Hi, So if I understand this correctly, you want to modulate the signals with this mixed mode in order to have a more robust signal & a lesser bandwith consumption?
Sounds great... What about an error-correction pipe for these more important Wo_E bits? So that the degradation does not affect them directly, and loosing bits randomly is not that hard... Maybe when using 16QAM or 4QAM as modulation for the entire stream, so that you keep the same bandwith consumption by donoring the free bandwith due to the more efficient modulation to a specific error correction (e.g. Viterbi FEC 4/5 - a low resource equivalent to a CRC8 on each 48 bits, as used in one test!) to these bits. So, as far as I got, the Codec2 - Signal with 1125Hz bandwith is based upon 14 carriers in QPSK => approx. 80,3Hz bandwith per carrier. (including guard bands). So your new configuration uses 10 QPSK carriers (approx. 80,4Hz) for the needed bits and 2 16QAM carriers (approx. 85,7Hz) for the ''loss-tolerated'' ones. When using 12 x 16QAM as a uniform modulation istead of mixed modulatin, your result is 1028,6Hz -> so far 9,375 % (!) bandwith is saved. Enough for a FEC 12/11, but this Viterbi mode is not even used in radio applications of direct sight, as it is not enough. such as DVB-S2. (Maximum specified by DVB consortium is FEC 8/9)... I know that the main priority is that these essential bits are transmitted correctly. So to find out how far they are loss-tolerable, just try to reduce the carriers up to 10 QPSK ones with 80,4Hz each, so that you only transfer these Wo_E bits! Maybe the codec signal would still work with these 804Hz! You have to try it with the FDMDV-modem... Another method to harden a signal may be spread spectrum. Which works completely counterwise, but could increase the bandwith useable in another way while being clearly recievable. For example, a GSM channel has 200kHz (!) width, and is segmented into 8 timeslots. So a wide-spread signal sent in small bursts requires much lesser signal strenght in order to be recieved propely. If you compare a cellphone to a radio, you have to confess that the power consumption for sending a signal of the cell is much, much lower... But this method, adapted to 2Hz SSB channels, would work in the other direction of your work to shape the signal bandwith, meaning that all stations in the channel are forced to transmitt with Codec2 and have to sync their timeslots (which could be done by a radio when it marks ' it's ' timeslot as occupied by sending a CTCSS tone in it's timeslot. So if we take Dan's method and are capable shape the signal to approx. 1kHz, and have a 2kHz channel - we could use a Syncronous TDMA for duplex usage, (double the bandwith and split the time - I recommen the frame lenght (40ms!) for a timeslot)! In case of a 10kHz CB channel, this means 10 radios can use the frequency simultaneously without interrupting each other, making a maximum of 5 duplex connections or 10 simplex connections - or whatever combined... To be back on topic: transfering the degradation ability of analog signals to digital ones is very challenging, as they have never been designed for. This is like wanting to port the 2300Hz tone bug into the ISDN-PSTN. A variable (lower) high/low edge steepness could solve bad signal problems (on cost of more noise). For me, it seems as if adding some error corrrection while optimizing the modulation, to decrease the main output signal (and then apply spread spectrum upon it!) could create this degradation ability... Greetings, Netzblockierer Am 15.01.2013 18:43, schrieb Daniel Ankers: > Hi, > > On 15 January 2013 16:59, Netzblockierer > <[email protected] > <mailto:[email protected]>> wrote: > > Hi Dan, > > as Codec2 is a digital codec, a drop-down of quality like on > analog transmission seems hard to archieve. Normally, a good > squelch-filter with amplifier would work better instead of trying > to modulate it again as an analog narrowband signal. > > > Some of the bits have less of an effect on the signal quality than > others. Jimmy KG4SGP did some testing documented at > http://www.kg4sgp.com/codec2-bit-study.html on the 1400bps codec, and > I've done some separate testing on the 1200bps codec. Errors in some > bits affect the quality of the speech without affecting the > intelligibility, while errors particularly in the Wo_E bits cause > major problems. One of the things which at lot of people don't like > about digital voice modes is that they don't degrade gracefully like > analogue modes such as SSB and it would be nice to be able to take > advantage of features like bit importance to make trade offs which > bring back some of that degradation ability. > > Instead of 16QAM, I would prefer 4QAM, as it is much more robust > than 16QAM in case of an angular phase shift. That's why 64QAM and > 256QAM only work fine in a stable connection and 16QAM is standard > in DVB-T (64QAM in DVB-C/S). > > > The current modem uses 14 x QPSK carriers, which I think has similar > robustness to 4QAM. If this was changed to 10 x QPSK + 2 x 16QAM with > the less important bits in the 16QAM carriers then we would cut the > bandwidth requirement from 1125Hz to 975Hz with (possibly) a minimal > loss in quality over the same path. > > Regards, > Dan MD1CLV. > > > ------------------------------------------------------------------------------ > Master SQL Server Development, Administration, T-SQL, SSAS, SSIS, SSRS > and more. Get SQL Server skills now (including 2012) with LearnDevNow - > 200+ hours of step-by-step video tutorials by Microsoft MVPs and experts. > SALE $99.99 this month only - learn more at: > http://p.sf.net/sfu/learnmore_122512 > > > _______________________________________________ > Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------------ Master SQL Server Development, Administration, T-SQL, SSAS, SSIS, SSRS and more. Get SQL Server skills now (including 2012) with LearnDevNow - 200+ hours of step-by-step video tutorials by Microsoft MVPs and experts. SALE $99.99 this month only - learn more at: http://p.sf.net/sfu/learnmore_122512
_______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
