RE: [digitalradio] The digital throughput challenge on HF
Why? Because your digital voice QSO sounds like noise to SSBers? From: John Becker [EMAIL PROTECTED] Reply-To: digitalradio@yahoogroups.com To: digitalradio@yahoogroups.com Subject: RE: [digitalradio] The digital throughput challenge on HF Date: Sun, 27 Aug 2006 20:29:05 -0500 You should try DIGITAL VOICE once. There is 1000% chance that your QSO will be QRM'ed At 04:43 AM 8/25/2006, you wrote: The problem is QRM. Consisting of PACTOR, MFSK, OLIVIA, PSK31, and on 30 meters also SSB signals coming on frequency during your qso. Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
You should try DIGITAL VOICE once. There is 1000% chance that your QSO will be QRM'ed At 04:43 AM 8/25/2006, you wrote: The problem is QRM. Consisting of PACTOR, MFSK, OLIVIA, PSK31, and on 30 meters also SSB signals coming on frequency during your qso. Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
There is always lot of emphasis on getting signals out of the noise. From my 9-months hands on experience using pskmail over a 1500 Mile path from Spain to Sweden and from Montenegro to Sweden, noise is not the problem on the amateur bands, as there is always a frequency you can use which gives you fair signal/noise performance. The problem is QRM. Consisting of PACTOR, MFSK, OLIVIA, PSK31, and on 30 meters also SSB signals coming on frequency during your qso. Normally that qrm is stronger than the server I am working. FEC does nothing to cure that. MFSK e.g. is inferior to PSK63 in that case, provided I use a 100 Hz filter. The only answer is ARQ. Pskmail will wait until the intruder is gone and finish the transfer when the qrm goes away within a certain time frame. My experience has also shown that it is very difficult to keep a frequency for more than say 30 minutes. Normally it won't take more than 10 minutes before a deaf, dumb and blind operator will spoil your fun with a 9+20 signal on top of you. 73, Rein PA0R -Ursprüngliche Nachricht- Von: digitalradio@yahoogroups.com Gesendet: 24.08.06 18:24:46 An: digitalradio@yahoogroups.com Betreff: RE: [digitalradio] The digital throughput challenge on HF Oh my...you are right about the baud rates of MT63...I was going from memory and I have the written down. The only problem with on-the-air testing is that you Never HAVE THE SAME CONDITIONS and you can do that with a simulator...but then the military always has a fly-off or shoot-off. I have no doubt that the sound card modes can work down into the noise. G4HPE(?) did some testing which is on the RSGB site under emergency communications. He used a software simulator and on-the-air testing. MFSK16 worked way down into the noise. KN6KB in this presentation on SCAMP to the DDC in Nov of 2004 showed his simulator measurements of P I/II/III and I believe MT63. I have the chart from the presentation that I will send to anyone interested. I have to agree that MT63 or anything else needs to have further development by knowledgeable individuals. Being more of a project manager type than technical (programmer) type, over the years I have specialized one on tactical HF antennas and overall communications than specific data modes or programming...I know a little C and C++ ;-). I would hope that some hardware mode better then pactor might come along but also would like to see single and multiple soundcard modes develop that might rival Pactor speeds. 73, Walt/K5YFW http://pa0r.blogspirit.com Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
How does the crest factor relate to the mean power vs the peak power? It doesn't seem correct to add 3 to that figure to come up with the crest factor. Patrick has the peak and mean power for the various modes listed in the documentation for Multipsk, but I am not clear how to convert them to crest factor. My understanding is that the peak power and average power of a rectangular wave is 1. It can't be correct to add 3 to that value to come up with 4, can it? And MT-63 which has a peak to average of 10 times has a crest factor of 13? If you want to broadcast a message from one to many, then the only practical alternative is to use a non-ARQ mode, typically with a large amount of FEC. While this is done on amateur frequencies for sending a bulletin, calling CQ, and having a roundtable, if your goal is to have accurate messaging, then I don't see any option other than a good ARQ system. If Clover II would have worked better, I would have considered keeping my HAL P-38 board. But it was not that good with weak signals. Also, the P-38 had serious problems with Pactor back then. I remember someone later criticizing me for not using a 386 computer for the card. But at that time the 386 was barely even invented and 286 machines were state of the art. 73, Rick, KV9U Mark Miller wrote: At 10:33 PM 8/23/2006, you wrote: I am not very knowledgeable on CRF (Crest Factors). Can you give us an idea of converting peak