PDP-11/34a Cleanup and Repair

[glitch]

You can get the full bootable TU58 XXDP image with all my diags on it here: https://ak6dn.github.io/PDP-11/TU58/ as 11XX_9.DSK

Yup, already have it pulled down -- thanks a million for providing prebuilt TU58 images, that has saved me so much time and trouble vs. making custom ones either with PUTR or booting XXDP on E11 and initializing blank TU58 images (last time I did it, SIMH didn't support writing to TU58, I don't know if that has changed)

 

[glitch]

Still not 100% stable, I had a random lockup during extended runs of diagnostics this evening. Probably need more backplane cleaning: I had several lockups over the day, and sometimes a failure of the programmer's console to grab the Unibus afterwards, until I reseated boards. I suspect it's oxidization from storage, the RL01s that were paired with this machine had a lot of it, to the point that I had to De-Oxit some of the front panel *lamp* sockets :/

 

[AK6DN]

Still not 100% stable, I had a random lockup during extended runs of diagnostics this evening. Probably need more backplane cleaning: I had several lockups over the day, and sometimes a failure of the programmer's console to grab the Unibus afterwards, until I reseated boards. I suspect it's oxidization from storage, the RL01s that were paired with this machine had a lot of it, to the point that I had to De-Oxit some of the front panel *lamp* sockets :/

Highly likely oxidation of the board/backplane connectors. I have a very similar 11/34A setup, BA11-K box processor and one expansion DD11 9slot backplane.
Every now and then the DL11-W that is configured for TU58 fails. Or the console fails. Or the inter-backplane jumper fails with a hung bus.
So pull out cards, use deoxit wipes on the connector fingers. Or sometimes just extract/insert a card. Starts working again for many, many months.
I typically power it on once every couple of weeks and run basic diagnostics. So far has survived the last ten years with basic biweekly testing.

 

[glitch]

Highly likely oxidation of the board/backplane connectors. I have a very similar 11/34A setup, BA11-K box processor and one expansion DD11 9slot backplane.
Every now and then the DL11-W that is configured for TU58 fails. Or the console fails. Or the inter-backplane jumper fails with a hung bus.
So pull out cards, use deoxit wipes on the connector fingers. Or sometimes just extract/insert a card. Starts working again for many, many months.
I typically power it on once every couple of weeks and run basic diagnostics. So far has survived the last ten years with basic biweekly testing.

I was thinking about flushing the backplane with MG Chemicals Electrosolve -- I've used it on other backplanes that I couldn't wash, with good results, but none of those were wire wrap backplanes. Do you think I'm better off doing that, or applying De-Oxit sparingly into the connectors? Thinking of maybe saturating some heavy cardstock with De-Oxit and plunging it into the backplane connectors, I really wouldn't want it dripping down into the wire wrap and becoming a persistent mess/dust catcher.

 

[bigfix]

In my times as a DEC fieldservice guy we washed backplanes with lots of freon spray - now this is forbidden...

We only used goldwipes on the module fingers, never sprayed it into the contacts
Together with dust this may create a sticky mess over time

Goldwipes were polyphenyl ether dissolved in trichloroethane in the past - not sure what is used today

We just inserted the "goldwiped" modules multiple times, to get the lubrication transfered
Soaked cardboard may work - I never tried this
Lubrication is important to reduce wear and seal the surface against environmental corrosion

We also found that backplanes with black connectors had much more contact issues than those with green blocks

I remember that the worst backplanes were those on 11/70 MK11 boxes, you could clean those forever until they worked reliable

My test was usually to run a system test like EQKC on 11/70, put a small AM radio nearby and knock all the modules with a screwdriver handle
Or even better position a flat screwdriver on the metal handle "knobs" and hit it with another handle - this showed all contact flaws
You hear immediate when the test failed while knocking

 

[glitch]

We also found that backplanes with black connectors had much more contact issues than those with green blocks

Good to know -- I assume you mean the dark green vs. black-black blocks, and not the later light green blocks seen in e.g. third-party stuff (like ELCO green). Mine are dark green, tin finish on the fingers, except the middle slot, which is gold finish (9-slot backplane).

My test was usually to run a system test like EQKC on 11/70, put a small AM radio nearby and knock all the modules with a screwdriver handle
Or even better position a flat screwdriver on the metal handle "knobs" and hit it with another handle - this showed all contact flaws
You hear immediate when the test failed while knocking

Hah, that's an awesome suggestion! I will have to try that!

