Re: [Freetel-codec2] FDMDV Time Domain Duplexing
With regards to the Whitebox switching times: Right now, it is the case that the PLL parameters are the same for both tx and rx on the same frequency. This works on 144/440/900MHz bands, but not 50MHz. The next version it will be the same for 50MHz, too. The CMX991 touts itself as usable for "APCO Project 25 (P25) Phase 1 and Phase 2 TDMA: TIA-102.CAAB" and I know of this project which has been working on using it for TDD (with the ADF4351, too) [1]. The switching time is going to depend on how quickly we can program the SPI registers (upto 5MHz clock); but sadly the datasheet is light on details about how much time it takes for each configuration change to take effect, so I don't have hard numbers yet. [1] http://www.skagmo.com/page.php?p=projects/21_superhettrx On Sun, Apr 20, 2014 at 7:47 PM, Bruce Perens wrote: > > > > On April 20, 2014 7:18:07 PM PDT, David Rowe wrote: > > >Bruce - I don't think TDMA will require PLL locking, tx and rx would be > all on one frequency. > > Not sure this is always the case for Whitebox. And we are looking at > low-phase-noise DDS chips, and they still have a VCO. If you know of a > really good NCO chip without one we'd like to know. > > Thanks > > Bruce > > -- > Sent from my Android device with K-9 Mail. Please excuse my brevity. > > > -- > Start Your Social Network Today - Download eXo Platform > Build your Enterprise Intranet with eXo Platform Software > Java Based Open Source Intranet - Social, Extensible, Cloud Ready > Get Started Now And Turn Your Intranet Into A Collaboration Platform > http://p.sf.net/sfu/ExoPlatform > ___ > Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
Ross, Great to hear about more people interested! Just a few thoughts on using existing radios, I agree it would be nice to make something that was backward compatible with existing radios, although it's also going to be at a great loss to spectral efficiency especially if you make the time slot smaller. Think how many new hams would get involved if there was cheap hardware around that didn't involve mucking with lots of wires and radio incompatibility. Not too long ago I built a softrock txrx which was pretty simple for what it does, it can't be that hard to build something like that made for VHF/UHF that has a built in ADC/DAC and something like a STM32... Just a thought anyway.. Also codec2 will probably always have some tweaking going on, I figure if something gets built that supports 1400bps or maybe 2400bps full duplex, it'll get used, if codec2 is less than that you have some extra bits to play with; extra FEC or RSSI or header info? Regards, Daniel Daniel Mundall On Mon, Apr 21, 2014 at 5:17 AM, Ross Whenmouth wrote: > Hi, > > In between work and study, I have also been thinking about TDD for > Codec2 on the shorter wavelengths. > > GMSK or C4FM is not only power efficient, but generally speaking, it is > also compatible with all those "9k6 data ready" VHF/UHF transceivers > that are out there. More people will probably connect their PC soundcard > or a STM32F4 discovery board into their existing radio than will go out > an buy a special experimental radio just to play with Codec2. In light > of this, it might be a good idea for at least one of the TDD operating > modes to be compatible with as many existing transceivers as possible > (eg a DC free line code with generous T-R switching delay allowances). > Then, when people like it, and build or buy new Codec2-compatible > transceivers, repeaters can be configured to use the advanced modes that > can cram lots of voice and data into a 25kHz channel (executed properly, > this could also alleviate some of the APRS people have with poor > collision avoidance with Ax.25). > > > Before anyone rushes off a writes a (TDD) protocol for Codec2, how > likely is the frame size/frame rate/etc of Codec2 to change in the near > future? > > > 73 > Ross Whenmouth ZL2WRW > > > -- > Start Your Social Network Today - Download eXo Platform > Build your Enterprise Intranet with eXo Platform Software > Java Based Open Source Intranet - Social, Extensible, Cloud Ready > Get Started Now And Turn Your Intranet Into A Collaboration Platform > http://p.sf.net/sfu/ExoPlatform > ___ > Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
