CBM PET 4032 Restoration

[JammyJules]

Hello all,

My first time posting, and my first retro computer restoration project.

I recently bought a PET 4032 which had been in storage for sometime. The owner assured me there probably isn't much wrong with it (he is a collector, and said he was selling it to make room).

Externally it's in very good condition, but the monitor metal housing is rusted, and the circuit board for the monitor is very dirty.

Anyway - cut to the chase - it doesn't display a picture. There is nothing at all - even if I turn up the brightness and sub brightness full the tube doesn't glow. There is life in the tube heater (excuse my terminology!) - as you can see the neck of the tube light up when it's powered on. I disconnect J7 from the motherboard and tweaked the sub brightness and focus and turn it on and a small dot appeared in the center of the screen which gradually faded out. Since then - I have had nothing. No static electricity can be felt across the front of the tube.

I bought an oscillascope and checked J7 for HSYNC and VSYNC - they appear to be present. None of the caps on the monitor PCB appear to be bulging or leaking from initial inspection. I just bought an HV discharge wand (the thought of this frightens the life out of me) of which I am wating for delivery. My plan is to discharge the tube (if there is anything in it) and remove the monitor PCB for further inspection.

I am just wondering if anyone has any advice on how to proceed ? Should I be expecting the tube brightness to show without J7 connected ? Or assuming my scope tests on J7 are correct, should I see a glow even if there is no video signal genereated by the motherboard (8032090) ?

Thanks in advance for any advice.

 

[Hugo Holden]

While the voltage for the CRT is very high, the current is low. Also when the CRT is charged up (and the power off) the amount of energy stored in it is relatively low, and the shock you receive from it has about the same mJ energy as a single zap from a car or lawn mower's spark plug terminal. Not enough to kill you. most people just over-react from the fright of it. Most of the charge in the CRT dissipates overnight, because of the non-zero reverse leakage of the EHT rectifier. If you don't go under the big anode cap you cannot come in contact with the stored charge anyway.

 

[dave_m]

Should I be expecting the tube brightness to show without J7 connected ? Or assuming my scope tests on J7 are correct, should I see a glow even if there is no video signal genereated by the motherboard (8032090) ?

Thanks in advance for any advice.h

The video board needs the 20 KHz horizontal drive from the main board to generate the high voltage, so if the horizontal and vertical (60 Hz) signals are present, you can connect J7.
You may want to check to see if the +18 VDC voltage regulator (VR2) is working in the video board. Without the +18V, nothing will work on the video board. It is powered by 21 VAC from the large transformer in the PET (pins 9 and 11 on the transformer).

However the +18V also powers the CRT heater that you may be seeing.

 

[MiaM]

Note that there are the 4032 PETs that are the 40 col version of the 80xx series, with a 12" monitors. Those use about 20kHz h-sync like dave_m mentioned above.
But there is also 4032 PETs that I call "fake 4032" as they are really a 3032/2001N-32 with upgraded ROMs. There use 15kHz h-sync like a regular TV.

For the latter for testing purposes you can use a simple adapter (connected either to the monitor port or the user port) to connect a regular composite monitor / TV, or you can (possibly needing inverters for one or both of the syncs?) connect a CGA compatible monitor.

In theory an EGA compatible monitor should work with the 12" PETs but I don't know if anyone has ever tested that.

Either way start with checking if you get any signals on the sync outputs and the video output from the digital/main PET board. If so continue trouble shooting the monitor. Otherwise I would disconnect the monitor (IIRC you have to desolder/cut the AC feed from the mains transformer) and just put the monitor away while you trouble shoot the main board, to free up some desk space. Possibly remove the main PCB, transformer and the large smoothing capacitor from the case to make it more easy to handle while trouble shooting.

A strong++ recommendation is to remove any mains filter! They have the same problem as the RIFA capacitors. If they go bad you end up with the smell of severely burnt and poisoned ginger bread, and anything that the smoke touches will smell awful for ages. Thankfully most of a PET is metal so it's easy to clean. When this happens to anything using a plastic case or whatnot you just have to dump it in your garage, in a parts/junk car in your yard or so and let it sit for a few weeks until the smell is mostly gone.

