Hi David
1) when I say flat VHF channel, I should have added more detail ! - I
mean AWGN. Trans-horizon tropo paths are generally AWGN around 2m and
70cm . There is some mobile SSB work around. that is usually fast flat
fading . 6m work can be anything that the band serves up, depending on
the directivity and front to back of the antennas. Doppler is usually
less than 1% of the symbol rate.
2) OK ! . A good RF speech processor will do it. They limit and clip
hard at RF. Alot of modern smaller SSB radios have rather poor
performing audio compressors....
I can also make recordings for yo and also verify using my usual IQ
recorders here. (they generate 32bit 'stereo' wave files - the new wave
file version that spports gigabit size files )
3) several hundred hours sounds excessive, but OK !
4) Most modern SSB radio will pass 2.8 kHz -3dB. 2.4 minimum. The first
(main) lobe of MSK as you know is wider (about 20%) than QPSK, - the
2nd lobe is about -26dB, and so the chopped off 2nd lobe doesnt hurt
too much (compared to QPSK where chopping off the 2nd lobe hurts
because the 2nd lobe energy level is only about 14dB ish down ) . But
the PAPR certainly increases from 1.0 when the 2nd lobe is chopped.
For coherent 4FSK and QPSK, the power efficiency should be the same, and
there would be, given coherent demodulator implementations, similar
performance.
I do not agree with you that " If you lose even a dB in performance you
might as well go to non-coherent 4FSK"
because the differences with non coherent FSK and coherent FSK are
substantial down at the 10e-02 end of the error curve.
For the purposes of the discussion, let's assume a linear demod for the
input to the modem.
The WSJT crew have considered 8 GFSK to be their optimum for FT8. IE -
Not too far left on the shannon capacity bound curve. As you know there
is about a 1 dB advantage on Eb/No at R=0.5 for n=8 versus n=4.
-glen
On 7/7/2020 7:57 AM, David Rowe wrote:
Hi Glen,
OK so you want to work on reducing PAPR, with the goal of improving
performance on VHF channels. Comments/questions:
1/ What do you mean by a "flat VHF channel"? Flat fading or just AWGN?
2/ Can you pls post a wavefile of a demodulated SSB signal with a PAPR
of less than 4.5dB? Given a sine wave is just 3dB so I'd like to hear
what that sounds l ike - and measure the PAPR myself.
3/ Writing, testing, and integrating a new modem into FreeDV (and
freedv-gui) is a several hundred hour project. I do that exercise
perhaps once a year.
A MSK wavefile running the same bit stream as FreeDV 700D will have a
main lobe bandwidth of 1.5*1400 = 2100 Hz. Have you determined if that
will pass through SSB radios without any performance hit? If you lose
even a dB in performance you might as well go to non-coherent 4FSK.....
4/ .... we already have 4FSK FreeDV modes - 2400A & 800XA, see
README_fsk.md.
5/ By "patch" you mean Pull Request - right? If you get that far I'll
help you design the ctests we'll need to see as well.
Cheers,
David
On 6/7/20 7:10 pm, glen english wrote:
Hi David
The goal is to improve FreeDV so that it is at least equal with SSB on a
flat VHF channel.
On HF, the performance of FreeDV is better than SSB mainly because it
can both a) work at low SNR and b) can mitigate the effects of a
frequency selective channel (which reduces speech intelligibility) .
At the moment, FreeDV is a bit behind on a flat channel. This is because
the PAPR is costing some ERP. PAPR on SSB on VHF with a good RF speech
processor is about 4.5dB (or less!). If the ofdm modem is pushed, the
performance doesnt improve of course because the BER starts to suffer
(on a non zero BER scenario IE marginal conditions.
I thought about working on getting the PAPR down on the FreeDV OFDM
modem. The last day I have done some revision and 8dB might be doable,
but 6 to 7dB will only be doable with computational intensity, which the
SM1000 does not have. That still leaves us a few dB in PAPR behind flat
channel SSB. Flat channel SSB is OK down to about -3dB SNR for a good
operator . Actually it is really hard to quantity exactly what the SNR
is because of the time varying power. (IE -3dB wrt to PEP is the usual
statement).
So now I think probably back to making the MSK modem work. I will also
put in a patch for a band limited clipper for the OFDM modem as soon as
I get the build environment... building.
cheers
glen
On 6/07/2020 3:52 pm, David Rowe wrote:
Hi Glen
Happy to comment, but there's quite a lot of material in your post below
and I'm unclear on what you are trying to achieve.
Can you pls tell me what the goal of your proposed work is?
Cheers,
David
On 5/7/20 6:43 pm, glen english wrote:
I've been looking at an coherent MSK modem for Codec2. I have not done
any coherent MSK modems since the late 90s.
Using an SSB radio , at something like 1600 bps , there will still be
appreciable PAPR on MSK due to bandwidth constraint. probably about 3dB
PAPR on a 2400Hz wide radio. When bandwidth is constrained, the
constant envelope signal is no longer constant envelope...
The is quite obvious using MSK441, the meteor scatter mode in WSJTX. It
operates at 2000 bps, and struggles a bit on bandwidth with most non-SDR
types . PAPR ends up around 4.5dB . Also the phase response of non flat
radios is usually aweful. A channel equaliser turned on to model and fix
your radio
Now...
OFDM doesnt need a specific radio equaliser, that's built in of course,
the uglyness of most radios can be obscured. .
And if the only complication from the existing OFDM modem of David's is
the need to constrain PAPR, I think that perhaps constraining the PAPR
of David's existing modem , rather than a MSK modem from almost scratch
, might be more overall beneficial to everyone (HF as well) . I am
thinking an incompatible change to the existing modem to reduce PAPR to
something like 4.5dB might be the best of both worlds. yeah it might
cost a 0.5dB but oh well, you probably lose that now with PAPR
constraints. The existing 2FSK or 4FSK modem should also be brought up
to working.
David, what are your candid thoughts ?
glen.
On 6/23/2020 6:16 AM, David Rowe wrote:
Hi Glen,
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
--
Glen English
RF Communications and Electronics Engineer
CORTEX RF
Pacific Media Technologies Pty Ltd trading as Cortex RF
ABN 40 075 532 008
PO Box 5231 Lyneham ACT 2602, Australia.
au mobile : +61 (0)418 975077
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
