I reconnected the data bus pins, and disconnected all the MA0..11 pins (in
case it's DMAing into memory when it should not be). Nope, same issue. Won't
boot Serial Disk, corrupts the upper six bits of the loader.
Only bus left is the memory data MD0..11, but that can't be shorted because
on the card it has only inputs. Also the IOT address is decoded from it and
those select properly. Still thinking it's DMA, will try watching the break
request and other DMA lines for activity.
Nothing unexpected there.
But at last I found something... I think. Despite it not making sense, I
disconnected the top 8 bits of MD0..7 (thus not only disconnecting the bus
receivers on the card, but also completely deselecting it). Now the system
booted and runs normally with the card plugged in! So, I figured either one
of the 8640 receivers at E3, E20 (page 5 of the 10-page schematic) is leaky,
quite possibly E20 which handles bits 4-7 the "troublemakers" from before,
or the 8136 at E11 which appears to be just a multi-input AND/OR gate combo
to select IOT x60x/x61x and was working with scope loop.
I gradually reconnected lines until it started failing to boot and wiping
out the loader again... I am so tired of toggling that loader in! At least
it's only 26 (octal) words.
Turns out MD4 was being pulled down weakly (to a volt or two) by something
when it should have been pulled up. Wiggling and flexing the card caused it
to work, intermittently. But I could not find it even with close visual
inspection. I suspected a tiny tin whisker somewhere...
So I crossed my fingers and tapped the pin with a cliplead from the 25 amp
+5 volt supply. Figured I had nothing to lose at this point! ... and
apparently did clear the short :)
Now that line looked normal just like the other 11 memory data bus lines.
OS/8 restarted with no problem, too.
Started the diagnostic AJRLAC (the "Diskless Controller Test"). Immediately
indicated a hard failure on bit 10! Oh $@#%. Now what did I do...
But it just took a minute to pull the extender card and sure enough I had
just made a bad solder joint reconnecting that pin and it had come apart
with the flexing ;)
Fixed that... running four passes without an error so far. Dare I touch the
middle of the board again?
Yep... flexing in both directions, no failures!
I think I got it! Make that eight complete passes with the extender removed
and the board in the cage.
Running AJRLIA on a scratch pack now. Initally I got a very occasional
seek/tracking error (Command Reg B 1017 or 1117) once per pass on each
drive, but it's lessening with "exercise"...
It's 90F in the computer room too, which may be above spec for an RL02
anyhow. Just finished Pass 0002 on Drive 1 without errors :)
-Charles
