Commodore PET 4016

[SiriusHardware]

That's quite an epic thread Colin linked to in #39 - I was even briefly involved in it at one stage.

One fundamental problem is that VPP (pin 21) on a 2716 is actually a power pin, not a chip select pin as it is on the original ROM - so in these machines the CS3 pin of the character ROM is tied to the mysterious INIT line along with many other clear / enable lines on various ICs and the INIT line is tied to +5V via a medium value resistor - that's not a 'hard' enough power connection for a 2716 which requires direct +5V on pin 21 when in read mode, so I definitely wouldn't expect it to work as a drop-in ROM replacement in that particular position.

You can get away with using a 2716 as a replacement for the EDIT ROM because in that position on the PCB, pin 21 of the socket happens to be tied to hard +5V.

There was an oft repeated suggestion in that thread which I would have thought would work, and that is to bend out pin 21 clear of the socket and tie it to hard +5V - however it seems that some people did try it to no good effect although one or two said it did work.

I think that as well as the EPROM needing to be provided with a 'proper' +5V supply to pin 21, it also needs to be at the faster end of the speed range for a 2716 - if both of these conditions can be satisfied then maybe it will work.

I would try it as a lash-up with a 'fast' 2716 and if it works, do a discreet mod on the underside of the PCB to disconnect pin 21 of the socket from the INIT line and connect it to +5V. Even if you then put a real ROM back in the socket, it should still work because (as per the EDIT ROM) commodore did sometimes wire pin 21 of these ROMs to hard +5V.

[SiriusHardware]

Regarding the lower - case conundrum, there is a specific address line (A10) on the character ROM which toggles either between upper case characters and graphics characters, or between upper case characters and lower case characters, depending on the machine version, 'Graphics' or 'Business'.

The line going to the character ROM A10 (marked 'Graph' on the diagram) is also connected to the 6522 pin 39 (CA2) and to a blobbed 'bubble link' on sheet 10 (CRTC) left hand side.

The link can be left open so that only the 6522 pin 39 and the character ROM A10 are connected to the 'Graph' line, in which case surely the CA2 pin is used in output mode to select lower case /upper case / graphics.

Or, the link can be closed in which case the 'Graph' line and the ICs connected to it are pulled high by R6 and the state of the 'Graph' line is made available to the CRTC controller's LPSTB pin, inverted via UD2 pins 3,4. With the bubble link made the pulled up state of the 'Graph' line can also be overridden by a forced 'low' on the 'PEN STROBE' input, which we will assume for the moment is not important. I would assume the CA2 output on the 6522 is also strong enough to drive the line low, overriding the R6 pullup, if it wants to.

To begin with, power off and meter out the connection between the 6522 pin 39 and the character ROM A10 pin. If that connection is intact follow it further to the 'bubble link' and tell us whether that link is made or open, and does it look as though the state of the link has ever been disturbed since the PCB was originally built?

[SiriusHardware]

On your other similar machine, is that link made, or open?

[ScottishColin]

They're both Graphics keyboards - I took a copy of the original working ROM and burnt that to a 2716. Which seemingly works in the 4032 (I've not exhastively tested it) but not in the 4016. I may try another 2716 to see what difference if any it makes.

I'm certain that the 4016 is full of RAM errors but the screen issues are making it impossible for me to read them.

I'm still of a mind to fix the 4016 before I start with the high resolution graphics board right now.

In slightly better news, I have put the 6522 back in and the boot screen (although corrupted) is now upper case.

Yet still strangely, only the Tynemouth boards (both diagnostics and their ROM/RAM emulator) work. None of the Romulator, a 6502 nor nor the Sven Peterson Diagnostic clip (link below) produce anything on the screen. I'll start to look at the 6502 and CRTC sockets I think next.

https://github.com/svenpetersen1965/PET-Diagnostic-Clip

I've checked voltages in the usual places (RAM chips, 6502 pin 8, ROM chips etc) and then they all seem fine. If anyone has any help regarding checking voltages elsewhere (voltage regulators for example), I'd like to look there.

