Saturday, July 21, 2012

Digital Radio Mondiale on WiNRADiO



The WR-G31DDC comes with a built-in Digital Radio Mondiale (DRM) decoder. Unfortunately it costs $49.95 USD Euros to unlock. Now it can be said that if you're willing to shell out for this receiver, then you can afford to pay to unlock the DRM part. Then again, it only decodes DRM. I coughed up $49.95 USD for the Virtual Sound Card instead, and pipe the VSC into a free external decoder.

There are free or at least cheaper and more general purpose virtual sound card drivers out there, but for me the integration into the receiver GUI made it about as convenient as the DRM decoder, while being more general purpose.

The Dream DRM receiver for Windows is most happy with an input signal that is 48 kHz mono, with an average power of -60 dB. -40 dB is the upper limit on its power chart, and anecdotally I've found that Dream loses sync when the signal is that strong, so it's probably clipping.

How to get that -60 dB target required some trial and error. I ended up with slow AGC , target level of -21 dB. An alternative strategy, if the signal experiences really fast fading, is to just impose a manual constant gain, and let Dream's post-FFT channel estimation do its work unassisted, but picking that gain require some trial and error.

For tuning and RF demodulation, I use upper sideband, 10 kHz of output audio bandwidth, and the upper sideband demodulator placed 5 kHz below the (suppressed) carrier.

Some comments on Dream settings:
  • MLC correction = 4: Supposedly there isn't much performance improvement with increased iterations, but given my circumstances I probably need as much advantage as possible.
  • Bandpass filter = ON: Given the interferer that bleeds into 9800 kHz, a BPF is a good thing, and I can live with some slight attenuation.
  • Modified metrics = ON: According to this, performance is impaired when switched on and there is no interferer present, because the Viterbi decoder may be fed overly pessimistic SNR measurements. Well, I've got interference, so that's not an issue.
MLC and BPF will lead to increased CPU utilization, but I don't really notice it. I've got a 2.6 GHz AMD Phenom II quad core - not exactly cutting edge - and typical utilization is about 35% across all cores.

No comments:

Post a Comment