• Please review our updated Terms and Rules here

What did I do to my PDP-8 today.

This seems to want a light touch via open cell foam, from what I can tell, much like what's used for many 8-Track cartridges. It seems like you could use felt as well, but only if you had a light spring force applied. These cartridges have foam glued to aluminum. I'm also seeing felt glued to spring (beryllium copper?) material online for 8-Track tapes. Hmm...

Jack, if you discover you have a Minidek, I'll be curious to know if you have the original 1351 or another 1371 like mine. I don't think I have schematics for the 1351 interface, but I do have one doc on Bitsavers relevant to the 1351. The one in Stuttgart is connected to a Straight-8 and Klemens does demo it from time to time.
 
Excellent, I'm glad to help. Thanks for digging up the 8/m connector part numbers; I've made a note of these. My 8/m is at home, not at LSSM, and it's a bit difficult to dig into as it's under a couple of S-100 chassis at the moment, but I'm pretty sure its power connectors are still ok. It runs and doesn't make any burning smells, anyway. ;) If you end up ordering these, please let me know if they fit ok.
The Mate-N-Lok connectors arrived from Mouser and seem to be the proper parts for the H740 supply in my PDP-8/M. I'm eager to get the system running and haven't yet built a replacement cable between the supply and backplane but did fit both sides of the shells on the opposing parts (see photos), and confirmed that the individual pins from Mouser fit properly on the existing pins on the supply and cable.
The previous post with TE Internal part numbers and Mouser links is here: https://forum.vcfed.org/index.php?threads/what-did-i-do-to-my-pdp-8-today.77577/post-1250571

Original DC Harness plugged into new Mate-N-Lok shell without socket pins installed. The new bright white shell in this photo is TE Internal #: 1-480274-0 (Socket Housing)
H740 connector new PS old Cable.jpg

New Mate-N-Lok shell without socket pins installed plugged into old connector on the H740 supply. I accidentally broke one of the latches on the old connector while cleaning up the system before powering up.
H740 connector new Cable old PS.jpg

I was originally planning to replace both sides of the Mate-N-Lok connector and pins with the new parts because of the burned pin damage on the +5V pin and broken latch, but have since implemented another fix based on Roland's suggestion: routing a pair of 12 AWG wires for +5V and GND between the C7 terminals in the H740 supply to the quick connect terminals on the backplane.

Photo of connections from the C6 terminals in the H740 supply. There's a pair of 12 AWG wires crimped in a yellow #10 ring terminal that is intended for a single AWG 12 to 14 wire. It was a tight fit but it crimped okay. Red and black heat shrink tubing has been fitted over the yellow ring terminals to better insulate the back side of the connection (because at first I wasn't sure how the wires would be routed) and for esthetic appearance. I still need to find a worm grommet to line the sharp edges of the metal where the wires exit from the H740 compartment.
5V cables H740-side.jpg

Photo of additional pair of red wires for the +5V connection to the backplane. Also added two black wires for GND which looks the same. New 12 AWG wires on the left use "piggy back" style quick connect lugs". The original +5V terminal on the right is a bit burned and will be replaced with a new one soon.
5V cables backplane-side.jpg
 
Working on getting ready for VCF. Decided to bring my 8/I this year. Fired it up and stuck memory bit. Swapped a G22x and good to go
It was only pretending to be good to go. Occasionally got funny FFT spectrum but assumed operator error or bug. With adding a status display the number changed randomly when it should be constant. Found the low 6 bits of MQ going to zero after a little bit with machine halted. After much poking tracked it down to the solder joint between the bus strip on the backplane and the wirewrap pin was intermittent. Added a wirewrap wire to the next 5V and now it seems to be working. Hopefully more pins won't start doing this.
 
Worked on the Console Serial disk server program. Did some testing on REAL HARDWARE. I am close to the point where I could try booting but I need to make an RK05 size disk image with the console serial disk handler installed as SYS and then transfer that to the RaspPi. Lots left to work on and probably bugs to squash.
 
Here is a short video of a Terasic DE0 Nano board with an Altera Cyclone IV FPGA pretending to be a PDP-8/e using the OpenCores PDP-8 implemented in VHDL.
It runs a 4 word diagnostic which increments AC with a delay to make it observable.
The program it runs is:

0: 7001
1: 2101
2: 5001
3: 5000

The contents of the top-most 7 bits of AC are displayed in the left 7 LEDs. The right-most LED is the RUN light.

The OpenCores VHDL implementation is at: https://opencores.org/projects/pdp8

 
Here are some photos of the Mate-N-Lok power socket on the H724 power supply in my new PDP-8/e.
Below is a photo of the associated connector on the power harness connecting to the Omnibus.

IMG_20220630_215755041.jpg

IMG_20220630_215831155.jpg

Both photos show that the 5V pin got hot sometime in the past and some oxide has formed on the contacts. I am not worried about the browning of the nylon connector bodies.