So I cleaned all edge connectors today, first with Q-tips and alcohol, then sprayed some De-Oxit on a lint free wiper and gave all the edge fingers a liberal application. Everything inserted *much* easier, of course. I did find that the DL11-W had some nasty crud on the contact fingers that required burnishing with the #1007 Singlex eraser and re-cleaning with alcohol.

Unfortunately, after that, the system won't come up -- I suspect really that the failure happened at the end of testing last night, when it mysteriously trapped to 4 during a memory exerciser run (after 20+ passes without issue) and a partial printout about CSR access. It was stopped with "777560" on the display, which is of course console SLU. Today, after the cleaning, when I turn on DC power, the RUN light flashes on, then goes out, and the machine halts with "040000" on the programmer's console display.

I pulled the system down to a minimal configuration: CPU, bootstrap/terminator, programmer's console, one RAM board, bunch of green handle dual height NPG grant cards, the DL11-W, and a final Unibus terminator. Same symptoms. Swapped every item I had spares of, which is everything except the two CPU boards, no change. Pulled everything and cleaned the backplane connectors with cardboard saturated in alcohol (removed a lot of gray debris!), then relubricated with De-Oxit on cardboard. Still no change. I can deposit to the console SLU and get characters on the terminal. I can deposit to memory. The Unibus is not hung. If I deposit to nonexistent memory (e.g. past 777566) I get the expected bus fault. Checked all three power rails, all are within spec. Checked ACLO and DCLO, both are logic high.

I toggled in the usual "jmp 0" and "br ." tests, the processor stays running (or at least RUN lamp on). I toggled in the memory sizer, and it never seems to complete. It doesn't trash registers or the program at 001000.

I am pretty sure this is a CPU board set fault: there have already been two failed ICs (a 7400 back in 2017, and the failed 74174 yesterday). Does anyone have pointers? I traded my spare 11/34a set to conmega a few years ago to fill an empty 11/04 chassis, so I can't just swap.

 

[glitch]

Update: I thought to re-enter the memory sizing program, manually set up the PC, and try to single-step it. Looks like I have a stuck bit in the program counter.

 

[glitch]

Looks like we're narrowed down to the scratchpad itself, the associated ALU, or maybe the output mux. All processor registers that are stored in the SPM have bit 12 stuck. The PSW does not, which is implemented in separate latches, but flows through the same output data path to the Unibus, so we know the output path past the muxes is fine. Thinking I will desolder + socket two of the DM85S68 register files and swap them, to see if the fault moves with the chip.

 

[glitch]

Socketed and swapped E39 and E29, stuck bit follows register file; therefore, bad register file! Spares ordered.

 

[bigfix]

Good to know -- I assume you mean the dark green vs. black-black blocks, and not the later light green blocks seen in e.g. third-party stuff (like ELCO green). Mine are dark green, tin finish on the fingers, except the middle slot, which is gold finish (9-slot backplane).

Yes, the dark green ones
Some people said that the tin finish contacts have a tiny gold contact point - but I did never see this and believe it is only a story...

Also we were told that the goldplated contacts are only good for about 50 insertion cycles - then the gold is worn off
I guess with lubrication this could last much longer

 

[glitch]

Yes, the dark green ones
Some people said that the tin finish contacts have a tiny gold contact point - but I did never see this and believe it is only a story...

Heh, yeah, got into it with a guy about that years ago, took pictures under the microscope of a destroyed connector I pulled some pins out of. Didn't change any minds "It's worn off from heavy use...or you removed it to try and make a point"

Also we were told that the goldplated contacts are only good for about 50 insertion cycles - then the gold is worn off
I guess with lubrication this could last much longer

I could see that, with the contact force in DEC backplane connectors. Contact lubrication definitely helps.

 

[thunter0512]

I was thinking about flushing the backplane with MG Chemicals Electrosolve -- I've used it on other backplanes that I couldn't wash, with good results, but none of those were wire wrap backplanes. Do you think I'm better off doing that, or applying De-Oxit sparingly into the connectors? Thinking of maybe saturating some heavy cardstock with De-Oxit and plunging it into the backplane connectors, I really wouldn't want it dripping down into the wire wrap and becoming a persistent mess/dust catcher.