Colin.

[ScottishColin]

That link is made. But I think it's a variant of the motherboard as there are several factory fitted changes made to it that don't exist on the 4016.

Colin.

Originally Posted by SiriusHardware

On your other similar machine, is that link made, or open?

[SiriusHardware]

Right, I had missed the fact that you didn't have the 6522 in at the moment - obviously now that it is fitted, it is providing a 'show upper case' signal from the 6522 to the character generator ROM.

[ortek_service]

I'd also not realised that the 6522 switching between the two Char Gen Modes via its A10 was removed.

Although, confusingly the Tynemouth Diagnostics had been showing both Upper and Lower case, so not sure if 6522 was fitted at that point?
Or whether there is another way of controlling that.
I also wonder how this Diag board copes with 'Graphics' PET's, without lower-case? Maybe there is a (DIL Switch) option to just use Upper case?

I also wonder what the CharGen's A8 & A9 do? - Unless it actually outputs 4 bytes of pixels per character, so why it needs to be quite fast.

As it is looking like several of the Main (D)RAM Memory IC's have faults, it may be worth considering socketing these - particularly bits 3,2 & 1, where faults appeared to be. And testing these by substitution with a working PET that has some of these socketed. That may help with things working a bit better, if there is RAM corruption in program workspace etc.
I presume the 4016 only has 8x 4116's fitted, but might be upgradeable to 32KB by fitting sockets to unused bank, if not already done . may be able to re-patch to that bank, before resolving faults with originals.

[ortek_service]

Regarding 2716's and Vpp (supply) input current, although this is a 'power supply' input it isn't the main supply and current from it is mainly drawn when it is at the programming-voltage & programming-pulses are applied.
When it is at 5V, in read-mode, the supply current is much-less. This one (Not actually the 27C16 but a standard NMOS 2716) says 5mA max: https://www.futurlec.com/Memory/27C16-150.shtml

I thought a CMOS 27C16 should be much-less - like the supply current. But I haven't found a 27C16 datasheet yet that gives this current at Vpp=5V.
CMOS EPROM's were also often much-faster, but unfortunately with the 27C16 it seems many of these are the slowest 450ns like the original.
Using a 28(C?)16 EEPROM may also work / may be faster, if your programmer supports these.

If not want to cut tracks on the PET board (albeit on underside), then could always put a wire-link under the EPROM between Vpp & Vcc and bend Vpp pin up so it doesn't go into the socket - Probably neater than stacking IC sockets, to isolate this.

[SiriusHardware]

INIT isn't only connected to the Char Gen pin 21, but to the clear / reset inputs of a shedload of other ICs as well. If anything places a significant extra load on this line the single pullup resistor may no longer be 'strong' enough to keep it properly pulled high.

Regarding trying to use a 2716 in that positition - definitely try a direct connection from pin 21 to +5V (as a lash-up initially) and also definitely a faster 2716 than the one which has already been tried (faster = a lower nS value after the main IC part number).

[ScottishColin]

OK - maas removal and replacement of 244 chips and the video RAM 2114s with known good chips has edged me forwards.

But still not there.

It looks like all the onboard RAM is knackered (by the way it's a 4016 which has been upgraded to 32k so is fully populated).

A boot of the Tynemouth diagnostics chip is much clearer but not there yet.

A boot with the Tynemouth ROM/RAM board gives me a better screen but it still has issues - see both .ZIP files. I'm guessing that there's still something wrong with the video section of the motherboard but having replaced the following I'm not sure where we gop next:

244s:
UE10
UE9
UE8
UB5
UB4
UB9
UB10

2114s
UC4
UC5

Also I have noted that C34 and C35 have been replaced at some point in the past so I may remove one leg of them and check them.

All ROMs are out of the motherboard right now.