I measure a voltage drop of about 15 mV with only a handful of boards installed:
  1. front-panel
  2. M8330
  3. M8310
  4. M8300
  5. M837 (ext memory)
  6. Rolands 32 kW memory
  7. M8320 (bus load)
The socket appears to warm up about 10 degrees above ambient temperature.

I have tried to clean the contacts with contact spray, but it made no difference in voltage drop or temperature.

I have a few options:
  • ignore the oxidization;
  • use some mechanical abrasive (fine grit sand paper?) to remove the oxidization;
  • cut the 5V wire on both socket and connector wire harnesses, find a suitable red hookup wire, crimp the contacts for both socket and plug onto the wire, replace the oxidized contacts with the new one and finally solder the new wires to the old harnesses with the solder joint protected by heat shrink;
  • replace the entire red 5V wire in both power supply and Omnibus harnesses with new contacts crimped on;
The last two would require some suitable red wire. I have modern hook-up wire (made in China) which looks to have sufficient thickness to carry the current but the insulation is unknown.
What wire was used in these harnesses? The wires in the harnesses have become rather stiff so likely the plasticiser has leaked out with age. Doug mentioned Teflon wires in the context of the H724 power supply fan wires. Teflon has no plasticiser so it could be ordinary PVC.

Does anyone know the type of wire that was used in these harnesses and also what the correct insulation material is?

Any recommendations on how to deal with the problem?

Thanks
Tom
 
Both photos show that the 5V pin got hot sometime in the past and some oxide has formed on the contacts. ...
I measure a voltage drop of about 15 mV with only a handful of boards installed...
I think this is a chronic problem with the +5V harness in a number of DEC computers. I don't think any repairs to that specific connector will affect the root cause, which is overloading the current rating of the connector contact. I can recall reading of similar issues with:
1. Roland Huisman's PDP-11/10
2. Dave McGuire's PDP-8/E
3. my PDP-8/M
I have a few options:
  • ignore the oxidization;
  • use some mechanical abrasive (fine grit sand paper?) to remove the oxidization;
  • cut the 5V wire on both socket and connector wire harnesses, find a suitable red hookup wire, crimp the contacts for both socket and plug onto the wire, replace the oxidized contacts with the new one and finally solder the new wires to the old harnesses with the solder joint protected by heat shrink;
  • replace the entire red 5V wire in both power supply and Omnibus harnesses with new contacts crimped on;
The last two would require some suitable red wire...
What wire was used in these harnesses?
I'm not sure about specs for the wire that DEC used. For the solution mentioned below I used red and black AWG 12 stranded wire purchased from my local Home Depot store. For stranded wire it was very stiff but I was able to successfully route the wires inside the chassis.

In my 8/M my first plan was to do your option 4, and I even ordered new connector shells and contacts. Dave McGuire has a post that identifies the correct part numbers of shells and contacts:
...and there are links to Mouser and DigiKey in this post (for you, links to the pins are valid, shells are for the 8/M):

Any recommendations on how to deal with the problem?
I think the best solution is to ignore the oxidization on the contacts (option #1 above) and add a pair (or more) of additional AWG 12 wires from the power supply to the backplane, in parallel with the existing single +5V wire through the Mate-N-Lok, but the additional wires of course not running through the Mate-N-Lok.
Roland describes it for a PDP-11/10 in this post:
and I copied Roland's solution in my 8/M:

Fortunately the 11/10 and 8/M share the same power supply so the fix was easy for me. In the 8/E you'll need to find a creative path to route wires from the power supply enclosure to the backplane quick connect tabs. A convenient tie point in the power supply is the large +5V filter cap. Use of piggy-back quick connect terminals enable doubling-up connections on the backplane side. 8/M backplane already has a double quick connect terminal for each power connection, which is helpful to not move the overloading problem to the backplane connection.
 
I think this is a chronic problem with the +5V harness in a number of DEC computers. I don't think any repairs to that specific connector will affect the root cause, which is overloading the current rating of the connector contact. I can recall reading of similar issues with:
1. Roland Huisman's PDP-11/10
2. Dave McGuire's PDP-8/E
3. my PDP-8/M

Yes, that is my experience too. These pins were fried in many of my machines too. Also the 11/20. The pins were probably rated for the current once they were new. In my 11/20 all boards including expansion chassis were powered by one single pin. So just like the other machines I doubled the GND and +5V connections.

In the 8/E is it a bit harder to place extra wires. So in that case I would prefer to replace the 5V and gnd pins. A bad 5V connector is less dangerous, a floating GND can be dangerous for the boards. A floating GND below 0 can cause a higher voltage on the TTL chips. That depends on the load on the negative line...
 
At LSSM we've found that a bit of DeOxit D5 on the contacts of those connectors cools them down quite a bit. But, no amount of chemical assistance can compensate for under-specced connectors, of course.

