Thanks for your comments. I made those measurements starting at 10:15 am today, after the radio was off since about 11 pm yesterday.
Today I switched on the K3 after several hours off-time.
I read the following temperatures immediately after turn-on:
Amb = 77.1°F (25°C), PA = 33°C and FP = 28°C. Here again the temperature gradient between PA and FP was 5°C. This indicates that there is obviously a misadjustment in the temperature calibration (I didn't yet find it in the CONFIGURATION) if there is such a thing at all. Why should PA be higher than FP, when the transmitter has not been activated at all.
Repeating the test after more than an hour off-time with Amb = 77.3°F (25°C) the PA and FP values were the same as just described above.
BTW, my K3 draws 0.91 A at 13.9 V (according to the display), which means 12.6 W power dissipation in receive mode.
I notice the same 1°C (or 1 K, to be precise) drop of PA when transmitting into a 50 Ohm power attenuator (50 W adjusted and reading 60 W on my calibrated home made power meter) only for the duration of making a reading on the display. This might be due to supply voltage dependency of analogue circuitry including the AD-converter (discrete or in the MCU).
I didn't yet find the temperature sensors and their associated circuitry in the K3 schematics.
I would like to check the temperature stress on the transistors in the final amplifier.
For this of course I would need to know the thermal resistance of the heat sink as well.
Thanks for sharing your experience with the radio.
73, Robert (W4BCZ, aka HB9BCZ)