Time to rebulid the pdp8/L

[czunit]

I have a pair of pdp8/L's here that I picked up years ago and it's about time I got around to restoring unit #2. Unit #1 is missing somewhere here, I got it to the point where it ran basic programs and I think it's in the closet underneath the pdp8/E.

Anyway, first step was cleaning up the key switches and replacing several broken toggle switches. I needed six of them, had three with broken pins and realized I could just print a couple of them on my Ender 3 pro. Hint: Do not use PLA, it's far too weak for the small pivots. PETG however works *very* well and now all of my switches are fixed and more importantly close reliably (no bounce or excess noise, cleaned the miscreants with Deoxit.)

This evening I checked out the power supply. Unplugging the it from the CPU and putting it on a very well insulated towel I powered up the supply and checked the voltages.

Color What is What should be Purpose
Blue/orange 10.21v Panel power Pretty unregulated
Grey/blue 0 -15 if locked Panel lock Would be -15 if unlocked, +5 locked
Orange 11.21 8v +5 input Unregulated, ok
Red 5.21 5 +5 regulated Regulated, ok
Blue -17.60 -15 Neg 15 power Unregulated, ok.
Green -35.17 -6 WAY OFF What is this? BAD!
Yellow -35.05 -30 Broken? Turns out ok ish.
Brown -.329 ? Mem reg base Unknown

Wire color	What it is	What it should be	Notes
Bl/Or	10.21	Unknown	Panel power
Gr/Bl	-15 locked	5 unlocked	Panel lock
Orange	11.21	8	+5 input unregulated
Red	5.21	5	+5 regulated output
Blue	-17.60	-15	-15, unregulated
Green	-35???	-6	Odd, not supplied by PS
Yellow	-35	-30	-30 unregulated
Brown	-3.29	?	Looks like from another card

So the only odd one is the -6 volt one. It is not tied to the power supply but goes to a pair of transistors as a regulator. Might get power from the regulator board next to it, not sure.

Anyway, any thoughts on the -6 line? Does the other card (G826) supply power to this?

I'm pretty certain at least one or two of the G220 register cards are sad, is there a troubleshooting guide to what components tend to fail and need replacement?

C

 

[daver2]

Welcome...

Did you have any dummy loads on the output of the power supply or not when taking the measurements?

The -6V (varies) is not too helpful by DEC - See page 5-4 of document http://bitsavers.org/pdf/dec/pdp8/pd...LmaintVol1.pdf...

EDIT1: I assume you have seen the data sheets and schematics over here http://dustyoldcomputers.com/pdp-com...odules/g.html?

Assuming your -6V is derived from a G822, the associated data sheet states that it should be supplied with 11V from a 715 power supply and produce an output of between -5V to -7V. Not sure where your -35V comes from in this case. Of course, I am still looking for the details relevant to the 8/L...

EDIT2: Ah, found a load of useful stuff on (PDF) pages 28 and 30 of document http://bitsavers.org/pdf/dec/pdp8/pd...chem_Feb70.pdf.

EDIT3: Just out of interest, when you were measuring the -6V line, what was the other probe of your voltmeter connected to?

Dave

 

[czunit]

I think it gets its' power from the board next to it (the core memory board power supply) as it doesn't seem to get anything from the transformer other than the two transistors on a heat sink. Need to solder a temp wire to it to test it in circuit. All tests were made open circuit (no load, board out) with respect to the black wire which is neutral.

In circuit I see a voltage of -12.40, which once again doesn't seem to help much. The manual says you should check memory voltage between memory supply + and memory supply -, need to find out where those are...

 

[Roland Huisman]

Maybe you like to read my PDP8/L adventures on this forum... I had a LOT of work to get it going again. But it passes all maindecs now and is still working after the restoration.

Regards, Roland

 

[czunit]

Very helpful Roland. Quick question: Where did you test for the memory stack voltage? That might be problem #1 with my unit right now.

 

[m_thompson]

You can also read the restoration blog for the PDP-8/L at the RICM here.

 

[czunit]

*nod* I read the RICM one. I might have different problems but the number of blown flip flops is probably going to be similar. One thing I do have going for me here is a large box of parts from a pdp8/I that I took apart 30+ years ago (it was one of two and had a forklift driven through its backplane so it was very sad). But I'd rather find and fix problems than just swap away....

 

[tradde]

That is always the best way or you might just end up frying another part/board in the process.

