PDP-11/45: RK11 III

Okay, the M9202 bus jumper arrived, and like the 2-foot BC11 cable, the occasional timeouts go away when this is installed. Hantek digital scope also arrived, so I decided to throw it on the backplane for a closer look at the SSYN and timeout signals. The results were pretty interesting. Here's a capture of an RKCS access triggering a timeout glitch with the M902. The yellow trace is BUS A SSYN L (taken from C12J1 on the 11/45 backplane), and the blue trace is UBCB TIMEOUT (1) H (taken from D12U1):

What's interesting is that with the M9202 in place, the SSYN waveform shape on RCKS accesses is not significantly different -- and the timeout glitch still ocurrs from time to time, but at a reduced amplitude:

If the problem had been one solely of lumped loads on the bus, I would have expected the fix to manifest as a waveform difference, and for the glitches to have disappeared. These observations steered me back toward my original (less plausible?) supposition -- the the 74123 one-shot in the Unibus timeout logic in the CPU was flaky, and particularly sensitive for some reason to SSYN pulses of 568ns. Adding some extra bus length via a BC11 or the M9202 moves the timing by a nano or two off the troublesome period, and reduces the magnitude of the glitches.

So I went ahead and clipped out the suspect 74123, and put in a socket and a fresh part. Bingo! Timeout glitching was eliminated entirely. Here's a trace after the 74123 was replaced. This trace looks different because with the timeout glitch fixed, I could no longer use it to trigger the scope -- instead I had to trigger on the trailing edge of SSYN, so we see both RKCS and non-RKCS bus cycles. In any case, the timeout glitching is now gone:

So that's a nice result -- I think the new scope is going to be pretty useful. The rather extreme sawtooth on the falling edge of SSYN on RKCS accesses still looks pretty bad to me, even though it is no longer triggering timeouts. I might try swapping out the M105 address decoder on the RK11, which generates this signal, and see if the integrity here is improved. All for now!