 

[JammyJules]

While the voltage for the CRT is very high, the current is low. Also when the CRT is charged up (and the power off) the amount of energy stored in it is relatively low, and the shock you receive from it has about the same mJ energy as a single zap from a car or lawn mower's spark plug terminal. Not enough to kill you. most people just over-react from the fright of it. Most of the charge in the CRT dissipates overnight, because of the non-zero reverse leakage of the EHT rectifier. If you don't go under the big anode cap you cannot come in contact with the stored charge anyway.

Thanks for that - I will try the wand with more confidence !

 

[JammyJules]

Note that there are the 4032 PETs that are the 40 col version of the 80xx series, with a 12" monitors. Those use about 20kHz h-sync like dave_m mentioned above.
But there is also 4032 PETs that I call "fake 4032" as they are really a 3032/2001N-32 with upgraded ROMs. There use 15kHz h-sync like a regular TV.

For the latter for testing purposes you can use a simple adapter (connected either to the monitor port or the user port) to connect a regular composite monitor / TV, or you can (possibly needing inverters for one or both of the syncs?) connect a CGA compatible monitor.

In theory an EGA compatible monitor should work with the 12" PETs but I don't know if anyone has ever tested that.

Either way start with checking if you get any signals on the sync outputs and the video output from the digital/main PET board. If so continue trouble shooting the monitor. Otherwise I would disconnect the monitor (IIRC you have to desolder/cut the AC feed from the mains transformer) and just put the monitor away while you trouble shoot the main board, to free up some desk space. Possibly remove the main PCB, transformer and the large smoothing capacitor from the case to make it more easy to handle while trouble shooting.

A strong++ recommendation is to remove any mains filter! They have the same problem as the RIFA capacitors. If they go bad you end up with the smell of severely burnt and poisoned ginger bread, and anything that the smoke touches will smell awful for ages. Thankfully most of a PET is metal so it's easy to clean. When this happens to anything using a plastic case or whatnot you just have to dump it in your garage, in a parts/junk car in your yard or so and let it sit for a few weeks until the smell is mostly gone.

Brillant - thank you.

 

[JammyJules]

Just an update - it seems to be getting worse! My signals have gone on J7, and scoping the RW of the CPU seems to be held high. I have a diagnostice board coming in the post ...

 

[daver2]

Welcome to VCFED.

Don't forget to check the VIDEO signal on J7. You need this - as well as VDRIVE and HDRIVE - to produce a picture...

No video, no picture!

Dave

 

[Hugo Holden]

Note that there are the 4032 PETs that are the 40 col version of the 80xx series, with a 12" monitors. Those use about 20kHz h-sync like dave_m mentioned above.
But there is also 4032 PETs that I call "fake 4032" as they are really a 3032/2001N-32 with upgraded ROMs. There use 15kHz h-sync like a regular TV.

This is why it is good when the OP takes the time to post a picture of the pcb they are working with, both the computer mobo and the VDU's pcb. There are a lot of PET board variations inside the same cabinets, it is all very mix & match and that leaves everyone, like the fortune teller, trying to peer through cloudy crystal balls.

 

[dave_m]

Note that there are the 4032 PETs that are the 40 col version of the 80xx series, with a 12" monitors. Those use about 20kHz h-sync like dave_m mentioned above.
But there is also 4032 PETs that I call "fake 4032" as they are really a 3032/2001N-32 with upgraded ROMs. There use 15kHz h-sync like a regular TV.

The OP did post that the assembly part number was 8032090. He has a 12" system with a 6545 CRT Controller. The Horizontal Sync frequency should be 20 KHz.

 

[JammyJules]

That's a good point - attached is my PCB. As a bit of background, this is the first time I have ever tried anything like this. I've worked in IT all my life (I'm 57) and understand the concepts behind the hardware, but have never delved into it deeply. I bought a scope specifically for this purpose - so it's the first time I've used that. It's also the first time I've tried to read a schematic. I've been downloading specs for the various chips - and I have labelled a few of them as you can see. I am finding it facinating tracing the logic through each. I'm using ChatGPT to help me diagnose issues (it likes to jump to conclusions it would appear!).

