Hi Tony!

That smaller transformer isn't the main switching transformer, you wouldn't get 800W out of it anyway, it's too small, but more importantly you wouldn't get over 60A through such tiny pins...
That's the auxiliary power supply transformer, it generates those 5V standby and what ever other voltages the control boards needs.
This is clear since that failed IC is just near that smaller transformer.
Also, those two diodes ( i am guessing they are diodes ) near that standby transformer on the secondary side, are there to rectify the 5V standby output.

The main power transformer is that larger one, sort of an R-core transformer, and it's winded on the secondary side with copper foil rather than copper wire, and that's totally to be expected, you have a high frequency power supply to allow you to downsize the magnetics, but you also have a rather high current on the secondary, and high frequency means high AC resistances on the windings due to skin effect, on high frequency the current tends to concentrate near the surface of the wire, so you use copper foil in stead of copper wire, just because the foil is very thin, so the current uses most of the thickness of that copper ( so it's more efficient ).

The rectification part of the main 12V output is done by those 6 devices on the larger heatsink ( in between those output inductors ), they look like mosfets, not diodes ( took some captions from your video and had a closer look ), and you would expect that, because rectifying 60A with normal diodes ( even Schottky ones ) is way too wasteful, so the power supply uses mosfets for synchronous rectification, which is a way more efficient rectification.
Those SOT23-5 devices near them seam to be some sort of gate drivers needed for fats switching ( so lower switching losses ).

I cannot be sure but from what i am seeing in the video, that Optocoupler seams to handle the main 12V output feedback part, so i am guessing that the 5V standby supply is regulated on the primary side and the outputs uses cross-regulation to get the needed voltages ( obtained by a fixed turn ratio between the outputs and the regulated winding ).

I also cannot be sure because it is not clear from the video where that larger main transformer is connected to it's phase shifted bridge converter ( done by those 4 mosfets near the dead TOP IC ), either by some traces on the top side ( or even an inner layer ), or maybe through those wires coming out of the transformer and going somewhere near the main bulk capacitor on the HV side, and at 390V you wouldn't need a very large current, so, that could be it.

09:58 - I am pretty sure the big yellow one is the switchmode transformer for the 12 V / 66 A, the 6 mosfets acts as a active rectifier. The other switchmode transformer is just for the 5 V / and local stuff.

If the top is missing from the TOP IC, something is wrong

I really appreciate your thorough explanations of what you are doing, and why. Thanks

Fuses actually doing their job!

20W for +5Vsb are quite reasonable for a server: it has to supply the entire out-of-band system management sub-system
even if implemented as one big asic it has quite a lot of functions, most notable an entire network stack with some sort of webhost; often it also emulates virtual drives to the main system bios - running all of this just on 5V requires quite some amps; look at single-board computers like raspberryPi - they also often require power supplies around 20W

9:27 - It's not just optocouplers; class Y capacitors are also allowed to connect between the primary and secondary sides of the PCB to reduce common mode noises (by providing a controlled return path).

Also, the board cutout is not meant to stop arcs between primary and secondary. Its purpose is to increase creepage distance, so dirt, grime, or moisture cannot easily form a conductive path across the surface. It can also help reduce arcing risk if the primary side reaches a high enough voltage.

Thanks! very much for the great video and walk through of whats what. Have a good weekend

Thank you for this!!! I have been working on an Agilent psu for a while now with no 5v standby. I thought it came from the 5v module, but now I know there is an offline 5vsb driver! Every card module worked outside the unit and the unit was supplying them power.

high voltage risk of electrical shock
If those electricians could read, they'd be very upset.

"yes, it's working!"........where have I heard that before...🙂
Nice work, love the investigation.

4amp 5stb is probably also used to power the LOM system in servers, which actually is often a full (linux powered) computer on its own thats used so you can remotely control the server.

I've always loved this industrial equipment, so different from domestic stuff :)

Related to discharging the big cap, Dave from EEVBlog taught me that the multimeter can be used for this if you switch it to resistance mode. After a few seconds you can switch it back and check the voltage.
Any decent multimeter will be protected against high voltages in all modes.

Thank you Tony, really liked your explanation of the snubber circuit!

It isnt just two fans in series, it it a counter rotating fan which has two fans spinning into the opposite direction

Great work 👌👏

Thanks. Great and informative.

The 6 output MOSFETs are used to balance the current output to be equal to the second PSU when they are in current balanced mode in the server, Where half the power is pulled from one PSU and the other half from the other. Or to turn off the output to be in hot-standby.

14:55 LM393A is not a regulator at all (you may have been thinking of LM317/337), it's a dual comparator!

As for the fuse, 2 A seems like a lot if all it's powering is the standby power supply. (It needs like what, 0.1 A tops?) Maybe that's some kind of code instead? Or maybe it ought to be 0.2 A?