Power measurements with Spectrum analyzer vs broadband meter

[cncjerry]

So let's say you measure a signal's power with a spectrum analyzer and the 20Mhz fundamental is -10.8dbm followed by harmonics of -52.7dbm, -70.2dbm and -75.3dbm, -67.6dbm and -71.9dbm.  What would and should a broadband power meter read?  The sumation of the fundamental with the harmonics?  I don't have an accurate power meter, have been thinking of building one using one of the AD chips and an STM32F4 and display but was wondering how you would calibrate it? 

The same question applies to my signal generator, an HP 8648C, which I assume generates a fundamental at the power level selected.  I was thinking about calibrating the power meter using a table of power levels and frequencies but then I thought about this issue.  But I guess it also comes down to what you want to measure.  I would generally be measuring the fundamental with a power meter so using the signal generator would get me close enough, or would it?  It would also depend on sine vs square vs triangle, no?

I was also wondering about people I know that have used power meters to determine the insertion loss of a low pass filter.  Since the filter has more impact on the harmonics, wouldn't the insertion loss really not be anywhere near correct (on a low pass filter)?  In my example above, if a broadband meter is reporting the sum of the fundamental with harmonics within its range, it would report a higher insertion loss on a low pass filter than actual because the harmonics are now attenuated, no?

Opinions?  (ha, sort of dumb asking this group for opinions as I will get them without the question  )

Thanks

[T3sl4co1l]

Check out Parseval's theorem, and the spec's operator's manual probably discusses this.

In short: it doesn't matter much unless the signals are close together in amplitude.  If they are significantly different (>10dB), the effect is small.  This comes from geometry: you're calculating the RMS sum over all frequencies, so the stronger signals are most heavily weighted.  RMS is just the Pythagorean theorem, evaluated for as many dimensions as there are frequencies.

Tim

[G0HZU]

A lot depends on the technology used for the power meter detector. Some are (much) more sensitive to harmonics than others.

eg  some broadband diode based detectors are linear so they will be very prone to uncertainty/errors from harmonics. eg a harmonic at -40dBc can give approx +/-2% uncertainty of the power measurement here. -30dBc can give about +/-6% uncertainty in the measurement.

Some Analog Devices chips are similar, eg the log amp based detector chips can give issues with uncertainty if there is a harmonic at -40dBc although in both cases a lot depends on the order of the harmonic.

[G0HZU]

The other thing to be wary of is the overall measurement uncertainty of the spectrum analyser. This will be affected by its input VSWR and its overall flatness wrt frequency and a whole load of other factors.

Also, on a side note, if you are interested in measuring harmonics, the analyser's ability to measure harmonics accurately depends on the performance/linearity of the front end of the analyser. This is because it will generate its own harmonic content in its front end. Ideally, the harmonics self generated inside the analyser need to be >15dB lower than the harmonic level you are trying to measure otherwise the measurement uncertainty starts to creep towards 2dB or worse.

[Fank1]

Assuming  !! that the instrument has a totally flat response.
Convert the individual readings from dbm to watts add all of the levels together and convert back to dbm.

[cncjerry]

I am going to play around with the ADI development boards I have to see what happens with a square wave vs a sinus.  I have 4 different detectors I think, three on one board and a couple of 8207 boards.

But then again, the RMS of a square wave is calculated differently than a sinus so I'll have to think about that as well.  My analyzer for the test will be my 8566b as it is the most recently calibrated.  I also have an 85685 preamp that has a comb generator for calibration so I would think it would be pretty flat.

[RadioNerd]

The best choice for true-RMS power measurements are thermocouple or thermistor based sensors. They basically measure the heat produced by the input signal applied to a 50 ohm resistor, They are therefore inherently independent from the measured signal shape.
The measurement error is then directly proportional to the ratio between the power of the fundamental component and the total (RMS) power of the harmonics