The PET was the first computer I ever touched, which then began my journey into an IT career which has spanned ~35 years. I've been watching YouTube videos like Adrians Digital Basement and been intrequed to see if I could try to restore something myself. As the PET has a special place in my heart - it seemed like a good candidate.

I have diconnected the power to the CRT board and diconnected J7 while I'm running these tests.

I have done some more scoping around. The CPU appears to be stopping after about 1 sec. The address lines show activity for that second and then hang.

The 74154 in UE12, pin 17 goes LOW, but 20 stays HIGH. ChatGPT suggest this is a problem with a gate between pin 17 and the CE on the kernal may be bad. I have traced the CE to pin 3 of a 74LS00B1 (UD7) - I need to check that logic next.

I have ordered some chip sockets and various chips should I need to replace any. If have a desoldering gun, but read somewhere that it potentially could pull traces / via's off the board - is that correct ? Is it better just to snip the pins and desolder using a fine tip on a soldering iron ?

Thank you all for the input so far - I really appreciate it.

 

[daver2]

If UE12 (74154) pin 17 pulses low - what pin 20 are you checking? Are you checking pin 20 of UE12 (74154) or the ROM at UD6?

If you are checking on UE12 - that is a waste of time. If you are checking on UD6, then that would indicate a problem with the PCB track between UE12 pin 17 and UD6 pin 20. However, this is only valid if the CPU is consistently executing instructions...

UD7 is the EDIT ROM (address range $E000 to $E7FF) and is associated with gates within UE5 and UE14 - but nothing to do with UE12 pin 17 or UD6... I would ignore Chat GPT (or any other dumb piece of AI come to that).

I would not assume that the CPU has stopped if the address bus looks like it is static. Check the 6502 CPU pin 7 (SYNC) for signs of pulsing. If this pin is not oscillating then the CPU has definitely halted.

If in doubt, ALWAY check CPU pin 7 for pulsing...

Dave

 

[dave_m]

JJ,
Note that the kernal ROM in position UD6 is mislabeled as 'Editor' but has the kernel part number, and the Editor ROM is mislabeled as 'kernal'. I assume the actual correct parts are installed in the right positions as the kernel ROM is soldered in place.

There are a lot of IC sockets in your board, more than on most PETs. Have you wiggled them a bit and reseated them? There may be some oxide on the sockets. The sockets will help with troubleshooting.
-dave_m

 

[Hugo Holden]

I'm using ChatGPT to help me diagnose issues (it likes to jump to conclusions it would appear!).

GPT-5 has undergone a series of proprietary upgrades, which only those in the inner design circle know about.

Previously it was a basic LLM (large language model a form of auto-complete) and the operating theory of LLMs does not support the ability of the AI to reason, however, after the reasoning modules were added, the AI can now reason and quite successfully too.

For example it can now correctly answer a question like this where reasoning is required: Under what condition could the following statement be true ? All of my Boa Constrictors are on the Moon, and all my Boa Constrictors are on Mars and all my Boa Constrictors are at the bottom of the Ocean. And now the AI can do something that a basic LLM could not do, reason out the answer and reply: I have no Boa Constrictors, (essentialy using vacuous set theory)

It can tackle many problems now that are not on the internet or in any textbook and it can solve original problems from first principles. I have tested with many unique problems that are not on the net or in any textbook and it does very well and as you know occasionally makes errors. The manufacturers are now selling/marketing it, based on its Reasoning Ability, the more $ you pay the more reasoning ability you get.

So the AI has in fact become quite a Transformative Tool and is now longer a "regurgitator of internet data" that many believe it to be. That is what it once was, that has now changed.

The AI also improves if you activate the memory settings and before every interaction it reviews all of your past questions and exchanges, unfortunately, for privacy reasons it has no enduring memory so it cannot pass information to any third party, so it will appear to behave differently if you start an interaction with a new instance of it, on a new computer. ChatGPT can study images, real time video and even extract spacial data from 3D video too, can OCR text, is multi-lingual, it can write original software in most computer languages.

