Recently got some cheap, semibroken HP 3456A DVMs. Two of them had no display or only displayed unrecognizable symbols.
After I had replaced the obviously bad caps in both PSUs, it turned out that one processor board had a short between 5V and GND. "Careful" application of 5V @ 5A from my PSU somehow solved the problem and the processor board was working again. The other processor board wasn't shorted but wasn't working either. So I started to add DIP sockets for the ROM, RAM and the MC68A00 CPU on both boards.
By systematically swapping CPUs, RAM and ROM between both boards, I found that one set of the old Mostek MK 36xxx ROMs had leaked and a MK4118A 1kB RAM chip was broken. So I designed a replacement using a salvaged AT49F512-90 from a broken graphics card, not realizing that computer inside the HP3456A was old and horribly slow and the 49F512 relatively new and thereby fast.
Of course I failed, and after checking my PCB layout and soldering for errors, I decided that I need to use a logic analyzer. So I designed an adapter for the 6800, so that I could connect my trusty HP 16500C to it and check the timings.
From the documentation I saw that the 6800 loaded the program entry point from address 0xFFFE and FFFF. So I set the HP 16557D to trigger on one occurrence of 0xFFFF at the address bus with 96% post store and transfered the results by FTP to my computer for analysis (HP 16500x are horribly slow btw.) and converted them by using the la2vcd program. (Logs are in the attachment of the second post, needs GTKWave).
As it turned out, the AT49F512 has access times of around 50ns, while the original ROMs had access times of appr. 270ns. This fast EEPROM somehow causes the old and slow 6800 to fail, as initially seen on addr 0xD8A8, at which there is some conflict on the data bus.
After I had realized that I was a bit mad about the lost time and failed design. But I continued and used some old 27C512 I had lying around, erased them in a ghetto style UV EPROM eraser, build from a 253,7 nm low pressure Hg vapor carp pond disinfection lamp and an electronic ballast salvaged from a broken fluorescent lamp.
(Be careful, the 253,7 nm UV is very harmful for the eyes, I put the 27C512 on top of the lamp and then put a book on top of the box. Then I connected the lamp to my isolation transformer, switched it on and left the room for 15 min, ozone emission is significant, too)
After I programed it in a Willem programmer, I first tested it using a breadboard and it worked.
So I designed the second board that directly plugs into the DVM, using the address lines from a MK36xxx and those from U6 (74LS138) below it. /OE is also supplied from U6. Schematics and PCB stencil layouts in A5 are supplied in the attachment of the second post.
Maybe this helps other guys to fix their HP 3456A faster.