This program generates an FM modulation, with RDS (Radio Data System) data generated in real time. It can include monophonic or stereophonic audio.
PiFmAdv modulates the PLLC instead of the clock divider for better signal purity, which means that the signal is also less noisy. This has a great impact on stereo as it's reception is way better.
For the PLLC modulation to be stable there is one additional step to do. Due to the low voltage detection the PLLC frequency can be reduced to safe value in an attempt to prevent crashes. When this happens, the carrier freqency changes based on the original GPU frequency. To prevent this, we can easily change the GPU freqency to match the safe frequency. Now when due to the low voltage detection the PLLC frequency changes to safe value, nothing happens as the normal value and safe value are the same.
Simply add gpu_freq=250
to /boot/config.txt
.
It is based on the FM transmitter created by Oliver Mattos and Oskar Weigl, and later adapted to using DMA by Richard Hirst. Christophe Jacquet adapted it and added the RDS data generator and modulator. The transmitter uses the Raspberry Pi's PWM generator to produce VHF signals.
It is compatible with both the Raspberry Pi 1 (the original one) and the Raspberry Pi 2 and 3.
PiFmAdv has been developed for experimentation only. It is not a media center, it is not intended to broadcast music to your stereo system. See the legal warning.
PiFmAdv, depends on the sndfile
library. To install this library on Debian-like distributions, for instance Raspbian, run sudo apt-get install libsndfile1-dev
.
PiFmAdv now uses the SoX Resampler for high-quality resampling. On Debian-based distributions, run sudo apt install libsoxr-dev
to install the dependency.
PiFmAdv also depends on the Linux rpi-mailbox
driver, so you need a recent Linux kernel. The Raspbian releases from August 2015 have this.
Important. The binaries compiled for the Raspberry Pi 1 are not compatible with the Raspberry Pi 2/3, and conversely. Always re-compile when switching models, so do not skip the make clean
step in the instructions below!
Clone the source repository and run make
in the src
directory:
git clone https://github.com/Miegl/PiFmAdv.git
cd PiFmAdv/src
make clean
make
Then you can just run:
sudo ./pi_fm_adv
This will generate an FM transmission on 87.6 MHz, with default station name (PS), radiotext (RT) and PI-code, without audio. The radiofrequency signal is emitted on GPIO 4 (pin 7 on header P1).
You can add monophonic or stereophonic audio by referencing an audio file as follows:
sudo ./pi_fm_adv --audio sound.wav
To test stereophonic audio, you can try the file stereo_44100.wav
provided.
The more general syntax for running Pi-FM-RDS is as follows:
pi_fm_adv
All arguments are optional:
--freq
specifies the carrier frequency (in MHz). Example:--freq 87.6
.--audio
specifies an audio file to play as audio. The sample rate does not matter: PiFmAdv will resample and filter it. If a stereo file is provided, PiFmAdv will produce an FM-Stereo signal. Example:--audio sound.wav
. The supported formats depend onlibsndfile
. This includes WAV and Ogg/Vorbis (among others) but not MP3. Specify-
as the file name to read audio data on standard input (useful for piping audio into PiFmAdv, see below).--pi
specifies the PI-code of the RDS broadcast. 4 hexadecimal digits. Example:--pi FFFF
.--ps
specifies the station name (Program Service name, PS) of the RDS broadcast. Limit: 8 characters. Example:--ps RASP-PI
.--rt
specifies the radiotext (RT) to be transmitted. Limit: 64 characters. Example:--rt 'Hello, world!'
