A Python radar module for simulating pulse-Doppler returns, generating range-Doppler maps (RDMs), and forming synthetic aperture radar (SAR) images. Designed for radar engineers and students who want to build intuition for how RDMs and SAR images are formed, how waveforms affect resolution, and how DRFM electronic attack techniques appear in the RDMs.
pip install rad-labgit clone https://github.com/JohnNehls/rad-lab
pip install -e ./rad-labfrom rad_lab import rdm, Radar, Target, Return, barker_coded_waveform
radar = Radar(
fcar=10e9,
tx_power=1e3,
tx_gain=10 ** (30 / 10),
rx_gain=10 ** (30 / 10),
op_temp=290,
sample_rate=20e6,
noise_factor=10 ** (8 / 10),
total_losses=10 ** (8 / 10),
prf=200e3,
dwell_time=2e-3,
)
waveform = barker_coded_waveform(10e6, nchips=13)
return_list = [Return(target=Target(range=0.5e3, range_rate=1.0e3, rcs=1))]
rdm.gen(radar, waveform, return_list)
Other available waveforms: uncoded_waveform, random_coded_waveform, lfm_waveform.
For additional RDM examples see apps/rdms,
the gallery,
or the API docs.
rad-lab also supports stripmap and spotlight SAR image formation from point-target scenes:
For more SAR examples see apps/sar or the gallery.
Many radar subsystems are demonstrated as standalone scripts in apps/exercises. Each file builds intuition for one concept and can be run directly. Topics covered include:
- Range equation
- Pulse-Doppler processing
- Waveforms and cross-correlation
- Ambiguity function
- Datacube processing and windowing
- Keystone formatting
- Detection theory
- Linear arrays and monopulse
- Stripmap and spotlight SAR
Every figure these exercises produce can be browsed in the gallery.
The RDM and SAR simulators in rad_lab use a few standard simplifications.
They are noted here so it is clear which physical effects the simulator
deliberately omits.
- Stop-and-hop (start-stop) propagation. Within a single pulse the
radar and target are treated as stationary; motion happens only between
pulses. Round-trip delay and carrier phase are evaluated once per pulse,
at the pulse-transmit instant, and the matched-filter template is a
perfect replica of the transmitted pulse. This is the standard
pulse-Doppler / SAR signal model (e.g. Richards, Fundamentals of Radar
Signal Processing, §8). Consequences: no intra-pulse range walk and
no Doppler time-scaling of the echo — all target motion appears as the
pulse-to-pulse phase progression
-4π f_c R(t_m)/c. - Point scatterers. Targets are ideal points with a scalar RCS; no glint, no extended-target spread.
- No propagation medium effects. No atmospheric attenuation, no ionospheric dispersion, no multipath.
- Ideal receiver chain. Linear, time-invariant, with thermal noise modeled from the receiver noise figure and operating temperature.
Contributions are welcome. Please fork the repository and submit a pull request.
This repo uses pre-commit for ruff linting/formatting and the fast unit tests. After cloning, install the hooks once:
pip install pre-commit
pre-commit installRuff and the unit tests then run on every git commit. The slower apps
regression runs in CI (see .github/workflows/python-app.yml), which runs the
full suite on main/dev pushes and on pull requests, and gates merges into
main via branch protection.
To run the test suite manually:
python -m pytest tests/ -v # unit tests (fast; apps regression is deselected)
python -m pytest -m apps # apps regression: run every apps/ script headless,
# comparing stdout against tests/app_baselines/ and
# each figure against tests/app_baselines/figures/After an intentional change to a script's output or plots, refresh both the stdout and image baselines with:
RADLAB_UPDATE_APP_BASELINES=1 python -m pytest -m appsNew app scripts are picked up automatically. Seed any randomness
(np.random.seed(0)) so the stdout and figure baselines are stable, or name
the file *_no_test.py to have the regression skip it. Figure comparison is
RMS-pixel based, so a matplotlib upgrade may require a baseline refresh.
This project is licensed under the GPL-3.0 License - see LICENSE for details.