Hi,
Kirsty sent the amplifier to me, and I can confirm that it was indeed crossover distortion, as the earlier voltage measurements suggested. The culprit was T3, which had a dead short between collector and emitter. This is a later version which uses a "
T2" rather than an OC71. I replaced it with an AC128.
The noise floor was reasonable, but I decided to replace the Lockfit transistors anyway. This made it about 6-9dB quieter, and the BC108s look the part. The hardest thing was finding exactly the right shade of green sleeving for the base leads! The noise floor is about 1mV, depending on how you measure - the spec says -60dB w.r.t. to 1.5 watts, which is 4.7mV RMS, so it's now well within spec. I thought it was worth paying attention to this as the set is used quietly late at night.
Otherwise, everything else was fine. It uses the blue Mullard caps that have a near-100% reliability record in transistor sets, so it came as no surprise that they measured fine (and were left well alone). I tightened up the contacts of the 5-pin connector and went over a few solder joints, finalised the settings of the presets, and double-checked all was well a couple of days later.
The photos show the crossover distortion on the 'scope, and the amplifier connected to my test box, which might be of interest to those who service these sets. Using parts from a scrap chassis - including the wiring - and a spare die-cast enclosure, it's very easy to test an amplifier without needing a radio to connect it to. The BNC on the top left is the input signal, the 3 pots are volume, bass and treble, power goes in on the red and black 4mm connectors, and the speaker output leaves via the BNC on the top right - this goes to a 'scope or distortion analyser or whatever else you might need. The toggle switch chooses between 15 or 30 ohms impedance (and it has a centre-off position for open-circuit). It's very simple, but I wish I'd built it years before I did!
Hope this is of interest,
Mark