[SOLVED] Accuracy of Power Measurement Techniques

Good day,

I am generating 16 GHz pulses using a mixer connected to a AWG and LO and I measured the power using a spectrum analyzer in time domain.
I encountered several comments from someone with more experience but I could not get further explanation from him.

I am a new in this field and I would like to ask the opinions of gurus in this field about his opinions.

He suggested me to:

1. each instrument is has phase noise so I should lock the phases of the instruments to minimize phase error. A maximum of about 3 dB error may occur if i did not do that. The AWG 10 MHz REF_IN is connected to the LO's REF_OUT but not to the spectrum analyzer.

2. The pulse width is 3 us and the PRF is 100 Hz. The RBW of 10 MHz is insufficient to accurately determine the power of such pulses.

I would like to know more about his comments but I am still doing a lot of hunting and I cannot really find the answer to these questions.
I might be searching in the wrong documents.

Would anyone be kind enough to shed some light or point me to the right direction please?

Thank you in advance.

if you are measuring power rise time, that is a scalar quantity. I do not see why phase coherence is important to you for the measurement. It would only mater if you were integrating the signal over many "events" of "dwells". Hooking the sources to the same reference will not hurt anything, but it also will not help.

As far as measuring pulse characteristics, spectrum analyzers are not the best thing to use. Diode detectors and log amplifiers give good power vs time measurements. If you only want to know the actual power, use a power meter and correct for your duty cycle.

thank you for your reply.

yes, i wish to measure the power during the rise time.
and sorry, what do you mean by many "events" of "dwells"?

we have a CW power meter but the same guy told me that it cant measure pulse signals.
Correcting it with the duty cycle sounds promising.
I will read up on that.

Thank you.

Nezs said:

thank you for your reply.

yes, i wish to measure the power during the rise time.
and sorry, what do you mean by many "events" of "dwells"?

we have a CW power meter but the same guy told me that it cant measure pulse signals.
Correcting it with the duty cycle sounds promising.
I will read up on that.

Thank you.

Click to expand...

Agree w/ biff on the phase-sync'ing... probably won't help much in this case. As for your power conversion using duty cycle, the concept is very straightforward.

Example: You have a 1 watt output signal and run it into a CW (averaging) power meter.
If your signal is on 100% of the time (100% duty), then the CW meter will read an average power of 1 watt.
If your signal is on half of the time (50% duty), then the CW meter will read an average power of half the power, 0.5W (1W * 50%).
If the signal is on 10% of the time, the average power detected by the meter will be 10% of 1W, so 100 mW.

Be sure to understand the minimum power that your power meter will accurately read. You should stay ~10 dB above that level to ensure that the power meter is really picking up energy from the pulse and isn't confusing it with the energy detected from noise in the system. You should be able to see a linear change in duty vs. detected RMS power... if it becomes non-linear, you've got a problem somewhere (possibly signal compression, for the high end, or swamping out due to noise contribution, at the low end).

If you trigger your spectrum analyzer off of your 100 Hz source, then you should be able to get a direct power measurement... which could also be used to correlate the power meter readings.

understood. that is certainly pretty straight forward..

If you trigger your spectrum analyzer off of your 100 Hz source, then you should be able to get a direct power measurement... which could also be used to correlate the power meter readings.

Click to expand...

i see.. so I can time the power measurements to match my pulses. I'll check if I can do that on my power meter.

Thanks again biff44 and enjunear.

The pulse width is 3 us and the PRF is 100 Hz. The RBW of 10 MHz is insufficient to accurately determine the power of such pulses.

Click to expand...

The spectrum analyzer should be able to follow the pulse envelope in zero span mode. 10 MHz RBW refers to a rise-time in in a 50 ns order of magnitude. Even with a video bandwidth of e.g. 1 MHz you still get the peak peak level of a 3 us pulse. By design, the magnitude accuracy of a SA won't be much better than 1 dB, so for highest accuracy, a power meter seems still the better alternative.

Hello,

After reading the manual on using the SA and power meters, I finally carried out my experiments..
I have the results and I would like to ask for some opinions please.

Firstly, my pulses are up-converted using a mixer with LO of 14 GHz (CW) and 2.8 GHz (pulses).

Using the SA, i obtained -30.65 dBm for 16.8 GHz using time domain (zero-span).
Using the power meter, with duty cycle correction and frequency calibration factored, i obtained -32.18 dBm.

When I measured the pulses using the power meter, are the unwanted/noise power included the result?
I am using a broadband, diode sensor (unlike the SA where i set the frequency to 16.8 GHz).

I am planning to measure the output of TWT amp using these pulses (with a 70 dB attenuator of course).
So, I am worried about unwanted signal from the mixer/spurious noise power calculated into the power meter's results with or without the TWT.

I sincerely thank you for your opinions and time..

Can you expect (by specification), that the power meter is averaging pulses linearly in the present pulse power range? What's the specified measurement uncertainty? I fear, it can be difficult to get more reliable results at this low duty cycle.

Thank you for your quick reply.
According to the specification, for the power meter, the chart is linear from -70 to -30 dBm.
As for the sensor's linearity, it is <1.8% with a rectangular probability distribution.
And you are right, as I lowered the duty cycle, i saw changes in the results although nothing else was changed.

I guess I am comfortable with +/- 1 dBm but i am rather concern that the -32 dBm includes power from other frequencies e.g. especially noise/spurious signal from the TWT.
I was told it is a problem for broad frequency bandwidth power meters (by the same guy who gave me advice on phase syncing and power measurements using SA).

If you use thermal power meter, the power should be higher if there are other spurs or harmonics. But in your situation, the power meter value is less than spectrum analysis.
Which type sensor of power meter are you using?

I am using a diode sensor but at the moment, without the TWT.
Since I am not familiar with microwave techniques, would be wise to be a bit prudent first.

I have confirmed that for powers meters, any unwanted signals are averaged into the measurement.
Looks like I would need to rely on the SA.

Thanks to everyone for their help!

Looks like I would need to rely on the SA.

Click to expand...

I think, the additional dynamic information given by the SA is important. As long as linear operation can be assumed, you are able to verify/calibrate level measurement and pulse response separately.

Agreed. Thanks again FvM for your advice!

Have a wonderful day.