Why the DMR 4FSK modem is less optimal than David Rowe's 4FSK modem

[Adrian Musceac]

Hi,

In this post: http://www.rowetel.com/?p=4650 David was asking the question why the 4FSK modem used by the DMR standard seemed to be performing so poorly, especially as compared to his ideal 4FSK modem.

I was curious myself, so I implemented both and analysed them using my modified version of Gqrx, known by some people as gqrx-digital.

The results can be seen in this video: https://www.youtube.com/watch?v=hLeJvqrFrS8

As you can see, the DMR modem uses a sensitivity of PI/sps, which leads to a lot of ISI, about 8 dB worth, compared to the optimal 4FSK modem which uses a sensitivity of 2*PI/sps. The reason why the DMR modem does this is twofold:

One is bandwidth efficiency. But the most significan one is the fact that DMR radios use the same FM modulator for both digital and analog voice. Using double sensitivity would lead to analog voice being too wide for standard FM channels in commercial bands.

Of course, for me as a developer, I can create two Gnuradio flowgraphs, so I can use one modulation type for digital, and another one (proper) for analog FM transmission.

Hope this is useful for someone.

73,

Adrian YO8RZZ

[Adrian Musceac]

Hi Phil,

Of course, you are right, but Trellis is computationally more
expensive than a straight change of the modem parameters. The DMR
software stack is huge, compared to what a small community of largely
open source enthusiasts can come up with. (my problem with DMR as a
radio equipment standard is the use of a closed source audio codec,
not the modem)
It would be even better if someone was to write a pure SDR
implementation of DMR and pit that against David's 4FSK modem :) Real
performance measurements would convince enverybody of the merits of
it.

Cheers,
Adrian

On 9/12/17, Phil Frost <ind...@bitglue.com> wrote:
> Wouldn't the ISI largely (theoretically, completely) be ameliorated by
> trellis decoding? And I'm not talking about FEC, but rather incorporating
> the ISI into the expected Cartesian coordinates for each symbol.
>
> I did some experimentation with this technique on BPSK31, and I was able to
> get a little more performance in simulations with a matched filter and
> trellis decoding than fldigi and PSKCore got with their unmatched filters
> designed to minimize the ISI. Though in the case of BPSK31 the ISI isn't
> all that bad, so there less than 1 dB of a gain to be had. Perhaps with DMR
> there's more potential?

>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
>> _______________________________________________
>> Freetel-codec2 mailing list
>> Freetel...@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/freetel-codec2
>>
>

[Steve]

The problem, as I see it, is the DMR deviation is too narrow for the first two tones. All the power is in the carrier, and there are only two sidebands.

What would happen if you had eight sidebands and no power in the carrier? ¡Ay, caramba!

[Adrian Musceac]

Hi Steve,
Using 8*PI/sps would lead to no absolute gain vs. 2*PI if using a dumb
FM demodulator. Tested this myself over the air. It would only waste
more bandwidth. In fact, this is the reason why high bitrate modems
don't use M-FSK. For example, I want to do H264 video streaming. I
would only consider modems from QPSK upwards (QAM64 being the best
candidate). This is for bandwidth considerations from amateur radio
regulations point of view :)

Cheers,
Adrian

[Adrian Musceac]

On 9/12/17, Phil Frost <ind...@bitglue.com> wrote:
> I did some experimentation with this technique on BPSK31, and I was able to
> get a little more performance in simulations with a matched filter and
> trellis decoding than fldigi and PSKCore got with their unmatched filters
> designed to minimize the ISI. Though in the case of BPSK31 the ISI isn't
> all that bad, so there less than 1 dB of a gain to be had. Perhaps with DMR
> there's more potential?
>

Hi,
On a related note, I've also done some tests between Codec2 voice
frames and Opus voice frames. For the tests, Codec2 bitrate was 1300
bits/s while Opus was 19.4 kbit/s
Results can be seen here: https://www.youtube.com/watch?v=C9107EA3vNE

As far as I'm concerned, Codec2 does amazingly well at voice frequency
reproduction (I used a synthesized voice on purpose as I believe it
makes it even more difficult for the vocoder.)
However, I can see a purpose for Opus, when combined with H264 video,
since we are already using a lot of bandwidth for the video frames and
the Opus codec does not really add much to it. (the good thing about
Opus frames is that they can be reused to multiplex exactly 8 Codec2
frames into timeslots which can be useful when going TDMA).

Cheers,
Adrian YO8RZZ

[Steve]

If you compare say packet radio 2FSK modulation index of 1.36, you can see that the carrier power decreases as it is spread to the FM sidebands. In this chart, that is 4 sideband pairs.

Compared to these ultra-narrowband (modulation index less than 1) 4FSK modulation indexes of .135 for the inside pair, and .405 for the outside pair (deviation from center).

[Steve]

Anyway, what I think the hobby needs is to replace the FM with a nice I/Q modulator and do PSK or QAM, direct and zoom into the 21st century :-)

DMR, DSTAR, etc, is like AM and CW, in that all the power is wasted in the carrier. We need a carrier-less mode on VHF. (But if you do use FM, at least use a tone deviation that is a multiple of the symbol rate.)

[Adrian Musceac]

Hi Steve,

I'm sorry, I'm not sure what you mean by carrier-less mode. We
certainly need an information carrier, you can't send information
otherwise, even in I/Q systems. What happens to the carrier in the
modulation process determines the shape of the waveform (whether the
frequency changes, the phase or the amplitude).

As far as PSK goes, you might be interested in this video:
https://www.youtube.com/watch?v=LUV2zF2pbYU
No QAM yet, at least not publicly available...

73,
Adrian

On 9/13/17, Steve <coupay...@gmail.com> wrote:

> --
> You received this message because you are subscribed to the Google Groups
> "digitalvoice" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to digitalvoice...@googlegroups.com.
> To post to this group, send email to digita...@googlegroups.com.
> Visit this group at https://groups.google.com/group/digitalvoice.
> For more options, visit https://groups.google.com/d/optout.
>