Ebay Board 2 : The Mystery of the failed 6 chip to 2 chip TOS upgrade

[Elethiomel]

As promised in my last thread where I fixed a "for parts or spares" board, here's part two.

My Ebay board came equipped with 6 chip TOS 1.0. Luckily I have a 2 chip 1.04 set so I thought I'd spend 20 minutes upgrading the board.

I first removed the existing 6 chip set. I then used copious amounts of flux and some desoldering braid to clean off the old 256K pads and soldered the 1M A16, A17 and CE pads. The next part was a bit of a PITA. This particular board (a C070789-001 REV B), didn't come with U68 installed. Luckily I had one spare from my donor board. I installed a socket, popped in the chip and the ROM and booted up...

I get a white screen for about 4 seconds. It then goes black for 2 and then white again and the cycle repeats.

Here's the list of debugging steps and tests I have done:

Reseating the all the socketed chips
Replacing the GLU with a known good one
Replacing the MMU with a known good one
Replacing the two chip ROM set with an Exxos TOS 1.02/1.04 adaptor board kit that plugs in to the 2 chip ROM sockets.
Replacing the the ROMs with a two chip TOS 1.02 ROM set
Undoing the upgrade completely, verifying the system works 100% and redoing the upgrade. I've done this 3 times.
I stole another 74LS11 from another system that doesn't really need it as it uses a TOS 2.06 decoder.
I've verified the 74LS11 works by testing it with my Minipro which correctly detects the chip type and passes it.
I've verified that the 256K pads are not connected.

To make sure the 74LS11 was at least connected, I consulted the schematics. It's getting power and has proper ground. Only one of the gates on it is used. 3 pins are correctly routing to the ROM0/1/2 pins on the GLU. The other pin connects to the CE 1M.

I found someone having a similar issue on the atari-forum, but he nuked his ROMs by putting them in backwards and no solution was found. https://www.atari-forum.com/viewtopic.p ... 4&start=25

Here's where things get a bit weird.

It can boot to a diagnostic cart. Carts aren't booted directly by the CPU, they're detected by TOS and jumped to early in the boot. At the very least, the decode logic has to be working in some way that both of the chips are at least decoded for some small part of the bottom part of the ROM.

The diagnostic cart also detects the correct TOS version which is to be expected if it can read the bottom of the ROM.

The diagnostic cart fails to CRC the ROM correctly which further leads me to believe that the decoding is arsed. In fact, it gets a CS error which I'm assuming is a bus error occurring and the diag cart catching that.

20220501_015238.jpg

So, before I spend a week tracing out each and every connection on this system, I thought it prudent to ask if anyone had any experience with this and knew what the solution was :D

I could of course just install a 6 chip 1.04 set. Exxos has them in stock, but where's the fun in that?

[DoG]

Is the three solderblobs set correctly?

[Elethiomel]

Is the three solderblobs set correctly?

Yup, they seem to be. All are set to 1M. I've cleaned the area and redone this 3 times

[exxos]

Are the ROMs original ? Tried on a known working machine ?

[DoG]

And they are sitting in Lo-0 and Hi-0? Otherwise maybe post a picture of the area might help.

[Elethiomel]

Are the ROMs original ? Tried on a known working machine ?

I have 2 original sets of TOS 1.02 and one original set of TOS 1.04
I also have your 2015 Tos 1.2/1.4 dual TOS adaptor board

All of these work in another machine. None work in the Ebay ST.

[Elethiomel]

And they are sitting in Lo-0 and Hi-0? Otherwise maybe post a picture of the area might help.

20220501_014219.jpg

All three bridges verified by sounding out vias on both sides of the bridge.
Also confirmed that the 256K bridges are not shorted.

bridges are shinier in real life and have been redone several times to make sure they're good.

[frank.lukas]

The Atari Service Diagnostic Rev 4.4 does not know about TOS 1.04 in two chip so the behave is normal and not an error !

[Elethiomel]

Nah, it's definitely an error. The cart doesn't know if there's 6 or 2 chips, so all it does is assume there's 6. Either way, it still reads the same address space.

Here's the same ROMs booted in another machine. Notice the lack of CS errors.

20220501_194008.jpg

[PhilC]

I think I’d start by metering out all of the address lines from the ROM to the cpu. It’s likely low ROM areas are fine as it boots to diag cart but something is corrupted after that.

Also the black foreground and white border also points to ram, so give the address lines and mad lines a text. I’m presuming ram tests ok with the cart?