Mike wrote:

Dave Platt wrote:

In article ,
Mike wrote:

Hello,
has anyone come across any projects or info on building your own DAB
radio.

This is my (limited) understanding:

How it's sent: The audio (or whatever) signal is split and sent using
Quadrature Amplitude Modulation over many narrow band carriers (each
carrier at a different frequency): Ie it's spread spectrum. Error
correction Coding is also used, etc.
To decode it: This is where I'm confused; Can I build a receiver
which will receive the whole collection of narrow band carriers, then
feed it into a computer that will do some FFT?



Yes, that's usually how it's done. The approach I've seen is to grab
the whole slice of the RF section, downconvert to a convenient IF (via
superhet or direct conversion), then feed the whole signal into
something which can use DSP to detect and demodulate the individual
carriers.


Yeah, that's the sort of idea I thought might work.

A common approach for Digital Radio Mondiale broadcasting (which has a
relatively narrow RF bandwidth) is to use a single-sideband receiver
architecture of some sort, downconverting the DRM signal from RF to a
same-width slice of audio spectrum, then feed it into a PC via the
PC's sound card, and doing the DSP work using the PC's CPU.


That's a good idea! I think DRM has more potential than DAB at least
from a technical standpoint, as it will go further (less retransmission,
less power needed). But I'm not really an expert.

The commercial high-fidelity DAB systems used in the US may very well
have a bandwidth too wide to allow this simple approach to be used (I
haven't studied them in any detail) but the same basic principles
would apply. Block-convert the RF range you want to an
easily-captured IF range, do an A-to-D on it, and do the rest of the
processing digitally in a CPU or dedicated DSP chip.


(Sorry, I should have mentioned that I'm in the UK).
I've just found a bbc 'paper' linked to from wikipedia, but now
hosted on web.archive.org at:
http://web.archive.org/web/200404010.../paper_21.html

that says a DAB station would occupy up to 1.54MHz of the spectrum,
with the individual carriers spaced 1kHz apart.
I was hoping a PIC might be able to decode it. Most seem to be able to
manage 40MHz, which might not be fast enough... although a pair of them
sharing the task (one getting the raw 1s and 0s, the other doing stuff
with them) might work.

Take a look at http://www.gnu.org/software/gnuradio/ for a look at
some open-source software-radio techniques.


Thanks for the hints and advice,
this is the first time I've used a public newsgroup for years
and I'm pleased by a) how helpful it's been and b) that it's still
alive and well. I'll try to help someone else in return.

Thanks Dave! (Your reply's really been helpful!!)

Mike Smith
lionfish _A_T_ uwcs.co.uk

What you need to do is well beyond the capabilities of even a 40MHz PIC.

You need to sample the data at a good bit over 2x 1.54MHz -- somewhere
between 2.5x and 3x would probably be good, depending on what kind of a
tradeoff you wanted to make between front-end filters and ADC sample
rate. Once you've done that you need to synchronize to the data and do
a 2048-point FFT 1000 times a second. That's going to take on the order
of 40 million multiplies per second -- and there's no way a PIC is going
to be able to do one multiply per clock cycle.

You _might_ be able to do this with a good fast DSP chip -- expect to
need one that clocks at over 100MHz. You could also do this with a
fairly modest FPGA, which may be the best way to do it from a technical
standpoint for a short production run or prototype, but which may be a
bit difficult for a hobbyist -- although once you were done you'd be
ready to go into consulting.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com