-Dave
 
The connector pins themselves were not the problem. The problem is the 40+ years of time passing. The best solution if you want to keep it original is to replace the pins every 15 years or so. Of course that is a silly suggestion. When I bought new housings a while ago, DigiKey suggested related items and one of those was gold plated pins and sockets. Replacing the tin plated pins with gold plated would most likely fix the issue forever. (Ok, maybe not forever, but it would push the issue out hundreds of years.)

Now I am anxiously awaiting photos of someone's connectors with gold pins!
 
About a month ago I spent a week working on not one but two Straight 8's (not mine). At the end of the week we (thanks Bob and Malcolm!) had a sort of running machine. The second machine had been modified by using some empty slots and reusing some slots originally used by the memory parity option. No idea what those mods were for but if it was my machine I would spend the effort to document and remove the changes. Both of these machines were newer than mine and were probably built in 1968.

The last few days I've been using my 8/e to debug the console serial disk project. I posted an update on the handlers thread.,
 
I am back to repairing PDP-8a Programmer's Panels. Whatever was inside of the counterfeit 7475 chip that I used probably had the normal Vcc and ground pin locations instead of the odd locations for a 7475. It shorted Vcc to ground and popped the +5V breaker in the power supply. I replaced the 7475 with an ancient National Semi 7475 I found in in my collection, and I am now back to making progress with the repairs. If you enter 7777 in the SR it displays 1133. If you do a LA, the address goes to 7777. I think that means the shift registers are OK, and it is just a display logic problem. The STATE LED doesn't work, and the AC LED is intermittent. If I replace the LEDs with modern ones they will likely be too bright, so I will need to change resistors too.
 
I have been fixing PDP-8/e MM8-EJ core memory boards - specifically G111 Sense/Inhibit boards.

One of the boards came with my new PDP-8/e, the other one I bought from Jack.

The first had a problem with bit 9 and was caused by a faulty quad NAND a 74H40 (E25).

Jack's board appeared to have a number of bits stuck high and low with only 6 functional when I first did a quick go-nogo test when I received it. Strangely yesterday 5 of the faulty bits had "healed" and I was left with a single error - bit 2 stuck low. They turned out to be a faulty OR gate in a SP384N (E14). I originally tested Jack's boards as I received them, but yesterday washed the X/Y and Sense/Inhibit boards in Isopropyl alcohol aided by a soft brush. It is possible that the "healing" of 5 bad bits was simply removal of some conductive "gunk".

Interestingly in both the replaced 74H40 and the SP384N the remaining other gates were fully functional.

I am getting much better and more confident removing ICs using my desoldering station to the point that I no longer use sockets when I replace ICs.

My next adventure will be fixing two 8655 UART boards. One transmits perfectly, but receives all zeroes instead of the actual character typed. The other transmits, but some sent bits are "hard-wired" to a level ignoring the actual bit value I try to send.
 
Good news - a gentle application of alcohol is always helpful, whether shaken, stirred or brushed!
Keep in mind that in that mixed context environment, one kind of IPA (Iso Propyl Alcohol) goes on the board while the other kind of IPA (India Pale Ale) goes in the person cleaning the board.

;-)
 
One of the M8655 UART boards is proving a bit more challenging as I don't have the two suspect parts in my parts drawers. One is a 74175 which is relatively easy to get and is at least available on this continent (Melbourne versus Perth), but the other is a SP380N a quad NOR with weird pin-out. These are ICs E24 and E30 on the M8655.

I can't find much about it on-line, but there are some Ebay sellers who claiming it is made by Signetics. One of the other M8655 boards has two DEC 8640 in the same position so they appear to be equivalent.

Strangely I cannot find this SP380N in any of the Signetics data books. I found a data sheet online which shows it as a quad 2 input NOR but the scanned pages show no manufacturer and the details are minimal.

Does anyone know of a scanned Signetics data book which lists this SP380N (or DEC 8640)?

Also are there any suitable alternative replacements?

Tom
 
Last edited:
Suspect that what you are looking for is a National Semiconductor DS8640 Quad NOR Unified Bus Receiver. That's consistent with the use of E24/E30 to gate the data signals off of the backplane and the pin-out appears to match the schmatic.
 

Attachments

  • DS8640 Quad NOR Unified Bus Receiver.pdf
    108.2 KB · Views: 10
Strangely I cannot find this SP380N in any of the Signetics data books. I found a data sheet online which shows it as a quad 2 input NOR but the scanned pages show no manufacturer and the details are minimal.
Try here:

There was/is also a more modern replacement for the DS8640, I think, or perhaps it was only for the DS8641 (transceiver). Never seen it IRL, though. I think it was essentially invented after it's popularity had passed.

DEC AFAIK used either the SP380 and the DS8640 depending on availability, which was roughly related to the date of manufacture. I think they also selected and rebranded some of the parts.

Vince
 
Back
Top