redsea RDS decoder
Decoded RDS groups are printed to the terminal as line-delimited JSON objects
or, optionally, undecoded hex blocks (
-x). Please refer to the wiki for
input data formats.
Redsea can be used with any RTL-SDR USB radio stick with the
rtl_fm tool, or any other SDR via
csdr, for example. It can also
decode raw ASCII bitstream, the hex format used by RDS Spy, and audio files
containing multiplex signals (MPX). These use cases are documented in
You will need git, the liquid-dsp library, libsndfile, and GNU
autotools. On macOS (OSX) you will also need XCode command-line tools
Clone the repository (unless you downloaded a release zip file):
$ git clone https://github.com/windytan/redsea.git $ cd redsea
$ ./autogen.sh && ./configure && make
$ make install
It is also simple to later pull the latest updates and recompile:
$ git pull $ ./autogen.sh && ./configure && make clean && make $ make install
For a slower machine it can take some time to compile the TMC support. This can
be disabled (
If you only need to decode hex or binary input and don't need demodulation,
you can compile redsea without liquid-dsp (
By default, a 171 kHz single-channel 16-bit MPX signal is expected via stdin.
The simplest way to view RDS groups using
rtl_fm -M fm -l 0 -A std -p 0 -s 171k -g 20 -F 9 -f 87.9M | redsea
Please refer to the wiki for more details and usage examples.
radio_command | redsea [OPTIONS] redsea -f WAVFILE -b, --input-bits Input is an unsynchronized ASCII bit stream (011010110...). All characters but '0' and '1' are ignored. -c, --channels CHANS Number of channels in the raw input signal. Each channel is demodulated independently. -e, --feed-through Echo the input signal to stdout and print decoded groups to stderr. -E, --bler Display the average block error rate, or the percentage of blocks that had errors before error correction. Averaged over the last 12 groups. For hex input, this is the percentage of missing blocks. -f, --file FILENAME Use an audio file as MPX input. All formats readable by libsndfile should work. -h, --input-hex The input is in the RDS Spy hex format. -l, --loctable DIR Load TMC location table from a directory in TMC Exchange format. -p, --show-partial Show some multi-group data even before they've been fully received (PS names, RadioText, alternative frequencies). partial_ will be prepended to their names. This is good for noisy conditions. -r, --samplerate RATE Set sample frequency of the raw input signal in Hz. Will resample (slow) if this differs from 171000 Hz. -t, --timestamp FORMAT Add time of decoding to JSON groups; see man strftime for formatting options (or try "%c"). -u, --rbds RBDS mode; use North American program type names and "back-calculate" the station's call sign from its PI code. Note that this calculation gives an incorrect call sign for most stations that transmit TMC. -v, --version Print version string and exit. -x, --output-hex Output hex groups in the RDS Spy format, suppressing JSON output.
Formatting and filtering the JSON output
The JSON output can be tidied and/or colored using
$ rtl_fm ... | redsea | jq
It can also be used to extract only certain fields, for instance the program type:
$ rtl_fm ... | redsea | jq '.prog_type'
- Linux or macOS
- For realtime decoding, a Raspberry Pi 1 or faster
- ~8 MB of free memory (~128 MB for RDS-TMC)
- C++11 compiler
- GNU autotools
rtl_fm(from rtl-sdr) or any other source that can output demodulated FM multiplex signals
Can't find liquid-dsp on macOS
If you've installed liquid-dsp yet
configure can't find it, it's
possible that XCode command line tools aren't installed. Run this command to fix
Can't find liquid-dsp on Linux
Try running this in the terminal:
Bug reports are welcome. Be prepared to check back with GitHub occasionally for clarifying questions.
Also, if a station in your area is transmitting an interesting RDS feature
that should be implemented in redsea, I would be happy to see a minute or
two's worth of hex data using the
Redsea is released under the MIT license, which means it is copyrighted to Oona Räisänen OH2EIQ yet you're free to use it provided that the copyright information is not removed. (jsoncpp and iconvpp have their own licenses.) See LICENSE.