Hi David,
On 22-05-12 03:14, David Rowe wrote: > Hello Kristoff, > > Hmmm, I can see how inverting audio would be a trap when driving a FM > system - it would deviate the wrong way. When I started porting the DSP-code from Jonathan to the ARM platform I was working on (as, to be clear, most of the DSP code in the gmsk modem is NOT written by me; but is just a port of Jonathan's code to a processor without FPU), I did wonder why the "invert bits" option was in there. That was until I discovered that is exactly what the sound-card of my laptop does :-) > Couple of questions: > > 1/ What other parameters is GMSK sensitive to, for example how about the > mapping of sound card audio levels to FM deviation,, of HP/LP filtering > of PC sound card audio? Actually, gmsk-encoding a signal on a FM radio is very simple. If you want to send a "0", you send +0,5 V (or +16384 for a 16 bit audio card) to the FM modulator. If you want to send a "1", you set the voltage to -0,5 V (-16384@16bits). Reception works in the same way, you simply look at the value you receive from the other side. If it positive, you concider it to be "0", if it is negative, you concider it to be a "1". (This is one of the reasons why I use the "gmskmodem" as a tool to explain people about digital communication. It's a very simple concept to explain that almost anybody can understand). Anycase, the end result is that this gives the signal quite some robustness to code. I set the audio-mixer to "roughly" halfway and that works pretty good in most cases. It is not very sensitive at all. Of course, setting the values to small will make it more subsable to noise, and modulating to much risks going over the 50% modulation index, but -overall- the system is quite forgiving. I haven't experimented with changing the setting of the audio mixer. I just set them to all flat and that seams to work OK. Note that this actually comes from the setup that has been tested by VA3UV of the "freestar" D-STAR network. They use this in a much more production enviroment then myself as they have D_STAR repeaters using this. I think they are better placed to comment on the details in this. Perhaps I can try to get him involved in this discussion. Other issues I have had where USB fobs timing-issues and cards that cannot do not playback or record in mono. I did find out almost all USB audio fobs DO support 48 Khz, so that is OK. (but I did found it quite some of them do NOT support 8 Khz, as strange as it might sound). Ah. I once heared somebody has an issue where the "invert" bit needed to be different for receiving then for sending. That's why the latest version of the code now has an option to set the invert bit different in either direction. > 2/ On a PC based VHF system, is there a we can include tuning tools so > people can make sure their transmitted signals will be compatible (e.g. > waveforms deviating the right way)? I understand specifying a certain > USB sound card is one way. Althou my modem software does not have it, it should be possible. In D-STAR, the frame-syncronisation pattern (that sites just behind the bit syncronisation) is a known pattern and it will be different if the "invert" bit is set incorrect. Perhaps this is something Peter can make sure we have the same result in the "VHF codec2 bitformat". > 3/ What is the relationship with GMSK and channel bit rate on a PC-FM > radio system? For example does halving the bit rate get us a 3dB > performance increase in terms of BER? Well, in theory, it should. Creating a 2400 baud modem would just involve 3 things: - change the low-pass filters for reception and the gauss-filter for transmission - send the same bits for 20 "audio-samples" instead of 10. (remember that the audio fob runs at 48000 Khz sampling) - addapt the PLL loop that does timing and syncronisation of the received signal. But that's one of the things that I try out in real life. What works best? 1400 bps speech in a 2400 bitstream with a FEC of 2/3 (but with the 3db power gain due the smaller channel-width) 1400 bps speech in a 4800 bitstream with a FEC of 1/3 My speculation that the latter would be better, not because of the better FEC, but because having more space to do interleaving; which could make the signal better deal with fading. Anycase, this is just speculation. And there are even more factors that also come into play: frequency, propagation (is fading an issue on that frequency?), limited bandwidth in that band (e.g. for 29.1-29.2 Mhz in the 10 meter, or the 4 meter band), etc. > Thanks, > David 73 Kristoff - ON1ARF ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
