Please see this question: Why samples per symbols $\geq 2$ in the GNURadio Constellation Modulator block?. The first answer here says "filtering the signal, you have bandwidth expansion". But look at the figure in this question: Relation between Bandwidth and Baud-Rate for 8-PSK - it looks like there is much less bandwidth required for the RRC filtered signal than the square wave / sinc. So why does the first question have an answer that says that increased bandwidth is necessary and, taking it a step further, why does sps increase, as was the actual question there?
Asked
Active
Viewed 103 times
1 Answers
0
I wouldn't really call it "bandwidth expansion". If you do 2 samples per symbol, from a purely theoretical point of view (i.e. using perfect infinitely long sinc pulses), you could do with exactly half of the Nyquist bandwidth of the output signal.
However, RRC is not infinitely long sincs, so you get excess bandwidth.
Marcus Müller
- 30,525
- 4
- 34
- 58
-
In comparison to what? Isn't the default (non-filtered) square waves in time? So you have less BW with a filtered version, which is essentially my question – Ken - Enough about Monica Mar 25 '21 at 16:10
-
no, you're not comparing it to these, but to the theoretically sufficient minimum bandwidth, as I wrote. These are never "the default"; there's no system that would choose them. – Marcus Müller Mar 25 '21 at 17:35