Hello, everyone!

I wonder, is it incorrect to use IIR filters for NRZ FSK signals? As I understand, IIR filter comes with a frequency-dependent delay and introduces synchronization issues. Am I incorrect in that?

You're right that the phase response of IIR filters typically makes them unsuitable for use in modulators or demodulators. This often leads to BER performance degradation caused by distortion, which can be quantified. If the remaining performance meets your specifications under expected channel conditions, you might be able to make do with it, but usually, the degradation is too significant.

Thankfully, simulating this scenario is relatively straightforward, allowing you to determine if it will be effective for your needs or not. For what it's worth, it's quite uncommon to encounter IIRs used in this context.

Hi,

Where would this filter reside?  After the demodulator?

No, it was thought to be before demodulator.

In that case, I suspect an IIR filter might be OK, but I am not an expert.  Any filtering after the demodulator can add ringing to an NRZ signal, so an IIR filter (e.g. Butterworth) might not be appropriate there.

Oh, ok. Thank you for your reply! I will try and look more closely at what happens.

Could you please explain some advantages of filtering after demodulator?

Sure.  FM demodulation includes differentiation of phase.  The frequency response magnitude of a differentiator is proportional to f.  The noise power passed is proportional to f^2.  So it behooves the designer to only pass the bandwidth of the signal and then have a sharp cut-off. (Such a filter for an FM demod used to be called a "roofing filter"). The simple differentiators often used, such as central difference, pass a lot of high-frequency noise.

Of course, if you put a sharp filter at the demod output, this will affect the NRZ data.  I don't have much experience with FM demods for data, so I'm not sure how filtering is typically handled.  No doubt there is a lot of literature out there about it.  One way to avoid ringing would be to use a window function as the filter (e.g. triangle or Kaiser), but then you would not have a flat passband response.

Thank you very much for the response!

You're correct that the phase response of IRR filters usually prevent them from use in modulators or demodulators.   There is usually BER performance degradation due to the distortion which can be measured.   If the remaining performance is within your spec in the expected channel conditions then you can get away with it, but usually the degradation is too much.

Fortunately this sort of thing is pretty easy to simulate, so you can assess whether it'll work for you or not.  FWIW, it's very rare to see IIRs used in this context.

Then, I guess, FIR is preferable? Is it common practice to avoid filters altogether and search for a better demodulation technique?

Filtering is necessary to maximize SNR and BER performance, e.g., matching the pulse shapes and/or spectrum in the modulator and demodulator.   Doing this with an IIR typically distorts the phase.   A symmetric FIR has a linear phase response but allows matching the transmitted signal.