UK Vintage Radio Repair and Restoration Discussion Forum

UK Vintage Radio Repair and Restoration Discussion Forum (https://www.vintage-radio.net/forum/index.php)
-   Vintage Computers (https://www.vintage-radio.net/forum/forumdisplay.php?f=16)
-   -   Non-working Commodore PET 3016 (https://www.vintage-radio.net/forum/showthread.php?t=174829)

ScottishColin 4th Jan 2021 9:33 pm

Re: Non-working Commodore PET 3016
 
Not much in the way of readings this time.

UH7
Pin 3 - solid 8Mhz @ 55%
Pin 7 - no reading

No readings on any of the other suggested pins.

Colin.

SiriusHardware 4th Jan 2021 10:08 pm

Re: Non-working Commodore PET 3016
 
Have a look at pins 2 and 6 of the same IC (UG5) - what frequencies (if any) do you see on those pins?

(Incidentally, I assume you meant UG5 in post #41, not UH7)

ScottishColin 4th Jan 2021 10:47 pm

Re: Non-working Commodore PET 3016
 
I did mean UG5 - apologies

Pin 2 - rock solid 4Mhz @51%
Pin 6 - rock solid 2Mhz @51%

Colin.

Timbucus 4th Jan 2021 11:12 pm

Re: Non-working Commodore PET 3016
 
I have work early tomorrow so will have to drop out now but, I wonder if the lack of the 1Meg main Processor clock is because it is dying due to a fault on the CPU or a short on it somewhere, or if this circuit has a wait state system that suppresses that clock - I will need to study the circuits a bit closer as I have never worked on a PET.

It may not seem like it but, we are making good progress - the clock net is both the most complex and most important part of any circuit so proving it is acting as it should is 50% of the problem usually.

Timbucus 4th Jan 2021 11:19 pm

Re: Non-working Commodore PET 3016
 
Without that 1Mhz signal I do not think any of the phases will be generated as it feeds U5 the shift register NAND inputs on pins 1 and 2 (It is a 74164) so something is likely suppressing it as the 8,4 and 2Mhz is running it is likely not the Chip UG5.

SiriusHardware 4th Jan 2021 11:33 pm

Re: Non-working Commodore PET 3016
 
Right, that's interesting. As you (Colin) may realise, that IC (UG5) takes in the 16Mhz from the main clock and divides it by two several times. Each successive division is available from the QA, QB, QC and QD outputs. Of these, only the QA (8Mhz) and QD (1Mhz) outputs are used in this circuit but your last measurements demonstrated that the QB (4MHz) and QC (2MHz) outputs are also working - also that the frequency measurement feature on your meter is working very nicely too.

However, the QD (1Mhz) output of that IC either does not seem to be working, or it is being prevented from working by something that it is connected to. As (bad) luck would have it that 'CLK1' signal goes to quite a few different areas on the rest of the mainboard.

The 'Flag' at the end of the CLK1 signal line coming out of UG5 pin 7 on the drawing (with the numbers [1,5,7] inside it) indicates that this 'CLK1' signal goes to other points also marked 'CLK1' on circuit drawings 1, 5, and 7 (this is circuit sheet 6 that we are using at the moment).

It's probably a bit late to go hunting off down those signal paths tonight, so I'll leave you with one final check from me:-

With power off, meter on its lowest ohms / resistance range, what resistance do you measure with

-Black probe on UG5 pin 8, Red probe on UG5 pin 7?
-Black probe on UG7 pin 7, Red probe on UG5 pin 16?

(This check is to see whether the 'CLK1' line is shorted to, or has a very low resistance to, either 0V or +5V).

I'm guessing that UG5 is not in a socket?

Edit: Again, cross posted with Tim.

SiriusHardware 5th Jan 2021 12:23 am

Re: Non-working Commodore PET 3016
 
One final follow-up for tonight, I've looked at where CLK1 does go to. As Tim observed it goes to the clock input on the 6502 but it also goes to a couple of other places as well.