Hi, In between work and study, I have also been thinking about TDD for Codec2 on the shorter wavelengths. GMSK or C4FM is not only power efficient, but generally speaking, it is also compatible with all those "9k6 data ready" VHF/UHF transceivers that are out there. More people will probably connect their PC soundcard or a STM32F4 discovery board into their existing radio than will go out an buy a special experimental radio just to play with Codec2. In light of this, it might be a good idea for at least one of the TDD operating modes to be compatible with as many existing transceivers as possible (eg a DC free line code with generous T-R switching delay allowances). Then, when people like it, and build or buy new Codec2-compatible transceivers, repeaters can be configured to use the advanced modes that can cram lots of voice and data into a 25kHz channel (executed properly, this could also alleviate some of the APRS people have with poor collision avoidance with Ax.25). Before anyone rushes off a writes a (TDD) protocol for Codec2, how likely is the frame size/frame rate/etc of Codec2 to change in the near future? 73 Ross Whenmouth ZL2WRW -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform ___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
It looks like the available NCOs don't get high enough in frequency. I think we need 350 MHz to over 2 GHz. On April 20, 2014 7:47:46 PM PDT, Bruce Perens wrote: > > > >On April 20, 2014 7:18:07 PM PDT, David Rowe wrote: > >>Bruce - I don't think TDMA will require PLL locking, tx and rx would >be all on one frequency. > >Not sure this is always the case for Whitebox. And we are looking at >low-phase-noise DDS chips, and they still have a VCO. If you know of a >really good NCO chip without one we'd like to know. > >Thanks > >Bruce > >-- >Sent from my Android device with K-9 Mail. Please excuse my brevity. > >-- >Start Your Social Network Today - Download eXo Platform >Build your Enterprise Intranet with eXo Platform Software >Java Based Open Source Intranet - Social, Extensible, Cloud Ready >Get Started Now And Turn Your Intranet Into A Collaboration Platform >http://p.sf.net/sfu/ExoPlatform >___ >Freetel-codec2 mailing list >[email protected] >https://lists.sourceforge.net/lists/listinfo/freetel-codec2 -- Sent from my Android device with K-9 Mail. Please excuse my brevity.-- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
On April 20, 2014 7:18:07 PM PDT, David Rowe wrote: >Bruce - I don't think TDMA will require PLL locking, tx and rx would be all on >one frequency. Not sure this is always the case for Whitebox. And we are looking at low-phase-noise DDS chips, and they still have a VCO. If you know of a really good NCO chip without one we'd like to know. Thanks Bruce -- Sent from my Android device with K-9 Mail. Please excuse my brevity. -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform ___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
Hi Daniel, Yes a SDR radio would be a good way to build this. Bruce - I don't think TDMA will require PLL locking, tx and rx would be all on one frequency. If required, a NCO (i.e. DDS) can change frequency instantly. Re PA linearity, it would be nice if the tx power can ramp up reasonably gently (say over 1 symbol period), to prevent splatter. This might not require a linear PA. There are already a few 100ms delays in FreeDV and no one has noticed, PTT is like that. However if you built a full duplex system (rather than a single channel repeater) it might be quite noticeable. Did you regenerate the filter coefficients (Octave script) for your higher sample rate? I would have thought that 100 Symbols/s wouldn't require a higher sample rate than Fs = 8000 Hz. So it's probably an implementation bug. I would suggest prototyping this using the Octave simulations of the fdmdv modem. Not sure about starting/stopping the demod. Maybe. Am wondering what effect this will have on the filter memories. Try inserting 0 symbols in the stream of symbols being fed to the transmit filter and see what the rx symbols look like. See if the mod waveform is nice and smooth as it starts/stops. You could do this first before changing the symbol rate, so you are only changing one thing at a time. TDMA systems usually require guard times between tx and rx as well - another good idea to make your bursts several symbols long, the guard times could be one symbol long, then the tx and rx will all have identical timing, just feed 0 symbols to the tx when listening