order call 800·234·0386.

Available in better bookstores and computer stores
everywhere. Or call 800-428-5331, ext. PJOI.

Que@

Contact Advertiser Directly

VERY HIGH' PERFORMA

ct:

Processors, Memory, and Display Adapters
The X24 High performance processor
•
•
•
•
•

12 or 16 MHz 80286 with NO WAIT STATES!
Small size ("XT" height and length) passive bus design
1 to 4 Mbyte 0 wait state dynamic memory
Fully "AT" compatible Award BIOS
Runs DOS versions 2.2 and later, Xenix and OS/2

The X24 combines the best of motherboard and backplane designs in a too% AT compatible system. Incor·
porating a t 6 MHz 80286, the X24 processor is designed to operate with the PC Tech Advanced System Moth",·
board, which contai ns the peripheral interfaces (hard disk, floppy disk, two serial ports and a parallel port). The
X24 processor can also be used wijh other totally passive bus backplanes. Most criti~1 components including
the microprocessor and up to ~ ooeslstMes of fast memory are contained on a single PC size plug-in card . This
allows the processor and mwn system memory to be serviced or upgraded without disturbing other peripherals
such as serial ports and disk drives.

PC Tech X24 and ASMB
The PC Tech Advanced System Motherboard
• Built in "IDE" interface for AT interface type hard drives
• Fully AT compatible floppy disk support for 3.5", 5.25" drives, capacities of 360k, 1.2m and 1.44m
• Two serial ports and one parallel port
• 8 total expansion slots PC/XT/AT compatible (4 slots have 32 bit bus)
The PC Tech Advanced System Motherboard is designed to complement PC Tech's X24 and X32 high performance processor cards. It contains the mass storage interfaces necessary for a complete system, plus the basic I/O required in most systems . Extra care has been given to FCC compliance by design.

34010 Monochrome Graphics Adapter II

PC Tech Mono-II

• Up to 384k bytes display memory
• Up to 2 Megabytes program memory
• Software is RAM based, allowing complete operating software
replacement and timing re-programming from the host bus
• 34010 program loader included. Assembler, debugger, and C
compiler available.
• Full hardware and software CGA, MDA and Hercules emulation
• Single bit shared memory bit-map with optional resolution up
to 2048 x 1536 (736 x 1008 standard)
• Very high resolution COLOR version available
• Custom 34010 software development available

The TMS34 01 0 is a true general purpose graphics processor. PC Tech makes the total processing power of the 34010 available to both programmers and end users. Our 34010 Monochrome Graphics
Adapter is designed to allow programming from the PC{XT/AT host bus. You can completely replace our 34010 software wijh yours to directly hamess the incredible image processing power of the TMS
34010 for your application. We make a complete set of development tools available. including an assembler, C compiler, program loader, 34010 debugger. and PC interface tracer/debugger. Our stand·
ard product includes support fO( extended CGA, MDA and Hercules emulation as well as a host addressable graphics bit· map. We also support and recommend the oolS graphics interface standard (from
Graphic Software Systems) for applications development as an anernative to native 340tO software development. Ready to run drivers are available for most major applications software packag es as well.

Custom Designs Available
PC Tech will license most products for non-exclusive manufacture. We will also customize any of our designs to better meet your needs on our in-house CAD systems.
All of our standard products are available in private label versions.

About PC Tech
PC Tech has been designing , manufacturing and marketing high performance PC related products for over three years. Our standard product line includes processor,
memory, and video products. All products are designed , manufactured and supported in our Lake City, Minnesota facilities.

Designed, Sold and Serviced By:

High resolution fractal produced
on the PC Tech COLOR 34010
907 N. 6th St., Lake City, MN 55041
(612) 345-4555 • . (612) 345-5514 (FAX)
PC, XT, AT, oolS, Hercules, and Award BIOS are trademart
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 : 2013:01:25 15:18:15-08:00
Modify Date : 2013:01:25 19:09:54-08:00
Metadata Date : 2013:01:25 19:09:54-08:00
Producer : Adobe Acrobat 9.52 Paper Capture Plug-in
Format : application/pdf
Document ID : uuid:11e977a2-4ee6-462f-995b-a4bccbae4a4e
Instance ID : uuid:d556ea78-dd5f-4c45-a526-fae5af855852
Page Layout : SinglePage
Page Mode : UseNone
Page Count : 100

EXIF Metadata provided by EXIF.tools

Micro_Cornucopia_#53_May90 Micro Cornucopia #53 May90

Micro_Cornucopia_#53_May90 Micro_Cornucopia_%2353_May90

Navigation menu

Versions of this User Manual:

Views

Navigation