I have lately been giving my Mark IIIs a break by running my ST-70 again. I have noticed that my ST-70 power transformer runs hotter than those in my Mark III amps - not much, but noticeably so. The ST-70 is on a top shelf same as was my Mark III amps "looking at the ceiling" as Bob puts it (sounds like Making Whoopee but never mind!).
I have fairly sensitive hands, brought on by 20+ years of software development - I can not touch my ST-70 power transformer for any extended period, it is too hot.
So... I went "researching" on this subject on the World-Wide Web and came up with the following (from various sources):
1. Many tube amplifier power transformers tend to run hot - some were actually DESIGNED to run for extended periods at temperatures around 100 deg. Celsius (that's 212 Fahrenheit for you ex-Colonials south of the Canadian border )
2. The heat is not caused by "bad transformer design" as some people out there on the WWW think, but is merely an unavoidable consequence of a transformer having to deliver several B+ voltages of 300 - 450 volts, and at the same time also driving high-current tube heaters
3. Some famous classic amplifiers, for instance the Marantz 5 or one of the earlier McIntosh amps (can't quite remember which) ran just as hot as the average ST-70 or Mark III power transformer
4. Some amplifier power transformers don't take too kindly at being cooled down (forced-air cooling etc) as it is contrary to the design engineer's parameters - obviously it won't damage it but core temperature was taken into account when the transformer was designed, so a cooled-down transformer is probably not running exactly at its design specs... (for the non-engineers, electrical conductors will change resistance in relation to temperature - carbon-based conductors decreasing with rising temperature, but most metallics including copper will increase)
Do anyone care to comment about these observations? I'm afraid it is so many years since I left engineering school that I have forgotten the maths involved in calculation Q and efficiency in transformers (been in software far too long, actually), not that I really care - the above does seem logical / plausible to me.
-- JunkyJan, BC Canada