We want to stabilise a 40MHz RF voltage to ~1 part in 10^6 over a time scale of a few hours by using power detector to feedback on a voltage controlled attenuator. We will be working at a constant power which we will calibrate externally, so we don't mind about the absolute accuracy of the detector, only its stability. To get this level of stability we expect to have to temperature stabalise the detector and PI circuit. Ideally, we'd like to design the circuit so that it doesn't need to be stable to more than ~20mK. But, obviously, if we had to, we could take this down to more like the 1mK level.
Looking at the datasheets, it seems the AD8361 is the only power detector which is specked to go down to 40MHz. Looking at the datasheet, the offset drift is ~ 500uV K^-1 which, for a 400mVrms input, gives a drift of 1.5e-4 K-1. The fractional gain change is ~2.5e-4K-1. So, these are in the right area for what we want, but still a little higher than we'd ideally like.
Other chips like the ADL5501 seem to have better temperature stability but have the coupling capacitor built onto the chip and so don't go down to as low frequencies. However, it's a little hard to tell from the graphs how quickly their performance falls of at the lower frequency range.
So, for this application, which chip would you recommend using?