PDP-11 /04 and /34 CPU tryouts: FAIL and FAIL!

[intabits]

Having previously tested or repaired the power supplies, console interfaces and memory boards of my PDP-11/04 and 11/34, it was time to try inserting the CPU boards into them.

Just before powering it on I noticed something on the 11/04's CPU board (M7263) that I had stupidly neglected to look for much earlier.
Some general corrosion, but especially bad around E32, where it appears that a trace may have been corroded away.



But after cleaning it up, and consulting the schematics, I'm 95% sure that it was a factory trace cut, and therefore the corrosion was not quite as bad as previously feared, so I decided to power it up.

The Run light was stuck on, and I'd read that this could be the result of the NPG or BGx grant chains being incomplete. While checking the NPG chain, I found a small broken part of the backplane
connector.



Unfortunately, it appears this break may have allowed -15v into the NPG out pin of the CPU.
So that put a halt to further testing of the 11/04 for now.

Turning my attention to the 11/34, I just plugged the power supply into the mains socket, which resulted in sparks and melted power plug pins. So there was an obvious short circuit in the power
supply mains input. There were no issues last time it was used.



So ended my ambitious attempt to quickly get two working PDP-11's on the same day.

I'm also wondering if the far more experienced PDP-11 people can tell me if an '04 or '34 CPU should be operable from the console with just the following: M7263/M8266+M8265, M7847, M7859 and M9302 (and G727A grant cards)?
Or is the M9301/M9312 (or anything else) also required?

Video:-

[radiogeek381]

My PDP11 04/34/45/55 processor handbook says that the 16K MOS memory configuration looks like this:

1. Rows 1 and 2 slots A-F CPU
2. Row 3 Slots AB -- M9301.   Slots C-F a "QUAD SPC" -- so a simple controller
3. Row 4 Slots AB -- M7850   Slots C-F quad SPC
4. Row 5 Slots A-F MS11-F or MS11-J
5. Rows 6, 7, 8 are for HEX SPCs
6. Row 9 Slots AB -- M9302.  Slots C-F QuadSPC

I don't remember the rules for continuity grant cards.  And frankly, all this is pretty fuzzy for me.

[intabits]

Thanks, but I'm trying to do this incrementally.
So I'm looking for the *minimum* configuration that will enable some operation of the CPU from the console, and would like to know if the board complement I outlined above is sufficient.

[radiogeek381]

The figure hinted that this *was* the minimal configuration -- alternative was with core memory.

However, there is no need for the SPCs in 3:C-F, 4:C-F and 6,7,8

Whatever the minimum is, it will have the two processor cards, a memory, and something to ensure grant continuity/termination and interrupt termination, won't it?

The 9301 looks like it is required.   -- among other things it provides termination and likely provides the minimal boot code.

From http://www.bitsavers.org/www.computer.museum.uq.edu.au/pdf/EK-11034-UG-001%20PDP-11-34%20System%20User%27s%20Manual.pdf  Section 1.2 the minimal configuration appears to be

* CPU KD11-E or KD11-EA
* CONSOLE KY11-LA or LB
* Bootstrap/Terminator M9301
* MS11 or MM11(core) memory
* M8264 SACK timeout module (?!)
* M7850 Parity controller.

[intabits]

(sorry, been busy trying to connect a PC to my hot water service)

Yes, but I guess that they mean a minimum usable system, which would nearly always include a M9301 or M9312 bootstrap card (my /04 and /34 have one of each).

My expectation is that it should be possible to operate the CPU by toggling in simple programs from the console, and so not need a bootstrap card, and I've seen nothing yet that definitively states that is or is not possible.

This termination business is also confusing. I'm sure I've read (but not sure where) that the CPU provides termination at its end of the bus. So with an M9302 at the other end, I don't see why a M9301/M9312 would also be required for termination.

And yes, the M8264 is surprising (never heard of it till now) because the M9302 already provides SACK turnaround. That list you referenced also includes the M9302, so listing both seems even stranger.

I also doubt that an M7850 is mandatory - my two systems don't have one.

It would be nice to hear from people with working systems that could confirm any of this.

When I fix the fault in the /34 power supply and try the CPU, maybe I'll know a little more.

[radiogeek381]

40 years can do a number on one's memory.

Are you sure that the CPU terminated the bus?  This would surprise me for two reasons:

1. UNIBUS was not series terminated -- the terminators were typically something like
140/370 ohm (140 down, 370 up?) parallel terminations.  It is not a good idea to place
parallel terminations near the source.  (They only kill the reflection back at the source.)
2. There were lots of unibus extenders and "long" unibus chains in customer land.  That's
why the terminator module (with and without ROMs) was in the catalog.  I seem to recall
UNIBUS extenders that were longer than 12 feet (!).

But then again, I don't have a detailed understanding.  It may be that this could be answered
by someone even older than me.?

