PDP-11/34A bring up

[Tony Duell]

> 
> I am the new keeper of the PDP-11/34A that Jack Rubin rescued a while
> ago and wrote about here,
> 
> http://decpicted.blogspot.com/2010/01/pdp-1134a-data-systems-design-dsd-880.html
> 
> I took it on a road trip from Chicago back to St. Paul after VCFMW
> in September.
> 
> I've been doing a lot of cleanup on it and finally got to the point
> where I could power it on (just the CPU box) this weekend.
> 
> I think now I need to learn about Grant Continuity ;-)

I think you're right there :-)

> 
> There is a M9302 terminator installed in the last slot (left most when
> looking from the front of the machine) and also an M9312 in slot 4
> (amoungst the CPU and cache cards).
> 
> Two of the original boards are removed from the backplane...  the DSD
> 808830 controller and the DILOG DU130 tape controller.  They were in
> slots 12 and 13.

Almost certainly both of those would be DMA devices ('NPR' in DEC's 
terminologiy, meaning 'Non Processor Request') . As a result, the NPG 
(non provessor grant) jumper will have been removed from the backplane 
wiring on those 2 slots.

> 
> I then also have an RL11 on hand but it is not currently installed in
> the machine.
> 
> When I power up the machine, it immediately lights the RUN light on the
> KY11-B programmer's console.  No matter what I do from that console, I
> cannot get it to exit RUN or print anything to the serial terminal.
> 
> However, if I remove the M9302 terminator (a trick I found on some web
> page), then sure enough, I can HALT it, the RUN light goes out and I can
> do CTRL+BOOT and the serial terminal will spring to life with a register
> dump and the '@' prompt.
> 
> I'm pretty sure that my problem is the empty slots 12 and 13 where boards
> used to be and should now have Grant Continuity cards installed instead...
> but I am curious why pulling the M9302 makes it "work".  What is the
> mechanism at play there?

I would agree with all that. You've done the first test that I always 
suggest -- Pull the M9302 (and replace it with an M930 if you have one).

)OK, as to what';s going on. This is is basically what should heppen when 
a Bus Request (BR) or Non Procesor Requeest occurs : 

1) Periphral asserts the appropraite BR on NPR line

2) Procesosr (actually the arbiter if you're pedantic) asserts the 
associated Grant line, procided (in the case of an BR) the priority level 
is high enough. Note that the grant lines are not bussed along the Unibus 
slots. Instead, the grant lines from th processor/arbiter) go to the 
grant inputs on the first Unibus slot. The grant outputs of that slot go 
to the grant inuts of the next slot, and so on.

3) A device which is not issuing a request at that level will simply pass 
the grant on, and make it's grant outputs the same state as the grant 
inputs. That's what a grant contuity card does (in a trivial way by 
simply connecting the inputs ot the outputs)

4) The device that did issue the reuest does not pass the grant on (it 
keeps its grant outputs deasserted) so that no device s further from the 
processor can repsond ot the grant. It then asserts SACK (Select 
ACKnowlddge) on the Unibus to say that it's accepted the grant. I think 
at this point the arbiter seasserts the grant lines, the peripohertla has 
control of the Unibus. 

Now, normally, a grant should never get as far as the terminator at the 
far end of the Unibus. Because the grant was issued in response to  a 
requexst from a device, and that device should intercept the grant and 
not pass it on. 

However, IIRC there can be a condition where things don't quite 
synchornise properly. Under those conditions, hte grant is not 
intercepted, as a result, the SACK line is never asserted and I think the 
machine hacks. So the M9302 terminator has a little bit of logic to 
assert SACK if it sees a grant. 

But that birings in aother problem. The grant lines, unlike the other 
Unibus signals, are active high. So if the grant chain is broken at some 
point, th grant inputs of all devices after th break will seem to be 
asserted. Those grants will get to the terminator, which asserts SACK, 
and the system hcngs for another reason. 

That's waht you're seeing. 

> 
> I also suspect that I may have to look at the backplane wiring for slots
> 12 and 13 to put back whatever DMA jumpering might have been modified for
> the two cards that used to be there--  or, at least for one of them as
> I can probably put the RL11 into one of those slots and it requires DMA.

It's a little wire-wrapped jumper between pins CA1 and CB1 of the Unibus 
slot. That is, the rearmsot 2 contacts on the right hand side (compoennt 
side of the PCB) on the 3rd edge conenctor from the rear of the machine.

Most of the grant contiuity cards used on 11/34s are the little almost 
square ones which go in connector D. Those connect up the 4 bus grant 
lines, but not the NPG line. There is a dual-height continuity card which 
goies into C nad D and completes all grants, but they're not as common.

A trick I've usef for testing is to cut a little strip of single-sided 
copper clad boardthat will go ito the C connector and connect pins CA 1 
and CB1 only. 

When I fit NPG jumpers by wire-srapping them on the backplane, I use red 
or orange wire-wrp wire ot make them easier to spot in the future.

And yes the RL11 is a DMA device, you must remove the NPG jumper from the 
slot where it's fitted

-tony