However, having said all the above, fixing something like a broken vintage computer is one of those things that takes many years of experience to do in a timely manner. While ChatGPT can even study the schematic if you present it as scanned image and even understand the operating theory and the firmware too, it won't make a good or efficient service technician right off the bat. It is like Elon Musk who remarked that his Robot Optimus would make an excellent Surgeon but forgot the obvious. Maybe after a decade of training and 1000's of operations where the Robot like the Human, gets to see all the vagaries of human tissues between individuals and has finally seen all of the things that can go wrong in its hands, go wrong , remembers all those and figures out how better to deal with them when they happen.

So I agree with @daver2 about not using the AI to help you fix the PET board, but not because the AI is "dumb" but because it does not have the decades of experience in the practical repair of digital electronics, especially relating to vintage computers, picking up on the many clues, some subtle and some not so subtle in a diagnostic process while you are scoping your computer board and making various diagnostic tests. You are much better off not using ChatGPT for this particular application/problem and instead relying on @daver2 and other members here, to help resolve the problem.

About de-soldering IC's I copied this I wrote from another thread:

I'm not sure of your level of experience de-soldering IC's, so I'll give some tips before its too late. Many people working on PET boards have damaged the pads, tracks & plated through holes and create extra troubles as they go. If you remove IC's you have to be very careful of the pcb itself, the IC's must be sacrificed for less risk of pcb damage. While certain de-soldering tools and methods can be used to save the IC's, there is more risk of thermal and mechanical pcb damage to the pcb. These particular IC's at their age, are not worth saving.

Use very sharp point small sized needle nose cutters such as the Hozan N-31, and cut each pin off close to the IC's plastic body. Then add fresh multicore solder to each pin and pad, this helps. Then , with small forceps hold each pin, only attempt to pull the pin out of the hole after the solder is fully melted with the iron. If you apply any force before that, it can pull the pad & tracks off the pcb and damage the plated through holes too. Once the pins are removed, clear the holes with the sucker one by one, and adding some fresh slolder can help here too. The best solder suckers are the large single shot ones made by Hakko, then clean up the area with IPA and cue tips, then very carefully inspect the pads & tracks for any damage under good light and magnification, then fit IC sockets. 2114's are available from many suppliers, buy a few from a couple of suppliers. Even unused parts can be defective sometimes.

Of course it is notorious too that people don't follow advice exactly, so they think the old cutters they have will do, but they are too big and blunt and excessive force is put on the pins & pads, or they don't have a good temp controlled soldering iron and good multicore Tin/Lead based solder, or do bad things like adding liquid fluxes, or have a useless solder sucker they think will work, and the pcb gets damaged anyway. If you don't have good tools or experience in this area, its much better to buy the tools, then practice on waste double sided pcb's removing DIL IC's for a while, before you attempt any IC removal on the actulal PET board. It is always a shame when a PET gets injured.

It is a shame when a more precious part like a soldered in original ROM has to be ruined in favor of saving the pcb though, when it might turn out the ROM was ok. The only saving grace is that you can recreate them in UVeprom form from the files on Zimmers. Though with an excellent sucker it may be possible after adding fresh solder to free the soldered in ROMs for testing, without cutting its pins, and without excessive pcb damage. When the hole is as clear as possible of solder, you can move the pin fron side to side in the hole, to free each one before attempting to remove the undamaged IC. You need to get a good ROM reader/writer, the GQ-4x works well for 2532 and 2716 ROMs. If any original non-socketed ROM's were successfully desoldered, you need to remove all excess solder from their pins with wick, as you would be fitting them into new sockets later.

 

[JammyJules]

JJ,
Note that the kernal ROM in position UD6 is mislabeled as 'Editor' but has the kernel part number, and the Editor ROM is mislabeled as 'kernal'. I assume the actual correct parts are installed in the right positions as the kernel ROM is soldered in place.

