I was recently trying to repair the H7826 in my DECstation 5000/125. I'm not sure that you can find official schematics for DEC power supplies. I was just looking through
http://bitsavers.trailing-edge.com/pdf/dec/ and found some schematics for various boards, but not power supplies. Though H7826 has been reverse engineered at
https://robs-old-computers.com/2018/09/23/h7826-power-supply-repaired/, with some older posts with additional info.
The older post at
https://robs-old-computers.com/2018/01/01/update-on-h7826-power-supply-repair/ shows the input section, which is relevant if the power supply shuts down. Note the Q2 and Q3 SCRs at
https://rjarratt.files.wordpress.com/2017/12/5019572.png. Those are two TO-92 case components on the small daughterboard between the big input filtering caps and the power MOSFET heat sink for driving the transformer. Both are crowbars, shorting out the low voltage power supply for the input section. When activated they also cause ISENSE for the SMPS control IC to rise via Q1, causing an even more immediate shutdown of the SMPS. The SCRs that was triggered keeps conducting until the power supply is switched off and capacitors had time to drain, so you need to either wait or short out that supply (at C24) to be able to try to turn on the power supply again.
Q2 gets the shutdown signal via the optocoupler closest to the edge of the board, from circuitry on the secondary side. That's the one shutting down my power supply.
Q3 seems to be triggered by overcurrent, via sensing on the primary side.
https://rjarratt.files.wordpress.com/2017/12/h7826-primary.png shows the circuit around that daughterboard. Note how C24 (small capacitor right by one of the big ones) charges from high voltage DC via R7 and R8. Current drain there must be low at first to allow C24 to charge. The SMPS IC is in low voltage shutdown. When voltage exceeds a threshold, the SMPS IC starts to oscillate, and then the power supply can power itself via a winding on the transformer.
I found I can simply inject power at C24, via a resistor to protect the crowbars, while the power supply is unplugged and see MOSFET drive waveforms at the power MOSFETs. I found C24 was bad. With a good capacitor there, feeding power to C2 via a 220 ohm resistor and no other source of power there, you'll see the power supply turn on and off as the capacitor charges and discharges, due to SMPS IC low voltage shutdown.
My problem is that shutdown signal going to Q2. Both probing the circuit and considering how quickly it happens, that doesn't seem to be triggered from the secondary side. But how is it triggered then?
I've defeated that shutdown by connecting the shutdown line from the optocoupler to ground via a 220 ohm resistor. Then the power supply sometimes starts, and always starts after being switched off unless locked out via the crowbar, allowing me to switch it back on so it stays running. It's enough power to run an RD54 plus its fans, but 12 V is just over 13 V and 5 V is 4.5 V. More load does not help.
It's odd that probing around the circuit board fibreglass, not touching any copper, I see voltages! I also measured 0.02 mA in current measurement mode. Maybe the board is covered in conductive dirt or degrading and starting to conduct?
Of course other capacitors are suspect. The 1800 µF 35 V narrow tall brown ones leaked in two other power supplies, but seem fine here. The voltages are wrong like I said but there isn't significant ripple.
Be careful regarding "ground". The "ground" of the primary side control circuit (negative side of C24) is at a dangerous negative DC voltage relative to AC neutral. Secondary side is safer, but note that ground connections to the circuit board happen via the screws which have lock washers and a silver area around them. If you take the board out of its case, you've severed ground connections to it.
