• Please review our updated Terms and Rules here

PDP-11/34a Cleanup and Repair

glitch

Veteran Member
Joined
Feb 1, 2010
Messages
5,342
Location
Central VA
I picked up a PDP-11/34a after VCF East this year, it's one I actually worked on back in 2017 for its previous owner, who sold it to me as he wanted the space it required back for other machines. I of course jumped on the opportunity to buy it, since it'd be a much more capable Unibus machine than my PDP-11/10. I put it on one of the benches this week to clean it and get any parts ordered ahead of VCF MW, so that they're hopefully in by the time I'm back. I have some upcoming customer work that'll be easier to complete on the PDP-11/34a vs. the 11/10. Started with taking panels off and dusting out the cobwebs:



It's pretty much how I remember it being in 2017, just with a little more dust and spider houses. I will be going through the H765 power system again, and disconnected it first:



I'm going to add a fuse for the 5411086 board (the one in my PDP-11/10 shorted a bridge rectifier and blew traces off the board), replace some of the known problem capacitors (so long, Sangamo!), generally clean everything out, etc. The power distribution board had a lot of greasy, caked-on dust, so it got a hot soapy water bath and a trip through the drying cabinet. I'll test and repair the regulator modules out of chassis, I made harnesses for that when I worked on the PDP-11/10's power system.

The 9-slot expansion backplane has a fault in it, I didn't get it working in 2017 and just left it disconnected, since it wasn't needed at the time. I want to try and isolate the fault and repair it, though I've heard it's sometimes bad wire wrap pin plating or bad soldering from DEC, and that it's often a nightmare to find under the wire wrap jungle! Backplanes, from the bottom:



All of the boards will get a thorough cleaning and get some likely problem caps replaced -- I already found one epoxy sealed Sprague 8uF electrolytic that's starting to leak past the epoxy. The chassis itself will get a deep cleaning, since I don't expect to have it this far apart after it gets racked for a good long while. I may respray the metal programmer's panel surround, since it has little rust spots all over it. At the moment, the chassis is fully stripped down of unboltable parts:

 
OMG look at that transformer? Look at that wiring! They don't make them like that anymore.
 
I felt sad about having to replace the two white Sangamo capacitors in my 11/34A. They were so pretty.

They look nice, but they're pretty much universally junk now. You can feel the difference in weight between basically any Sangamo and someone else's equivalent part -- the Sangamos all feel empty now! I guess they used a more volatile electrolyte, or had poorer sealing or something. They must've been decent when new, or they wouldn't be in so much old stuff.

OMG look at that transformer? Look at that wiring! They don't make them like that anymore.

Indeed, quite inefficient :p At least it's relatively easy to work on!
 
They look nice, but they're pretty much universally junk now. You can feel the difference in weight between basically any Sangamo and someone else's equivalent part -- the Sangamos all feel empty now! I guess they used a more volatile electrolyte, or had poorer sealing or something. They must've been decent when new, or they wouldn't be in so much old stuff.
What did you use for their replacements? Have any before/after photos?
 
For the two small ones, I used modern radial low ESR caps we have in stock. Still have to find a replacement for the big guy.
 
For the two small ones, I used modern radial low ESR caps we have in stock. Still have to find a replacement for the big guy.
I found that all the large electrolytic caps are fine (LAB-8/e, PDP-8/e, NOVA 2, PDP-11/34 and PDP-11/05). These large (and expensive) caps can be reformed using a current limiting bench supply by slowly ramping up the voltage from 1V to the rated voltage in 1V steps. A current limit of 20 mA is a good starting point even for very large caps. After reforming these caps will be fine for a long time if you occasionally power on the machine they are in.

It is the small electrolytic caps that can leak or stop acting like a capacitor.
 
You're referring to an H744, right? Not the 5411086 (https://users.glitchwrks.com/~glitch/2019/06/05/pdp1110-psu-repair) where there's just the one moderately-sized axial.

Right, the H744. The H7441 had two leaking caps, but they were smaller radial caps, and the big electrolytic reformed fine.

I found that all the large electrolytic caps are fine (LAB-8/e, PDP-8/e, NOVA 2, PDP-11/34 and PDP-11/05). These large (and expensive) caps can be reformed using a current limiting bench supply by slowly ramping up the voltage from 1V to the rated voltage in 1V steps. A current limit of 20 mA is a good starting point even for very large caps. After reforming these caps will be fine for a long time if you occasionally power on the machine they are in.

It is the small electrolytic caps that can leak or stop acting like a capacitor.

Quite aware, I put a 100R 2W resistor in series with one of the bench supplies, set for 5-10% less than nameplate voltage rating, and monitor charge rate with a meter. I then discharge and monitor the rate, and repeat the charge cycle. As long as they behave reasonably and will hold up with no load, I keep using them!