There are a lot of IC sockets in your board, more than on most PETs. Have you wiggled them a bit and reseated them? There may be some oxide on the sockets. The sockets will help with troubleshooting.
-dave_m

Thanks for that observation - I will amend the labels for clairty.
I think I'll invest in some deoxit and give them all a clean.

 

[JammyJules]

GPT-5 has undergone a series of proprietary upgrades, which only those in the inner design circle know about.

Previously it was a basic LLM (large language model a form of auto-complete) and the operating theory of LLMs does not support the ability of the AI to reason, however, after the reasoning modules were added, the AI can now reason and quite successfully too.

For example it can now correctly answer a question like this where reasoning is required: Under what condition could the following statement be true ? All of my Boa Constrictors are on the Moon, and all my Boa Constrictors are on Mars and all my Boa Constrictors are at the bottom of the Ocean. And now the AI can do something that a basic LLM could not do, reason out the answer and reply: I have no Boa Constrictors, (essentialy using vacuous set theory)

It can tackle many problems now that are not on the internet or in any textbook and it can solve original problems from first principles. I have tested with many unique problems that are not on the net or in any textbook and it does very well and as you know occasionally makes errors. The manufacturers are now selling/marketing it, based on its Reasoning Ability, the more $ you pay the more reasoning ability you get.

So the AI has in fact become quite a Transformative Tool and is now longer a "regurgitator of internet data" that many believe it to be. That is what it once was, that has now changed.

The AI also improves if you activate the memory settings and before every interaction it reviews all of your past questions and exchanges, unfortunately, for privacy reasons it has no enduring memory so it cannot pass information to any third party, so it will appear to behave differently if you start an interaction with a new instance of it, on a new computer. ChatGPT can study images, real time video and even extract spacial data from 3D video too, can OCR text, is multi-lingual, it can write original software in most computer languages.

However, having said all the above, fixing something like a broken vintage computer is one of those things that takes many years of experience to do in a timely manner. While ChatGPT can even study the schematic if you present it as scanned image and even understand the operating theory and the firmware too, it won't make a good or efficient service technician right off the bat. It is like Elon Musk who remarked that his Robot Optimus would make an excellent Surgeon but forgot the obvious. Maybe after a decade of training and 1000's of operations where the Robot like the Human, gets to see all the vagaries of human tissues between individuals and has finally seen all of the things that can go wrong in its hands, go wrong , remembers all those and figures out how better to deal with them when they happen.

So I agree with @daver2 about not using the AI to help you fix the PET board, but not because the AI is "dumb" but because it does not have the decades of experience in the practical repair of digital electronics, especially relating to vintage computers, picking up on the many clues, some subtle and some not so subtle in a diagnostic process while you are scoping your computer board and making various diagnostic tests. You are much better off not using ChatGPT for this particular application/problem and instead relying on @daver2 and other members here, to help resolve the problem.

About de-soldering IC's I copied this I wrote from another thread:

I'm not sure of your level of experience de-soldering IC's, so I'll give some tips before its too late. Many people working on PET boards have damaged the pads, tracks & plated through holes and create extra troubles as they go. If you remove IC's you have to be very careful of the pcb itself, the IC's must be sacrificed for less risk of pcb damage. While certain de-soldering tools and methods can be used to save the IC's, there is more risk of thermal and mechanical pcb damage to the pcb. These particular IC's at their age, are not worth saving.

Use very sharp point small sized needle nose cutters such as the Hozan N-31, and cut each pin off close to the IC's plastic body. Then add fresh multicore solder to each pin and pad, this helps. Then , with small forceps hold each pin, only attempt to pull the pin out of the hole after the solder is fully melted with the iron. If you apply any force before that, it can pull the pad & tracks off the pcb and damage the plated through holes too. Once the pins are removed, clear the holes with the sucker one by one, and adding some fresh slolder can help here too. The best solder suckers are the large single shot ones made by Hakko, then clean up the area with IPA and cue tips, then very carefully inspect the pads & tracks for any damage under good light and magnification, then fit IC sockets. 2114's are available from many suppliers, buy a few from a couple of suppliers. Even unused parts can be defective sometimes.

