• Please review our updated Terms and Rules here

Restoring a DEC PDP-11/05

If I recall correctly reversing the cable gives all LED on. Since there isn’t power in the cable it isn’t that dangerous.

Ok so PP is always 0. That could explain why you have all LEDs off. What a out the clock signal at pin CC and the 74193 counters? Is there a clock and are the two 74193 counting correctly?
 
Here is what I have for pin PP with the ribbon cables installed between the front panel:

View attachment 1243980

I was also thinking about the cables this weekend, it occurred to me that the 40-pin Berg plug does not appear to be 'polarized' and can be inserted both ways. Drawing BCO08R, which shows the pin orientation references markings on the shield side of the cable but does not specify if the shield should be on the side facing out the front or back. I didn't want to try reversing the cable for fear of messing up something, but it's just a little thought that popped into my head. (I'm notorious for spending a lot of time troubleshooting only to find the issue was something very simple)
The principle we've come to know as "Murphy's Law" is closely related to unkeyed cables. The "real" Murphy is reputed to have said "If something can be installed incorrectly, eventually it will be."

 
If I recall correctly reversing the cable gives all LED on. Since there isn’t power in the cable it isn’t that dangerous.

Ok so PP is always 0. That could explain why you have all LEDs off. What a out the clock signal at pin CC and the 74193 counters? Is there a clock and are the two 74193 counting correctly?
Looks like you are correct. All LEDs, including the RUN LED illuminate when I reverse the cable.
Nothing on pin CC, though. Here is what I found on the 74193s, known as E6 and E8:

Clock input (pin 5) on E6:

E6 pin 5.PNG

Clock input (pin 5) on E8:
E8 pin 5.PNG
 
Ok. Then the problem is not the display. You need to get a board extender to raise the control board so that you can probe it. Douglas Electronis used to have suitable modern extender boards. I am not sure if still is the case.
 
Ok. Then the problem is not the display. You need to get a board extender to raise the control board so that you can probe it. Douglas Electronis used to have suitable modern extender boards. I am not sure if still is the case.
I checked the website and they offer a variety of PCB extender boards. The test equipment and tools section of the manual calls for three W984A dual-width extenders. Douglas Electronics offers two types of dual-width extenders, one of which is 8 inches high and the other is 5 inches high.

Just out of curiosity, as I am fairly new to working with Unibus PDPs, is there a 'most probable' cause for why the front panel wouldn't be receiving a clock from the CPU? (This is also directed towards anyone else who may have experienced similar issues. I'm not saying that it would be the exact same problem, but it's nice to have an idea as to where to check first)

I did some reading in the manual and it seems that the clock and microprogram functions come from the 7261, while the front panel itself is plugged into the 7260, which is where any data-path-related logic takes place. I almost thought that the CPU backplane could be removed from the case and still be connected to the power supply, which would at least allow access to the 7260, but with all the wire wrap pins, I thought it wouldn't be worth the risk...
 
You are right. The front panel is connected to the datapath. There is a 74150 that selects a bit to serialize over the connection to the frontpanel. It could be just that one that has a problem. But it is not very likely.

I think the problem is in the control circuitry since nothing at all happens when you toggle the switches and it shows RUN while if fact it should be halted.

At this point I think you should start look at the M7261. Is the cpu clock ticking? Does anything happen in the microprogam. A potential help here is a KM11 board which is plugged into the backplane. The KM11 is a board with four switches and a bunch of lamps. It is possible to single step the micro code and see what microinstruction it executes. There are replica KM11 designed by some people. http://gunkies.org/wiki/KM11-A_Maintenance_Set

I used a KM11 and an oscilloscope to fault find a 11/05. It took a while but I found the faulty 74181 after a while.

image0.jpg
 
I went with one dual width and one quad width extender, both from Douglas. The extenders can be a little finicky to line up when working with hex width boards. So having to deal with two instead of three is a little easier.

20200202_085559 small.jpg


The edge connectors on the Douglas boards are not ideal, but they work well enough to get the job done.

While you're waiting for the extenders, I would check the operation the main CPU clock, plus any downstream circuits you can access. The main generation circuit is on the M7261 board, which should be accessible with only the CPU boards in the system. This circuit is shown on page CONJ of the engineering drawings

Bear in mind that there are two revisions of the KD11-B CPU board set, as described here: http://gunkies.org/wiki/KD11-B_CPU. I have the earlier set, which is documented in the GT40 engineering drawings (DEC-11-HGTEA-A-D, "GT40 Graphic Terminal Engineering Drawings").
 
Looks like Mattis is a step ahead of me (which isn't surprising)!
 
I went with one dual width and one quad width extender, both from Douglas. The extenders can be a little finicky to line up when working with hex width boards. So having to deal with two instead of three is a little easier.

View attachment 1244059


The edge connectors on the Douglas boards are not ideal, but they work well enough to get the job done.

While you're waiting for the extenders, I would check the operation the main CPU clock, plus any downstream circuits you can access. The main generation circuit is on the M7261 board, which should be accessible with only the CPU boards in the system. This circuit is shown on page CONJ of the engineering drawings

Bear in mind that there are two revisions of the KD11-B CPU board set, as described here: http://gunkies.org/wiki/KD11-B_CPU. I have the earlier set, which is documented in the GT40 engineering drawings (DEC-11-HGTEA-A-D, "GT40 Graphic Terminal Engineering Drawings").