during the rx timeslot. So it could be . GGGG . were G is a guard symbol (no one transmitting), R is rx symbol, T is tx symbol. The guard times consume bandwidth overhead. Burst modems can be tricky, not sure how freq offset estimation and sync will work. Might be better to lock that out for now. Cheers, David On Sun, 2014-04-20 at 17:44 -0700, Daniel Mundall wrote: > David, > Thanks so much for your input! > That's the idea a super cheap VHF repeater(no duplexers), also > thinking of doing something with and SDR to allow one repeater to > manage like 10 channels in a 25khz of spectrum. But that's if I can > get the timing figured out. > Yes it does seem to be an issue on some radios... > > > If I can I'm trying to avoid noticeable delay, but it's an option if > all else fails.(I may well have to do that..) > One idea was to drop some of the carriers and run a higher symbol > rate, but for some reason I can't get the filtering to work right when > I change the sample rate above 16khz, and for anything higher than 100 > samples/second I seem to need a higher sample rate to keep the > noise/harmonics down. > > > 2. Great idea, I think I can get that to work! I doubt any clocks will > drift too much in that amount of time. > I guess the other thing is to have the first RX symbol after TX to > know what the last RX symbol was to compare that differential QPSK. > > > Would it be possible to just pause the demod program some how every > two symbols and revive it just where it left off? > > > 3. Thanks I'll give that a try. > Thanks, > Daniel va7drm > > Sent from Daniel's iPhone > > On Apr 20, 2014, at 5:03 PM, David Rowe wrote: > > > > Hi Daniel, > > > > That sounds very cool. Another nice application is single channel > > repeater work for VHF. Need a radio that can tx/rx switch fast > > enough. > > > > Couple of ideas: > > > > 1/ Have longer timeslots than 20ms, like say 200ms, estimate the rx > > symbol timing over that window. The timing is probably constant for > > 200ms anyway, as it changes only slowly. > > > > 2/ Have a square wave shaped window for the timing estimation, that > > is > > set to zero when you are transmitting, that way you get several of > > your > > rx timeslots included in the timing estimate. > > > > 3/ Test any changes by measuring the BER at an Eb/No point, as per > > the > > Octave simulations. It's really easy to introduce an error that > > will > > degrade the demod performance. > > > > Cheers, > > > > David > > > > On Sun, 2014-04-20 at 14:58 -0700, Daniel Mundall wrote: > > > Hello Guys, > > > I've been playing with the FDMDV modem to see if I can get it to > > > support Time Domain Duplexing(TDD), with the end of goal of making > > > full-duplex voice in around 2KHz. > > > > > > > > > Currently I have the modem configured at 100 symbols/second with > > > 14 > > > carriers spaced at 125Hz it works out to 2800bits/s. From my > > > testing > > > that end of things seems to be working pretty well. I'm planning > > > on > > > splitting this up into 20ms chunks (2 symbols at a time). > > > > > > > > > What I'm trying to figure out now is the best way to keep the > > > timing > > > when I never have more than two consecutive symbols. > > > Does anyone have a good idea to keep the timing correct while > > > having > > > 20ms interruptions every 20ms? > > > Thanks, > > > Daniel M
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
Hi Daniel, I like the idea of TDD for a repeater instead of FDD. On HF it would give you break-in, but at the cost of some mostly-wasted bandwidth. Since HTs are power-limited, a modem that does not require linear amplification might work better for them. GMSK or C4FM, etc. There is a lot of talk about "Class D" amplifiers that can be linear and power efficient, but building one and getting the spurious emissions more than 60 dB down is probably still a research project. Existing radios will have keying delays based on, among other things, the lock-up time for the synthesizer PLL. It may be that the same PLL makes the LO frequency for receive, which is different due to the IF, so the lock-up delay between T and R can be significant. Often this isn't even specified. Thanks Bruce On April 20, 2014 5:44:41 PM PDT, Daniel Mundall wrote: >David, >Thanks so much for your input! >That's the idea a super cheap VHF repeater(no duplexers), also thinking >of >doing something with and SDR to allow one repeater to manage like 10 >channels in a 25khz of spectrum. But that's if I can get the timing >figured >out. >Yes it does seem to be an issue on some radios... > >If I can I'm trying to avoid noticeable delay, but it's an option if >all >else fails.