C128D displaying random characters on startup screen (plastic case model)

[banman]

Hi rmzalbar,


Thank you. You were very accurate in diagnosing a bad CharROM chip.

Unfortunately my ageing oscilloscope has amn undetermined failure. I've tried a few things suggested by people who have this model, but no luck. The analog section is okay, it's the on screen digital display stuff. Basically no display at all. The front controller panel comes up with the binary error code 6000. I think that means it's the wave form display board playing up.
I wasn't able to check signals coming out from the IC's using an oscilloscope and I don't think that my logic probe would be great this type of thing. I could be wrong though.

I tried this method for checking the CharROM chip instead. I few web sites had details on creating a 2364 to 2764 ROM



adapted socket.

https://portcommodore.com/rcarlsen/cbm/ ... dapter.jpg

http://www.tabalabs.com.br/eletronica/r ... ad2364.gif

5eebEA4.png (8.57 KiB) Viewed 479 times

I made one up and tested it in Tl866 mini pro chip programmer comparing it to a known good ROM on the internet.


http://www.zimmers.net/anonftp/pub/cbm/ ... ters/c128/

The CharROM chip verified bad. I burned an EPROM 2764 and installed with a 2364 to 2764 adapted.

That seemed to solve that particular issue.

[banman]

Here is the final version mounted under the heat sink. I had to do a dodgy to get the EPROM not to strike the power supply. I also bent the heat shield can back a little bit to accommodate the larger EPROM and socket (28 pin 2764 as opposed to the 24 pin of a 2364).

I put the EPROM in a sacrificial socket and put them both in a vice and pushed the EPROM backwards bending all the legs at the lead frame back a few mm.

[rmzalbar]

In some cases I've done stacked sockets where there was room to do so, in others I did trace surgery on the PCB itself. I'm to see glad you've got it working. Are there any issues left?

[banman]

Hi rmzalbar,

Thank you!
The vertical bars are now gone, I appreciate yours and Gyro Gearloose's helpful troubleshooting hints.

There seems to be an issue which was present along with the bad CharROM chip.

In 128 mode I can type the ubiquitous Hello World program in. When I "List" back the program the line numbers come out as reversed characters instead of numbers. When I run the program it gets a bit flakey ending up in a general lockup. I do observe that the cursor blinks red instead of the normal green C128 scheme, only the cursor though.

Interestingly in C64 mode the same program is totally okay.
In C64 mode the whole system appears to be okay.

Running the dead test cartridge in C64 mode everything checks out fine. No issues at all.
However when I run the C128 dead test cartridge there are many ram chip errors. They do not appear at the same locations each time I run the test.


I'll post some pictures to try to characterise the problem.....

[rmzalbar]

Yep, that sounds like faulty address logic in a component that isn't used in 64 mode. It would be useful to know the differences in memory map between the c64 and 128 to start with. Does 128 do basic in the higher 64k-128k bank for example? Are the errors showing in the lower 64k bank also? Any hot logic ics? Have you tried freeze spray?

[banman]

Hi rmzalbar,


Apologies for the delayed response. I had to do a little bit of research.

Yes, I think you could be possibly right.

Before I carry on. Here is a picture of the C128 dead test cartridge running with the new CharROM EPROM chip installed.

This version of the C128 deadtest cartridge uses the C128's CharROM chip for displaying text. It appears to be work fine now.

The 2 CIA chips and SID chip read bad I believe because I don't have any loopback harnesses attached.

[banman]

The C128 deadtest cartridge test fails the ram tests. The C64 deadtest cartridge passes all the ram tests okay.

I did some searching and found this diagram representing the physical ram layout in the C128. Apologies for the poor image quality.

So the bottom row of green shaded ram chips represents the memory used in C64 mode. Looking at the physical C128 mainboard with the RF modulator orientated towards the top, these ram chips would be the closest to the bottom edge of the PCB board.

The 2 rows of ram chips are used in C128 mode.


I have been reading a little from a book...

Troubleshooting and repairing your Commodore 128

https://archive.org/details/troubleshoo ... 1/mode/1up

From the little I understand the memory management unit (MMU) can configure this ram in several different configurations depending on what's required of the system. The MMU seems to have been made to handle 256kb of ram for presumably a future model upgrade revision.

The top row of ram chips is used to store variables in C128 mode.

I would hazzard a guess that an issue may reside in the upper row of ram chips. That is if the deadtest cartridge is diagnosing correctly.

Maybe the issue of reversed characters instead of line numbers when I type a basic program in C128 mode lies in a faulty ram chip in the top row of ram chips.

[banman]

I have been watching a YouTube video from Retrorecipes



At about 22 minutes into the video Perifractic describes a technique to switch the 2 banks of ram around so that the top 64 kB of ram which is used for C128 mode can be used in C64 mode.
This would allow the factory C64 deadtest cartridge to check the upper memory bank U46-U53 as if it were the lower bank

Here's a screenshot of his setup.

Here's my take on it.

[Gyro Gearloose]

Serious question: people have faith in these "dead test" cartridges? They seem cumbersome and not very helpful. IMO the 64 is simply not a powerful enough computer to properly and reliably self-diagnose. No boundary scan, no system management bus, etc
It seems to me people spend more time troubleshooting the cartridge than the computer?

[rmzalbar]

Serious question: people have faith in these "dead test" cartridges? They seem cumbersome and not very helpful.

Well..

They're not normally cumbersome. On a black screen 64 you just plug it in and go. It's not guaranteed to help, but it only takes a few minutes to use and if you do get feedback from it, it can save you a lot of time. Of course, you do need to interpret the results correctly. It looks like its doing its job just fine in the above case.