Re: pet 2001 repair

On Mon, Jan 8, 2018 at 2:24 PM, Mia Magnusson <mia@plea.se> wrote:
> Den Sat, 6 Jan 2018 18:05:54 +0100 skrev Francesco Messineo
> <francesco.messineo@gmail.com>:
>> Edit: screen picture is not attached, here's a link:
>>
>> https://photos.app.goo.gl/zy6vBiuoYziDXa6J3
>>
>>
>>
>> Hi all,
>> I'm not progressing much on this 2001. I'm now able to test the 6550s
>> out of the PET and I've found two surely bad (also confirmed using
>> them as video RAMs).
>> Most of the times it just boots directly to an illegal quantity error,
>> but sometimes displays the basic prompt without the bytes free amount
>> as you see in this picture.
>> I've suspected a problem reading the video ram (writing just works,
>> otherwise it wouldn't clear the screen and display correct
>> characters), but running the system with a nop generator showed no
>> obvious problem on the bus transceivers or address decoders and
>> anything else I could test.
>> Screen scrolling looks a bit too "jumpy" compared to what I'm used to
>> see on 3032, but I really don't see any problem in the video ram to
>> data bus glue circuits.
>
> Does it scroll like two lines at a time?

no

>
> Not sure how a 40-column PET does, but if it is anything like a VIC-20
> or C64 then it has logical and physical lines. If it believes that the
> physical lines are linked together then it scrolls more than one
> physical line at a time.
>
> Illegal quantity, no shown amount of free memory and strange scroll
> would lead me to check the RAM. Not sure which adresses the BASIC
> floating point accumulator is at or where a PET stores the
> logical/physical line links, but as your machine atleast starts and
> tries to run basic it probably has working ram for zero page and the
> stack (i.e. adress 0-511). I'd check the ram for adress 512-1023.

I've been checking all the ram chips from the start. I've first used
the VRAM sockets of the 2001 itself and found two bad chips (they're
the very particular MPS6550 ICs) and then I completed my own test
setup where I could test all the rams better, including a 3-state test
in any (wrong) combination of the four chip select plus phi2 inputs.
So the ram looks fine. I've also run this 2001 with a nop generator
for a long time, making sure all addresses and selects reach all the
intended places, including the ROMs and the RAMs.
I can't really exclude a subtle problem on the data bus transceivers,
but the fact that the initialization sequence always can clear the
VRAM, print things and initialize the keyboard scanning (so travelling
on at least other 3 different ROMs) makes me conclude that the data
bus path to the ram must be working enough well.

>
> If all your ram is socketed I'd just try to move around the RAM chips
> and see what changes.

I think all the 6550 were swapped around on the first bank sockets at
least 10-15 times before I completed the out-of-pet 6550 tester :)

>
> (I haven't read anything about the old 65xx RAM chips as my oldest PET
> uses 2114. Not sure what size they really are, if they are 1k or more
> each then this probably doesen't apply).

6550 is 1kx4 as the 2114, but it has 4 chip selects (2 active high, 2
active low) and latch the address bus + selects + R/W on the rising
edge of phi2.
I've made a replacement PCB to allow substituting one 6550 with one
2114, I'm waiting for the PCBs to come back from USA.

I'm suspecting a bad ROM by now, but I'm still working at a couple of
good ways to test the 6540s now ;)

Best regards
Frank

       Message was sent through the cbm-hackers mailing list