In my opinion, the most uncertainty will lie with the 478A sensor and its limitations at 50 MHz rather than with the meters themselves. I'm not sure the old 432A meter itself is the weak link here especially if you use DC substitution via a calibrated >5 digit DVM.?
?
In my experience, you have to be careful using the standard 478A thermistor sensor at 50 MHz as the VSWR begins to creep up below about 70 MHz and there may also be a subtle wiggle in the return loss near 50 MHz.
?
This can very slightly affect the efficiency of the sensor where there is a wiggle as it becomes frequency dependent throughout the bandwidth of the wiggle. If you then factor in that the sensor efficiency of a 478A has to be known accurately if you want to use it for metrology then this system isn't ideal if you want to do this stuff outside a calibration lab at 50 MHz.
?
The best performance will come from a special (option) version of the 478A sensor that has low VSWR at 50 MHz and has known efficiency at 50 MHz.?
?
Despite all this I do still check my 0dBm 50 MHz source using a 432A and a 478A sensor every year or so here at home and I use the DC substitution method with a 6.5 digit DVM. I get very repeatable results. So I think the efficiency of the sensor is very stable vs time as I've had this 478A sensor for over 30 years now :)