Both 6520s are from a known working PET.

Colin.

[ortek_service]

Well that was a bit unexpected, changing all the (display memory?) LS244's as well as the 2114's. As the symptoms hadn't really indicated to me that these were at fault (eg OK Checkerboard, only so,e bits were at fault at specific locations - in the Main-Memory at least).
Did you test any of the removed IC's (eg in another PET), to confirm which / if any had failed?

I hadn't seen a Diag board screenshot, sincw the 6522 controlling the CharGen Mode had been re-fitted. So I wondered how-much that had improved the display?

In some ways it seems like the Diag Test is now actually worse!
As 'Tynemouth' etc is no-longer displaying correctly.
And it seems all the memory areas are now displaying all bits in error with the same '4FF0&Errors' after each 4K block. I can't recall exactly what this means - whether this as a stored and read(incorrectly) Bytes?
- As number too large for offset where error occurs. Although I note 13FC is shown for 1KB 8000-83FFh area.

So I'm wondering if something else has now failed / been accidentally damaged during the removal&replacement of all these IC's?

I did think it was worth just going for the 3 bits (3,2 & 1) of the DRAM's, first, being as they seemed to be main errors. And should only need to go for lower 16KB bank, providing extra 16KB bank haven'y gone s/c on any pins so dragging-down busses.
If the extra 16KB was added as an upgrade post-manufacture as a 4016, are these socketed so could just be removed for now? (Or maybe tested in another PET and maybe banks swapped over if fairly-easy to do on this board?).

I would expect (Electrolytic? previously-changed?) capacitors to be causing these issues (Or issues with the regulators - But no harm checking the rails with your 'scope, to look for excess ripple etc causing them to dip too-low (<4.5V, Ideally not dipping below 4.75V).

When you say all ROM's are out of the mobo, I presume that the CharGen one is still fitted? (And taken from the 4032, as should be the right one if both of these are the same variant?). Although presumbably the ROM's from the 4016(+16KB=4032?) could be checked in the 4032 if they are all socketed in that?

[ortek_service]

Originally Posted by SiriusHardware

INIT isn't only connected to the Char Gen pin 21, but to the clear / reset inputs of a shedload of other ICs as well. If anything places a significant extra load on this line the single pullup resistor may no longer be 'strong' enough to keep it properly pulled high.

Regarding trying to use a 2716 in that positition - definitely try a direct connection from pin 21 to +5V (as a lash-up initially) and also definitely a faster 2716 than the one which has already been tried (faster = a lower nS value after the main IC part number).

I did have a look at some 2816 EEPROM datasheets, and see that these were available as fast as 200ns. And the 28C16 down to 150ns.

Plus pin 21 on these is nWR rather than a Vpp pin (These 28(C)16's only require a (12V) 'Vpp' voltage (On a shared pin like nOE?) to erase them. So will program a blank-erased one with a 5V single supply. Hopefully there is sufficient protection from accidentally writing to them, when in-circuit? - particularly when the processor is in brown-out so it could execute invalid program instructions).
So hopefully nWR shouldn't draw much current / won't get taken low and possibly cause corruption of contents? which could make this more of a drop-in replacement for the CharGen.

And I believe most modern v.cheap IC-programmers support these as a minimum event if not higher-voltage EPROM's?

Although unfortunately it seems the 28(C)16 (Like the 27(C)16 and most EPROM's apart from larger + plastic-packaged non-windowed OTP ones) is no longer made.
However there are many places (inc. Amazon MP etc) selling old (stock?) ones (that may have been used, but can be erased).

So might be interesting to try one (I have a few, and recently picked-up some more at a local Radio-club B&B night, so might have to try one when I dig out a dead PET (Although original 2000-series with non JEDEC-std pinout RAM's - possibly ROM's?) I've got and been meaning to look at fixing one day.