Sangamo capacitors are a special case, their large can-type capacitors have some manufacturing flaw or something that results in them going dry when others of the same era don't. I've had to replace dozens, and now just replace without testing if it's a Sangamo. I had to recycle a huge NOS quantity of them a few years ago, they'd dried out just sitting on our stockroom shelves! 95% of our CDE, Sprague, MEPCO, etc. NOS capacitors are completely fine after a reforming, but not the Sangamos. Even if one did reform OK, I feel like it's not worth the trouble in something like the H765 power system, which is not exactly fun and easy to disassemble. Sure, they're expensive to replace, but way cheaper than blowing up something else because of a failure.
 
...
Quite aware, I put a 100R 2W resistor in series with one of the bench supplies, set for 5-10% less than nameplate voltage rating, and monitor charge rate with a meter. I then discharge and monitor the rate, and repeat the charge cycle. As long as they behave reasonably and will hold up with no load, I keep using them!

For one of the large 50000uF caps rated 40V when operated with a 100R 2W resistor you burn 16W when there is a short. It will light up your resistor, but more importantly it does permanently damage to the aluminium foils in the cap rather than giving it a chance to regenerate the oxide layer (or "heal") using very low current of say 20mA or even 10mA. Why not use the current limiting feature of your lab supply?
 
For one of the large 50000uF caps rated 40V when operated with a 100R 2W resistor you burn 16W when there is a short. It will light up your resistor, but more importantly it does permanently damage to the aluminium foils in the cap rather than giving it a chance to regenerate the oxide layer (or "heal") using very low current of say 20mA or even 10mA. Why not use the current limiting feature of your lab supply?

I do both. I have a safety limit set of approximately 500 mA on the supply (Lambda LPT-7202-FM). If it punches through with that combination it is a failed capacitor. I prefer quick failures on the bench to eventual failures in operation.

I also run at higher currents because I don't have time to wait all day for a capacitor to reform -- it would occupy a significant amount of bench space. We repair equipment like this as our main line of business.

I just bought the replacement for the Sangamo 31,000 uF @ 50V capacitor out of the H744 module, it was $35 shipped for a Sprague Powerlytic (never had one in good physical condition that didn't reform). If I have to spend $35 because every now and then I push a marginal capacitor to failure on the bench, I consider that money well spent!
 
Alright, power system back together! I stole a non-Sangamo capacitor for the H744 module out of another H744 module from a different system that needs repaired. It reformed fine. I plugged in grant cards for the SPC slots and brought the system up with a minimal configuration. Programmer's console wasn't operational, then I remembered this system is cabled wrong somewhere and the stripe goes toward the PSU! Came right up, examined memory, etc. Added a RL11 I bought years ago for my PDP-11/10 but never installed (no available slots at the moment). I'd run it through the controller diags at some point, either on one of RetroHacker_'s machines or when I had this 11/34a on my bench in 2017. I have the DL boot ROM in the M9302 bootstrap/terminator board, so I can boot directly. I was greeted with an unexpected sign-on from XXDP:

Code:
BOOTING UP XXDP-SM SMALL MONITOR


XXDP-SM SMALL MONITOR - XXDP V2.6
REVISION: E0
BOOTED FROM DL0
28KW OF MEMORY
UNIBUS SYSTEM

RESTART ADDRESS: 152010
TYPE "H" FOR HELP

.

...SM? Huh? Shouldn't I have the XM on a PDP-11/34a? I ran UPDAT to reboot and run the 11/34a memory management diags (FKTH??) and immediately got my answer:

Code:
*BOOT DL0:

FAILURE DETECTED IN MEMORY MANAGEMENT UNIT.
SYSTEM CANNOT COMPLETE BOOT PROCESS.
PLEASE USE XXDP SM BASED DIAGNOSTICS TO
DETERMINE CAUSE OF FAILURE



BOOTING UP XXDP-SM SMALL MONITOR


XXDP-SM SMALL MONITOR - XXDP V2.6
REVISION: E0
BOOTED FROM DL0
28KW OF MEMORY
UNIBUS SYSTEM

RESTART ADDRESS: 152010
TYPE "H" FOR HELP

.R FKTH??
FKTHB0.BIN

CFKTHB0  11/34 MEMORY MGMT. DIAG.

SR2 NOT TRACKING CORRECTLY
SR2 WAS EXPECTD TESTNO  ERRORPC
022034  022030  000012  022050


DID NOT LOCKUP CORRECT VIRTUAL ADDR.
SR2 WAS EXPECTD TESTNO  ERRORPC
022076  022072  000012  022124


DID NOT LOCKUP CORRECT VIRTUAL ADDR.
SR2 WAS EXPECTD TESTNO  ERRORPC

...and so forth. Looks like there's a bad bit in the SR2 register of the KT11 compatible MMU! This is a new development, my notes show that in 2017 the MMU diags, along with the rest of the CPU diags, were all running fine.
 
