Ok, have Chromatix exact suggestion up and running. Definite improvement. Still some garbage, but way less. The remaining garbage kinda looks like software issues to me. It's mostly snippets of good data in strange places. I had that before, but also a ton of random characters. Now I have occasional random characters, but not near like before.

I have enough of a monitor working that I can read memory. If I read the address in VRAM that has displayed a stray character, it shows that character to be there. That's not unexpected of course.

But when I read the corresponding address in the off screen buffer in SRAM, I also see that garbage character. Correct me if I'm wrong, but doesn't that suggest my code is the problem? It's mostly snippets of correct display data that, along with appearing where it should, appears in strange places. The garbage snippets are predictable (if I do the same things, I get the same, seemingly random, results).

My screen write code is fairly complex unfortunately. It implements scrolling (kinda works right now. Kinda). Thus, there is alot going on as far as managing where in the buffer to write, and where in the buffer to start drawing. I can easily imagine mishandling a 16 bit address in there somewhere. I should count how many characters away from the good stuff the garbage is, see if its a suspicious number like 256...

Anyway, things ARE better than they were, and I gained almost 8k of RAM at no cost. I'll keep you posted.

Oh, I should mention that my Clear Screen routine works perfectly. No garbage. Since it just writes spaces to the entire buffer, and doesn't need to deal with scrolling or line positions at all, this too suggests my code is the culprit . If it were a hardware thing, you'd think that the clear screen routine would be vulnerable as well. I realize miss-placed spaces would be invisible, but there is just enough random stray characters during other screen writes that I think it's relevant that the clear screen routine doesn't have them. I am gonna double check my address lines just to be sure of course

Software bugs are always a possibility, but now I think your hardware is reliable enough for you to start tracking them down. Good luck!

If you clear the screen AND the corresponding screen buffer in SRAM and then use your routines to write a character to a dedicated screen address (with varying addresses) you could insert a buffer search (still clear?) and stop if not. Then you might get an idea what routine under which circumstances does not operate properly.