17th Mar 2021, 10:23 pm | #1121 |
Octode
Join Date: Mar 2020
Location: Kitchener, Ontario, Canada
Posts: 1,294
|
Re: Non-working Commodore PET 3016
Colin, try changing trigger to channel 1, set trigger on rising edge. That might capture rare positive pulses that you don’t see when you trigger on channel 2.
|
17th Mar 2021, 10:37 pm | #1122 |
Octode
Join Date: May 2012
Location: Perth, Scotland
Posts: 1,803
|
Re: Non-working Commodore PET 3016
Nothing that I can see?
Colin. |
17th Mar 2021, 10:39 pm | #1123 | ||
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Quote:
Quote:
Alan |
||
17th Mar 2021, 10:58 pm | #1124 | |||
Octode
Join Date: May 2012
Location: Perth, Scotland
Posts: 1,803
|
Re: Non-working Commodore PET 3016
So if I got this straight, I need to buy 16 to be able to go to 32k? My de-soldering seems to be getting better - I've had plenty of practice now...
It's worth a thought I guess. I think I've provided lots of (potentially confusing) information tonight, so I guess let's see what the collective brains come up with. If it helps Sirius, I have 2 x spare new chips for UI10 and UI11 if you think they're suspect. Colin. Quote:
|
|||
17th Mar 2021, 11:06 pm | #1125 | |
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Quote:
Alan |
|
17th Mar 2021, 11:07 pm | #1126 | ||
Octode
Join Date: May 2012
Location: Perth, Scotland
Posts: 1,803
|
Re: Non-working Commodore PET 3016
And I'd have to remove all existing 16 - that right?
Colin. Quote:
|
||
17th Mar 2021, 11:11 pm | #1127 |
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Safest option I think because if there is a faulty 4108 in row 'J' you could still end up with the mixed character screen and you'd be none the wiser.
Alan |
17th Mar 2021, 11:17 pm | #1128 |
Dekatron
Join Date: Aug 2011
Location: Newcastle, Tyne and Wear, UK.
Posts: 11,553
|
Re: Non-working Commodore PET 3016
As I understand it, you can replace the 16 x 4108 (16K) with 8 x 4116 (16K), but you have to move the links around for that new configuration.
I also think that if we were able to identify one or two dud 4108s we could replace them directly with 4116s and leave the links the way they are. In this case, the system would only use one half of each of the substitute 4116s, believing that they were 4108s like all the rest. At the moment I don't feel we have sufficient proof of faulty RAM although I agree that all RAMs from that period have a deserved reputation for bad reliability. There are other areas of note - one is the appearance of the BD0-BD7 lines, a couple of which look odd, another is the apparent absence of any attempt to read from the system RAM even during initialisation. Another is the apparent absence of activity on the RAS and CAS lines. We'll have a think about how to go forward from here. |
17th Mar 2021, 11:37 pm | #1129 | |
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Quote:
Alan Last edited by ajgriff; 18th Mar 2021 at 12:07 am. Reason: Grammar |
|
17th Mar 2021, 11:59 pm | #1130 |
Octode
Join Date: Mar 2020
Location: Kitchener, Ontario, Canada
Posts: 1,294
|
Re: Non-working Commodore PET 3016
I just want to make some suggestions on setting up the scope to detect pulses that are not highly repetitive.
For TTL logic level signals like the pet, keep the trigger voltage at about 1.5v. Select the channel with the least active signal for trigger input. If the least active signal is mostly low, set the trigger to rising edge. If its mostly high, set the trigger to falling edge. Try setting the trigger mode to trigger instead of auto. If there really is no activity this will give blank traces as nothing will generate a trigger, but you can select auto so you see the active trace. |
18th Mar 2021, 12:05 am | #1131 | |
Dekatron
Join Date: Aug 2011
Location: Newcastle, Tyne and Wear, UK.
