You've done pretty well to get this far and it looks like various capacitors and resistors are the main cause of problems in your set.

The higher current 6.3v AC winding needs to feed the dial lamps and all the heaters except for the rectifier.

The lower current winding only has to power one heater - the rectifier so the current requirements are much lower. The HT is connected to the rectifier heater to prevent problems with heater - cathode shorts in the rectifier. Because they are connected together, there is no stress on the insulation. The 6.3v AC winding on the transformer is isolated from all the other windings so it can "float" at 250v DC without having any effects on the rest of the radio.

The low heater voltage (5.6v instead of 6.3v) could be due to your multimeter. Some multimeters only read average volts not RMS volts and so will read low. The other possibility is that your mains supply is only around 230 to 240v while the transformer is set to the 250-260v tapping.

To solve your varying volume problem your going to have to replace all the remaining resistors around the volume control circuit. If this does not solve it only then should you start digging around inside IFT2 again.

I've think I've homed in on IFT2 as the probable location of the volume-fade fault.

There are no original resistors at all left in the signal path, and the few between B+ and chassis all test correct. Voltages all seem stable.

Very consistently, the radio switches on after an overnight rest and comes to full volume. After about ten minutes, it gradually fades over about 30seconds to a much lower volume and stays there. Signal at the lower level is not distorted in any way - just very low.

I did what I probably should have done earlier and hooked up my 2-probe oscilloscope to watch what happens.

Probe B is between IFT2 pin 6 and the volume control, after the 100K resistor.

Probe A is on IFT2 pin 1, fed from plate of V2.

Scale is 1ms, where signal modulation is clearly visible.

At cold start,
A shows AC modulation range of 50millivolts,
B shows DC modulation range of 500millivolts.

When the fade kicks in,
A shows no change and stays at 50mv
B declines to 50mv - 10% of original amplification

B+ voltage does not change and bias voltages don't change, as far as I can see.

I'm guessing a short inside IFT2. Is there any other possibility?

Maven

I wouldn't jump to any conclusions yet without more tests. Your approach so far is exactly what is needed, an evidence based approach to fault finding.

Assuming you haven't made any typo's in your last post this is an interesting fault.
By your description, probe A is on the detector on the secondary side of IFT2 (L7). Probe B is on the primary side of IFT2 (L8). Your saying that you loose 90% of the signal on the primary side yet the signal on the secondary side doesn't change.

This could be due to IFT2 or more likely a fault before IFT2.
Did the anode voltage on V2 stay the same when the AC component dropped from 500mV to 50mV?

To make sure that the fault is in IFT2, have the CRO monitor pin 4 of L6 (signal grid of V2) on Ch A and Ch B monitoring the anode (pin 1 L8). If the grid signal doesn't change but the anode does then the fault definitely lies around the V2 / IFT2 section. I'd also check that the screen voltage of V2 (junction of R22 and C24) does not change when the fault occurs.
If the screen voltage, grid signal/voltage and anode voltage of V2 don't change when the fault occurs then that indicates the fault is most likely within IFT2. The most likely fault would be failing mica caps so try replacing those first. You've also considered a failure in C11? If this became leaky it would mess with the AGC biasing and cause exactly what you describe.

Normally when the volume snaps high it is the earthy end of the volume pot going open.

Do not overlook the AGC and that new caps can fail. If the AGC is loosing its control voltage then the the set may well go loud on a strong signal.

Marc

BIG APOLOGY - there was a typo in my previous post. It's the primary on IFT2 that doesn't change, and the secondary that drops 90%. I've edited the post to correct that.

I'm pretty sure there is nothing wrong around the volume pot. It still works normally and without any noise even on the faded input. AGC I'm not at all confident about and will check those voltages at next opportunity. The schematic does not convey the theory of operation, and there seem to be feedback paths all the way to IFT1 from the amplifier section.

All caps are new polypropylenes or silver micas and none are open or shorted.

Maven

I thought something was not quite right!

Your correction makes a lot more sense but I would still eliminate any easier to get at potential failures first. Silly things like a dirty valve socket (poor valve pin connection) and other strange things could also cause this problem.

That set has delayed AGC if I am reading the right circuit.

The AGC voltage is sourced from the pentode plate of V2 via c11 and fed to a diode plate of the same valve. The audio is derived from the other diode plate via the IFT

If c11 goes awry that will feed positive into the AGC and V1 will not be happy. The principal components in the AGC are the AGC voltage dividers R4 & 5 and the backbias resistors R16 &17. Caps 2, 6, 11& 25: C25 is positive to chassis. neg to CT.

Do make sure that repaired IFT coil wires are actually soldered well. If there is a dry joint the grid may swing wildly positive and the crack is it "flashing over" on the bad joint, giving enough track for it to operate.

DVM or CRO (High impedance meters) are the only thing with any hope of measuring AGC voltage (loading)

Marc

WA2ISE how do you remove the rivet holding the bottom capacitor and base together in the K-tran transformers that you are repairing at http://www.wa2ise.com/radios/repair.htm without destroying the rivet? If you destroy the rivet what do you use in its place when reassembling the transformer?

Thank you.
Jim