So -- any 11 geezers out there?  (You can tell them by the scars on the back of their hands.)

[intabits]

It seems you are correct, and I was laboring under a misapphrehension, so my apologies for causing you to doubt your memory and for causing general confusion.

I have spent quite some time trying to find the statement that indicated that the processor board(s) contains a terminator, but could not find it. I did find something that if read carelessly, might give that impression, so perhaps that's what happened (though my recollection was of something more definitive).

And unsurprisingly, there are no terminators in the M7623 CPU schematics.

Not related to 11/04 and 11/35, but it seems that the 11/10 & 11/20 (and possibly others?) actually do have terminators in the CPU.

While trying to understand how the processor should power-up into the halted state that I had expected, the descriptions of the M9301 lead me to suspect that the M9301 might be required to achieve this (with the proper wiring and switch settings). So it might be worth trying the 11/04 again with the M9301 installed, and hoping that the CPU hasn't been damaged by -15v getting into the wrong place.

I haven't yet looked into why the 11/34 power supply has become a short circuit on the mains.

[radiogeek381]

re: power supply shorting the mains --

During the 70's (and certainly after that) many power supplies had an integral decoupling/rf-supression/transient-sink network across the power line.  (CORCOM was a frequently used brand in the things I worked on.). With time, many of these have failed.  Of course, failing short seems to be the failure that manifests itself most clearly.

Does the 11/34 power supply have a suppression network across the input?  If so, it may need "replacing" (where "air" might be an alternative component, as a temporary test configuration only.)

[radiogeek381]

follow up to my own follow up --

I couldn't find an H777 supply schematic but I did see schematics for the H720 and others.  At the time, the company was quite small and the power supplies were likely designed by a team that worked on lots of different products. 

The H720 had caps from live and neutral to ground on both the line inputs and the line output connectors. 

These may have been packaged in a can around the connectors.  I'll bet they smell bad.

matt

[intabits]

Thanks, I'll be looking for stuff like that (and a loose screw, as this happened while the computer was on it's side. When previously powered on with no problems, it was sitting flat)

The H777 power supply is well covered in the available documentation:-

Theory of operation:-
http://bitsavers.org/pdf/dec/unibus/EK-BA11L-TM-001_Oct77.pdf

And the schematics are buried in the engineering drawings (p104)
http://www.bitsavers.org/pdf/dec/pdp11/1104/MP00019_1104_EngrDrws_Feb78.pdf

I haven't opened it up yet, but I'm expecting an easy fix, as there's not much to go wrong right at the input. (famous last words?)

[radiogeek381]

The drawing for A/C input assembly shows the EMI filter in the path to the primary.

Is it tripping the breaker or blowing the fuse?

The fuse is only in the path when the 5V regulator isn't on.  (When 5V is active, it closes a relay contact across the fuse.). Putting it there protects some parts of the 5V regulator, as if the output of the regulator isn't good, then the power path (through the transformer) to the regulator input stage will be current limited (to about 100 watts -- so not sure how much protection that's going to offer...

The breaker is rated at 10A which seems huge today, but back then ... 

[intabits]

The fuse didn't blow, but then maybe I removed the plug quickly enough. But the current limiter test indicates that the supply's input is practically a dead short.

The breaker was open the whole time, so no power was even getting to the regulators (probably - haven't verified that in the schematics yet)

Rather than speculating on the problem beforehand, I'll wait until I get around to opening it up, and start troubleshooting from there.

[intabits]

As expected, the PDP-11/34 power supply problem was easily found.

A bad EMI filter, for which I was able to find and obtain a suitable replacement, for a good price, the same day, on EBay!

Also uncovered, an error in the DEC documentation of the PSU's input section.

After the repair, I notice a label that casts doubt on whether this is really the 11/34, and that it may have been mixed up with my 11/04.

So before plugging in the CPU boards, I'll first have to verify that the correct backplane is fitted in this machine...

[intabits]

Having fixed the power supply short circuit, I've now investigated the labels that caused suspicion of a mix up of parts between the 11/04 and the 11/34 (there wasn't).

So finally I could try out the CPU in the PDP11/34, but no joy.
I tried a few variations, but again, no success.

These things have been taking up too much of my time, so I'm putting them aside for now.

I have a lot of other interesting junk to show, so future videos about those will mostly be just to show them, without attempting to power them on, as that takes too much preparation and research.

[RJSV]

40 years ago   I was thinking 60, the Qantel biz computers looked very very similar 'Pizza Square' circuit board, with 1 IC per square foot. And the board venders... They wanted cash-money to make and deliver those empty boards, to the Qantel factory.
  But real thing on my mind:
  Do you have the vintage enclosures, including printers ? Maybe it was teletype, first, after 1970 the players in mini-computers had 'screen text' with rudimentary discrete logic processors.
   I'd sure like to see the metal cabinets, as I have interest in 'presentation' dynamics.
-RJHayward