.--af
specifies alternative frequencies (AF). Example:--af 107.9 --af 99.2
.--pty
specifies the program type. 0 - 31. Example:--pty 10
(EU: Pop music). See https://en.wikipedia.org/wiki/Radio_Data_System for more program types.--tp
specifies if the program carries traffic information. Example--tp 0
.--dev
specifies the frequency deviation (in Hz). Example--dev 25000
.--mpx
specifies the output mpx power. Default 30. Example--mpx 20
.--power
specifies the drive strenght of gpio pads. 0 = 2mA ... 7 = 16mA. Default 7. Example--power 5
.--gpclk
specifies the General Purpose Clock (GPCLK) to use. Avaliable options are 0, 1 and 2.--gpio
specifies the GPIO pin used for transmitting. Can be specified multiple times to transmit on multiple pins at once. Default 4. Can be used in combination with--gpclk
. Example--gpio 20
.--pll
specifies the PLL to modulate. PLLC is used by default, PLLA doesn't seem to work on Raspberry Pi 4. Available options:a
andc
.--cutoff
specifies the cutoff frequency (in Hz) used by PiFmAdv's internal lowpass filter. Values greater than 15000 are not compliant. Use carefully.--preemph
specifies which preemph should be used, since it differs from location. For Europe choose 'eu', for the US choose 'us'.--ctl
specifies a named pipe (FIFO) to use as a control channel to change PS and RT at run-time (see below).--ppm
specifies your Raspberry Pi's oscillator error in parts per million (ppm), see below.--rds
RDS broadcast switch.--wait
specifies whether PiFmAdv should wait for the the audio pipe or terminate as soon as there is no audio. It's set to 1 by default.
By default the PS changes back and forth between PiFmAdv
and a sequence number, starting at 00000000
. The PS changes around one time per second.
The RDS standards states that the error for the 57 kHz subcarrier must be less than ± 6 Hz, i.e. less than 105 ppm (parts per million). The Raspberry Pi's oscillator error may be above this figure. That is where the --ppm
parameter comes into play: you specify your Pi's error and PiFmAdv adjusts the clock dividers accordingly.
In practice, I found that PiFmAdv works okay even without using the --ppm
parameter. I suppose the receivers are more tolerant than stated in the RDS spec.
One way to measure the ppm error is to play the pulses.wav
file: it will play a pulse for precisely 1 second, then play a 1-second silence, and so on. Record the audio output from a radio with a good audio card. Say you sample at 44.1 kHz. Measure 10 intervals. Using Audacity for example determine the number of samples of these 10 intervals: in the absence of clock error, it should be 441,000 samples. With my Pi, I found 441,132 samples. Therefore, my ppm error is (441132-441000)/441000 * 1e6 = 299 ppm, assuming that my sampling device (audio card) has no clock error...
If you use the argument --audio -
, PiFmAdv reads audio data on standard input. This allows you to pipe the output of a program into PiFmAdv. For instance, this can be used to read MP3 files using Sox:
sox -t mp3 https://www.linuxvoice.com/episodes/lv_s02e01.mp3 -t wav - | sudo ./pi_fm_adv --audio -
Or to pipe the AUX input of a sound card into PiFmAdv:
sudo arecord -fS16_LE -r 44100 -Dplughw:1,0 -c 2 - | sudo ./pi_fm_adv --audio -
You can control PS, RT, TA (Traffic Announcement flag) and PTY (Program Type) at run-time using a named pipe (FIFO). For this run PiFmAdv with the --ctl
argument.
Example:
mkfifo rds_ctl
sudo ./pi_fm_adv --ctl rds_ctl
Then you can send “commands” to change PS, RT, TA and PTY:
cat >rds_ctl
PS MyText
RT A text to be sent as radiotext
PTY 10
TA ON
PS OtherTxt
TA OFF
...
Every line must start with either PS
, RT
, TA
or PTY
, followed by one space character, and the desired value. Any other line format is silently ignored. TA ON
switches the Traffic Announcement flag to on, any other value switches it to off.
PiFmAdv is an experimental program, designed only for experimentation. It is in no way intended to become a personal media center or a tool to operate a radio station, or even broadcast sound to one's own stereo system.
In most countries, transmitting radio waves without a state-issued licence specific to the transmission modalities (frequency, power, bandwidth, etc.) is illegal.
Therefore, always connect a shielded transmission line from the Raspberry Pi directly to a radio receiver, so as not to emit radio waves. Never use an antenna.
Even if you are a licensed amateur radio operator, using PiFmAdv to transmit radio waves on ham frequencies without any filtering between the RaspberryPi and an antenna is most probably illegal because the square-wave carrier is very rich in harmonics, so the bandwidth requirements are likely not met.
