pysdrweb

Project to create a web interface to RTL-SDR tooling.

FM Radio Receiver

The first focus of the project is to make a simple, web-based FM Radio receiver that can be controlled entirely through a web browser. By visiting the main page, the user can listen in on an (FM) radio station using any compatible RTL-SDR dongle and change the frequency, as desired.

This receiver is backed by some driver, which implements the set of calls necessary to run the radio, i.e. calls to generate a raw PCM data stream that is then encoded on the fly. Usually, this driver is simply the RTL-SDR program: rtl_fm, which the server searches for by default. (Other driver types are planned for the future, but are not yet implemented.)

By default, python supports some (usually uncompressed) formats out of the box:

• WAV: Raw format used in CDs and other media.

If soundfile is also present (and installed correctly), then this can also encode into more formats:

• FLAC: Free Lossless Audio Codec, supported by most browsers with some possible compression.
• OGG: Container format, supported by browsers. This format usually compresses fairly well, especially compared to more raw formats. (The list isn't exhaustive, but the goal is to support more browsers, not as many formats. The code should be easy enough to add more formats, however, if that is ever really needed.)

If lameenc is also present, then this can also encode into still more formats like:

• MP3: Enough said.

Currently, this server appears to work on Firefox and Chrome. With HLS, Safari should also be supported, though this will require the app to stall for a bit to build up the appropriate HLS buffering.

Goal

The goal of this server is to support multiple formats and multiple clients simultaneously, without straining too many system resources; the hope is that this server is low-key enough to run on weaker hardware (yes, even with python).

Running the Server

Currently, the server requires rtl_fm to be installed and a valid RTL-SDR device to be plugged in.

The simplest way to run the server is:

pysdrweb_server -p 8080 -f 107.3M

This runs the server on port 8080 and listens (initially) on 107.3 FM.

If rtl_fm is not on the path, that can be passed in with:

pysdrweb_server -p 8080 -f 107.3M -m ${RTL_FM_PATH}

For the full set of options, it is easier to configure the server with a configuration file. A sample configuration file is shown below:

#
# Sample 'native' driver configuration file.
# 
---
port: 9000
# Uncomment to require auth to change the frequency.
# auth:
#   type: basic
#   # This indicates whether 'readonly' (GET) requests require auth or not.
#   ignore_on_read: true
#   # Each user is denoted by: <user>: <password>
#   users:
#     admin: admin

# Starting (FM) channel/frequency to listen on startup.
default_frequency: 107.3M
# Driver settings.
driver:
  # Path to rtl_fm, if not on the path of the server.
  rtl_fm: /usr/local/bin/rtl_fm
  # Optional.
  kb_buffer_size: 128

# HLS Streaming Options
#
# Uncomment below to enable HLS with various options. HLS is a protocol
# that handles streams of media by splitting it into 'chunks', and serving
# those chunks as static files. The client will piece these chunks together
# into a continuous stream; the chunks are advertised to the client as an
# "HLS Playlist", a special file (with .m3u8 extension) that points to each
# chunk.
#
# NOTE: HLS is currently needed to support Safari/iOS browsers.
#
# Uncomment to enable HLS:
hls:
  enabled: true
  # Number of chunks (distinct audio files) to store.
  chunk_count: 6
  # The length of each individual chunk.
  seconds_per_chunk: 10
  # The format of the chunk. If unspecified, this will be a sensible default.
  # format: flac

Then, to run the server:

sdrfm_server -c config.yml

This permits adding authentication as well.

Server REST API

Currently, the server supports a REST API that the main page invokes to render. The main calls are described below:

Method	Route	Description
GET	/api/frequency	Returns the current frequency the server is listening on.
POST	/api/frequency	Change the frequency the server is listening on.
GET	/api/procinfo	Returns any (stderr) logging for the current driver.
GET	/api/radio/audio.<ext>	Fetch the audio streamed from the radio.

Audio Route (More Detail)

Much of this server relies on /api/radio/audio.<ext> to function since this is the route that actually serves the audio file. This route uses the given <ext> extension to determine the format to stream in. The route also accepts an optional timeout parameter, telling the server how long to listen for (in seconds). If not passed, the server will continue streaming indefinitely.

To download 10 seconds of the audio in WAV format (assuming the server is running locally on port 8080) call:

curl http://localhost:8080/api/radio/audio.wav?timeout=10 > audio.wav

If a format is not supported, the response will indicate as such.

Authentication

By default, the requests are not authenticated. If configured, however, the requests can all require authentication using "HTTP Basic"-style authentication. When making a request, the browser will prompt for the username/password combination.

Other authentication may be added later.

How It Works

For simplicity, the RTL-SDR tooling includes rtl_fm, which streams FM audio data as raw, mono, 16-bit (2 byte) PCM data at some configurable frequency. For example, the following shell pipeline streams the frequency 107.3M to an MP3 file:

# Listen on 107.3 FM, sampling at 48kHz
rtl_fm -f 107.3M -M fm -s 48000 | \
sox -traw -r48000 -es -b16 -c1 -V1 - -tmp3 - > output.mp3

The sox command is used to convert raw PCM data into framed MP3 data. (It is possible to use ffmpeg in a similar vain as well). The quality is quite bad; we need to sample more quickly then resample down to a lower frequency:

# Sample at 200k, then resample down to 48k.
# Passing '-A fast' denotes the way to perform the resample.
rtl_fm -f 107.3M -M fm -s 200k -r 48k -A fast | \
sox -traw -r48k -es -b16 -c1 -V1 - -tmp3 - > output.mp3

This command can be run remotely and played (once) in the browser using:

# Like above, but pipes the output to port 8080.
rtl_fm -f 107.3M -M fm -s 200k -r 48k -A fast | \
sox -traw -r48k -es -b16 -c1 -V1 - -tmp3 - | \
socat -u - TCP-LISTEN:8080

This pipeline is cool, but has some obvious problems. In particular, the pipeline ends as soon as the browser stops listening for any more input. Also, only one browser/client can listen at a time. (Also, some browsers might not support this for some formats...) Once finished, the command must be executed again. Changing the frequency also requires rerunning or otherwise changing the command.

To address these issues, pysdrweb will buffer the PCM output from whatever source (currently rtl_fm) and encode it on the fly, which permits multiple clients and formats. This also permits changing the frequency more easily.

Future

A few low-hanging fruit features to consider adding:

• Scan for available channels.
• More sophisticated UI (Theming and Features)