# gmsk: suppression of higher-frequency spectral components

I study gmsk modulation and worked with gmsk implementation, BT = 0.5. I have decided to test with BT = 0.25, which is used for near Earth communicaiton.

Spectrum of output signal with BT = 0.5, oversampling ratio 4 and truncated length of the filter 3

I used the same simulation and plotted for BT = 0.25, oversampling ratio 4 and truncated length of the filter 3

Obviously, the -70dB suppression of higher-frequency spectral components is achieved at BT=0.5 (the first figure), but not at BT=0.25 (the second plot).

What can I change if I want -70dB for BT = 0.25?

EDIT 1 new spectrum for BT = 0.25. Matlab code is below ( edit 2)

• (ignoring any domain-specific background knowledge, we can ask ourselves:) You have three parameters, one of which is just an interpolation factor, and one you explicitly fix. Can you address your concern by varying the remaining one? If that's not possible, then your requirement is impossible. If it's possible, you found your solution. Commented May 23, 2022 at 9:57
• (suddenly re-discovering our domain knowledge:) 3 is a ridiculously short length for a filter than needs to achieve a suppression of 70 dB (which is very much, and unusually high in most practical applications). You see that in your top graph very well. Does that look more like a bell curve, or more like |sinc|? Commented May 23, 2022 at 10:01
• @MarcusMüller thank you for your comments. I will "play" with my simulation. The first one looks more like abs(sinc), the second one looks more like a bell Commented May 23, 2022 at 10:29
• and you know why, right? (this doesn't ask for experimentation, honestly, this requires you to think about why you're truncating, and what happens when you do that. It did not randomly fall from the sky that I said |sinc|.) Commented May 23, 2022 at 10:33
• @MarcusMüller FrHart64 how do you determine the suppression? is it the level of sidelibes? which part of the spectrum gives you it? Commented May 23, 2022 at 10:51