EEVblog #663 - Compucorp 322G Calculator Teardown

[EEVblog]

Teardown Tuesday
Inside a 1973 vintage programmable scientific desktop calculator.
The P2102 is a 1K x 1 bit SRAM, so looks like a 4 bit processor architecture.
Teardown photos: http://www.eevblog.com/2014/09/16/eevblog-663-compucorp-322g-calculator-teardown/

[Refrigerator]

I think you should replace all the caps because they do look like crap.  If there was one short there must be more. 
Ps: is that white crud leaked electrolyte ?

[Hade]

Quite disappointed that you couldn't get it fired up.
Great watching your fault finding approach though!

[coppice]

When that calculator was made HP was already selling their pocket sized scientific calculators. They were not programmable, but they offered a lot more functions than this machine seems to.

I was amused by Dave's idea of them using gate arrays. Does he really think gate arrays existed in 1972? Every gate counted back then. Youngsters have no clue, do they?  The earliest gate array like devices I know about were made in the late 70s, and gained no popularity for several years because of costs and performance limitations.

[Pentium100]

I agree that Dave should replace all of those caps. You can't measure the cap leakage with the resistance function of a multimeter (not for high voltage caps, anyway). I have stepped on this particular mine once when repairing a tube tape recorder. The caps were leaky, but that only showed up if I put at least 12V on them (with a series 1M resistor, the cap "resistance" turned out to be around 1M which is quite low for a circuit with 500k resistors).

So, the caps are probably leaky (but that only shows up with voltage).

Also, with that much power, something should be getting warm, a cap probably.

[rsjsouza]

Quick info on the calculator's chipset: here

[marshallh]

You don't see Burroughs panaplex displays around much

[German_EE]

Back in 1978 when I did the final year of my degree I designed and built a calculator using 74 series logic. This one however used Roman numerals so if you really wanted to know what MCMVI-LX was then I was your man. The total board area was about the same as Dave's calculator but it consumed about 5A when idle and about 7A when carrying out a multiple or divide.

As the display of the Compucorp device is totally blank I would check the driver chips. The display itself will probably have a filament without which there will be no output so that is another thing to check.

[robrenz]

You don't see Burroughs panaplex displays around much

I lived about a mile from the Burroughs plant in Downingtown PA where I think those were made. I made dies for the Styrofoam packaging for those displays way back when so I think they must have been made there. I made them for a contract packaging insert company that is why I am not sure.

[SeanB]

On that processor board likely one of the chips will be an ALU, one will be the register set, one the microcode decoder and the last will be the control signal generator. TI did that with later CPU chipsets, until they were able to fit all of it onto one chip with decent yield. Very likely the ROM chips are actually EPROM, likely either a 2708 or a 2716 ( depending if there is a -21V supply rail on the power supply) in an OTP package, with them being programmed during assembly with the code. If you only were doing less than 10k of a ROM you did this, later changing to a mask ROM which worked out to a few cents cheaper per part. Most expensive part of that ROM is the actual CERDIP package, which costs a lot more than the die inside. If you wanted the erasable version you simply just had a windowed steel top soldered to the carrier instead of the sealed top.

[rsjsouza]

...and a description of a very similar calculator, with disassembled pictures:
http://www.datamath.org/Related/Compucorp/324G.htm

[HP-ILnerd]

Great video, Dave!  Love watching you debug stuff.
Also, thanks for hauling out the 1052e.  It might be old, and none too fancy by modern standards, but it's nice to show people that it's still a perfectly usable tool.

[G7PSK]

At the end of the video just as Dave was saying there was no display as pressed the buttons I thought I saw a faint flicker on the display right hand side. At about 27 min plus a few seconds there is a glow on the last 2 digits of the display very briefly.

[edpalmer42]

The display will likely display a zero on powerup.  If you probe the leads that go to the display and look for the segment leads, you might see them change when you press a key.  That would help seperate display problems from logic problems.  I've seen those displays fail with age because the seal around the lead fails.

