rafirafi / dabs Goto Github PK

dabs - a DAB/DAB+ decoder
[version 1.0]

This program receives a DAB+ signal and plays audio data out of it.

It works on gnu/linux systems. Maybe others. You need alsa for audio.

This program is incomplete. I tested it only where I live. I didn't implement
many many features.

It is released in the public domain as far as I am concerned. I used libfftw,
so maybe it's not legal. I don't care.

You need libfftw, alsa, libfaad.

For USRP devices, you need libuhd.
For RTL-SDR devices, you need librtlsdr.

You will probably need to edit Makefile to change directories or whatever.
Contact me if you have problems: mailto:[email protected]

This program is hosted at: http://sedcore.eu.org/dabs/index.html

COMPILATION
===========

Edit Makefile maybe. Then just do 'make'. Should be it.

As written above you need some libraries. Install them with whatever is
needed in your system (apt in debian for example).

USAGE
=====

1. scan

Run:

    ./dabs -d -w -scan channels.txt

I have a shitty antenna, that's why I use -d -w. When it finds some data on a
frequency the program won't stop scanning that frequency before it gets
the setting data of the channel twice without reception errors in between.
In case of error, the process starts over from scratch. There is no 'oh, this
channel sucks, let's skip it'. The -d and -w help me to see if the scan is
moving or stuck on a frequency. The terminal is flooded with text but, well,
I can see what's going on and move the antenna around until I get a good
signal.

If the program does not find any data (every decoding attempt fails) then it
will skip the frequency.

That's the logic of the scan. Maybe it's bad. I don't know. I like it like
that.

When the scan is over, found channels are stored in the file 'channels.txt'.
Of course, you can give another filename.

2. listen

Run:

    ./dabs channels.txt

Commands to type when the program runs:

    l
        list all channels

    c <x>
        <x> is the number of the channel as seen with the command 'l'
        play channel <x>

    n
        play next channel (loop, the next channel of the last is the first)

    p
        play previous channel (loop, the prev. of the first is the last)

You can look at the function user_input in dabs.c, maybe this documentation
is not up to date (it should be!).

That's basically it.

You can also record and replay. Run './dabs -h' for more information.

Ah, the names of the channels are in whatever character set used by the
server. We don't convert to utf8 or whatever. Live with that.

There is a equalizer hardcoded. Edit audio/alsa.c, function put_alsa, remove
the call to filter() in there to disable the equalizer. (I should do something
better, no? like a command-line option or something. Bah, live with it.)

LIMITATIONS
===========

Only FIG 0.0, 0.1, 0.2, and 1.1 are used.

Only DAB+, not DAB.

Code was done with specifications version 2.1.1 (2017-01). They removed stuff
from previous versions I think, like in this specification there is only
"transmission mode I".

PAD/XPAD is not handled. So the program only does audio, no text or image
stuff.

Frequency offset correction is not good. The "fine" part is okay I would
say. The "coarse" is bad and will be improved in the future, depending on
time/motivation.

The time synchronization is done on the first symbol that is supposed to
contain no energy. It's very basic. Seems to be enough for my setup.

The scan stops when we have received setup data twice without reception
errors in between. Not sure if correct.

The synchronization for the superframe may be wrong. We simply suppose a
superframe starts at cif_id % 5 == 0 (see function audio_receive_frame).
We don't check actual data. That works for me. The firecode thing is not
used.

The Viterbi decoder is a "hard" one, not "soft". We probably lose some
dB in there.

The Reed-Solomon stuff is not used at all. If the Viterbi fails (i.e. the
CRC is wrong) then it's game over for the AU block.

Character sets are not handled. If text is not in ASCII, game over.

"Change flag" in FIGs is not cared at all.

Only "primary services" are used.

For USRP devices, no automatic gain control (AGC). The gain is set as 1 and
so be it. If too loud, game over. I don't know how to deal properly with gain,
so I don't do it. As simple as that. Yes, that sucks. (I'm talking about RF
gain, not audio gain. Audio gain is not done either. Could be, in a GUI or
something. Maybe at some point in the future...)

For RTL-SDR devices, AGC is enabled.

Audio is brutal. There is no fade-out/fade-in when some audio data is missing.
Actually, to me it's something good. In the DAB+ receiver of a car, when there
is no audio, there is fade-out and then fade-in when audio is back. I find
this behavior very annoying. I like the brutal cut of my program way more.
I also like when the AAC decoder does its funky ringing sounds when some data
is missing. I guess it's just me, but so be it. My program, my decision. Live
with that.

The Viterbi decoder is amazingly slow.

There could be a GUI. The program might be betterly verbose. It might also be
more user-friendly. When I saw sdrangel, I was wooo! But, you know, this is
a summer project. I get audio out of data received over-the-air, the code is
small, not much dependancies, I'm proud of myself, and that's it.

Settings not used by the DAB+ emitters around me have normally not been
implemented. The program should exit if such settings pop up. Get in touch
with me so I can include missing pieces in a future release. (Basically you
record a few seconds of the frequency you use and put the file somewhere. I
download it and implement what is missing.)