Well the repair and modification is done. Here's what it looks like now, from the side. The top would be to the right of the picture. The big capacitors are vertically above the transformer. The plate that screws to the back of the cabinet is at the bottom of the photo.
The owner got some information from KEF. They won't repair this model any more, but suggested C52 (labelled 5 in the photo) was changed, along with the two big capacitors. It was a useful suggestion as C52 is nominally 47 μF and it measured less than 1 μF and appeared to be discoloured. The two big caps were going to get changed anyhow.
C52 smoothes the supply to the relay that isolates the speaker. The supply to that is half-wave rectified from the main transformer. Label 3 on the photo shows where there was one pair of the 2.2 kΩ, 2W resistors. You may be able to see the outlines of where they were on the circuit board. The other pair would have been behind C52 in the photo, between 5 and 4. Those resistors were running at well over half their rating, even when the amplifier was on standby. C52 had been cooked.
Anyhow, to stop it happening again, I arranged an alternative supply to the ±15 V regulators. The regulators are 7815 and 7915 and you can see their heatsinks at 4. I wound 2 lots of about 65 turns on the transformer (1) to give around 2 x 15.5 V rms with centre tap, and rectified and smoothed that with the components at 2 to give ±20 V which is wired to the regulators.
It all seems to work fine.
I still can't work out why the four 2.2 kΩ, 2W resistors were there. With the ±15 V regulators no longer fed from the big capacitors, the ±55 V supply decays quite slowly, maybe taking a minute to get to ±20 V. The resistors would have brought that down to 10 seconds or so, but it would have been less than a minute if the regulators were also taking power from the big capacitors. I can't see why the manufacturer was so keen to get it to turn off fast. It's a subwoofer. No-one ever turns off a subwoofer at the mains or the switch on the back to silence the sound, as it doesn't work. All the mid-range and higher frequencies will be running. If the sound has already been muted, I can't see any problem with the capacitors going down slowly.
Now that the ±15 V supply is separate, that decays in less than a second. It could be that the amplifier does something funny if the main power supply decays while the ±15 V supplies still work. However, the relay would turn off in well under as second, so I can't see what the amplifier could possibly do that would make a 10 second discharge important, especially when it was those resistors that we such major cause of unreliability.
The other question that I have is what would the effect of C52 being effectively missing? Here is the circuit.
The "V" at the top right is the connection to one of the transformer windings, so D10 is half-wave rectifying that. The IC's output is a transistor that connects the bottom of the relay to ground, or not.
The relay has inductance, which is why D9 is needed, and I would have expected it have enough inductance to work fine without C52. Anyone got any experience with situations like that?
I never tried the amplifier until I had changed C52 and the big capacitors, so I don't know what had actually stopped it working.