TI-99/4a severe display problems

[Stig]

Hi. I got this cheap TI-99/4a the other day, but the screen is all messed up. Anyone know how to fix this?

 

 

 

 

[+OLD CS1]

I suspect VDP RAM problem, though it does not look like one of these masks here

 

https://www.ninerpedia.org/wiki/Troubleshooting

 

Maybe it is a stuck-on bit rather than a stuck-off.

[Tursi]

Probably still VRAM, since it made it to the title page. Looks like there's probably an address bit issue combined with a data bit issue

 

[+mizapf]

I can see this:

 

E (>45) → U (>55)

A (>41) → Q (>51)

N (>4E) → ^ (>5E)

I (>49) → Y (>59)

 

In the empty space we have "0 0 0 ..."  which means that at every second position in the Screen Image Table there is a >30 ("0") instead of the >20 for space.

 

Also, the character definitions have additional dots. The space character looks like 10 00 10 00 10 00 10 00.

[Stuart]

It suggests that the 4th bit is stuck at 1: xxx1xxxx, but it is not consistent - "TEXAS" for example is fine on the top line but corrupt on the bottom line, and some characters that should be affected are displayed OK. After it has warmed up for a bit, do you see the exact same screen as above each and every time you reset it?

 

I'd start by replacing the RAM chip labelled "3" in the photo on the link above and see what you're left with.

[+mizapf]

TEXAS is correct on the top line because every second character is changed, but only if the bit is not already set. That is, characters >50 - >5F will not change, while >40 - >4F will turn to >50 - >5F *if* they are on the affected columns.

 

So on the top line we have (T)E(X)A(S) → TEXAS

and on the bottom line we have T(E)X(A)S → TUXQS

 

[+mizapf]

Looks like this:

 

 

I added the following line to the VRAM write function:

 

if ((m_Addr & 1)==0)
           data |= 0x10;

 

m_Addr is the current video address, data is the value passed to the VDP. Theories?

[Stuart]

So is that the lowest order address line pulling that data bit high, but can't pull it low? I think I'd be looking out my box of spare 9918s as well ... 

[+mizapf]

The problem is with U105 (the chip marked by "3"); maybe setting the LSB address bit (A0 at the VRAM) to 0 forces the data input to 1. However, the address is set in two passes (7 bits, strobed by RAS and CAS), so this is not really convincing yet.