[ortek_service]

Originally Posted by ortek_service

>>
>>
I would expect (Electrolytic? previously-changed?) capacitors to be causing these issues (Or issues with the regulators - But no harm checking the rails with your 'scope, to look for excess ripple etc causing them to dip too-low (<4.5V, Ideally not dipping below 4.75V).
>>
>>

Sorry, too late for edit-window but I'd meant to type "I wouldn't expect (...) capacitors to be causing these issues"!

[Mark1960]

Electronwilly was offering TM2516s at 300ns back in December. It might be worth asking if he has any left if you think these might be fast enough.

[SiriusHardware]

All of our suggestions about suitable EPROM substitutes for the char gen ROM will hopefully prove useful in time but for the time being Colin is using a real char gen ROM while he attempts to fix the other problems.

Colin, I'm glad you did a blanket replacement of all of the 244s as we would have probably had to replace them one after the other anyway given their famously fragile nature - maybe in due course you can put them in a working machine one at a time and work out which ones really were faulty.

I will try to take a deeper look at this when I have time, which unfortunately I don't have a lot of at the moment - in the meantime I'm watching with interest, and you seem to have made a fair bit of progress in spite of our help.

[ScottishColin]

My 2716s don't seem to have speeds marked on them.

I do have one ST M2716A-AFI which is marked as "FAST" - if that helps. I may try that one.

As Sirius mentioned, I currently have an original Commodore Character ROM in place - that is the only ROM on the motherboard right now.

I will test the parts I have removed once I have it working.

Colin.

Originally Posted by Mark1960

Electronwilly was offering TM2516s at 300ns back in December. It might be worth asking if he has any left if you think these might be fast enough.

[ScottishColin]

Well we did replace all the 244s on the now-working 4032 so I thought I'd just get on with it now.

There were so few sockets on this motherboard and the new RAM that was installed was soldered direct to the motherboard sadly.

Having said that I've still replaced fewer chips (so far) than on my first PET...

Colin.

Originally Posted by SiriusHardware

All of our suggestions about suitable EPROM substitutes for the char gen ROM will hopefully prove useful in time but for the time being Colin is using a real char gen ROM while he attempts to fix the other problems.

Colin, I'm glad you did a blanket replacement of all of the 244s as we would have probably had to replace them one after the other anyway given their famously fragile nature - maybe in due course you can put them in a working machine one at a time and work out which ones really were faulty.

I will try to take a deeper look at this when I have time, which unfortunately I don't have a lot of at the moment - in the meantime I'm watching with interest, and you seem to have made a fair bit of progress in spite of our help.

[ScottishColin]

I've just measured some voltages that I'm not clear about - they're on P11 of the schematics here:

https://www.zimmers.net/anonftp/pub/cbm/schematics/computers/pet/univ/8032081-11.gif

They're all marked as unregulated voltages and I'm not sure I've measured them before and they're a bit high so I was wondering if someone could take a look and educate me about unregulated voltages.

I'm guessing what actually matters is the 'regulated' voltages but as these are a little high I was just wondering.

J10 pins 1&2 - marked on the schematics as -9v unreg - actually -12.10v
J10 pins 4&5 - marked on the schematics as +16v unreg - actually +20.83v

J11 pins 1,4&6 - marked on the schematics as +9v unreg - actually +11.01v

Thanks.

Colin.

[SiriusHardware]

Well, as you have at least one similar working machine the way to find out how 'normal' those voltages are is to measure the same points in that one. One reason they might be high is if the machine is configured to run from 220V rather than 240, if it even has that option.

In the main though the regulated voltages are the ones which really matter.

[ScottishColin]

Baby steps forwards. with a 6502 fitted in UB14 and the normal ROM fitted in UD6 and Dave's PET Tester fitted in UD7 it now boots.

It boots to the first screen and stays there with the kind addition of some speckled moving artefacts across the screen.

My recollection is that this means that it can write to video ROM but is erroring on reading it back. The 2114s in here are known good from the 4032 so I don't think it can be them.

Colin.