power/average power into CRF? Using powers, crest factor = Peak Instantaneous Power / Average Power. A more piratical way of measuring crest factor is (PEP/Average Power) + 3dB. I agree that ARQ has its benefits, but we still have to rely on the modem scheme. This was my point earlier, that we reach a limit because we are power and bandwidth limited. Because we are using HF frequencies, we pay a coding penalty. Also if we look at the broadcast nature of non-ARQ modes, it is apparent that they are much more efficient than ARQ modes. This does not mean that ARQ does not have its place, it certainly does. The more tools in the tool box the better. BTW I am an AMTOR OT myself. I remember when APLINK was used before unattended operation was allowed on HF. I miss keyboarding with AMTOR/PACTOR and CLOVER. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
I'm not sure that we can quantify ARQ vs. broadcast. One thing which has been over looked is that we think of ARQ as sending a packet(s) and the receiving station sends an ACK. If however each packet is numbered and contains a CRC number, then if the receiving station misses a packet (missing number) or does not resolve the CRC number, only then does it transmit and that is a NAK. In this condition, you could broadcast and only provide fills to those stations for specific packets. If two stations missed the packet, then if the re-transmission to one stations is copied by the second stations...so only one retransmission per packet would be needed even is several stations missed the same packet. ACK is out...NAK is in. Combine varying levels of FEC with ARQ NAKs will increase error free information without utilizing a lot of time with retransmissions. Walt/K5YFW -Original Message- From: digitalradio@yahoogroups.com [mailto:[EMAIL PROTECTED] Sent: Wednesday, August 23, 2006 8:17 PM To: digitalradio@yahoogroups.com Subject: RE: [digitalradio] The digital throughput challenge on HF At 04:29 PM 8/23/2006, you wrote: It in deed would. That is the reason Pactor and Amtor work so well. It's the AQR - even with the hi S/N needed. There is some value to ARQ, I wonder how we would quantify the advantage? In a point to point link I think it would be easy, but in a point to multipoint network, I think the value is significantly diminished. From an efficiency standpoint, broadcast modes like soundcard modes are very efficient. Point to point modes can be very reliable and very accurate, but very inefficient. I am not sure how one quantifies these differences. When it comes to speed and or throughput, we have the bandwidth, power, and coding barrier with which we much deal. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
Rick, My explanation was for sinusoids not rectangular waves, our radios transmit sinusoids. You are correct about rectangular waves they would have a crest factor of 1 in linear terms or voltage terms, and 0B in non-linear or power terms. Yes MT63 has a crest factor of 13dB. It is very high. Lets deified a few power terms. Average or mean power is what you get when you multiply RMS voltage and current Peak Instantaneous power or peak power is what you get when you multiply peak instantaneous voltage and current. Peak Envelope power or PEP is what you get when you average the peak power over one RF cycle. Crest Factor is normally given in terms of voltage and is equal to the peak amplitude of a waveform divided by the RMS value. In terms of power this is the 10log(peak power/average power). The relationship between peak power and PEP in a sinusoid is 3 dB. This is very easy to prove. The relationship between peak voltage and RMS voltage is the square root of 2. 20 log of the square root of 2 is 3dB. Lets use a more complicated example. Lets say we generate a sinusoid from two equal amplitude tones. A two tone test. On an oscilloscope we observe that the peak voltage is 1 voltage unit for each tone which makes the peak power 1 power unit for each tone. The peak voltage of the envelope is 2 voltage units , so the peak power is 4 power units for the envelope. Each tone's RMS voltage is 1/square root of 2 or approximately .707 voltage units. The average power is the RMS voltage squared or .5 units. The total average power of the two tones is .5 + .5 or 1 unit. When using N tones to produce an envelope, PEP to average power ratio is 1/N. In this case it is .5, which means that the PEP is 2 units. Here are the relationships Peak power of the envelope = 4 PEP of the envelope = 2 Average power of the envelope = 1. PEP to average ratio = 3dB Peak to average ratio = 6 dB Difference = 3 dB This same example can be worked with any number of tones. Patrick used two programs if I remember correctly to calculate the peak and mean power for the various modes listed in the documentation for Multipsk. The two programs were Cool Edit Pro and Sox. In Cool Edit Pro the peak value given in the statistics is PEP. What Patrick is giving you is the PEP to average ratio. I have proven this in Cool Edit Pro using the 2 tone example above. 