To clean the backplanes in my machines (PDP-8/E, PDP-11/34 and PDP-11/05) I first used compressed air to blow out any loose debris and dust and then soaked them in denatured alcohol (Metylated Spirit in Australia) for about 20 minutes while occasionally swirling the backplane in the liquid. I finished off by removing the backplane from the bath and blowing out all the alcohol from the backplane slots and exterior.

This treatment seems to nicely clean out any remaining debris and previously applied lubricants.

I would never spray any lubricant into the backplane slots as that will just combine with inevitable dust ingress into sticky goo.

Caig Labs ProGold applied in a very thin layer to the gold PCB edge connectors seems to work well. I first wipe it clean with denatured alcohol and after drying use the supplied brush to apply a thin layer of the ProGold on the gold fingers and spread and wipe any excess with Q-tips. This makes PCB insertion and removal noticeably easier and reduces wear on both contact surfaces.

The only drawback of ProGold is that it is rather expensive in Australia. On the other hand the small bottle I bought a few years ago is still half full and I used it on many PCBs. Quite likely there are much cheaper alternatives for the lubricating ingredient if one knows what to look for.

As to card stock soaked in De-Oxit - I haven't tried it but I would be worried that it will damage the backplane contacts and that it would leave excess dust-catching liquid behind.

 

[Terri Kennedy]

Yes, the dark green ones
Some people said that the tin finish contacts have a tiny gold contact point - but I did never see this and believe it is only a story...

Also we were told that the goldplated contacts are only good for about 50 insertion cycles - then the gold is worn off
I guess with lubrication this could last much longer

I wonder if, "back in the day" DEC Manufacturing had extender cards with ZIP-type sockets for final test. Given the profits in that era, it might have made more sense for them to just pick up a new DD11 backplane from the storeroom when one started wearing out. That could get expensive for things with dedicated backplanes (RK611, DH11) and would have been nuts for things like the 11/70 backplane. Although "stuff it full of boards and see if it runs" seems like a horrible way to do first test on an 11/70 processor card set just out of board assembly.

 

[AK6DN]

I wonder if, "back in the day" DEC Manufacturing had extender cards with ZIP-type sockets for final test. Given the profits in that era, it might have made more sense for them to just pick up a new DD11 backplane from the storeroom when one started wearing out. That could get expensive for things with dedicated backplanes (RK611, DH11) and would have been nuts for things like the 11/70 backplane. Although "stuff it full of boards and see if it runs" seems like a horrible way to do first test on an 11/70 processor card set just out of board assembly.

The manufacturing flow of DEC was to produce modules/boards that could and would be tested/validated on GenRad test systems, and modules/boards would be put into stock that passed these tests 100%.
The GenRad tester could access any signals that came to the edge fingers or offboard connectors and could drive/sample them as needed for each individual board test program.

When it came to final assembly/test for an actual system (be it an 11/05, 11/34, 11/40, 11/45, 11/60, 11/70 whatever) those systems were then built up using 100% tested good boards from inventory stock.
Stuffing a large backplane full of untested boards was never used as a strategy (even back in the modular CPU days of the PDP-8i PDP-8L etc) systems were built with pre tested modules.

From a GenRad catalog from 1973 (note the PDP-8e as the controller). GenRad was in Concord, MA, just nextdoor to DEC in Maynard, MA.

 

[pbirkel@gmail.com]

From a GenRad catalog from 1973 (note the PDP-8e as the controller). GenRad was in Concord, MA, just nextdoor to DEC in Maynard, MA.

High-speed paper tape! "Disk Memory (optional): A 2315 removable disk pack."

 

[glitch]

Always thought those Gen-Rad ATE systems were super cool! Don't know what I'd do with one

I super appreciate the "was actually there" field service and engineering advice, stories, best practices, etc. That is well before my time, and I always get the impression it's a lost experience nowadays -- companies just don't run like that anymore.

To clean the backplanes in my machines (PDP-8/E, PDP-11/34 and PDP-11/05) I first used compressed air to blow out any loose debris and dust and then soaked them in denatured alcohol (Metylated Spirit in Australia) for about 20 minutes while occasionally swirling the backplane in the liquid. I finished off by removing the backplane from the bath and blowing out all the alcohol from the backplane slots and exterior.