Posts: 11,553
|
Re: Non-working Commodore PET 3016
Quote:
The only other sure way to do it is to remove them all intact and use something else, like an Arduino, to independently test them. Or remove them all intact, fit sockets and if possible fit combinations of them to try to find a combination of 8 out of 16 of them which will work as 8K, if that is a valid combination in these machines. Then cycle through the remaining ones to find out which ones stop the machine from working, much as you did with yours. My suggestions before that:- Fit the NOP tester in place of the CPU and check again to see if there is activity on UI10 / UI11 pin 1. If so, then it just shows that the CPU is currently getting into a loop or a mode where it is just not attempting to read from system RAM, and we can disregard the lack of activity there when the CPU is fitted. Similarly, the RAS and CAS lines of the system RAM, pins 4 and 15 on each 4108 chip. Is there activity there during a NOP test? Lastly, still while running the NOP test, scope the BD0 to BD7 lines again to see what they look like. UI10 pins 11-12 UI10 pins 13-14 UI10 pins 15-16 UI10 pins 17-18 UI11 pins 11-12 UI11 pins 13-14 UI11 pins 15-16 UI11 pins 17-18 |
|
18th Mar 2021, 12:29 am | #1132 |
Octode
Join Date: Mar 2020
Location: Kitchener, Ontario, Canada
Posts: 1,294
|
Re: Non-working Commodore PET 3016
Maybe modify the NOP tester to use PHA, $48, push accumulator to the stack. This would make the 6502 keep writing to system ram in the stack page, and it would also read from ram as it cycles through the address space. This would then give more repeatable activity to system ram that could be traced with the scope.
D7 low, D6 high, D5 low, D4 low, D3 high, D2 low, D1 low, D0 low. |
18th Mar 2021, 12:48 am | #1133 |
Dekatron
Join Date: May 2008
Location: Derby, UK.
Posts: 7,735
|
Re: Non-working Commodore PET 3016
Even during a NOP test, the processor will still be trying to read addresses that fall within RAM (i.e. &0000 to &7FFF) half the time.
The data pins are hard-wired so the processor always sees &EA at every location it reads. When it starts up, both the low and high bytes of the reset vector at &FFFC and &FFFD respectively read &EA; so it begins executing from &EAEA. It reads NOP, of course, and continues on to &EAEB, &EAEC and so forth all the way to &FFFF, back to &0000 and up to &EAE9, then back around to &EAEA again, almost 8 times a second. If you trigger from the falling edge of A15 and move the trigger point near the middle of the screen, you will be able to see if RAS or CAS behave any differently before the change (while A15 is high and the processor is reading ROM / peripherals -- only the refresh logic will be accessing RAM) and after (while A15 is low and the processor is reading RAM -- both it and the refresh logic will be taking it in turns). EDIT: The PHA test sounds fiendish! It should certainly exercise the address bus lines a bit .....
__________________
If I have seen further than others, it is because I was standing on a pile of failed experiments. |
18th Mar 2021, 12:52 am | #1134 | |
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Quote:
Alan |
|
18th Mar 2021, 1:23 am | #1135 |
Dekatron
Join Date: Aug 2011
Location: Newcastle, Tyne and Wear, UK.
Posts: 11,553
|
Re: Non-working Commodore PET 3016
I like Mark's PHA test idea so if Colin is up for that, let's do that. in the meantime I'll get a 2716 or 28C16 programmed up with some diagnostic code and lash up an adaptor for it to go into.
I'll have to fit it into the hairline cracks between other stuff at work but I will try to have it rigged in a few days or so, then I'll try reading it as though it is a PET PROM and if that works (ie, proves the adaptor is working) I'll whizz it up to Colin. |
18th Mar 2021, 4:39 am | #1136 |
Octode
Join Date: May 2018
Location: Northampton, Northamptonshire, UK.
Posts: 1,440
|
Re: Non-working Commodore PET 3016
Regarding replacing the original TMS4108's like for like, then one major problem is that you need to get the right versions NL0 or NL1 suffix, to match the one being replaced. And if you've got 1 or more failures in both banks, then would have to source both types. Attached is a full datasheet on it, for reference, as can sometimes be difficult to find.
