FPGA Central

A common way of demodulating a 2-FSK signal involves shifting the complex
baseband up and down by the deviation, convolving each shifted result with a
rectangular filter, and differencing the magnitudes at the filter outputs.
This essentially places a sinx/x response at each tone frequency, and works
well considering its low complexity.

Now, suppose the signal is GMSK. The deviation is one fourth of the bit
rate, just like MSK. Would it be reasonable to do the same as above for
2-FSK, but replace the rectangular filter with the Gaussian data filter
instead? Is this done in practice? How would it compare to a
discriminator-based detector?

Thank you.

John

John Sampson wrote:
>
> A common way of demodulating a 2-FSK signal involves shifting the complex
> baseband up and down by the deviation, convolving each shifted result with a
> rectangular filter, and differencing the magnitudes at the filter outputs.
> This essentially places a sinx/x response at each tone frequency, and works
> well considering its low complexity.

This is the simplest incoherent detector.

> Now, suppose the signal is GMSK. The deviation is one fourth of the bit
> rate, just like MSK. Would it be reasonable to do the same as above for
> 2-FSK, but replace the rectangular filter with the Gaussian data filter
> instead?

That is not going to work well because of ISI.

> Is this done in practice?

No

> How would it compare to a
> discriminator-based detector?

The simple way for GMSK demodulation is the discriminator with backwards
adaptive threshold. This method loses about 1dB to optimal demodulation
for any reasonable BT value. The topic has been treated in the numerous
textbooks such as Proakis, etc.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com