Non-working Commodore PET 3016

[ScottishColin]

OK - problem number one (of perhaps many).

I have taken the keyboard apart again to get to the keyboard PCB and test continuity and I have a broken trace. If you look at the photo, you can see the right hand trace has gone. It managed to weld itself to the plastic keyboard chassis with I expect some prolonged damp. I can get continuity to the bare trace just before it turns, but I cannot get any continuity to the keys it is supposed to go to. All other wires check out to their keys.

Are there good ways to deal with these kind of issues?

Colin.

[SiriusHardware]

There are two types of 'repair paint' available and to be honest I would never use the metallic type (originally meant for track repairs) to revive the conductive 'pills' on keyboards. It's too hard, and when repeatedly crushed will just fall apart into metallic powder which will end up all over the place.

There is another type of conductive paint which is black, and dries with a rubbery and flexible finish and is usually marketed as 'remote control repair' solution - it is very similar to the rubber solution used for fixing bicycle inner tubes but obviously has a lot of carbon mixed into it. Because it is flexible, unlike the metal stuff, it is more squash-tolerant.

The rubber bits on the ends of the plungers are probably removable, they look very similar to the key contacts in my Atari ST and I have seen them used in PC keyboards of a certain vintage as well - so it might be possible to harvest a set of good 'dimples' from an old PC PS/2 keyboard of a certain age.

[Mark1960]

If some keys are not working and the ones that work are giving the wrong characters then its most likely to be a problem with the 74LS145, or one of the inputs to it.

Maybe you could make a list of which ones don’t work and which characters are wrong.

I don’t think you should paint anything, thats likely to be temporary at best.

[julie_m]

If you can carefully scrape away a couple of patches of the green solder resist over the track either side of the break to reveal bare copper, you should be able to solder a wire link between them and complete the circuit.

[SiriusHardware]

Are there good ways to deal with these kind of issues?

All you can really do there is to carefully scrape the green stuff off the track for some distance (half an inch?) on either side of the break, lightly tin the exposed sections of track and then lay down a fine piece of wire which you've already cut and shaped to follow the route of the missing bit of track. Solder the wire into place on both sides of the break so it replaces the missing piece of track.

By the look of it this needs to be very low profile (as thin and flat as possible) so use the thinnest piece of wire you can find - a single strand from a stripped end of 7-strand would be about right.

Make sure the replacement 'track' follows the same loop shape as the original obviously did, as it is clearly diverting around something like a support post which fits directly against the PCB in that area.

Crossed with Julie.

[Mark1960]

Originally Posted by ScottishColin

Are there good ways to deal with these kind of issues?

I think for that trace you can scrape about 3mm of track clear of solder resist at each end of the break and use a single strand from a multistrand cable to link the two ends. Keep the solder to a minimum and it should still allow the keyboard to be reassembled.

Edit: Is three a record for cross posting?

[Slothie]

I came to answer this but I see 3 of you have it covered!

[SiriusHardware]

its most likely to be a problem with the 74LS145, or one of the inputs to it.

Better let him fix that obvious physical fault first, however.

We had a PET thread about 2-3 years ago where the OP started off with a basically working machine with keyboard problems, only his fault was that the keys in row n and the keys on row n+1 generated the same characters as each other and the keys in row N+2 and N+3 generated the same characters as each other, so clearly the 4-bit binary input to the 74LS145 was advancing in steps of 2, or the 74LS145 thought it was. That did turn out to be the 74LS145.

Here we have what sounds like the output from the decoder being offset by +1 from the value being input to it - could still be a fault on the decoder I suppose, as I don't know what the internal architecture looks like.

[ScottishColin]

See attached PDF.

Colin.

Originally Posted by Mark1960

If some keys are not working and the ones that work are giving the wrong characters then its most likely to be a problem with the 74LS145, or one of the inputs to it.

Maybe you could make a list of which ones don’t work and which characters are wrong.

I don’t think you should paint anything, thats likely to be temporary at best.

[SiriusHardware]

That's almost a carbon copy of the fault I was remembering in #1528.

You press Q, get W. Press 'W', get W. Both Q and W are being interpreted as 'W'. Same pattern repeated as you go from left to right across the keyboard.

I'm going to say this is the 74LS145.

[Lloyd 1985]

Another fix for the broken track, that I have used myself (anyone remember the Sony ICF2001D that batteries ate?!), if you have any of the conductive silver paint You could use that to bridge the damaged track, as above, clean the green solder resist off over the break, then just paint it with a little of the conductive paint. If the break is really small you might get away with just bridging it with solder, which might last long enough for a test, but might crack with heavy use. I’ve also used the strands from solder wick / braid to bridge tracks too.

Regards
Lloyd

[ortek_service]

Originally Posted by SiriusHardware

There are two types of 'repair paint' available and to be honest I would never use the metallic type (originally meant for track repairs) to revive the conductive 'pills' on keyboards. It's too hard, and when repeatedly crushed will just fall apart into metallic powder which will end up all over the place.

