FreeDV Robustness Part 6 – Early Low SNR Results

Anyone who writes software should be sentenced to use it. So for the last few days I’ve been radiating FreeDV 700 signals from my home in Adelaide to this websdr in Melbourne, about 800km away. This has been very useful, as I can sample signals without having to bother other Hams. Thanks John!

I’ve also found a few bugs and improved the FreeDV diagnostics to get a feel for how the system is working over real world channels.

I am using a simple end fed dipole a few meters off the ground and my IC7200 at maximum power (100W I presume, I don’t have a power meter). A key goal is comparable performance to SSB at low SNRs on HF channels – that is where FreeDV has struggled so far. This has been a tough nut to crack. SSB is really, really good on HF.

Here is a sample taken this afternoon, in a marginal channel. It consists of analog/DV/analog/DV speech. You might need to listen to it a few times, it’s hard to understand first time around. I can only get a few words in analog or DV. It’s right at the lower limit of intelligibility, which is common in HF radio.

Take a look at the spectrogram of the off air signal. You can see the parallel digital carriers, the diagonal stripes is the frequency selective fading. In the analog segments every now and again some low frequency energy pops up above the noise (speech is dominated by low frequency energy).

This sample had a significant amount of frequency selective fading, which occasionally drops the whole signal down into the noise. The DV mutes in the middle of the 2nd digital section as the signal drops out completely.

There was no speech compressor on SSB. I am using the “analog” feature of FreeDV, which allows me to use the same microphone and quickly swap between SSB and DV to ensure the HF channel is roughly the same. I used my laptops built in microphone, and haven’t tweaked the SSB or DV audio with filtering or level adjustment.

I did confirm the PEP power is about the same in both modes using my oscilloscope with a simple “loop” antenna formed by clipping the probe ground wire to the tip. It picked up a few volts of RF easily from the nearby antenna. The DV output audio level is a bit quiet for some reason, have to look into that.

I’m quite happy with these results. In a low SNR, barely usable SSB channel, the new coherent PSK modem is hanging on really well and we could get a message through on DV (e.g. phonetics, a signal report). When the modem locks it’s noise free, a big plus over SSB. All with open source software. Wow!

My experience is consistent with this FreeDV 700 report from Kurt KE7KUS over a 40m NVIS path.

Next step is to work on the DV speech quality to make it easy to use conversationally. I’d say the DV speech quality is currently readability 3 or 4/5. I’ll try a better microphone, filtering of the input speech, and see what can be done with the 700 bit/s Codec.

One option is a new mode where we use the 1300 bit/s codec (as used in FreeDV 1600) with the new, cohpsk modem. The 1300 bit/s codec sounds much better but would require about 3dB more SNR (half an s-point) with this modem. The problem is bandwidth. One reason the new modem works so well is that I use all of the SSB bandwidth. I actually send the 7 x 75 symbol/s carriers twice, to get 14 carriers total. These are then re-combined in the demodulator. This “diversity” approach makes a big difference in the performance on frequency selective fading channels. We don’t have room for that sort of diversity with a codec running much faster.

So time to put the thinking hat back on. I’d also like to try some nastier fading channels, like 20m around the world, or 40m NVIS. However I’m very pleased with this result. I feel the modem is “there”, however a little more work required on the Codec. We’re making progress!

4 thoughts on “FreeDV Robustness Part 6 – Early Low SNR Results”

  1. HI David,

    MOST IMPRESSIVE!!

    Is an ‘echo’ mode feasible?

    Remote ‘hears’ a message and transmits it back?

    It would be fun to set up some unattended addressed remote nodes.

    John

    1. Hi John,

      Sure it’s feasible, would you like to help implement this feature?

      I also have some hooks from running system commands in response to text strings received, for example getting a callsign could make it “post” to a web site to give presence and spotting information. I need some help here with the web server side.

      Thanks,

      David

  2. Hi David,

    Yes! I have the time if you have the beer!!

    Not much fancy ham gear here, but I do have some static IPs!

    And RF test equipment and other possibly useful stuff.

    John

    You can view my garden conductivity real time at: nhfn.net/cond/cond.html. This could pretty easily include received messages? Also, see nhfn.net/cond/r301.txt for one of my line’s real time telephone line activity.

    Lots of fun :)

Comments are closed.