73, Mark N5RFX At 09:12 AM 8/24/2006, you wrote: How does the crest factor relate to the mean power vs the peak power? It doesn't seem correct to add 3 to that figure to come up with the crest factor. Patrick has the peak and mean power for the various modes listed in the documentation for Multipsk, but I am not clear how to convert them to crest factor. My understanding is that the peak power and average power of a rectangular wave is 1. It can't be correct to add 3 to that value to come up with 4, can it? And MT-63 which has a peak to average of 10 times has a crest factor of 13? If you want to broadcast a message from one to many, then the only practical alternative is to use a non-ARQ mode, typically with a large amount of FEC. While this is done on amateur frequencies for sending a bulletin, calling CQ, and having a roundtable, if your goal is to have accurate messaging, then I don't see any option other than a good ARQ system. If Clover II would have worked better, I would have considered keeping my HAL P-38 board. But it was not that good with weak signals. Also, the P-38 had serious problems with Pactor back then. I remember someone later criticizing me for not using a 386 computer for the card. But at that time the 386 was barely even invented and 286 machines were state of the art. 73, Rick, KV9U Mark Miller wrote: At 10:33 PM 8/23/2006, you wrote: I am not very knowledgeable on CRF (Crest Factors). Can you give us an idea of converting peak power/average power into CRF? Using powers, crest factor = Peak Instantaneous Power / Average Power. A more piratical way of measuring crest factor is (PEP/Average Power) + 3dB. I agree that ARQ has its benefits, but we still have to rely on the modem scheme. This was my point earlier, that we reach a limit because we are power and bandwidth limited. Because we are using HF frequencies, we pay a coding penalty. Also if we look at the broadcast nature of non-ARQ modes, it is apparent that they are much more efficient than ARQ modes. This does not mean that ARQ does not have its place, it certainly does. The more tools in the tool box the better. BTW I am an AMTOR OT myself. I remember when APLINK was used before unattended operation was allowed on HF. I miss keyboarding with AMTOR/PACTOR and CLOVER. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.orgTelnet://cluster.dynalias.org Other areas of interest: The MixW Reflector :
Re: [digitalradio] The digital throughput challenge on HF - peak and mean power
Hello Rick, the peak and mean power for the various modes listed in the In Multipsk, you have the ratio between the average power and the peak power. The peak power is obtained if you have no band base windowing and if you are always transmitting one carrier at a given time (RTTY, MFSK16...). If you have one carrier (PSK31, for example) but a windowing or many carriers at a the same time (CMT/HELL for Video ID or MT63), you ratio is less than 1 (close to 1 for PSK31 and close to 0 for CMT/HELL for Video ID or MT63). It means that at the maximum, remaining linear, you cannot transmit much power: for aexample, 10 watts in MT63 for a 100 watts standard transceiver. This supposes that the AF is a sine. The crest factor of a sine is about -3 dB compared to a square wave (AF signal completly saturated). But there is no reason to transmit an AF square wave. There would not be a decoding gain (because the power considered by the decoder would be the one of the fondamental which would be -3 dB compared to the total power) and the signal would be found on all the AF spectrum (fondamental and harmonics). So the ratio to be considered is the one considered by Multipsk (taking into account that the figures are not very precise, but sufficient for comparizons). 73 Patrick - Original Message - From: KV9U To: digitalradio@yahoogroups.com Sent: Thursday, August 24, 2006 4:12 PM Subject: Re: [digitalradio] The digital throughput challenge on HF How does the crest factor relate to the mean power vs the peak power? It doesn't seem correct to add 3 to that figure to come up with the crest factor. Patrick has the peak and mean power for the various modes listed in the documentation for Multipsk, but I am not clear how to convert them to crest factor. My understanding is that the peak power and average power of a rectangular wave is 1. It can't be correct to add 3 to that value to come up with 4, can it? And MT-63 which has a peak to average of 10 times has a crest factor of 13? If you want to broadcast a message from one to many, then the only practical alternative is to use a non-ARQ mode, typically with a large amount of FEC. While this is done on amateur frequencies for sending a bulletin, calling CQ, and having a roundtable, if your goal is to have accurate messaging, then I don't see any option other than a good ARQ system. If