On sheet 1 - goes only to UC4, (The 6502) which is in a socket
On sheet 5 - goes to 4 * 74LS153 (UE3, UE4, UE5, UE6) - not in sockets?
On sheet 7 - goes to 3 * 74LS157 (UF3, UF5, UF6) - not in sockets?

ScottishColin 5th Jan 2021 1:47 pm

Re: Non-working Commodore PET 3016
 
Test results

UG5 pin 8->UG5 pin 7 - 0 ohms
UG7 pin 7->UG5 pin 16 - 0.7 M ohms


Socket information:

UF10 - 901447-10

UD9 - 901463-03
UD8 - ?
UD7 - 901465-02
UD6 - 901465-01
UD5 - empty
UD4 - empty
UD3 - empty

UC7 - 6520
UC6 - ?
UC5 - 6522
UC4 - 6502

Colin.

SiriusHardware 5th Jan 2021 2:12 pm

Re: Non-working Commodore PET 3016
 
This one:

Quote:

UG5 pin 8->UG5 pin 7 - 0 ohms
...looks very promising. If that reading is really true there is a short between some point on the CLK1 signal line and a nearby 0V. That would certainly kill the 1MHz output from UG5.

Although there are quite a few devices on that CLK1 line it would be very unusual for a semiconductor device to fail absolutely zero ohms short-circuit - several ohms or a few tens of ohms would be more typical, so my thinking at this stage is that you are more likely to have a physical short, like two component leads bent over and touching, or a solder bead or solder splash between the CLK1 signal line and a nearby 0V point.

Your mission (should you choose to accept it) is to follow the CLK1 track which sets out from UG5 pin 7 and follow it to every other place that it goes to on the board, all the while looking for possible shorts to nearby 0V / GND areas.

Afterthought: As the 6502 is in a socket, you may as well try removing it and seeing if that short between UG5 pin 7 and UG5 pin 8 clears when the CPU is out of the board. If it's still there with the chip out, put the 6502 back in and continue looking for a physical short between the CLK1 track and other tracks next to it.

Timbucus 5th Jan 2021 5:26 pm

Re: Non-working Commodore PET 3016
 
It is unlikely with 0ohms but, a rusty mark between two very close tracks or pins could create such a short as well so a thorough review of that track inch by inch is worthwhile - note there may be vias (joints between sides of the board) and anywhere it is on an IC pin it could split off both sides as well so like playing a text adventure keep a map marking Vias and Chip/Pin numbers so you explore all parts of the tree as it expands. It is easy to follow a path and forget to go and trace another.

Mark1960 5th Jan 2021 6:26 pm

Re: Non-working Commodore PET 3016
 
I would suggest to start with a close inspection of the pins and soldering around UG5 and the devices listed in Siriusí post #47. You have to start somewhere and a bent IC pin or solder splash on the clock line is most likely to be where the trace is not covered by solder mask.

On a large board like the PET you can sometimes get an idea of how close to the short you are by the difference between 0.1 and 0.2 ohms on the multimeter.

SiriusHardware 5th Jan 2021 7:10 pm

Re: Non-working Commodore PET 3016
 
Mark makes a good argument for inspecting the areas where the CLK1 line will not be protected by solder mask first but Colin may need the guidance to be a bit more specific.

The CLK1 line goes to the following pins on the following devices.

Pin 37 of UC4's socket (6502 CPU).

Pin 2 on each of UE3, UE4, UE5, UE6 (All 74153 ICs).
(On these pin 1, next to pin 2, is connected to 0V so a short between pin 1 / pin 2 on one of these four devices is a possibility).

Pin 1 on each of UF3, UF5, UF6 (All 74LS157 ICs).

On UG5 the CLK1 output comes from pin 7 which is next to the 0V supply pin, pin 8, so this is another area to check for a short.

All of these areas should be examined carefully on the top side of the PCB -and- on the underside of the PCB.

ScottishColin 6th Jan 2021 12:34 pm

Re: Non-working Commodore PET 3016
 
Gents - I had a good first look last night and cannot find any obvious shorts/pins touching/solder splashes but I will give it a better going over today.