 

[czunit]

Well, time to start testing. Without the riser it's a bit more complicated to see voltages and such but I looked at the schematic and I can see that mem + and mem - come out of G826 (the big regulator board) and go into the four G624's where they are filtered by a capacitor then sent on to the resistors that feed the core memory system. So the points across those electrolytic capacitors are perfect test points.

One capacitor hooks up to pins E2 and H2, easy to find. Just count the pins remembering that DEC doesnt use the letter G I O or (I think Q) and solder two wire-wrap test wires across the capacitor. Then hook up to a meter, plug in the G624, and test.

Voltage across the core is only about 12.4 volts, way too low. Small wonder it's not working. I'll check out the thermistor line on G826 to see if the thermistor is open, if not I'll try to find a full schematic for that board. It has a pot so it's possible I can adjust the voltage but with that big a difference (should be 22 volts or so) it's possible something's broken.

C

 

[czunit]

Update: Turning the pot on G826 does increase the voltage, about half a volt per turn. According to the manual that means the thermistor is in circuit, but cranking it up to 22.4 volts is going to require a lot of turns. Before I do that can someone check my work so far and ensure that I'm really reading MEM - and +, and that 22 volts is a proper amount? Either someone really cranked that pot down a long time ago or something is unusual...

C

 

[czunit]

Well, I cranked the pot up and can only get to 17.8 volts before it hits the end stop. Cranking it down to the other stop I get a low of 5 volts. So something is still wrong here, might be the thermistor, will check that next...

 

[czunit]

Hm. I have a lot of 4k core memory planes so I decided to get one out and check it. One of them has a thermistor on the outside, reading it shows about 390 ohms. The one on the core plane in the unit is 700 ohms. Bit different, but they do seem to exist. Popping the spare core memory into the 8 shows a voltage of 11.8 volts while the first one showed 12.4 volts. Bit different, but not much. So it looks like the thermistors are at least there.

So what is going on?

In the "picture is worth 100 words" here's a picture of the 8's memory system, the spare core board, and my wires for sensing the memory voltage.

https://i.imgur.com/wV84nut.jpg


C

 

[m_thompson]

Well, I cranked the pot up and can only get to 17.8 volts before it hits the end stop. Cranking it down to the other stop I get a low of 5 volts. So something is still wrong here, might be the thermistor, will check that next...

The Maintenance Manual says that the Memory Supply+ should be adjustable between -1V and -12V, and that should yield a difference between the Memory Supply+ and Memory Supply- of -18V and -29V. What is the range of the voltage on pin BM2 of the G826 flipchip?

It also says that the thermister resistance should be 330 Ohms, so your stack with a 700 Ohm thermister could cause problems.

The actual voltage is not as important as the current going through the core. The Memory Supply voltage is adjusted to result in 320 mA going through the cores, and the termister adjusts the voltage/current as the core stack warms up.

 

[czunit]

Hm. This is interesting....

So I've been measuring the voltages with one of the G624's modified with wires coming out from the capacitor across Mem+ and Mem-.

First thing is I need to simplify. I'm pulling all four of the G624's so the memory voltage is not going anywhere near the sense or inhibit drivers. With those out, I see a voltage across G785's green and yellow wires of of 21.15 volts. That should be the regulated mem power (+ to -) and is reasonable. But that is still a voltage of -33.95 (yellow) and -12.7 volts (green) with respect to ground. Odd, but there it is.

So when I have the G624's in, I get the lower voltage. Now an interesting observation was that when I monitor the voltage with the 624 in I get some very different readings.
A26 21.16
A25 12.6
B26 33.7
B25 21.16

Now since I'm checking the voltage on the same capacitor on the G624 I may be seeing a half voltage from the inhibit drivers but that doesn't make a lot of sense, especially as why I am seeing the whole 33 volts on one of them, proper voltages on two of them, and the 12.6 volt side on the third.

Very odd. What's next up the line from these G624's. Probably the G221/228's. Maybe I'll remove all of those and see how things look.

Ok, setting the voltage to 22.4 volts using the trimmer pot I see a voltage of -39 and -16 on the -6 and -30 lines. Granted it's -22.4 volts between the two but it is weird. This matches the voltage across the capacitor on B25.

I think I have to find my other 8/L and check it out.

 

[czunit]

In the meantime I dragged out the other 8/L. Sure enough it's pretty nutty as well, haven't run it in about 5 years. But it's a different kind of nutty (major registers are insane) so I should be able to figure something out here.

