Re: [Freetel-codec2] 4FSK modem performance

[glen english]

Hi Brady ahh yes I think I know where you numbers were- assumption I think is in correct for dmr radio sensitivity. They can do quite alot better than your assumption. But, it is a reasonable start. Remember also the coding performance on DMR is strong. (Your green line does not take this into account- (it is pretty much like a vertical cliff due to the Iterative Turbo block product code) The noise figure for a 'good' DMR radio will be about 4dB to 5dB  approx. The better ones will  do about -123dBm for quasi error free, and 12dB SINAD in narrow band of about -125dBm (~ 7dB CNR)  I think about 9dB CNR  is quasi error free in the turbo radios.  On 7/03/2016 5:21 PM, Brady O'Brien wrote: Glen, I think the extra perf we're expecting comes from the tone spacing and RRC filtering in DMR. For ideal non-coherent 4FSK, the tones should be spaced Rs apart. The tone spacing of DMR is 1296 Hz, as you've said above. David wrote it up here: http://www.rowetel.com/blog/?p=4650 . Thanks, Brady O'Brien On Sun, Mar 6, 2016 at 10:29 PM, glen english <g...@cortexrf.com.au> wrote: Hi David I too am a fan of 4FSK. Just the right amount of complexity/ tradeoffs. How are you doing 10dB better than DMR (which also uses 4FSK) when the symbol rate is 1/4 that of DMR  (implying 6dB would be more like it) ? different rolloff ? Peak deviation of DMR is ± 1.944kHz. "half" deviation is ± 648 Hz. or D = 3h/2T regards On 7/03/2016 10:08 AM, David Rowe wrote: > Hi Adrian, > > Modem performance and trade offs like RF bandwidth is something I have > been studying, e.g. this one compares various modulation schemes for > digital voice: > >    http://www.rowetel.com/blog/?p=4663 > > and many other posts.  There seems to be a lot of "low hanging fruit" in > the area of modems - opportunities to significantly improve our > communications systems with a small amount of effort. > > Yes for FSK you need Rs Hz (symbol rate) between the tones to get the > best BER performance at a given SNR.  Many services, such a C4FM, DMR, > use a smaller spacing and cop the performance hit, presumably to save > bandwidth. > > Yes mPSK can be very bandwidth efficient, especially with high order > constellations.  However it's hard to get the theoretical performance of > PSK in practical implementations.  For example to avoid phase estimation > using DQPSK means double the bit error rate of coherent PSK. > > I've found it's very easy to get "ideal" (bang on theoretical) > performance from FSK modems.  They go through crappy PAs.  The receivers > don't need a linear signal path or AGC. For the VHF FreeDV modem and > SM2000 I've settled on non-coherent 4FSK at 1200 symbols (2400 bits) per > second.  It's 5kHz bandwidth, and performance is just 2dB off coherent > PSK.  The combination outperforms analog FM and DMR by > 10dB. > > You can actually put unfiltered BPSK through a non linear amplifier, for > example use an XOR gate for the "mixer" to reverse the phase of the > carrier.  This can be friendly to other users of spectrum if you broadly > filter say 10Rs wide to reduce way out of band emissions. > > I've found this PDF from Atlanta Signal Processing to be a great overview: > >     http://www.atlantarf.com/FSK_Modulation.php > > Cheers, > > David > > On 07/03/16 07:37, Adrian Musceac wrote: >> Hi David, >> >> Since you mentioned working on a 4FSK modem, I have started to model >> one using Gnuradio and test it over a noisy channel. >> My question is: since to maintain a reasonable bandwidth of the signal >> you have to do heavy pulse shaping and not use orthogonal freqencies, >> how do you manage a reasonable BER? >> By my simulations, 4FSK requires orthogonal tones in order to still >> get a reasonable decode error with a noisy channel. That creates a >> signal with a huge bandwidth, especially compared with DQPSK which for >> less bandwidth handles noise a lot better. Of course QPSK requires >> high linearity of the amplifier, which is why my next step is to >> investigate PI/4 DQPSK which is used as modulation scheme by the Tetra >> standard. >> >> >From my experiments so far, given a fixed bandwidth and a noise >> spectrum with the same power, the ranking from best to worst in BER >> is: DBPSK > DQPSK > 2FSK, GMSK, 4FSK >> >> Can you confirm this using you mathematical knowledge? >> >> Cheers, >> Adrian >> >> ------------------------------------------------------------------------------ >> _______________________________________________ >> Freetel-codec2 mailing list >> Freetel-codec2@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/freetel-codec2 >> > ------------------------------------------------------------------------------ > _______________________________________________ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- - Glen English RF Communications and Electronics Engineer CORTEX RF & Pacific Media Technologies Pty Ltd ABN 40 075 532 008 PO Box 5231 Lyneham ACT 2602, Australia. au mobile : +61 (0)418 975077 ------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://makebettercode.com/inteldaal-eval _______________________________________________ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2 ------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://makebettercode.com/inteldaal-eval _______________________________________________ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2 -- - Glen English RF Communications and Electronics Engineer CORTEX RF & Pacific Media Technologies Pty Ltd ABN 40 075 532 008 PO Box 5231 Lyneham ACT 2602, Australia. au mobile : +61 (0)418 975077

------------------------------------------------------------------------------
Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
http://makebettercode.com/inteldaal-eval
_______________________________________________
Freetel-codec2 mailing list
Freetel-codec2@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/freetel-codec2