Of course it is notorious too that people don't follow advice exactly, so they think the old cutters they have will do, but they are too big and blunt and excessive force is put on the pins & pads, or they don't have a good temp controlled soldering iron and good multicore Tin/Lead based solder, or do bad things like adding liquid fluxes, or have a useless solder sucker they think will work, and the pcb gets damaged anyway. If you don't have good tools or experience in this area, its much better to buy the tools, then practice on waste double sided pcb's removing DIL IC's for a while, before you attempt any IC removal on the actulal PET board. It is always a shame when a PET gets injured.

It is a shame when a more precious part like a soldered in original ROM has to be ruined in favor of saving the pcb though, when it might turn out the ROM was ok. The only saving grace is that you can recreate them in UVeprom form from the files on Zimmers. Though with an excellent sucker it may be possible after adding fresh solder to free the soldered in ROMs for testing, without cutting its pins, and without excessive pcb damage. When the hole is as clear as possible of solder, you can move the pin fron side to side in the hole, to free each one before attempting to remove the undamaged IC. You need to get a good ROM reader/writer, the GQ-4x works well for 2532 and 2716 ROMs. If any original non-socketed ROM's were successfully desoldered, you need to remove all excess solder from their pins with wick, as you would be fitting them into new sockets later.

Wow ! That's an amazing response !

I use a paid ChatGPT for my job - it is very impressive for code - and therfore I just naturally used it to debug the PET. I take all your observations on board and it is all much appreciated. I do have a good soldering iron, and I am experienced in soldering / desoldering - although not on anything as old as this. I am going to remove the PCB and PSU from the case today - I was waiting for my anti-static mat to arrive. This should make it easier to work through the logic.

 

[JammyJules]

If UE12 (74154) pin 17 pulses low - what pin 20 are you checking? Are you checking pin 20 of UE12 (74154) or the ROM at UD6?

If you are checking on UE12 - that is a waste of time. If you are checking on UD6, then that would indicate a problem with the PCB track between UE12 pin 17 and UD6 pin 20. However, this is only valid if the CPU is consistently executing instructions...

UD7 is the EDIT ROM (address range $E000 to $E7FF) and is associated with gates within UE5 and UE14 - but nothing to do with UE12 pin 17 or UD6... I would ignore Chat GPT (or any other dumb piece of AI come to that).

I would not assume that the CPU has stopped if the address bus looks like it is static. Check the 6502 CPU pin 7 (SYNC) for signs of pulsing. If this pin is not oscillating then the CPU has definitely halted.

If in doubt, ALWAY check CPU pin 7 for pulsing...

Dave

I will check CPU pin 7 - thank you.

 

[JammyJules]

Hi all,

I have pulled the motherboard and put it on the bench.

As a re-cap (no pun intended), the 6502 appears to run for a second or so and then stop.

The reset pin looks good - it is LOW for ~1 sec before staying HIGH.
The clock in is a solid 1mhz 5v all the time the power is on. Clock out is the same.
The SYNC shows some activity and then stays HIGH.
All the RAM chips have the correct volatages (+12, +5, -5).
I have beep tested all the address & datalines from the cpu to the following components : 74154, 2 x 74LS244, 6520 - they all look ok. I've also beep tested all the CE lines for the roms back to 74154.
I ran the 6502 through a NOP tester and it appears to be working.

I orderd a romulator weeks ago which still hasn't arrived - does anyone have any advice what I could test next ?

Thanks in advance.

 

[JammyJules]

I also now have a T48 Programmer, so I can read / write / test chips.

 

[daver2]

A beep test is not as good as the NOP generator and an oscilloscope. This will test the logic gates as well as the PCB tracks.

Do you have an oscilloscope though? On the assumption you checked the 1 MHz clock, you must have?

The likelihood is RAM or ROM faults causing the CPU to go into the weeds shortly after powering up (or resetting).

What ROMulator have you ordered, and does it support my PETTESTER ROM?

Dave