I'm developing an application with my SDR Board and I've managed to produce a calibration table to verify the output power of the transmitter for every TX configuration of antenna gain and IQ digital word, for a given center frequency.

I.e., by setting the transmitter to that fixed center frequency, I've changed the TX gain and then transmitted a sinusoidal signal at different IQ amplitudes, using a spectrum analyzer to obtain the power of the transmitted signal.

So far so good, but I've now reached some problems and I can't seem to find a clear a solution for them online. I'm just an hobbyist, I have no previously experience with RF whatsoever.

The first problem is how should I properly use the RBW (resolution bandwidth) of the spectrum analyzer within my measurements.

For instance, if I measured an output power of 10dbm with a RBW of 100Hz, does that mean that within those 100Hz, there is a total of 10dbm power? Because if that is correct, then how can I correctly measured the power at the frequency of interest (which is inside that bin)? Surely those 10dbm will also account with the power from the frequencies close to the frequency of interest and inside my bin. What is the formula that I can use to correctly measure the output power at the given frequency with some error associated?

The second problem is related to what happens to the power when I change the window applied to the signal. In particular, I have my measurements performed for a CW setup, but I intend to transmit pulses and, at every time, I must know an estimative of the power at the carrier frequency, as well as some estimative of the total power of the signal.

These pulses are just simple amplitude modulation pulses, e.g: gaussian pulses, sinc pulses, hermite pulses, hard pulses, etc...

The way I thought about this problem is that it is just the application of another window to the already rectangular window that I use when I'm obtaining the initial power results. So, my idea was that since I know the shape of the pulses that I'm transmitting, i.e., I know the windowing functions, I should somehow be able to convert the measured power into the new power that is actually sent with the pulse transmitted.

However I can't find any formulas or any theory that would enlight me and I've been stuck in this problem for a few days now. I could just perform the power calibration for different pulse shapes but that would take me ages.

Can someone enlight me with some papers/theory? Is my approach incorrect? What should be my approach then?

Thanks in advance.

  • \$\begingroup\$ Let's start with this: Why are you measuring the power? This may have some effect on what you should be doing. The CW power is in some sense the real power. You may not need to directly measure anything else unless you are trying to pass some regulation or something. In that case, you need to know how the regulatory agency will judge compliance, and make the same measurement that they will use. \$\endgroup\$
    – user57037
    Commented Jan 27, 2019 at 18:53
  • \$\begingroup\$ It's not about regulations. I'm using the SDR for NMR (Nuclear Magnetic Ressonance) applications and it is important that I know the pulse power at the frequencies of interest. \$\endgroup\$ Commented Jan 27, 2019 at 18:55
  • \$\begingroup\$ Can you update your question to explain more about what modulation you will be using? Are you going to use some form of frequency (or phase) modulation during your pulses? Or are you just pulsing a single frequency on and off? Any amplitude modulation (other than turning the signal on and off completely)? \$\endgroup\$
    – user57037
    Commented Jan 27, 2019 at 19:06

1 Answer 1


I would set RBW > 0.35/rise time = f-3dB 10~90% to optimize SNR as RBW limits the BW. The SA then measures peak of each spectral component. Do you want to measure RMS, pk or average RMS in 1 burst?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.