Anyway it does change the values in the memory buffer when I advance the memory address with examine so something's working in the memory region (unlike the first 8/L) so let's check voltages (see table below)

Core voltage is a nice 22.69 and I see the same effect as the first 8/L when I move the G624 around. So that's normal-ish. However it gets the -22.69 by summing the -10.5 and the -33.3. This is different from the first one which is summing the -39 and -16 volts. Why, I don't quite know: I might try pulling the G624's from this second 8/L and putting in the memory power controller board from the first 8/L to see what it does.

Wire color	What it is	What it should be	Notes
Bl/Or	7.35	Unknown	Panel power
Gr/Bl	-15 locked	5 unlocked	Panel lock
Orange	8.78	8	+5 input unregulated
Red	4.8	5	+5 regulated output
Blue	-16.4	-15	-15, unregulated
Green	-10.5	-6	Memory +
Yellow	-33.3	-30	Memory -
Brown	-.5	?	Crowbar circuit

 

[czunit]

False lead. Deleting.

 

[czunit]

Here's the status so far:

System turns on, Loading an address of 7777 sets all bits in the Memory address register.
Clicking examine shows nothing in Memory buffer
Clicking examine again clears MA (loops to zero) nothing in MB
Clicking examine increments MA to 1, then 10, then 11, then 100... nothing in MB
Loading address 0 then setting all switches to 1 then clicking deposit shows nothing in MB

So to start with, what boards are used to set bits in the memory buffer from the key switches?
To answer my own question it looks like the board in B08 is the one that toggles MB_LOAD based on a signal called TP2.

 

[czunit]

Ok, I got the extender board which makes checking things a *LOT* easier. Basically the +5 line is good, the -15 is a bit lower (-16) and the memory voltage is 22.5 volts on the nose.

So on to testing why the MB isn't loading:

According to the schematics, the line MB_LOAD is the toggle for the M220's for the memory buffer. First step was to check the M220 to see if MB_LOAD and PC_LOAD were low or high. Sure enough PC_LOAD was low, MB_LOAD was high (+3v). So is it the M220's?

Pulled all of the M220's out except one. Tested. MB_LOAD high.
Pulled the last M220 out of the extender board. Tested. MB_LOAD High.
On page 11 of the Volume 2 pdp8/L manual I see that MB_LOAD is from a M617 on Slot B08. Pulled the M617. MB_LOAD is now zero.
Put in all six M220's. MB_LOAD is low. So the signal is coming from M617 and not from a short in one of the M220's. Good.

Now what's going on? Put M617 in B08 into the extender board. The schematic says that K2 and L2 are both connected to a +3v source (10? What is 10?) and M2,N2 are connected to a line called TP2 (what's this). However checking I see that K2 is reading +4v, L2 is reading 0, M2 and N2 are both reading high (3v)

Well, that's interesting. On a 4 input NAND I believe it's if all 4 inputs are high the output is low. So L2 being low would drag the output high all the time. Which is what we see.

The question is what is sourcing K2 and L2? If the schematics say they are tied together, why are they not reading the same voltage?

Note: I am assuming that K2 in Dec flip chip speak means side 2 (the one with the chips on it), pin K (which is read as K in VUTSRPNMLKJHFEDCBA when read from the front of the board left to right. Is this correct?)

Progress, which is good. I'm tempted to try just jumpering K2 to L2 but I think that is wrong somehow.

 

[vrs42]

Now what's going on? Put M617 in B08 into the extender board. The schematic says that K2 and L2 are both connected to a +3v source (10? What is 10?) and M2,N2 are connected to a line called TP2 (what's this). However checking I see that K2 is reading +4v, L2 is reading 0, M2 and N2 are both reading high (3v).

I think you've got K1/L1 confused with K2/L2. If would be expected for L1 to be a simple inversion of K1.

The question is what is sourcing K2 and L2? If the schematics say they are tied together, why are they not reading the same voltage?

Note: I am assuming that K2 in Dec flip chip speak means side 2 (the one with the chips on it), pin K (which is read as K in VUTSRPNMLKJHFEDCBA when read from the front of the board left to right. Is this correct?)

Side 1 is the component side, and side 2 is the solder side.

Progress, which is good. I'm tempted to try just jumpering K2 to L2 but I think that is wrong somehow.

Yeah, don't do that .

Vince

 

[czunit]

Ok, thanks! Flip chip is a bit maddening, for example the single side boards (much of the G logic) only have pins on one side, which is the solder side. Is there a version of the Logic guide that explains exactly what pin is what? Does the lettering order flip as well if it's a single side vs. double side board? Ug.

C