uhd_sample_recorder

Record and (optionally) run FFT on I/Q samples from an Ettus SDR. Samples can be compressed as written; FFT may be run on GPU (using Vulkan, including RPi4).

example usage

See https://files.ettus.com/manual/page_transport.html for notes on tuning USRP transport --args.

$ ./uhd_sample_recorder --args num_recv_frames=1000,recv_frame_size=16360,type=b200 --file test.zst --duration 2 --rate 2.048e6 --freq 101e6 --nfft 2048

FFT points can be output along with samples. This example records 1s of samples at 10.24e6sps at 108MHz, writing compressed samples to test.ci16.zst and uncompressed FFT points to fft_test.raw. Then, the points are plotted to an image output by plot_fft.py.

$ uhd_sample_recorder --args type=b200,num_recv_frames=1000,recv_frame_size=16360 --duration 1 --rate 10240000 --freq 108e6 --gain 10 --type short --file test.ci16.zst --fft_file fft_test.raw --nfft 256
$ pip install numpy matplotlib
$ ./plot_fft.py

Vulkan FFT support

Requires a Vulkan compatible GPU.

Running on a Raspberry Pi4

This works on both Raspberry Pi OS (Debian) and Ubuntu.

1. Add dtoverlay=vc4-kms-v3d-pi4 to /boot/firmware/config.txt (/boot/config.txt on Raspberry Pi OS).
2. Make sure your user has access to the /dev/dri/renderD128 device (e.g. sudo usermod -a -G render ubuntu).
3. If running Ubuntu 22.04 or later, proceed to the Build section. If running Raspberry Pi OS, you will first need to update your Vulkan/MESA drivers - use https://github.com/jmcerrejon/PiKISS to Configure and install Vulkan using the main branch first then proceed to the Build section.

build

./bin/install-deps.sh
./bin/build.sh
./bin/test.sh