Dead PET4032

[Ethan Dicks]

On Mon, Oct 5, 2015 at 2:52 PM, william degnan <billdegnan at gmail.com> wrote:
> Look for inconsistencies in voltages and pulses to and from RAM.

Yep.

> You may need RAM at the lowest memory locations to boot.

Absolutely.  The startup routines will be unhappy if zeropage is not there.

You can check for -5V as well, since the 4116 DRAMs need that.
There's a small circuit on the board with a 7905 to produce that.

>  Hooking up the keyboard
> at least to eliminate a missing keyboard as a reason the system does not
> boot

Nope.  As Tony Duell mentioned, it's a passive matrix.  The system
will not notice a missing keyboard.

> and you can issue commands to search a disk drive even if the display
> is faulty.

You can issue disk commands headless, but if you don't get the piezo
"twiddle", you probably aren't getting through POST to a READY prompt
anyway.

You _could_ be at a TIM prompt, I've seen that for some flavors of
broken PETs.  The video works, but something is bad in RAM or ROM to
screw up BASIC initialization.

> At least if you have a IEEE drive attached you can watch for
> the the drive to respond when the 4032 is powered on, for signs of life
> from the computer.

The machine does not autoboot.  The only signal that can affect the
disk drive is RESET, and that doesn't require the IEEE signals to be
initialized in software.

If you can get power to the video, you can pull the Chargen ROM and
you'll just see blanks for regular chars and solid blocks for inverse
video - like the cursor.

With the older models with TTL video circuits, they would display
chars if the ICs in the video section were working even if the CPU
wasn't, but the 4032 has a 6545 CRTC which is initialized by the ROMs.

On a machine that mostly worked but with dead video, I'd expect to see
strobing on the keyboard matrix every 1/60th of a second, and lots of
RAM and ROM enable strobing constantly.

One of the old hardware debugging tools was to take a real 6502 and
bend out the data pins and run wires from them to power and ground to
"spell out" $EA, for NOP (like this guy did...
http://www.grappendorf.net/projects/6502-home-computer/the-microprocessor).
Then back in its socket, the 6502 would run up the address bus and you
could watch the accesses flood over the address space, looking for
stuck select/enable bits

I have replaced dead 2114s (video SRAM), dead RAM (4116 DRAM and 6550
SRAM) and every once in a while, a dead ROM.  That's where I'd poke
around first, depending on what symptoms match what you are seeing.

-ethan