• Please review our updated Terms and Rules here

M8192-YB fails BEVENT test

S

shirsch

Veteran Member
Joined
Aug 17, 2008
Messages
873
Location
Burlington, VT
After the first one turned out to be a dud, I picked up a second M8192-YB w/ FP chip and installed it in my modified (22-bit) Heath H-11 backplane along with a QBone to emulate memory, storage, console and ltc interrupt. First order of business was to run the ZKDJB2.BIC CPU test for 11/73. It fails the BEVENT register checks starting at test code 01166. If I suppress the BEVENT test by setting the appropriate bit in the configuration register it passes everything else. All other tests (memory, cache, memory mapping, etc) pass 100%. The ZKDJ test requires that that halt-on-trap be disabled by inserting jumper W9 on the CPU board. Problem is that the M8192 documentation calls out W5 for this function and lists W9 as a disable for the BEVENT reqister. I've tried all combinations of W5 and W9 with no change in the (failing) result. Before I assume the CPU has a hardware defect, I wanted to double check that I'm not doing something dumb to cause the problem. I have verified that the BEVENT signal appears on the bus and tried (individually) both the 50 Hz. signal generated by the QBone and the 60 Hz. signal available from the Heath power supply.

Advice?
 
The test that generates error code 1166 - checks that with a processor priority of 5 during the execution of the command "SOB P1,." with the initial value of register P1 equal to 177777, there will be an interrupt from the line clock (interrupts from line clock has been enabled).

This also means that the B9 jumper must be set.

And to pass the test as a whole, jumper B5 must also be installed.
 
Note to self: Next time I'm convinced that a signal is "present on the
bus" this should be verified at every applicable connector. Turns out
that while BEVENT was connected through vertically on the backplane
between the four B and four D connectors, never the twain shall meet.
There was continuity from the power supply BEVENT only to the B
connectors and, indeed, the PCB layout and x-ray pictorial show no
evidence of a trace between the sets. I added a jumper to bring the D
connectors into the fold and - bingo - everything is working correctly.

2.11 BSD boots and the ZKDJB2 CPU test now passes.

At the time my original H-11 (not the 'A' model) was sold they were using
the quad-height M7264 LSI-11/03 CPU exclusively. I'm betting that it
picked up BEVENT from the 'B' connectors. I'll further wager that the D
connectors will show continuity on an H-11A backplane (which definitely
has a different part number).

Interestingly, the tests pass without W9 installed. This makes sense, since the docs indicate that W9 inserted disables the BEVENT register.
 
You must log in or register to reply here.
Back
Top