• Please review our updated Terms and Rules here

PDP-8/L restoration - PSU problems.

MattisLind

Veteran Member
Joined
Sep 30, 2013
Messages
1,158
Location
Stockholm, Sweden
I have been looking at restoring one of our PDP-8/L machines. A while ago I brought up the PSU on the bench. Cleaned it and reformed the big capacitors. One of the rectifier bridges had been stolen so it had to be replaced. Then I used a variac and a dummy load. When raising the input volatge with the variac I was surprised to see that as the input voltage was rising across the regulator the output increased above 5V. It actually reached 5.7 volt until it started regulating down to 5V. I was a bit confused since although TTL usually has a absolute maximum around 7V it is probably not so good that there is a spike at turn on. There were some reports that PDP-8/L have had problems with their power supplies in the past.

I did find found the PSU schematics, but the regulator board was a black box.

rIGdw49.png


No schematics was found on the 54-08047 so I traced it out myself and simulated in CircuitLab. Directly from reading the schematic though it is evident that the left transistor half of the matched pair need above 5.7V to start conducting at all. It will then bias Q2 which will then eat up the base current from Q3. Is this really a good design?

YBP2dMJh.jpg


5oQUhAN.png


There is a crowbar circuit but since it is using the same zener as the regulator itself it will trip above 6.3 V or so (the simulation is consistent with actual measurements). A bit to much for my taste at least. I have seen rumors that there are no crowbar in a PDP-8/L but there is one, indeed.

I didn't have the proper SPICE parameters but used more standard PNP and NPN models. Anyhow there were the same type of spike shown.

ifrNxoPh.png


My plan is to modify the regulator board so that it regulates better. I did a simulation in CircuitLab where I replaced the 5.1V zener with a 3.9V zener. A voltage divider was inserted on the right hand side of the matched pair. From the simulation it looked much better. There are no spike when the input voltage is ramped up and the crowbar now trips at somewhere around 5.3V.

Next step is to test this in the real circuit to see if works equally well.
 
That looks very similar to the behavior of the H-724 power supply in a PDP-12.

I looked at the schematics for the G824 module. It looks very similar, indeed. I can't see anything of the current limit and fold back that are present in the PDP-8/L though. It was just a quick glance so I can be wrong.
 
I would be real careful with playing around with the threshold of the crow bar, know what you’re saying about trying to keep from hitting the five volt bus too hard but I always tend to think sometimes people get too worked up about keeping the five volt bus at exactly five volts. A little bit over five volts makes up for the losses in the distribution and connectors so by the time you get onto a circuit card you may not have five volts any more. When crow bars fire they have a bad habit of sometimes shorting the SCR and have been known to also damage the pass transistors so you don’t want it going off any more then necessary, and in the event of a catastrophic failure in the regulator that usually turns out to be a shorted pass transistor the full unregulated voltage is dumped on the output so you would be looking at something well above five volts.
Once the crow bar fires its intended to blow F3 and with that it will also remove the gate voltage from the SCR so it should be ok but I have had too many crow bars that once fired the SCR shorts but that’s just me, so it’s just my opinion and not to take away from what you’re doing but I would not mess with the threshold of the crow bar.
 
@Qbus: Maybe you are right. In my opinion 6.4V crowbar triggering is far to high (the simulation give the same result as the actual circuit tested with a bench supply). In other supplies I worked with the crowbar trigger around 5.5V which I think is more reasonable. With the design of this PSU were there are a spike in the 5V so that it reaches 5.7V when the supply capacitor is charging at startup they just added a diode forward drop to make it not false trigger. Thus making the crowbar trip around 6.4V...

By adjusting down to a 3.9V or 4.3V zener I expect crowbar to trigger at 5.4V or 5.8V. I'll give it a try but I keep in mind what you said about short circuit SCRs. In this case they have a 0.1ohm in series with the resistor limiting the current a bit. The 2N4441 is rated at 80A surge current which is exactly what you get with 8V and 0.1ohm...
 
Maybe it’s just my paranoia but I have had nothing but misery with SCR being every time they somehow get the gate turned on it manages to short out the device if there is nothing on the cathode. Like a three legged self-destruction device!
 
I noticed that there are at least one more version of the regulator board. I have reverse engineered the one in my 8/L #905 schematics here and pictures in a zip file.

I wander if the crowbar is designed for tripping at 6,3V to 6,5V to avoid tripping if the 5V rail reaches 5,7V at startup. Looking at the schematics it looks right that is should trip at this voltage.
 
Maybe it’s just my paranoia but I have had nothing but misery with SCR being every time they somehow get the gate turned on it manages to short out the device if there is nothing on the cathode. Like a three legged self-destruction device!

I happily replace SCRs every time the crowbar trips if it saves tens or hundreds of TTLs or other components.
 
The absolute max supply voltage for the Vcc bus on TTL is 7.0 Volts. It is recommended that it operated between 4.75 and 5.25 Volts.

I am aware of the abs max rating. My thinking is that it might be a bad idea to expose 50 year old TTL chips to voltages at 5.7V at above which is what the PSU will output momentarily at startup. Or maybe it is a quick way to weed out marginal chips?
 
Question: what’s more stressful? Operating vintage electronics at lower then originally rated voltages, at rated voltage or slightly higher voltages? On very old radios with tubes I tend to run them around 10% below the original line voltage keeping in mind that the modern line voltage is slightly higher then what was around at the time of their manufacture and want to try to extend tube life. Others run sets at way reduced voltages to keep from stressing the sets. But a vintage computer is something from maybe the late sixties or early seventies when higher line voltages were common so I run all the vintage computer stuff at regular line voltage with the idea of reduced voltage may produce more stress on power supplies then the design voltage. Something about reduced voltage will require more current to accomplish the same power and its current that’s where the real work occurs in both power supplies and distribution. It may be antidotal but I have noticed that most damage occurs when turning on systems and would speculate that perhaps this occurs because of reduced voltage and increased current for a short period of time and perhaps that’s what pushes devices into failure? Or maybe it’s just the insulation of the gates that fail due to the application of voltage at startup? The real question is what pushes gates into failure upon startup?
Attached is a picture of my 1938 Navy HRO receiver, at almost eighty years old it’s still working great. My two PDP-11 systems are half that age and would like to think that they may be around just as long.


38hro.jpg
 
I am aware of the abs max rating. My thinking is that it might be a bad idea to expose 50 year old TTL chips to voltages at 5.7V at above which is what the PSU will output momentarily at startup. Or maybe it is a quick way to weed out marginal chips?

Don't laugh!

I was once told by our HVAC SME (Subject Matter Expert) that our buildings were allowed to heat up to 90F at night to expose marginal circuit packs.

Since he had already shown that he had no idea how to work a wet/dry bulb hygrometer, much less what it was for, I just took it all with a grain of salt.
 
I've been told that it's the thermal cycling that stresses things. Note that incandescent light bulbs usually blow when first turned on. When devices go from room temperature to powered on temperature, their component materials expand rapidly and at different rates. This stresses the points where one material is physically connected to another material and over time those connections fail. When things are powered off, they typically return to room temperature at a much slower and therefore much less stressful rate.

You've probably heard of this lightbulb:

https://en.wikipedia.org/wiki/Centennial_Light

The key here is probably that it has remained powered on for most of it's working life.
 
The crew at UMN Duluth had a crowbar problem with H-724 power supply in their PDP-12. Reforming the output caps for about a week, and putting a significant dummy load on the power supply reduced the voltage overshoot at power on and stopped tripping the crowbar. The power supply has been working fine for the last few weeks.
 
Back
Top