There is another type of conductive paint which is black, and dries with a rubbery and flexible finish and is usually marketed as 'remote control repair' solution - it is very similar to the rubber solution used for fixing bicycle inner tubes but obviously has a lot of carbon mixed into it. Because it is flexible, unlike the metal stuff, it is more squash-tolerant.

The rubber bits on the ends of the plungers are probably removable, they look very similar to the key contacts in my Atari ST and I have seen them used in PC keyboards of a certain vintage as well - so it might be possible to harvest a set of good 'dimples' from an old PC PS/2 keyboard of a certain age.

Well in that video you'd linked to, he does use an expensive Chemtronics keyboard repair kit, but it was silver (in colour certainly) paint (although it seems you have to add a liquid solvent to it and mix it up first). But, yes, it does ideally need to be quite flexible.

I had also heard of remote repair kits containing small discs of self-adhesive metal foil.

But, it does seem he'd got it wrong that the keys should be very low resistance when pressed, and just cleaning them had brought the resistance back down to low-enough to make them work.

[SiriusHardware]

If the break is really small you might get away with just bridging it with solder

Experience tells me that even if I repair a hairline / invisible crack with just a blob of solder the solder, being soft, will eventually crack in the same place especially if it has to withstand the might of an enthusiastic typist, as in this case. A track break, no matter how narrow, should always be patched with wire (which can bend if the board bends) and not just with solder.

it does seem he'd got it wrong that the keys should be very low resistance when pressed

Didn't he try it, though, after cleaning all the pills and the PCB fingers? Maybe the circuit does expect lower than the >1K he was seeing.

[ortek_service]

On track repairs, it is much better to use thin wire - like that on wire-wrapping spools, with enamelled copper (often used for 'wire mods' bodges), or very thin solid core wire with insulation partially or fully-removed (But can also salvage a strand of normal stranded wire, if you haven't got any to fine solid-core hand).
- rather than try to bridge gaps with solder or conductive paint that can crack away.

And I often lacquer / clear adhesive over them, if there's several closely spaced ones, to prevent them shorting etc.
Or to prevent them getting damaged again.

I did once glue a snapped-in-half VCR panel back together, wire linking all the fine tracks, and it even worked OK including for many years afterwards.

[julie_m]

That looks consistent with a fault on UC9; a 74LS145 BCD decoder. You could try comparing a 'scope trace on pin 12 against pins 13-15 (the inputs, which take a binary value from 0000 to 1001; pin 12 is the units, and this might be the interesting one), looking out for anything that isn't a clean high or low logic level. Also try probing pins 1-7 and 9-11 (the outputs; one of which will be low depending on the input value, the others should be high). If one of them is never going low, or more than one at a time is low, that would be highly suspicious.

[ortek_service]

Originally Posted by SiriusHardware

Didn't he try it, though, after cleaning all the pills and the PCB fingers? Maybe the circuit does expect lower than the >1K he was seeing.

No he didn't on that video (and also made the mistake of testing it on the (< 200R) continuity beep range to start with).
From finding the circuit had 10k pull-ups to +5V, and that the 6520 has 0.8V max. threshold for guaranteed logic-low, I'd worked out that upto 2k resistance should be OK.

[SiriusHardware]

Hmm, good spot. I really liked his little animation of the operation of the key scan circuit though - explained it much more quickly than we could have done by trying to explain it verbally. I think Colin's current problem, after he sorts that eaten track, is very likely to be the 74LS145.

[SiriusHardware]

Expanding on Julie's thoughts in #1535 I suggest this, only when that track fault has been fixed.

----

Scope UC7 pins 2, 3, 4, 5

Scope UC9 pins 15, 14, 13, 12 - should have the same waveforms on as the above.

Then scope the following pairs of pins together, always with the first pin of each pair on the upper trace.

UC9 pins
1 and 2
2 and 3
3 and 4
4 and 5
5 and 6
6 and 7
7 and 8
8 and 9
9 and 10

[ortek_service]

After trying to find an actual PET keyboard schematic via google, I eventually found this: http://www.6502.org/users/andre/peti...ds.html#layout

And looking at Colin's '1 response for both keys in a pair' pdf, it can be seen that the Scan Rows 0 to 9 are only incrementing 2 at a time.


So it does rather look like an issue with bit0 of the BCD code on the 74LS145 is not working.
And there's still a chance of a bad connection to the 6520 PIA (or even that itself to be a 'bit' dodgy!)

'scoping the KEYA line at UC9(74LS145) Pin 15 ('A' input) / the UC7(6520) Pin 2 (PA0), should hopefully see if that is switching OK at both ends and therefore working OK at least - before suspecting/condemning the 74LS145
- Which unfortunately has open-collector outputs, and the pull-up resistors are on the other side of the keyboard matrix,
So you would need to attach a 100k etc. pull-up to +5V onto 'scope probe connection, to check each of the 74LS145's outputs.

But might as well fix the broken track first, if no doubt that's on a fairly important scan line.

[SiriusHardware]

Which unfortunately is open-drain

I missed that little detail

Last time we ran into this problem the OP didn't have a scope so we never got as far as trying to analyse the problem, we just went for a chip change which fortunately did solve it.