Looking like it has to be the SR2 latch at E50, or the KT mux at E67. Leaning toward the mux as I was able to run @AK6DN MEMX and ZMSD?? without error from XXDP-SM. ZMSD?? refuses to relocate, but MEMX says it's relocating and testing all installed memory.
 
That works! Onto the next weirdness. This is what I get when running ZMSD??

Code:
@DL0


BOOTING UP XXDP-XM EXTENDED MONITOR


XXDP-XM EXTENDED MONITOR - XXDP V2.5
REVISION: F0
BOOTED FROM DL0
064KW OF MEMORY
UNIBUS SYSTEM

RESTART ADDRESS: 152000
TYPE "H" FOR HELP !

.RUN ZMSD??
ZMSDD0.BIN

 CZMSDD - MS11L/M MEMORY DIAGNOSTIC
   NO CACHE AVAILABLE
    32K OF UNIBUS PARITY
    64K WORDS OF MEMORY TOTAL

                        MEMORY CONFIGURATION MAP
                             16K WORD BANKS
                1       2       3       4       5       6       7
        012345670123456701234567012345670123456701234567012345670123
ERRORS
CPU MAP 1111
INTRLV  ----
MEMTYPE PP
CSR     0000
PROTECT PP
RELOCATION NOT POSSIBLE
RELOCATION NOT POSSIBLE
END PASS #     1
RELOCATION NOT POSSIBLE
RELOCATION NOT POSSIBLE
END PASS #     2
RELOCATION NOT POSSIBLE
RELOCATION NOT POSSIBLE
END PASS #     3
RELOCATION NOT POSSIBLE
RELOCATION NOT POSSIBLE
END PASS #     4

Is this expected behavior? My 128KW of RAM is spread across two half-populated MS11-L boards (why is beyond me), and above I'm running with just one. No matter which one is in, or if both are in, I get only 32K of parity detected, and it tells me relocation is not possible. I've readdressed both and ran them alone, also moving them to other MUD slots. No change. If both are in, they show up on the same CSR, despite being strapped for different CSRs.

@AK6DN MEMX sees all of the RAM as parity, and correctly identifies both CSRs.
 
I bought an Emulex UC18 Unibus dual SCSI controller a month ago, one showed up for a really good price (for Unibus SCSI...it was still $500, which hurts!). This evening, I boot it for the first time:

Code:
BOOTING UP XXDP-XM EXTENDED MONITOR


XXDP-XM EXTENDED MONITOR - XXDP V2.5
REVISION: F0
BOOTED FROM DU0
124KW OF MEMORY
UNIBUS SYSTEM

RESTART ADDRESS: 152000
TYPE "H" FOR HELP !

.

Incredibly, it's slower than the UC07 on QBus with XXDP :p Had to turn the chassis over and remove a NPG wrap, I thought I already did on this slot, back in 2017, since it had a green dual-height grant card in it. I must not have, it had the factory yellow wrap on it (I always do blue or red to make them stand out).
 
I bought an Emulex UC18 Unibus dual SCSI controller a month ago, one showed up for a really good price (for Unibus SCSI...it was still $500, which hurts!). This evening, I boot it for the first time:

Code:
BOOTING UP XXDP-XM EXTENDED MONITOR


XXDP-XM EXTENDED MONITOR - XXDP V2.5
REVISION: F0
BOOTED FROM DU0
124KW OF MEMORY
UNIBUS SYSTEM

RESTART ADDRESS: 152000
TYPE "H" FOR HELP !

.

Incredibly, it's slower than the UC07 on QBus with XXDP :p Had to turn the chassis over and remove a NPG wrap, I thought I already did on this slot, back in 2017, since it had a green dual-height grant card in it. I must not have, it had the factory yellow wrap on it (I always do blue or red to make them stand out).

On my same XXDP disk that has memx.bin on it there is also an mscp.bic test that does simple testing of a the UC17/18 controller, and does read performance testing if there is any media attached.
So it will tell you how many blocks/second you are getting on reads.
 
On my same XXDP disk that has memx.bin on it there is also an mscp.bic test that does simple testing of a the UC17/18 controller, and does read performance testing if there is any media attached.
So it will tell you how many blocks/second you are getting on reads.

Awesome! I'll move it over to the system at some point and figure out how slow it is. I think the slowness under XXDP is due to how XXDP handles MSCP, which is compounded by how the Emulex SCSI boards handle controller resets (SCSI bus reset + runs onboard diags). I'll have to update my XXDP media/images...
 
Awesome! I'll move it over to the system at some point and figure out how slow it is. I think the slowness under XXDP is due to how XXDP handles MSCP, which is compounded by how the Emulex SCSI boards handle controller resets (SCSI bus reset + runs onboard diags). I'll have to update my XXDP media/images...
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
 
Back
Top