Ed

[kwass]

...and a description of a very similar calculator, with disassembled pictures:
http://www.datamath.org/Related/Compucorp/324G.htm

The 324G isn't just a similar calculator it's exactly the same calculator with the prog 1/2 switch slot populated.  It was common practice back in the day to buy a 322G and make it into a 324G by cutting a hole in the top template and putting in your own slide switch.

I have most of the Compucorp 300 series manuals on my website: http://www.wass.net/manuals/index.htm along with the service manual for the 326.   Dave, this should give you all (most?) of what you need to debug the 322G I think:  http://www.wass.net/manuals/Compucorp%20326%20Service.pdf

Also, I have working versions of all these machines if you need me to take some measurements, let me know.

-Katie

[coppice]

Very likely the ROM chips are actually EPROM, likely either a 2708 or a 2716 ( depending if there is a -21V supply rail on the power supply) in an OTP package, with them being programmed during assembly with the code. If you only were doing less than 10k of a ROM you did this, later changing to a mask ROM which worked out to a few cents cheaper per part. Most expensive part of that ROM is the actual CERDIP package, which costs a lot more than the die inside. If you wanted the erasable version you simply just had a windowed steel top soldered to the carrier instead of the sealed top.

Your dates are way off. The 2708 didn't exist in 1972. Even later when it did exist it was always in a windowed ceramic package. The market for cheap OTP devices came much later.

The 2708, 2716 and probably all the MOS devices in the calculator needed +12V, +5V and -5V. The early EPROMS needed (I think) +28V for programming, but you wouldn't find that rail in a product.

Ceramic packages were not the expensive part of an EPROM until much later. In 1972 an EPROM die cost a fortune, which totally swamped the cost of the windowed package. Only later, when silicon prices collapsed, did the package become a key issue. in 1972 the key issue was maximising the packaged yield from the precious good dies. Ceramic was better than plastic for this. For comparison, look in an HP calculator of that vintage. They used a selection of weird, wonderful and horribly expensive ceramic packages.

[TheMG]

I wouldn't be surprised if a short/breakdown has developed inside the display and taken out the display driver along with it.

I've repaired quite a large number of Bendix-King aircraft transceivers that have similar gas-discharge displays and the display module (consisting of the display itself and the driver ICs) are the most common failure item. Sometimes you could actually see where breakdown had occurred inside the display if you looked closely.

I'd still go ahead and try replacing all the capacitors in case one is breaking down, doesn't hurt to try, but at this point I'd say there's a more than 50% chance the display is at fault.

[EEVblog]

I'd still go ahead and try replacing all the capacitors in case one is breaking down, doesn't hurt to try, but at this point I'd say there's a more than 50% chance the display is at fault.

That's the vibe I get too. Need to do some more testing.

[EEVblog]

Dave, this should give you all (most?) of what you need to debug the 322G I think:  http://www.wass.net/manuals/Compucorp%20326%20Service.pdf

Awesome, thanks!

[Flump]

I liked the diagnosis and fault finding
but would it not have been better to use
an ESR meter on the caps or were they the wrong type ?

Shame you didn't get the full repair but you had a good go 

[VK3DRB]

An excellent article on Compucorp and their technology...

http://www.oldcalculatormuseum.com/d-compucorp.html

Looking at the board, EMI/EMC were not a big concern in equipment back in the 70's. Even if it was working at a couple of hundred kHz, if the slew rate of the digital switching was high enough and the currents high, it could cause interference to AM radio.

Dick Smith's dodgy System-80 "computer" is a classic example. Virtually no shielding, slow logic, antennas everywhere and it radiated like Chernobyl on steroids. Wiped out the ABC TV on 52MHz. Smith might have been lucky the radio inspectors (remember them?) were too busy policing the federal government's CB radio licensing scam.