HP 6824A fails while filming my latest episode (Patreon)

My HP 6824A power amp just failed while I was filming the latest video. I had spent the last few days wiring up all the external connections of the mechanical Soyuz clock, mainly to my box of Apollo switches and indicators.

It takes an amazing number of connections to entirely run this simple clock. Three Cat5 cables connect to the 19 pins of the connector at the back.

Then it goes to a type 66 punch block, where I can patch them up to the various buttons and lights in my box. 

I was ready to give it a go, but could not get the clock going. The clock drive signal has to go through my new punch block and a few switches. I thought I made a wiring mistake, but I could not find any. 

I finally discovered that my HP 6824A power supply/amplifier had gone bonkers. That's the instrument I use to amplify the timing pulses so they can drive the clock solenoids, the one with the large white meter.

It was pegged at the negative end of the voltage, and happily outputting -74V no matter what.

Piece of cake to repair I thought, the power transistor on the negative rail must have blown in a short and connected directly to the -74V unregulated voltage. No need to film this. So up on the bench it goes. It's basically a bipolar power supply, a positive and a negative supply joined together but driven in mirror push-pull fashion. A power supply turned into an  audio amplifier of sorts. It's a gift from a local viewer by the way, thanks Vincent H. !

Surprisingly, all the power transistors checked out OK. I then opened up the voltage regulating feedback loop (you can do that easily on the back of the instrument by removing the local sense strap) so I could see if the feedback chain was working. And it seemed to work too, reacting to sense input almost as you'd think it should. I tested all the stages, and they all seemed to work, going up and down with my sense input. Except that there were large offsets - although it is hard to tell from the schematics around which DC voltage it should swing at each stage. After Q3, right about in the middle of the long feedback amp, it really got quite negative on average, which was suspicious. 

I checked the supply, transistors, diodes, resistors and caps around that stage and everything seemed OK. I was not making any progress, and not understanding what I was measuring, so I removed the transistors in the error amplifier around Q3 to see where the bias levels stood without any active components. Sure enough, a load resistor after Q3 that was connected to +140V on one side was reading -56V on the other side, with seemingly nothing active attached to it at the other end. It's test point 18 in the schematic below, above and to the right of my finger.

Well, attached to almost nothing. There was still transistor Q12, at which I am pointing in the picture. It's part of the current limiting scheme, and not really part of the feedback loop. Plus it is in the positive rail, which appeared to function normally, not in my seemingly diseased negative rail. So I had left it alone, as it should not do anything in normal operation.

In the schematics above, you can identify the power transistors in the - rail (top) and + rail (bottom). They are stacked in a parallel arrangement (to the left of my hand, 3 PNPs on the negative side and 2 NPNs on the positive side). Q12, which I point to, measures the current in the positive rail after the power transistors, across a small power resistor, and turns on if current is too high, then sends the signal back into the feedback loop to turn the NPN power transistors off to limit the current. Which is a bit unusual. Usually you'd just directly short the power transistor itself, as it is done by Q5 on the negative rail. 

So I became suspicious and removed transistor Q12, an innocuous looking, garden variety generic NPN. Sure enough, it had turned into a 500 Ohms resistor between emitter and collector. It was sending a large part of the output rail voltage back into the feedback chain, and making it a positive feedback loop instead of a negative feedback loop, sending the whole thing to its negative rail. 

And yay, without Q12, my voltage at the end of the Q3 load resistor turned back to positive territory. I put all the transistors back in except Q12, and the supply worked perfectly again. 

Q12 is a 2N3417 50V transistor, still available today, but I replaced it with an HP branded 2N3904 40V transistor I had on hand, close enough. I checked that the current limit worked, and adjusted it to the required 1.2A value (there is a pot for that). Which did not require much readjustment, so the replacement transistor was a good choice. I re-calibrated the whole thing while I was at it.

And voila, the atomic time setup works again. The clock ticked right away. Back to filming the main part. 

That's what you get when working with old test equipment. It tends to die on you at the most annoying moment. And take you on a few unexpected side trips. Keeps you on your toes I guess.

Marc