Clover II would have worked better, I would have considered keeping my HAL P-38 board. But it was not that good with weak signals. Also, the P-38 had serious problems with Pactor back then. I remember someone later criticizing me for not using a 386 computer for the card. But at that time the 386 was barely even invented and 286 machines were state of the art. 73, Rick, KV9U Mark Miller wrote: At 10:33 PM 8/23/2006, you wrote: I am not very knowledgeable on CRF (Crest Factors). Can you give us an idea of converting peak power/average power into CRF? Using powers, crest factor = Peak Instantaneous Power / Average Power. A more piratical way of measuring crest factor is (PEP/Average Power) + 3dB. I agree that ARQ has its benefits, but we still have to rely on the modem scheme. This was my point earlier, that we reach a limit because we are power and bandwidth limited. Because we are using HF frequencies, we pay a coding penalty. Also if we look at the broadcast nature of non-ARQ modes, it is apparent that they are much more efficient than ARQ modes. This does not mean that ARQ does not have its place, it certainly does. The more tools in the tool box the better. BTW I am an AMTOR OT myself. I remember when APLINK was used before unattended operation was allowed on HF. I miss keyboarding with AMTOR/PACTOR and CLOVER. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links [Non-text portions of this message have been removed] Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
Oh my...you are right about the baud rates of MT63...I was going from memory and I have the written down. The only problem with on-the-air testing is that you Never HAVE THE SAME CONDITIONS and you can do that with a simulator...but then the military always has a fly-off or shoot-off. I have no doubt that the sound card modes can work down into the noise. G4HPE(?) did some testing which is on the RSGB site under emergency communications. He used a software simulator and on-the-air testing. MFSK16 worked way down into the noise. KN6KB in this presentation on SCAMP to the DDC in Nov of 2004 showed his simulator measurements of P I/II/III and I believe MT63. I have the chart from the presentation that I will send to anyone interested. I have to agree that MT63 or anything else needs to have further development by knowledgeable individuals. Being more of a project manager type than technical (programmer) type, over the years I have specialized one on tactical HF antennas and overall communications than specific data modes or programming...I know a little C and C++ ;-). I would hope that some hardware mode better then pactor might come along but also would like to see single and multiple soundcard modes develop that might rival Pactor speeds. 73, Walt/K5YFW -Original Message- From: digitalradio@yahoogroups.com [mailto:[EMAIL PROTECTED] Sent: Wednesday, August 23, 2006 8:38 PM To: digitalradio@yahoogroups.com Subject: Re: [digitalradio] The digital throughput challenge on HF To be onest, Walt, I don't see Rick's claim of such a good performance level for MT-63. If you look at his presentation on comparing several modes with Pactor, at: http://winlink.org/Presentations/RFfootprints.PDF he seems to suggest that all the non-ARQ sound card modes (e.g, PSK-31, MT-63) will only work down to about 0 db S/N. He also claims that P1, 2, and 3 all go to about -5 before they shut down. I am not as pessimistic as his data shows, but real world (not necessarily simulator) tests seem to put P modes well into the noise. I wish the RSGB would have another series of tests with some additional new modes and compare with P modes. I believe that if you check the baud rates of MT-63, you will find them to be even lower than 31 baud. MT-63 1K at only 10 baud and MT-63 2K at 20 baud. So I agree that it should be able to work quite well at 40 baud and 80 baud and even 160 baud under good conditions. I asked about this in the past and got the impression that this would be hard to do. But then again, Patrick was able to take PSK31 and increase the baud rate a great deal for a faster mode. A pipelined ARQ MT-63 mode (or something along those lines) running at several speeds that can auto switch on the fly would be a major coup for amateur radio. When word throughput is used, I consider the following: cps (characters per second) is about equal to wpm (words per minute) times 10 so 10 cps ~ 100 wpm This is because one uses an average of 5 character words and a space which comes to six total characters per word and it also makes it easy to calculate in your head. 73, Rick, KV9U DuBose Walt Civ AETC CONS/LGCA wrote: Rick, You asked...Why do the Pactor modes work so well? They have the same bandwidth, power, and fairly similar coding to sound card modes. Is their coding something that can not be implemented on current sound cards in terms of the modulation? Here may be part of the answer... Measurements made by Rick, KN6KB, when