(I may well have to do that..) >One idea was to drop some of the carriers and run a higher symbol rate, >but for some reason I can't get the filtering to work right when I >change >the sample rate above 16khz, and for anything higher than 100 >samples/second I seem to need a higher sample rate to keep the >noise/harmonics down. > >2. Great idea, I think I can get that to work! I doubt any clocks will >drift too much in that amount of time. >I guess the other thing is to have the first RX symbol after TX to know >what the last RX symbol was to compare that *differential QPSK.* > >*Would it be possible to just pause the demod program some how every >two >symbols and revive it just where it left off? * > >3. Thanks I'll give that a try. >Thanks, >Daniel va7drm > >Sent from Daniel's iPhone > >On Apr 20, 2014, at 5:03 PM, David Rowe wrote: > >Hi Daniel, > >That sounds very cool. Another nice application is single channel >repeater work for VHF. Need a radio that can tx/rx switch fast enough. > >Couple of ideas: > >1/ Have longer timeslots than 20ms, like say 200ms, estimate the rx >symbol timing over that window. The timing is probably constant for >200ms anyway, as it changes only slowly. > >2/ Have a square wave shaped window for the timing estimation, that is >set to zero when you are transmitting, that way you get several of your >rx timeslots included in the timing estimate. > >3/ Test any changes by measuring the BER at an Eb/No point, as per the >Octave simulations. It's really easy to introduce an error that will >degrade the demod performance. > >Cheers, > >David > >On Sun, 2014-04-20 at 14:58 -0700, Daniel Mundall wrote: > >Hello Guys, > >I've been playing with the FDMDV modem to see if I can get it to > >support Time Domain Duplexing(TDD), with the end of goal of making > >full-duplex voice in around 2KHz. > > > >Currently I have the modem configured at 100 symbols/second with 14 > >carriers spaced at 125Hz it works out to 2800bits/s. From my testing > >that end of things seems to be working pretty well. I'm planning on > >splitting this up into 20ms chunks (2 symbols at a time). > > > >What I'm trying to figure out now is the best way to keep the timing > >when I never have more than two consecutive symbols. > >Does anyone have a good idea to keep the timing correct while having > >20ms interruptions every 20ms? > >Thanks, > >Daniel Mundall VA7DRM > > >-- > >Learn Graph Databases - Download FREE O'Reilly Book > >"Graph Databases" is the definitive new guide to graph databases and >their > >applications. Written by three acclaimed leaders in the field, > >this first edition is now available. Download your free book today! > >http://p.sf.net/sfu/NeoTech > >___ Freetel-codec2 mailing >list >[email protected] >https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > > > > >-- >Start Your Social Network Today - Download eXo Platform >Build your Enterprise Intranet with eXo Platform Software >Java Based Open Source Intranet - Social, Extensible, Cloud Ready >Get Started Now And Turn Your Intranet Into A Collaboration Platform >http://p.sf.net/sfu/ExoPlatform >___ >Freetel-codec2 mailing list >[email protected] >https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > > > > >-- >Start Your Social Network Today - Download eXo Pl
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
David, Thanks so much for your input! That's the idea a super cheap VHF repeater(no duplexers), also thinking of doing something with and SDR to allow one repeater to manage like 10 channels in a 25khz of spectrum. But that's if I can get the timing figured out. Yes it does seem to be an issue on some radios... If I can I'm trying to avoid noticeable delay, but it's an option if all else fails.(I may well have to do that..) One idea was to drop some of the carriers and run a higher symbol rate, but for some reason I can't get the filtering to work right when I change the sample rate above 16khz, and for anything higher than 100 samples/second I seem to need a higher sample rate to keep the noise/harmonics down. 2. Great idea, I think I can get that to work! I doubt any clocks will drift too much in that amount of time. I guess the other thing is to have the first RX symbol after TX to know what the last RX symbol was to compare that *differential QPSK.