Paul Anderson wrote:
You might want to pull the mos memory and get
everything down to one
backplane. I don't know who made the mos, but it might be causing
I've also seen the backplanes and the foam under them cause weird
shorten the bus as much as possible and see if things work better.
At Paul's suggestion, this afternoon, I got the essentials stuffed into
Packed in there was the Front Panel, CPU/EAE, 2x4K Core sets, Mem
Ext/Timeshare, RX8E, RK8E set, M8650 Serial, and bus loads board. It
all just barely fit -- there is one free slot.
After checking to make sure that nothing was interfering or touching
between the boards, I powered up the 8/e with baited breath.
It powered up OK, no smoke.
I restarted OS/8 from 07605, and spun up the RK05, and ran RLKFMT. It
went through the write pass with no problems (as usual), and made it
through a good chunk of the read verify pass, but failed at exactly the
same place as it did before (as soon as the disk address went to 1xxxx).
I did a ZERO RKA0:, and then tried to do DIR RKA0:, and it gave the
director error as usual.
I then fired up DHRKAE (Diskless Controller Test), and it failed in the
exact same place, TST81, the first data break transfer test.
So, whatever is going on acts exactly the same with a relatively
minimized system packed onto one Omnibus.
Next step is to put the system back the way that it was, and then tack
some wires onto the RK8E on M7105 (Major Registers) board, to monitor
the LOAD and CLOCK signals on the 74161 counters that make the up
current address register, to see if there is a transient on the clock
line, as David H. observed on his system. If the CLOCK signal does have
a transient on it after the LOAD pulse has finished (the load of the
current address register is clocked in by the clock signal), the
transient could cause the counter to increment by one before any DMA
transfers start, which could create the symptom that I am setting.
Don't know if I'll have time to do this today, but if not, perhaps
I will post updates as information is available.
Thanks again to all for the suggestions.