RF SoapySDR timestamp advances 2x too fast with RTL-SDR most likely due to float vs. complex sample miscount

[lambertch]

Title:
SoapySDR RF plugin: timestamp advances 2× too fast with RTL-SDR (float vs complex sample miscount)

Description

When using an RTL-SDR with the SoapySDR RF backend, the transmit (and potentially receive) timestamp is advanced exactly twice as fast as it should be.

Root cause

In lib/src/phy/rf/rf_soapy_imp.c

*secs      = floor(timeNs / 1e9);
*frac_secs = (timeNs % 1000000000) / 1e9;

but a comparison with wall clock shows that time advances exactly twice as fast, most likely caused by timeNs twice as large because a complex sample is counted twice.

[Radiomix2000]

How did you manage to work SRSRAN with RTL-SDR?

[lambertch]

By asking ChatGPT/Gemini to give me the code I need to prototype some radio SLAM algorithms, where I only care about PSS/SSS.

[eabase]

@lambertch

... to give me the code I need to prototype some radio SLAM algorithms, where I only care about PSS/SSS.

Anything you can share?

Also, the MULSPAD "paper" looks cool. Where is it?

[lambertch]

https://arxiv.org/abs/2402.15756v3