I took a look at the CONJ print today. If I'm reading it correctly the clock outputs come from 74H40 ICs E10, E64, and E84. I took the same exact measurements on all 3. Ch1 is on pin 6 (Clock L), and Ch2 is on pin 8 (Clock H):

7261 E84_64_10 pin 6.PNG
 
Looks like there i no clock. Now you need to figure out why.. First I think that you should check on the oscillator itself. At E39:8.

Then if there is nothing here there are multiple signals that could turn off the clock.

Screenshot 2022-07-27 at 07.04.44.png

Check all those and see if anyone has a level that blocks the clock generation. If not then the problem is E39 itself. Otherwise you have to follow the signals upstream to understand what is going on. As an example for why the clock is stopped is that the CPU is waiting for a Unibus transaction to complete. An MSYN has been issued and it is waiting for a SSYN.

Actually, I had a similar problem with stopped clocking. A constant that was used as the address of the switch register was wrong so the unibus access to the switch register never got an SSYN. The machine hanged just as the microprogram tried to access the switch register. Pressing start releases this hang. The problem was a bad 74181 in the data path.
 
BTW. Sometimes it can be useful just to probe with the oscilloscope and find signals that are sitting on levels that are not really valid TTL. An input that has a level around 2V means that there is no output driving it. That is normally not a good thing. Unless it is supposed to be unconnected or if there are three-state drivers involved. Luckily the PDP-11/05 has no three-state drivers in it. Another thing to watch out for is OC wire-oring. DEC used this quite frequently. It can be a bit confusing at times. In this case there are these current probes that might be helpful. I have no such thing and have never tried it.
 
It had occurred to me that my scope was having a grounding issue and I wasn't getting the best measurements. I fixed the issue and retook said measurements of the circuit shown on 'CONJ':

Starting with E75, pin 1,2,5 had no signal on them. Pin 4 (CH1) and Pin 6(CH2) did:
E75 pin 6, 4.PNG

E38, showing inputs on Pin 11(CH1) and Pin 12(CH2):
E38P11,12.PNG

E49, Pin 9(CH1), and Pin 10(CH2):
l
E49P8,9.PNG

E39, Pin 8(CH1):
E39P8.PNG


E57, I got the same measurements on Pins 11 and 12 as Pin 8 on E39.
E65, I have to go back and take a look at that one, I neglected to take notes on it...

E64, P1(CH1) P2(CH2):
E64P1,2.PNG

E64, P4,5(CH1) P6(CH2)
E64P4-5,6.PNG

E64 P8(CH1)
E84P8.PNG

E84, P1(CH1) P6(CH2)
E84P1-4,6.PNG

E84 P8(CH1)
E84P8.PNG
 
Per request, here are the readings on a 500 ns/div scale:

E75- Nothing on pins 1,2, and 5. Here's what was on pin 4:
E75P4.PNG

and on pin 6:
E75P6.PNG

E38- pin 11 on CH1; pin 12 on CH2; nothing on pin 13 out:
E38P11,12.PNG


E49- pin 9 on CH1; pin 10 on CH2:E49P9,10.PNG
pin 8:
E49P8.PNG

E57- pin 11 on CH1; pin 12 on CH2:
E57P11,12.PNG

pin 10:
E57P10.PNG

E65- pin 1 on CH1; pin 2 on CH2:
E65P1,2.PNG

pin 3 on CH1; pin 12 on CH2:
E65P3,P12.PNG
 

Attachments

  • E57P11,12.PNG
    E57P11,12.PNG
    61 KB · Views: 5
Right. Then there are clocks! check the signals on E10, E64 and E84. What is the state of CK OFF signal?

In my experience 7440 buffers are prone to fail.

Just curious, what is the bandwidth of your scope?
 
I'm using an Analog Discovery 2, by Digilent. 30MHz bandwidth. Here is what I found on E10, E64, and E84:

E10- Pin 1,2,4,5 on CH1, Pin 9,10,12,13 on CH2: (1,2,4,5 are shown tied together, as well as 9,10,12,13)
E10P1-5,9-13.PNG

Pin 6 on CH1 and Pin 8 on CH2:
E10P6,8.PNG

E84- Pin 1 on CH1 and pins 2,4,5 are tied together on Ch2:
E84P1,2-5.PNG

Pin 6 on CH1 and pin 8 on CH2:
E84P6,8.PNG

E64- Pin 1 on CH1 and 2 on CH2:
E64P1,2.PNG

Pin 4/5 (CONC CKOFF L) on CH1 vs. E75 pin 2 on CH2 (CONC CKOFF H):
E64P4-5 vs E75P2 CKOFF.PNG

Last but not least, pin 6 on CH1 and 8 on CH2:
E64P6,8.PNG

So there are clocks? I honestly expected to see a more square wave than what was seen. Might have to see if my probes are compensated properly. before the last post, I had an 'a-ha' moment when I realised that I had both probes on the 10x setting and not the 1x.
 

Attachments

  • E10P6,8.PNG
    E10P6,8.PNG
    71.9 KB · Views: 1
  • E10P6,8.PNG
    E10P6,8.PNG
    71.9 KB · Views: 2
  • E10P6,8.PNG
    E10P6,8.PNG
    71.9 KB · Views: 2
Okay. I suppose that’s good to hear. I currently do not have my schematics in front of me, but I recall microprograms and the like being on the same board. Also, before I place an order for the riser cards, is there anything on the backplane itself worth checking out?

That aside, I just want to give a big thank-you to Mattis and everyone else so far : )
 
Back
Top