* *Would it be possible to just pause the demod program some how every two symbols and revive it just where it left off? * 3. Thanks I'll give that a try. Thanks, Daniel va7drm Sent from Daniel's iPhone On Apr 20, 2014, at 5:03 PM, David Rowe wrote: Hi Daniel, That sounds very cool. Another nice application is single channel repeater work for VHF. Need a radio that can tx/rx switch fast enough. Couple of ideas: 1/ Have longer timeslots than 20ms, like say 200ms, estimate the rx symbol timing over that window. The timing is probably constant for 200ms anyway, as it changes only slowly. 2/ Have a square wave shaped window for the timing estimation, that is set to zero when you are transmitting, that way you get several of your rx timeslots included in the timing estimate. 3/ Test any changes by measuring the BER at an Eb/No point, as per the Octave simulations. It's really easy to introduce an error that will degrade the demod performance. Cheers, David On Sun, 2014-04-20 at 14:58 -0700, Daniel Mundall wrote: Hello Guys, I've been playing with the FDMDV modem to see if I can get it to support Time Domain Duplexing(TDD), with the end of goal of making full-duplex voice in around 2KHz. Currently I have the modem configured at 100 symbols/second with 14 carriers spaced at 125Hz it works out to 2800bits/s. From my testing that end of things seems to be working pretty well. I'm planning on splitting this up into 20ms chunks (2 symbols at a time). What I'm trying to figure out now is the best way to keep the timing when I never have more than two consecutive symbols. Does anyone have a good idea to keep the timing correct while having 20ms interruptions every 20ms? Thanks, Daniel Mundall VA7DRM -- Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/NeoTech ___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2 -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform ___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2 -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
Re: [Freetel-codec2] FDMDV Time Domain Duplexing
Hi Daniel, That sounds very cool. Another nice application is single channel repeater work for VHF. Need a radio that can tx/rx switch fast enough. Couple of ideas: 1/ Have longer timeslots than 20ms, like say 200ms, estimate the rx symbol timing over that window. The timing is probably constant for 200ms anyway, as it changes only slowly. 2/ Have a square wave shaped window for the timing estimation, that is set to zero when you are transmitting, that way you get several of your rx timeslots included in the timing estimate. 3/ Test any changes by measuring the BER at an Eb/No point, as per the Octave simulations. It's really easy to introduce an error that will degrade the demod performance. Cheers, David On Sun, 2014-04-20 at 14:58 -0700, Daniel Mundall wrote: > Hello Guys, > I've been playing with the FDMDV modem to see if I can get it to > support Time Domain Duplexing(TDD), with the end of goal of making > full-duplex voice in around 2KHz. > > > Currently I have the modem configured at 100 symbols/second with 14 > carriers spaced at 125Hz it works out to 2800bits/s. From my testing > that end of things seems to be working pretty well. I'm planning on > splitting this up into 20ms chunks (2 symbols at a time). > > > What I'm trying to figure out now is the best way to keep the timing > when I never have more than two consecutive symbols. > Does anyone have a good idea to keep the timing correct while having > 20ms interruptions every 20ms? > Thanks, > Daniel Mundall VA7DRM > > -- > Learn Graph Databases - Download FREE O'Reilly Book > "Graph Databases" is the definitive new guide to graph databases and their > applications. Written by three acclaimed leaders in the field, > this first edition is now available. Download your free book today! > http://p.sf.net/sfu/NeoTech > ___ Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 -- Start Your Social Network Today - Download eXo Platform Build your Enterprise Intranet with eXo Platform Software Java Based Open Source Intranet - Social, Extensible, Cloud Ready Get Started Now And Turn Your Intranet Into A Collaboration Platform http://p.sf.net/sfu/ExoPlatform ___ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
