Model 4P screen repeated picture in 80 column mode

[oktology]

[some of yall will have seen this on the Tandy Discord server]

My Model 4P is finally booting, thanks to some aboveground repairs to the floppy drives and Ira’s generous floppy disk service, but it does this in 80 column mode:


In 64-column mode, it works fine:

I’ve adjusted the capacitor on the front of the logic board, and this is as good as it gets. The picture in 64 col is rock-solid, no flickering, no rolling, no tearing.
Onboard RAM test passes. Can’t run Model 3/4/4P diagnostics disk (Model 3 ROM images consistently fail CRC check, a separate issue that I’ve been advised may b due to misaligned floppy drives).

Where should I start looking?

 

[Bruce Tomlin]

Model 4P, 80 column screen won't sync, it's probably the varcap that adjusts the PLL for the 80 column timing. Apparently Tandy cheaped out and used a PLL with R/C parts instead of just using a second crystal. The varcap is in the middle of the logic board, I remember that it's a pain in the ass to get it out where you can adjust it.

 

[oktology]

it's probably the varcap that adjusts the PLL for the 80 column timing

Already tried that, unforch. All that would do is give me scrambled noise or the picture above, through the whole range of adjustment.

 

[Bruce Tomlin]

Then it isn't simply out of tune. I recall that Ardian had a 4P where he explained how the PLL worked for 80 columns because his was more broken than that, see if you can find that video. But it should have absolutely zero to do with the CPU or software, just a bad hardware design. If it works in 64 columns, it's just a sync problem.

 

[fritzeflink]

TRS-80 Model 4P 80-column crash repair (Part 1)

 

[Bruce Tomlin]

It looks like the necessary signal is 12.6745 MHz, so if you have some way to supply that instead of the PLL, you might have a better fix. A real crystal and a CMOS 40106 could do it.

The hard part is getting the right crystal value. (Double-check the actual value if you do this!) I see Mouser has a 12.67917 mini-can in their selector, non-stocked. But there's a 12.688MHz in a small ceramic SMT which you could hook up with bodge wires, for less than a dollar and ~900 in stock. That's probably close enough for the CRT circuitry to sync to.

 

[oktology]

A real crystal and a CMOS 40106 could do it.

Not a bad idea. Eliminates a source of unreliability, as well.