Here's a quick demonstration of how to create an arbitrary waveform file from a digital standard in Rohde & Schwarz's WinIQSim2 software, then resample and filter the waveform in Rohde & Schwarz's Arb Toolbox software. The output waveform was played on a Rohde & Schwarz arbitrary waveform vector signal generator.
Rohde & Schwarz's WinIQSim2 digital standard waveform creation software can generate DVB-S2X signals with whole number multiples of samples per symbol.
I needed to generate a DVB-S2X signal with a symbol rate of 118 MSym/s. My AWG was capable of playing waveforms up to 160 MHz RF bandwidth (200 MHz sample rate).
Therefore, without resampling, the maximum integer oversample rate I could use without exceeding the capabilities of the arb is 1 sample/symbol. Using a root raise cosine transmit filter with a rolloff factor of 0.2, the signal's occupied RF bandwidth would be approximately 142 MHz. To avoid aliasing, the AWG uses a 1.2:1 bandwidth to sample rate ratio, with a low pass filter to smoothly attenuate the signal. Therefore, the DVB-S2X's rolloff would be further attenuated by the AWG's digital low pass anti-alias filter if played at a 141.6 MHz sample rate.
To shift the low pass filter, we can generate the DVB-S2X signal with the RRC transmit filter at a high oversample rate and resample the waveform to AWG's maximum sample rate, extending the AWG's low pass filter cutoff frequency from +/- 70.8 MHz to +/- 80 MHz, reducing the rejection of the excess bandwidth required by the RRC transmit filter.
By rearranging when the resampling and decimation was performed, I was able to achieve slightly better EVM by performing the following:
• Generate the DVB-S2X waveform with a high oversample rate (10sa/sym) in R&S WinIQSim2 www.rohde-schwarz.com/us/software/winiqsim2/
• Resample and filter the waveform to the exact max sample rate of the AWG (1180 MSa/s → 200 MSa/s) in R&S Arb Toolbox www.rohde-schwarz.com/us/applications/r-s-arb-toolbox-application-note_56280-15443.html
This approach is also helpful for reducing the error when resampling with a ratio of M/N where M≈N.