I could not be held liable for any misuse of your own Raspberry Pi. Any experiment is made under your own responsibility.
PiFmAdv was successfully tested with all my RDS-able devices, namely:
- a Sony ICF-C20RDS alarm clock from 1995,
- a Sangean PR-D1 portable receiver from 1998, and an ATS-305 from 1999,
- a Samsung Galaxy S2 mobile phone from 2011,
- a Philips MBD7020 hifi system from 2012,
- a Silicon Labs USBFMRADIO-RD USB stick, employing an Si4701 chip, and using my RDS Surveyor program,
- a “PCear Fm Radio”, a Chinese clone of the above, again using RDS Surveyor.
Reception works perfectly with all the devices above. RDS Surveyor reports no group errors.
CPU usage is as follows:
- without audio: 9%
- with mono audio: 33%
- with stereo audio: 40%
CPU usage increases dramatically when adding audio because the program has to upsample the (unspecified) sample rate of the input audio file to 228 kHz, its internal operating sample rate. Doing so, it has to apply an FIR filter, which is costly.
The RDS data generator lies in the rds.c
file.
The RDS data generator generates cyclically four 0A groups (for transmitting PS), and one 2A group (for transmitting RT). In addition, every minute, it inserts a 4A group (for transmitting CT, clock time). get_rds_group
generates one group, and uses crc
for computing the CRC.
To get samples of RDS data, call get_rds_samples
. It calls get_rds_group
, differentially encodes the signal and generates a shaped biphase symbol. Successive biphase symbols overlap: the samples are added so that the result is equivalent to applying the shaping filter (a root-raised-cosine (RRC) filter specified in the RDS standard) to a sequence of Manchester-encoded pulses.
The shaped biphase symbol is generated once and for all by a Python program called generate_waveforms.py
that uses Pydemod, one of my other software radio projects. This Python program generates an array called waveform_biphase
that results from the application of the RRC filter to a positive-negative impulse pair. Note that the output of generate_waveforms.py
, two files named waveforms.c
and waveforms.h
, are included in the Git repository, so you don't need to run the Python script yourself to compile PiFmAdv.
Internally, the program samples all signals at 228 kHz, four times the RDS subcarrier's 57 kHz.
The FM multiplex signal (baseband signal) is generated by fm_mpx.c
. This file handles the upsampling of the input audio file to 228 kHz, and the generation of the multiplex: unmodulated left+right signal (limited to 15 kHz), possibly the stereo pilot at 19 kHz, possibly the left-right signal, amplitude-modulated on 38 kHz (suppressed carrier) and RDS signal from rds.c
. Upsampling is performed using a zero-order hold followed by an FIR low-pass filter of order 60. The filter is a sampled sinc windowed by a Hamming window. The filter coefficients are generated at startup so that the filter cuts frequencies above the minimum of:
- the Nyquist frequency of the input audio file (half the sample rate) to avoid aliasing,
- 15 kHz, the bandpass of the left+right and left-right channels, as per the FM broadcasting standards.
The samples are played by pi_fm_adv.c
that is adapted from Richard Hirst's PiFmDma. The program was changed to support a sample rate of precisely 228 kHz.
- 2015-09-05: support for the Raspberry Pi 2
- 2014-11-01: support for toggling the Traffic Announcement (TA) flag at run-time
- 2014-10-19: bugfix (cleanly stop the DMA engine when the specified file does not exist, or it's not possible to read from stdin)
- 2014-08-04: bugfix (ppm now uses floats)
- 2014-06-22: generate CT (clock time) signals, bugfixes
- 2014-05-04: possibility to change PS and RT at run-time
- 2014-04-28: support piping audio file data to PiFmAdv's standard input
- 2014-04-14: new release that supports any sample rate for the audio input, and that can generate a proper FM-Stereo signal if a stereophonic input file is provided
- 2014-04-06: initial release, which only supported 228 kHz monophonic audio input files
© Miegl & Christophe Jacquet (F8FTK), 2014-2017. Released under the GNU GPL v3.