Chirp Sounder 2

This software can be used to detect chirp sounders and over-the-horizon radar transmissions over the air, and to calculate ionograms from them. The software relies on Digital RF recordings of HF. The program can be run in realtime on complex voltage recorded into a ringbuffer, but it can also be run offline on a recording on a disk.

This is a new implementation of the GNU Chirp Sounder. This new version allows you to now automatically find chirps without knowledge of what the timing and chirp-rate is. You can still figure out the true distance if you have a GPSDO, as most sounders start at a full second.

Jens Floberg's masters thesis discusses the use of chirpsounder2 to make oblique ionograms using this software.

Installation

See dependencies.txt for instructions on how to build the dependencies (tested on Ubuntu 18 & 20) You need to compile the chirp downconversion library, which is written in C.

make

There is no packaging or other installation needed. You just run the scripts in place.

Python packages that are required: pyfftw, numpy, scipy, matplotlib, digital_rf, mpi4py, h5py. Tested on Python 2.7.17 and Python 3.6.9.

Here is an example of a startup script for kicking off all of the programs. There are scripts in the same directory that also show how to install chirpsounder2 as a service with Ubuntu:

https://github.com/jvierine/chirpsounder2/blob/master/examples/sgo/sgo_ski.sh

Usage:

1) Make a data capture with the rx_uhd that requires a USRP N2x0, a GPSDO, and a broadband HF antenna in a quiet location. I recommend using a 12.5 MHz center frequency and a 25 MHz sampling rate. I recommend using this simple C++ program instead of thor.py, which comes with Digital RF, because it is more lightweight, handles dropped packets consistently, and has no dependency to Gnuradio.

./rx_uhd

Alternatively, you can make a data capture with THOR (comes with DigitalRF), Here is an example command to kick off a recording:

thor.py -m 192.168.10.3 -d "A:A" -c cha -f 12.5e6 -r 25e6 /dev/shm/hf25

Tip: You can use a RAM disk ring buffer to avoid dropped packets on slower computers and hard disks. This is not necessary, as the chirp analysis will be okay with dropped packets. Here's an example of how you can use rsync to shovel a digital rf recording on the fly from a ram disk to a hard disk.

# copy digital rf from ram disk to permanent storage:
while true; do rsync -av --remove-source-files --exclude=tmp*
--progress /dev/shm/hf25/cha /data_out/hf25/ ; sleep 1 ; done

2) configure by copying the example1.ini to e.g., configuration.ini. Edit the file to make sure you have the right center frequency, sample-rate, data directory, and channel name. I've only tested 25 MHz sample-rate and 12.5 MHz center-frequency so far.

[config]

# channel name for the digital rf recording
channel="cha"

# the sample rate of the digital rf recording
sample_rate=25000000.0

# the center frequency of the digital rf recording
center_freq=12.5e6

# the location of the digital_rf recording
data_dir="/data_out/hf25"

# detection
threshold_snr=13.0

# how many chirps can we at most detect simultaneously
max_simultaneous_detections=5

# how sparsely do we search for chirps (1 .. N) 1 is slowest, but the most sensitive
# every Nth block is analyzed 
step=10            

# how many samples per block are coherently integrated on chirp detection
n_samples_per_block=5000000

minimum_frequency_spacing=0.2e6
# what chirp rates do we look for
chirp_rates=[50e3,100e3,125e3,500.0084e3]

# this is where all the data files are produced in
output_dir="./chirp2"

# what is the range resolution of the ionograms
range_resolution=2e3

# what is the frequency step of the ionogram
frequency_resolution=50e3

# what is the range extent around the strongest echo that is stored
max_range_extent=2000e3

# how many threads are used when chirp downconverting
n_downconversion_threads=4

3) Detect chirps on the recording. can be parallelized with MPI to speed things up if you have lots of CPUs and a very fast disk. If you don't have a fast disk, using too many processes may actually reduce performance due to trashing.

mpirun -np 4 python detect_chirps.py configuration.ini

4) Run find_timings.py to cluster together multiple detections of the same chirp to create a database of chirp timings

python find_timings.py configuration.ini

5) Run calc_ionograms.py to generate ionograms based on the timings that were found. Can be paralellized with MPI. Keep in mind that adding a lot of processes may be detrimental to performance, due to the 100 MB/s read requirement. If you have a slow disk, don't use too many processes here! Each MPI process is additionally multi-threaded, with the number of threads configured in the configuration file

python calc_ionograms.py configuration.ini

6) run plot_ionograms.py to create plots

python plot_ionograms.py configuration.ini

Programs

The software consists of several parts:

detect_chirps.py # this is used to find chirps using a chirp-rate matched filterbank
find_timings.py # this is used to cluster detections and determine what chirp timings and chirp rates exist
calc_ionograms.py # this is used to calculate ionograms based on parameters
plot_ionograms.py # plot calculated ionograms

Output files

The program creates several different kinds of output files.

chirp-%017d.h5 - Files created by detect_chirps.py, indicating that a chirp was detected in a block of data being inspected. The starting frequency, time, chirp-rate, chirp-time, and signal-to-noise ratio are recorded. Chirp-time means the virtual time at which the chirp started from a frequency of 0 hertz.
par-%11.3f.h5 - Files created by find_timings.py, which analyzes chirp-*.h5 files and determines what are the sounder parameters. By default, three independent detections of the same chirp at different times with consistent parameters to classify the chirp as real. This is to avoid false positives.
lfm_ionogram-%03d-%11.2f.h5 - Files created by calc_ionograms.py. These contain the ionogram itself. Optionally the chirp downconverted raw voltage can also be stored in order to allow the chirp to be reanalyzed with different spectral analysis settings.