Power Supply Board D0900848 was oscillating. Here is the procedure how the issue was fixed.

PCB schematic: LIGO DCC D0900848

0. Extracting the power board.

The top lid and the front panel were removed. Top two modules were removed from the inter-board connection.
Some of the SMA cables were necessary to be removed to allow me to access to the botttom power board.

1. D1~D4 protection diodes

Daniel asked me to remove D1, D2, D3, D4 as the power supply sequence is controlled by the relays.
This was done.

2. Power supply oscillation
Since the power supply systems are entagnled, the oscillation of the transister boosted amps had to be checked one by one.

2.1 VREFP (U5)

First of all, the buffering stage of the positive voltage reference (U5) was oscillating. Attachment 1 is the observed voltage at "VREFP" at D13.
The oscillation was at 580kHz with 400mVpp. This was solved by replacing C20 with 1.2nF. (0805 SMD Cap)

2.2 VREFN (U6)

Then the buffering stage of the negative voltage reference (U6) was checked. Attachment 2 is the observed voltage at "VREFN" at D16.
The oscillation was at 26MHz with 400mVpp. This seemed to have a different mechanism from the U5 oscillation. This oscillation frequency is
higher than the GBW of OP27. So there must be some spurious path to the transister stage. This amplifier stage is a bit unique.
The input is VREFP, but the positive supply is also VREFP. And the feedback path between R31 and C24 is very long. I was afraid that this oscillation
was caused by some combination of L and stray C by the long feedback path and the output to power VREFP coupling, although I could not reproduce
the oscillation on LTSpice. 

After some struggles, adding a 100pF cap between the output of U6 op27 (PIN6) and VREFP (PIN7) stopped the oscillation.
I think this changes the loop function and fullfills the stability condition. I confirmed by a LTSpice model that additional cap does not
screw up the original function of the stage at audio frequencies when everything is functioning as designed.

2.3 Positive supply systems (U10, U11, U12)

Even after fixing the oscillations of U5 and U6, I kept observing the oscllative component of ~600kHz at U10 (+21V), U11 (+15V), and U12 (+5V) stages.
Among them, U11 had the biggest oscillation of 400mVpp at the opamp out (Attachment 3). The other two had small oscillation like 20mVpp at the opamp outputs.
The solution was the same as 2.1. C50, C51, and C52 were replaced to 1.2nF. After the modification I still had the 600kHz component with 2mVpp.
I wanted to check other channels and come back to this.

2.3 Negative supply systems (U7, U8, U9)

Similarly the outputs of U7, U8, and U9 had the oscillation at 600kHz with 40~80mVpp. Once C35, C36, and C37 were replaced with 1.2nF,
I no longer could see any 600kHz anywhere, including U10~U12.

2.4 -24V system (U13)

Last modification was U13. It had a noise of 50mVpp due to piled-up random pulses (Attachment 4). I just tried to replace C63 with 1.2nF
and remove a soldering jumber of W1. There still looks random glitches there. But it's no longer the round shaped pulses but a sharp gliches
and the amplitude is 20mV each (Attachment 5). In fact, later I noticed that Q9 is not stuffed and W2 is closed. This means that the +24V external supply is
directly connected to +24AMP. Therefore U13 has no effect to the 24V suppy system.

3. Restoring all connections / final check of the voltages

Restore the middle and top PCBs to the intra-PCB connector board. Attach the front panel. Restore the SMA connections.

The missing soldering of the SMA cable (reported in the previous entry) was soldered.

Once all the circuits are connected again, the power supply voltages were checked again. There was no sign of oscillation.

All the above modifications are depicted in Attachment 6.