A couple of tricks I've read about for diagnosing faulty rather-similar 4116's in ZX Spectrums: First to feel if some are getting hotter than others. And remove any that are getting much warmer. You could also try checking the resistance of data pins to ground / each of the supplies, to check for any internal shorts in one (although if both banks are on the same bus, without isolation buffers, then there would be both ones in each bank on that particular data in/out line. But can't do on Address-lines, as all shared. However, 'scoping the datalines and checking for anomalies maybe better. The 2nd trick is to place a known-good DRAM IC on top of each of the board's DRAM IC's, and see if any differences occur. This is meant to work if IC is dead but not shorting. And probably works best if only one has failed. Note: It's probably wise to power-down each time, before holding IC on top, and would either need to use the same TMS4108 suffix type or a 4116. I've also seen Data Out etc. pins cut and then resoldered, to help isolate faults, but not too good for it looking fairly untouched. I think I have tried these techniques once, but Ididn't have too much luck. So I found it easiest to just socket all 8, and put known-good ones in to start with. Then swap the originals back in a few at a time, to find which ones were faulty. - Especially as I have acquired quite a few other the years very cheap from radio rallies - mostly new unused ones. Although this place list them at only 40p each (around what they eventually dropped to when new, before discontinuation made them retro-priced): http://www.diverse.4mg.com/compfull.htm (If they still exist / have these, as website is rather old) Being as the PET 6502 CPU only runs at 1MHz, compared to nearly 4MHz on Spectrum Z80A, then access time isn't as important. And so even the slowest (250 / 300ns?) ones should be OK. There's quite a lot of different make 4116 datasgeets here, for finding out what the suffix numbers mean in terms of access times (some have higher single number suffix for fastest speed rating / lowest access-time, rather than it being the start of the access-time: http://www.citylan.it/wiki/index.php/DRAM_16k_x_1 This also shows an interesting DIY ROM Board converter, to use a single EPROM: http://www.themadhowes.org.uk/pet/index.html Although they don't show a schematic or the underside layout of the veroboard. |
18th Mar 2021, 10:45 am | #1137 | |
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Quote:
Alan |
|
18th Mar 2021, 11:12 am | #1138 |
Nonode
Join Date: May 2004
Location: Halifax, West Yorkshire, UK.
Posts: 2,587
|
Re: Non-working Commodore PET 3016
Just one other thought that's occurred to me since mentioning the two 6520 PIAs earlier. The 2001-8 will boot without either PIA fitted and I think the 3016 might be the same. As a quick check Colin could test the PET without the PIAs fitted and it might even boot if all the 4108s (plus everthing else!) are working ok. Again I'm assuming that the PIAs are socketed.
Alan |
18th Mar 2021, 1:26 pm | #1139 |
Octode
Join Date: May 2018
Location: Northampton, Northamptonshire, UK.
Posts: 1,440
|
Re: Non-working Commodore PET 3016
It would seem that the 3000-series O/S was quite close in operation to the 2001, if someone had 'downgraded' their ROM's on a 4000-series to 3000-series ones for better games compatibility, here:
http://www.themadhowes.org.uk/pet/index.html And it seems that Commodore had generally always socketed the larger IC's at least, on their earlier computers. Although if DRAM's were socketed originally on that 4000-series one, you'd expected they would have been on the (older?) 3000-series - Especially if they can be swapped & links changed to fit 4116's and double the memory with 2 banks of these. Although Commodore did seem to rather randomly used several different motherboards of various vintages in the early series PET's. So worth a go removing PIA's at least, to see if anything different is observed. Especially if the board in it will also work with 2001 O/S. From my experiences of 4116 failures, I've found it's more common to find that > 1 has failed than just a single one (sometime over half of a bank). So that doesn't help with putting good ones on top, with hope that faulty one has gone open-circuit. Although the Spectrum's rather flakey DC-DC converter, to generate the +12V & -5V rails, seems to be often the cause of many 4116's failure when rails go missing / incorrect voltage. As you need most of a 4116 to work in these, for it to reset & display OK, then I've not found out if 4116's often fail with just some dead locations. I have found just some locations faulty on some TMS4532's (half-faulty 4164's), but that may have been when the links hadn't been set correctly to match the good half of the IC. And 4164's seemed much more reliable, although have had the odd one fail with output virtually s/c (as with some TMS4532's) |
18th Mar 2021, 2:55 pm | #1140 | |
Octode
Join Date: May 2012
Location: Perth, Scotland
Posts: 1,803
|
Re: Non-working Commodore PET 3016
Tried this - no difference (ie blank screen which has power).
Colin. Quote:
|
|