Also, I managed to go through the schematics and work out which pin goes to where all on my own (might not sound much to you but it's a minor miracle to me). Please don't stop pointing out the obvious though.

I removed the 6502, but still no difference in the UG5/pin7->8 measurement.

I'll report back later on.

Thanks again all.

Would it be right to say I should measure the resistance from UG5/pin7 to all other noted pins to see what results I get?

SiriusHardware 6th Jan 2021 1:10 pm

Re: Non-working Commodore PET 3016
 
Normally yes, that would verify that the signal path on CLK1 out from UG5 to all the other ICs was complete, but since it is currently shorted to 0V any measurement you make will at least partly confused / invalidated by the unwanted connection to the 0V track.

The priority is to find the short between the CLK1 line and 0V and clear it. You can't make any progress until you manage to find that, and unfortunately we can't physically assist you in any way from here. If it really came down to it and you didn't mind the risk of sending the main PCB back and forth in the post I'd be happy to try to locate that problem for you but I have faith in your patience and determination, and I am sure you can find it yourself, given time.

In spite of what I said earlier it is possible that that zero-ohm resistance down to 0V is actually inside one of the ICs which have pins connected to the CLK1 line, so you may eventually get to the point of having to unsolder and remove each of those ICs one by one to see if that clears the short.

Unsoldering ICs from double sided PCBs can be a delicate job, not without risk of damage to the PCB tracks, so I would try very hard to find a physical cause for the short before moving on to that stage.

ScottishColin 6th Jan 2021 5:03 pm

Re: Non-working Commodore PET 3016
 
Still looking.

Could I do continuity tests to see if the signal pops up anywhere unexpected?

Mark1960 6th Jan 2021 5:57 pm

Re: Non-working Commodore PET 3016
 
Probably best to concentrate on the clock signal short to ground.

You could try measuring resistance from each of the pins that should be connected to clock to ground at each of the ICs and see if one of them is very slightly lower, that could then indicate its closer to the short.

Still worth trying to trace the clock signal across the board. Use the meter to make sure you are still following the right trace when it goes under any ICs.

If you donít find a short itís probably time to start suspecting one of the ICs.

Instead of desoldering each IC in turn to see if the short is cleared, you could cut the IC leg for the clock connections only and lift it clear from the pcb. Checking the short to ground after each cut to minimize the repair needed later.

Any that are not shorted to ground could be linked back to the pcb later.

If all the clock connections are cut and the clock trace is still shorted to ground you could wire directly between the IC pins and bypass the pcb trace.

Personally I would start with the output of the divider.

It sounds drastic to cut the IC legs, but this is probably the easiest way to remove an IC from an old through hole plated board. Cut all the legs to remove the body, then remove each leg with a soldering iron and stainless steel tweezers, then remove solder from the hole with a solder sucker or desolder wick. If using the pump type solder sucker, remove the soldering iron tip from the board before you release the spring on the sucker to avoid damage to the pcb from shock of releasing the spring.

Donít remove the ICs yet, just cut the legs connected to clock, check for the short after each leg is cut.

SiriusHardware 6th Jan 2021 6:38 pm

Re: Non-working Commodore PET 3016
 
Quote:

Originally Posted by Mark1960 (Post 1327758)
just cut the legs connected to clock, check for the short after each leg is cut.

I'm slightly wary of this advice because, with only the wide part of the IC leg available on the top side of the PCB it will take a very fine and surgically sharp pair of cutters to cut straight through the leg without twisting it and potentially doing damage to top side tracks in the process. Plus, unless you are a real solder artist the after-repairs to the pins will look obvious and you don't really want that on a vintage PCB which at the moment is a bit scruffy but otherwise absolutely original.

I would favour complete removal of the device - the trick to doing that is to solder-suck as much of the solder out from the bottom side as possible, and if the hole doesn't clear straight away then actually put more solder on before trying again, because solder suckers work best when there's a good sized blob of solder to suck up.

