Hi all,
I run vintage computer exhibitions out of a UK university - we do big ones that are open to the public and also smaller, pop-up ones during our open days and in outreach events at local schools. Most of the machines I showcase are from my own personal collection, so as such are heavily biased towards UK legends of the 80s - Sinclair Spectrums, BBC Micros etc. But there's also plenty from your side of the pond too, Commodores, Ataris etc. I recently acquired my first Apple for the collection, a UK PAL IIe in absolutely glorious condition.
On receipt, the machine ran fine for the most part. The colours on the display were wrong, but I'm fairly sure this is a dying TCA650 PAL encoder chip, and I've got a replacement for that on order. That wasn't a show-stopper.
But now the bigger problem. I was installing an SDisk2 and all was going fine - I had got my first successful boot of DOS 3.3 from the device and was in the process of trying to get Robotron to load.
Suddenly, out of nowhere the keyboard (which had been working perfectly up until that point) went haywire. Some keys started producing completely wrong responses. Some keys stop producing a visible character at all (although I think some kind of signal is being sent because the cursor "reacts"). And some keys work fine.
If I do a complete run along the numeric keys I get
123..13890
Next row (alpha keys only) I get:
Q89.TYUIOP
Next row (again, alpha keys only):
TIUFGOJKG
And bottom row alpha keys:
JKCVBNC
(where a . indicates the key doesn't seem to produce anything printable, like I said, the cursor "reacts" so some kind of keypress is actually being registered, I think).
I have tried simple, obvious things like reseating the keyboard ROM and the keyboard encoder chip - no joy.
I am considering buying a new keyboard encoder chip - probably the one from JCM - but before I do, does anyone have any other suggestions?
I am pretty much hopeless with a soldering iron, incidentally - possibly capacitor replacements are within my ability but not much more. Certainly not replacing ICs (although pretty much everything looks socketed in this beauty, thankfully).
Pics below so you can enjoy my new friend. Sinclair Spectrum for scale. Motherboard pic might also be useful :)
I don’t think the problem is the keyboard encoder. When it is the encoder, all the bad keys are usually either on the same row or on the same column of the matrix. Yours are all over the place, which leads me to believe that it’s probably the keyboard EPROM in position F12 or its connections to the rest of the chips.
Interesting. Does anyone know where one can source a replacement EPROM? Had a quick Google to no avail.
You would just need a blank EPROM (27c16 on ebay) and a programmer device. And this file:
US/UK keyboard ROM
Maybe you can find someone near you - or here in the forum, who lives in the UK - who has a programmer. But if you're working on retro machines, you may already have one. Otherwise, maybe be worth considering to get one. They are also useful for testing and downloading ROM devices.
Following up myself here, but just looked at the bottom of the keyboard, and it looks like someone has done a repair job at some point. Or is this red wire normal? It looks like someone's fixed a track
Could this all be due to a botched repair that has failed?
PXL_20240726_134850820.jpg
Certainly someone worked on the board. Of course worth to double-check with a multimeter, if the traces and botch wires are still conducting.
However, if a connection on the keyboard fails, as CVT already pointed out, normally all keys in an entire row or column of the keyboard matrix fail. In your case, some keys did work - but they triggered other characters. That's not the usual symptom if a keyboard matrix connection fails.
Just stole a multimeter from our engineers and as far as I can see continuity is pretty much as it should be, between the pads connected with the red wires and in the area where the tracks look a bit scratched. The scratches are much worse-looking on the photo than in reality; the camera seems to have really picked them up.
I think I'm going to take your advice and treat myself to an EPROM programmer...
If you can get a hold of a second 2516 or 27C16 EPROM chip which is not empty, but mostly full, there is a simple test you can perform that will determine with high certainty if the problem is in the keyboard EPROM or somewhere else. This is based on the fact that with the current EPROM the following pair of keys produce the same character on the screen:
1 and 6
3 and 7
W and 8
E and 9
A and T
S and I
D and U
G and L
Z and J
X and K
M and C
Now if you replace it with this other EPROM, the characters you see on the screen will no longer correspond to the keys you press. However if every single pair of keys from above still produce the same character on the screen (even though it will be a different character than before), you can be virtually certain that the problem is not in your keyboard EPROM.
This approach works best with an EPROM chip which contains firmware, since the data resembles random data much more closely than if it contained characters.
What's wrong with 2716?
Who says there's something wrong with it?
So I invested in a EPROM writer and a 27C16 chip.
Burned the image and put the chip into the IIe.
No joy. Precisely the same symptom. Same wrong letters for the same keys.
Any ideas? Should I turn my attention to the keyboard encoder? Or should I be looking somewhere else?
PS: Incidentally, out of curiousity, is there any reason why one couldn't use a 28C16 EEPROM for this sort of thing?
An EEPROM (28c16) should also work. It has an additional write-enable pin, but that's pulled to +5V on the IIe mainboard, so it should work.
The data lines of the keyboard ROM directly connect to the 6502's data bus. There is nothing in between. Not even a bus transceiver, so nothing which could alter the character data.
A possibility is a simple connection issue (broken trace) between the 6502 and the ROM. You could check with a multimeter if each data line makes proper contact (MD0-MD7). However, the key issues you mention in your first post do not look like there was a broken data bus trace (this would result in some pattern, e.g. mismatching odd/even characters when MD0 was broken etc).
This mainly leaves two possible issues:
1. A broken trace between UF13 (the keyboard encoder) and the UF12 (the keyboard ROM).
2. A broken keyboard encoder itself.
So, you could check with a multimeter, if A0-A8 of the ROM made proper contact to the keyboard encoder (see schematics). If that's not it, then it's probably a broken encoder after all.
AIIe_Euro_KeyEncoder.jpg
If we expand the XY matrix position of each pair of keys that produce the same character on the screen and convert it to binary, we can clearly see that they all differ by exactly same bit, which must be the bad bit. The last pair is an exception, but I believe it's due to an error. I am just not sure if B1 is the highest or lowest bit.
Keys.png
So just concluding the mystery - after all that it was the keyboard encoder.
A replacement AY-5-3600-PRO and we're good to go.
For posterity, and if anyone in Europe is looking for a source of these, https://www.ebay.co.uk/itm/195380501147 (based in France) is a legitimate seller - or at least the one I bought works just fine. Turnaround to the UK was about a week.