Detecting Pulsed Signals with hackrf_sweep - Understanding Timing and Sample Collection

[UnseenBanana]

trafficstars

What would you like to know?

I'm trying to detect several pulsed signals using hackrf_sweep but they're not appearing in my spectrogram output. I suspect this might be related to timing mismatches between my pulse characteristics and hackrf_sweep's sampling behavior, so I'm trying to understand exactly how the sweep process works.

Context:

• I have pulsed signals that I wish to capture in my spectrogram
• For additional context, each pulse is about 0.5 ms, and repeats every 150 ms. BUT, I want this to work with pulses that have different durations and periods as well
• The command I used is hackrf_sweep -f 5000:6000 -a 1 -l 24 -g 24 -B

My Understanding So Far:

1. The sweep is initialized with hackrf_init_sweep() and parameters are sent to firmware via USB control transfer (HACKRF_VENDOR_REQUEST_INIT_SWEEP)
2. Firmware takes control of frequency changes and continuously streams data to the rx_callback function
3. The rx_callback processes the data stream and outputs it according to the specified format

Specific Questions for Pulsed Signal Detection:

1. FFT Processing and Timing:

   • How long does hackrf_sweep dwell at each frequency step?
   • With my 0.5ms pulses every 150ms, what's the probability of capturing a pulse per frequency step?
   • Is each FFT window computed from a single time snapshot, or averaged over multiple captures?

2. Output Format Understanding: In hackrf_sweep.c, I see the FFT output is written in offset chunks:

   // First chunk
   fwrite(&pwr[1 + (fftSize * 5) / 8], sizeof(float), fftSize / 4, outfile);
   // Second chunk
   fwrite(&pwr[1 + fftSize / 8], sizeof(float), fftSize / 4, outfile);
   

   • Are these chunks in correct frequency order for spectrum analysis?
   • Could short pulses be getting lost in the FFT averaging process?
   • This relates to issue #412 point 2: do I need to reorder these chunks to get a coherent spectrum? If this point has already been answered, please let me know where I can find the answer too!

3. Optimization for Pulsed Signals:

   • Would increasing dwell time (reducing sweep speed) improve pulse detection?
   • Are there better alternatives to hackrf_sweep for intermittent pulsed signals if I wish to cover large frequency ranges (like 5-6 GHz)?

My Concern: I suspect my pulsed signals might be averaged out by the FFT process, and understanding hackrf_sweep's exact timing and sample collection behavior will help me determine if the issue is timing-related or if I need a different approach entirely.

Any insights into these timing relationships and suggestions for optimising pulsed signal detection would be greatly appreciated!