For additional information on Level-0 product decoding, see the SAR Space Packet Protocol Data Unit Specification and the Sentinel-1 Level-0 Data Decoding Package.
sentinel1_decode is a C++ program/library for quickly decoding Level-0 Raw data from the Sentinel-1 satellite, and forming Level-1 SLC images. Currently, image formation for stripmap, IW swaths, and IW bursts mostly works for visual analysis, but won't have accurate phase and may contain some ambiguities. I am creating this as an education experience for myself, and because there isn't a good fast, simple, and robust program for decoding this data in a way that allows you to get the intermediate products and signals.
On my PC (Ryzen 9900x and 64GB DDR5 RAM), azimuth compressing a full stripmap image takes around 1 minute, and uses ~52GB of RAM. Azimuth compressing an IW mode swath takes approximately 2 minutes, and uses ~27GB of RAM. A burst takes about 30 seconds and uses ~20GB of RAM. Memory usage will go down as I move from double to float where possible. Decoding the raw data for a swath takes only a couple seconds.
Check out these other projects which were very helpful:
Rich Hall's sentinel1decoder (Python)
Rich Hall's sentinel1Level0DecodingDemo (Python/Jupyter)
jmfriedt's sentinel1_level0 (Matlab)
$ bin/s1_write --help
spectrogram [swath] [burst_num] [azimuth_line] [in_path] [out_path]
burst [swath] [burst_num] [in_path] [out_path]
swath [swath] [in_path] [out_path]
range_compressed_burst [swath] [burst_num] [in_path] [out_path]
range_compressed_swath [swath] [in_path] [out_path]
range_doppler_burst [swath] [burst_num] [in_path] [out_path]
range_doppler_swath [swath] [in_path] [out_path]
azimuth_compressed_burst [swath] [burst_num] [in_path] [out_path]
azimuth_compressed_swath [swath] [in_path] [out_path]
azimuth_compressed_burst_eccm [swath] [burst_num] [detection_threshold] [mitigation_threshold] [in_path] [out_path]
azimuth_compressed_swath_eccm [swath] [detection_threshold] [mitigation_threshold] [in_path] [out_path]
save_burst_as_cf32 [swath] [processing_level] [in_path] [out_path]
save_swath_as_cf32 [swath] [processing_level] [in_path] [out_path]
Scaling Options: [--norm_log|--norm|--mag|--real|--imag]The sample image data/points/point.dat is the VV data from S1A_IW_RAW__0SDV_20240813T095440_20240813T095513_055193_06BA22_1119-RAW.
$ bin/s1_write swath IW2 data/points/point.dat IW2.tif --norm
$ bin/s1_write range_compressed_swath IW2 data/points/point.dat RC_IW2.tif --norm
$ bin/s1_write azimuth_compressed_swath IW2 data/points/point.dat AZ_IW2.tif --norm
Some basic heuristic RFI mitigation is also available. I'm currently working on a better version that should work without as much guess work. The values in the commands below are probably good starting points for VV and VH, respectively.
$ bin/s1_write azimuth_compressed_swath_eccm IW2 13000 100 data/korea_patriot/korea_patriot_vh.dat data/korea_patriot_vh_iw2_eccm.tif --norm| Without ECCM | With ECCM |
|---|---|
 |
 |
$ bin/s1_write azimuth_compressed_swath_eccm IW2 15000 250 data/petro_interference/petro_interference_vv.dat data/petro_interference_vv_iw2.tif --norm| Without ECCM | With ECCM |
|---|---|
 |
 |
$ bin/s1_print --help
packet_info [packet_index] [path]
complex_samples [packet_index] [path]
swath_names [path]
index_records [path]
annotation_record [record_index] [path]
state_vectors [path]The sample image data/sample/sample.dat is the VV data from S1A_IW_RAW__0SDV_20240806T135224_20240806T135256_055093_06B68A_AE41
$ bin/s1_print print_packet_info 0 data/sample/sample.dat
Primary Header:
packet_version_number: 0
packet_type: 0
secondary_header_flag: 1
process_id: 65
process_category: 12
sequence_flags: 3
packet_sequence_count: 11157
packet_data_length: 18621
Secondary Header:
coarse_time: 1406987562
fine_time: 7757
sync_marker: 892270675
data_take_id: 225252800
ecc_number: 8
test_mode: 0
rx_channel_id: 0
instrument_configuration_id: 7
sc_data_word_index: 27
sc_data_word: 48835
space_packet_count: 240533
pri_count: 243384
error_flag: 0
baq_mode: 12
baq_block_length: 31
range_decimation: 8
rx_gain: 8
tx_ramp_rate_sign: 1
tx_ramp_rate_mag: 1605
pulse_start_frequency_sign: 0
pulse_start_frequency_mag: 12335
pulse_length: 1967
rank: 9
pri: 21859
swst: 3681
swl: 13979
ssb_flag: 0
polarisation: 7
temperature_compensation: 3
elevation_beam_address: 6
azimuth_beam_address: 385
calibration_mode: 0
tx_pulse_number: 6
signal_type: 0
swap: 1
swath_number: 10
num_quadratures: 11938
Operating Mode Info:
Data Format: D
BAQ Mode: fdbaq_mode_0
BAQ Block Length: 256
Test Mode: measurement_mode
Sensor Mode: interferomatric_wide_swath
Signal Type: echo
Error Status: nominal
Pulse Info:
Swath: IW1
RX Polarization: V
TX Polarization: V
Pulse Length: 52.4048
TX Ramp Rate (TXPRR): 1.07823
Start Frequency (TXPSF): -28.2515
PRI: 582.367
SWL: 372.428
SWST: 98.0692
RX Gain: -4
Range Decimation: 8
TX Pulse Number: 6$ bin/s1_print swath_names data/sm_sample/sample.dat
S1_100MHz_TXCAL: 80
S1_TXCAL: 250
S1_100MHz_CAL: 20
S1: 60701$ bin/s1_print state_vectors data/sample/sample.dat
$ bin/s1_print print_complex_samples 0 data/sample/sample.dat
# 50,000 lines later...
...
(-11.201,1.59988)
(1.59988,8.00034)
(-4.80058,-11.201)
(-4.80058,20.8031)
(4.80058,4.80058)
(17.6024,17.6024)
(4.80058,31.7194)
(-14.4017,8.00034)
(-11.201,20.8031)
(-4.80058,1.59988)
(1.59988,-4.80058)
(-8.00034,-4.80058)
(-4.80058,-14.4017)
(1.59988,11.201)
(1.59988,-4.80058)My goal is to not involve too many external dependancies; however, some are necessary. OpenMP and FFTW3 are required. The image writing is done via libtiff.
Everything necessary for compilation to work on MacOS, Linux, and Windows should be available through the use of the conda environment. Windows support is still WIP, however. Install it via:
conda env create -f environment.yml
conda activate sentinel1_decode
Once the dependencies are installed, simply run:
cd build
cmake ..
cmake --build build
# If on Windows:
# cmake --build build --config ReleaseNote the the binarys will be located in ./build/bin/ (or ./build/bin/Release on Windows).