I'd be concerned about the makeup of the denatured alcohol -- here, it's kind of a manufacturer-randomized cocktail of volatile solvents. There's often MEK included. It cleans stuff great, but I'd be concerned about damaging wire insulation, plastics, etc. with an extended soak. We sometimes use it to remove heinous amounts of flux from old PCBs, and have had it affect plastic components sometimes.

I wonder if perc wouldn't be a better saturation solvent? Here, it's available cheaply as CRC Brakleen in the red can (literally just pressurized perc). It's cheaper than buying from chemical supply houses. I don't know if it's restricted in .au though. I'm getting ready to remove some connectors from a very damaged scrap backplane (comms multiplexer, bare backplane came in a pile of scrap), if I remember I'll hose it down with Brakleen and see if it affects the wrapping wire insulation or connector plastic. We haven't had any issues with cleaning boards with it, but I've never used it on a wire wrap assembly!

I would never spray any lubricant into the backplane slots as that will just combine with inevitable dust ingress into sticky goo.

Yeah that was my main concern. Excessive oily stuff plus super high airflow seemed like a recipe for future issues!

Caig Labs ProGold applied in a very thin layer to the gold PCB edge connectors seems to work well. I first wipe it clean with denatured alcohol and after drying use the supplied brush to apply a thin layer of the ProGold on the gold fingers and spread and wipe any excess with Q-tips. This makes PCB insertion and removal noticeably easier and reduces wear on both contact surfaces.

I'll look into these, thanks! Currently I am battling the idiot "smart straw" thing all aerosol manufacturers seem to want to do now: my cans of De-Oxit leak as much from the valve/straw assembly as comes out the tube or spray nozzle, so I wrap a lint-free wiper around the top while I'm using it, and get "free" De-Oxit wipes. I ought to buy a big bag of normal aerosol tops and see if I can swap it for a normal one with detached tube.

As to card stock soaked in De-Oxit - I haven't tried it but I would be worried that it will damage the backplane contacts and that it would leave excess dust-catching liquid behind.

You have to choose narrow enough cardstock that it won't force the contacts past their normal travel, of course. It leaves very little chemical on the contacts, and the abrasiveness of the cardstock removes stubborn contamination. It comes out with dark gray lines where all the contacts scrub against it, especially in the first pass alcohol phase.

 

[m_thompson]

You have to choose narrow enough cardstock that it won't force the contacts past their normal travel, of course. It leaves very little chemical on the contacts, and the abrasiveness of the cardstock removes stubborn contamination. It comes out with dark gray lines where all the contacts scrub against it, especially in the first pass alcohol phase.

I use the packaging cardboard from men's dress shirts. It is usually the right thickness and has one side that is white and a little abrasive, so it shows the grey lines of oxidation that is being removed.

 

[glitch]

Got back from VCF Midwest yesterday evening, this morning the replacement register file ICs were waiting at the shop. The PDP-11/34a is back up and running confidence tests under XXDP!

 

[glitch]

PDP-11/34a continues to run well! I need to pull down the RL02 and check in on an issue...I am getting a squeak when heads load, and sometimes when seeking. No marks on the platter, I think it's actually the carriage bearings squeaking as the servo keeps them on track, but I want to be *absolutely* sure.

I started work today on my RK05s and controllers, with the controllers coming first. I'm going to be using the PDP-11/34a for initial testing. During RK11-D controller checkout, I discovered that the Unibus Out connection from the system backplane had high resistance on data bit 6! I had to polish it with crocus cloth to get a good connection -- remember, this has been cleaned at least twice within the last two weeks!

 

[glitch]

Burned a MSCP boot PROM for the M9312 from the archive @AK6DN has. The 11/34a's M9312 was full of boot PROMs, so I pulled the RK06/07 one, since I don't have any. The other boot PROMs are:

• 23-765A9 TU58 boot ROM
• 23-756A9 RK03/05 and TU56 boot ROM
• 23-751A9 RL01/02 boot ROM

The UC18 started, but it was so slow I thought it wasn't working! It must've taken 10 minutes to boot XXDP from a Bernoulli 90 cartridge. I ran the Emulex F.R.D. and told it to autoconfigure, and now it's booting in a much more reasonable amount of time! Maybe a minute to fully boot XXDP, which is pretty much in line with the UC07 on the PDP-11/23.