When you've got all the holes as clear as possible the chances are the pins will still be soldered where the wide parts of the pins rest on the top of the top side pads. What you must never ever do at that point is try to lever the chip vertically up and away from the PCB, if that is done it will come off alright, but probably with 2-3 of the top side pads still attached to it. Instead of trying to lever the chip upwards out of the PCB you have to push it firmly from side to side with a blunt wooden edge - the square end of the wooden handle of a wire brush or something similar is ideal. Push it gently back and forwards from left to right, left to right, and eventually the top side solder bonds will crack and allow the chip to be lifted out.

That said, I would like to suggest a possible compromise between my way and Mark's and that is to desolder only the pin which is of interest, move the pin around gently until the top side solder cracks and then apply pressure to the end of the pin from below to bend it upwards just enough to lift the wide part of the pin off its pad. Check that the pin is clear of all solder surfaces and going straight through middle of the the hole without touching the sides, do your check and if the chip is not guilty, reposition and resolder the pin.

Timbucus 6th Jan 2021 9:48 pm

Re: Non-working Commodore PET 3016
 
Just another hint on the de-solder - add a little bit of flux and new solder until it flows before trying to suck it out - it is counter intuitive but, it works.

SiriusHardware 6th Jan 2021 10:00 pm

Re: Non-working Commodore PET 3016
 
There's another method of narrowing down the area of a short - track cutting. I'd better say immediately that I don't favour using this method on a vintage PCB in a potentially quite valuable antique machine, but I will outline the method anyway.

Let's say you have followed the CLK1 line all the way from UG5 to every place that it goes and you still haven't been able to identify the short and you still don't know where it is. What you can then do is cut the CLK1 track up into sections by making very fine cuts across the tracks.

The broad principle is that you first do this at about the half way point of the CLK1 track's voyage across the PCB, making the original whole track 'A' into two electrically separate tracks which we will call 'B' and 'C'. Then, you measure from 'B' to 0V and from 'C' to 0V and you should find that now only one of the two is still shorted to ground.

For the purposes of this 'walkthrough', let's say you find that 'B' is no longer shorted to 0V but 'C' still is. You've now ruled out about half of the original whole CLK1 track and any devices which are connected to that half.

The next thing is to look at track 'C', the one which still has a short on it. Find the mid-point of track 'C' and make another cut to make two new track sections which we will call 'D' and 'E'. As before, check from 'D' to 0V and 'E' to 0V and find out which one is no longer shorted to 0V and which one still is.

At this point, still with only two track cuts made, you have narrowed the problem area down to just one quarter of the original track length and maybe just one or two devices which are connected to that section. If you have to, you can make further track cuts to narrow the problem area down to 1/8th, 1/16th and so on.

I've already said that I really don't like to do this to antique PCBs but if it has to be done then the best approach is to look for places where the track in question passes from one side of the PCB to the other through a 'VIA' and cut that through-hole connection by drilling through it with a mini drill bit only a tiny bit bigger than the via hole. This will remove the internal copper coating which connects the pad on the top side of the VIA to the pad on the bottom side. To restore the connection after you complete your fault finding and tracing, you pass a snug fitting bit of tinned copper wire through the via hole, solder it to the pads on the upper and lower sides of the PCB and clip off the excess wire above and below - this is a lot neater and less obtrusive than hacking and repairing a track on the PCB surface.

By the way, this can only be done on double sided PCBs, some modern 'multilayer' PCBs have even more track layers sandwiched in the middle of the PCB layers and if you drill out a VIA on one of those the connection from the VIA to the intermediate layers will be lost forever.

Refugee 6th Jan 2021 10:38 pm

Re: Non-working Commodore PET 3016
 
I would not de-solder anything to start with.
It is a low resistance short.
Without power on the board
Chop an old USB-1 cable and connect a 22 ohm resistor to the +5V wire and connect the negative to ground on the board. Plug the other end into a USB port/charger.
Connect the end of the resistor to the clock pin and measure the voltage to establish that it is not a dead short.
If you get 2 or 3 volts it should warm up the offending chip after a few minutes.


All times are GMT. The time now is 5:30 pm.

Powered by vBulletin®
Copyright ©2000 - 2021, vBulletin Solutions, Inc.
Copyright ©2002 - 2021, Paul Stenning.