working on SCAMP measured PI/II/III with the KC7WW HF Channel Simulator found that PIII at a -10 dB SNR had a slightly better throughput than MT63. MT63-2K has been measured by KC7WW using the same simulator that he sold KN6KB and found that MT63 needed a about a 5 dB better SNR than Pactor III to have the same throughput. The problem with MT63 is that it does not change its modulation dynamically as the SNR changes but Pactor III does. So when conditions are good, Pactor III screams. But when conditions are very poor, Pactor III is not that much better than MT63. Another thing is that MT63 doesn't use ARQ and Pactor does. Also, the modulation rate is lower than optimum for all of the HF bands...31 Hz. Research for the past 30 years has reveiled that a 45-50 baud modulation rate works very well on HF. Thus if MT63 kicked up its modulation rate and added ARQ, it might very well outperform Pactor III and low SNRs. If you added dymanic modulation changes to MT63, you might very well have a throughput of 400-800 WPM. A typed page is about 720 words. Copy some E-Mail into your word processor some time and do a word count...you might be surprised. 73, Walt/K5YFW Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups
RE: [digitalradio] The digital throughput challenge on HF
Research done by private research firms have addressed this problem. For medium length messages in the broadcast mode, they suggest heavy FEC. For longer messages they suggest less FEC and NAKs. For critical messages of any length they suggest FEC and ACKs. For many receiving stations, they suggest medium to heavy FEC. Their research grant was for a national/international national command structure type message for many receivers. The length of their message was not revealed. The research done is considered intellectual property. They did find that the number of bits recovered from a single tone carrier that is generally used is less than what can easily be obtained. I do know that the computers they used were all of the single Intel P4 category. The mode(s) they used were all developed on common sound cards as well as special hardware soundcards. Walt/K5YFW -Original Message- From: digitalradio@yahoogroups.com [mailto:[EMAIL PROTECTED] Sent: Thursday, August 24, 2006 9:13 AM To: digitalradio@yahoogroups.com Subject: Re: [digitalradio] The digital throughput challenge on HF [stuff deleted] If you want to broadcast a message from one to many, then the only practical alternative is to use a non-ARQ mode, typically with a large amount of FEC. While this is done on amateur frequencies for sending a bulletin, calling CQ, and having a roundtable, if your goal is to have accurate messaging, then I don't see any option other than a good ARQ system. 73, Rick, KV9U Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
[digitalradio] The digital throughput challenge on HF
If I were a company technology officer, of a company who's purpose was developing communications technology...or the technology officer for amateur radio, I would be very dis-heartened at the data protocols/modes/modems produces as well as the HF E-Mail applications developed. None are really as robust as the should/could be, none of the sound card modes have the throughput that they should and there are is no really good HF E-Mail program that is based on the capability of operating stand-alone without using the Internet. Walt, You have pointed out a basic principle with respect to data throughput. Throughput is a function of bandwidth, power, and coding. With amateur HF we are power, and bandwidth limited. The nature of the media we are opening in makes forward error correction a must, thus we suffer a loss of throughput because of coding. The very robust modes like MT63 and Olivia require interleaving and convolutional coding. Compare MT63 and Olivia with RDFT or amateur DRM. RDFT and DRM are great modes, but requires a fairly high S/N ratio. The challenge is there, but the solution is far from easy. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
And we must do more research and testing. The latest research from university and private research institutions and industry indicates that high data throughputs in a wide channel, such as a 10 KHz data channel, can be substantially more than the collective throughput of five 2 KHz channels. Additionally, recent research indicates that on one wide channel, there is the possibility of having more than just one specific data stream. Thus, if you could have a throughput of 128 Kbps in a 10 KHz channel, you might only be able to pass 9.6 Kbps in five 2 KHz channels (48 Kbps total). Additionally, depending on the actual total data throughput per epoch, it is probably that 6 or 8 QSOs might well all share the same 10 KHz space without QRMing each other and still obtain and aggregate of 128 Kbps throughput. Walt/K5YFW -Original Message- From: digitalradio@yahoogroups.com [mailto:[EMAIL PROTECTED] Sent: Wednesday, August 23, 2006 11:52 AM To: digitalradio@yahoogroups.com Subject: [digitalradio] The digital throughput challenge on HF If I were a company technology officer, of a company who's purpose was developing communications technology...or the technology officer for amateur radio, I would be very dis-heartened at the data protocols/modes/modems produces as well as the HF E-Mail applications developed. None are really as robust as the should/could be, none of the sound card modes have the throughput that they should and there are is no really good HF E-Mail program that is based on the capability of operating stand-alone without using the Internet. Walt, You have pointed out a basic principle with respect to data throughput. Throughput is a function of bandwidth, power, and coding. With amateur HF we are power, and bandwidth limited. The nature of the media we are opening in makes forward error correction a must, thus we suffer a loss of throughput because of coding. The very robust modes like MT63 and Olivia require interleaving and convolutional coding. Compare MT63 and Olivia with RDFT or amateur DRM. RDFT and DRM are great modes, but requires a fairly high S/N ratio. The challenge is there, but the solution is far from easy. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
Rick, You asked...Why do the Pactor modes work so well? They have the same bandwidth, power, and fairly similar coding to sound card modes. Is their coding something that can not be implemented on current sound cards in terms of the modulation? Here may be part of the answer... Measurements made by Rick, KN6KB, when working on SCAMP measured PI/II/III with the KC7WW HF Channel Simulator found that PIII at a -10 dB SNR had a slightly better throughput than MT63. MT63-2K has been measured by KC7WW using the same simulator that he sold KN6KB and found that MT63 needed a about a 5 dB better SNR than Pactor III to have the same throughput. The problem with MT63 is that it does not change its modulation dynamically as the SNR changes but Pactor III does. So when conditions are good, Pactor III screams. But when conditions are very poor, Pactor III is not that much better than MT63. Another thing is that MT63 doesn't use ARQ and Pactor does. Also, the modulation rate is lower than optimum for all of the HF bands...31 Hz. Research for the past 30 years has reveiled that a 45-50 baud modulation rate works very well on HF. Thus if MT63 kicked up its modulation rate and added ARQ, it might very well outperform Pactor III and low SNRs. If you added dymanic modulation changes to MT63, you might very well have a throughput of 400-800 WPM. A typed page is about 720 words. Copy some E-Mail into your word processor some time and do a word count...you might be surprised. 73, Walt/K5YFW -Original Message- From: digitalradio@yahoogroups.com [mailto:[EMAIL PROTECTED] Sent: Wednesday, August 23, 2006 1:57 PM To: digitalradio@yahoogroups.com Subject: Re: [digitalradio] The digital throughput challenge on HF Mark, Why do the Pactor modes work so well? They have the same bandwidth, power, and fairly similar coding to sound card modes. Is their coding something that can not be implemented on current sound cards in terms of the modulation? P2 is a variable DPSK mode and P3 is OFDM are they not? 73, Rick, KV9U Mark Miller wrote: You have pointed out a basic principle with respect to data throughput. Throughput is a function of bandwidth, power, and coding. With amateur HF we are power, and bandwidth limited. The nature of the media we are opening in makes forward error correction a must, thus we suffer a loss of throughput because of coding. The very robust modes like MT63 and Olivia require interleaving and convolutional coding. Compare MT63 and Olivia with RDFT or amateur DRM. RDFT and DRM are great modes, but requires a fairly high S/N ratio. The challenge is there, but the solution is far from easy. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
It in deed would. That is the reason Pactor and Amtor work so well. It's the AQR - even with the hi S/N needed. I got into Amtor in the early days when the KIT BOARD was over 500 bucks. Ask HB9AVK what he thinks of the AQR modes and Amtor in general. Or G3GPS. A lot of us old RTTY'ers played with Amtor way back then. John, W0JAB At 03:33 PM 8/23/2006, you wrote in part: Thus if MT63 kicked up its modulation rate and added ARQ, it might very well outperform Pactor III and low SNRs. If you added dymanic modulation changes to MT63, you might very well have a throughput of 400-800 WPM. A typed page is about 720 words. Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
Mark, If what you say was true, it would be easy to have sound card modes that compete with Pactor modes. From studies that I have seen, Amtor can work down around zero db S/N. Same with Pactor I. Some claim a bit below 0 db. In fact one recent test claimed that RTTY was better than PSK31 for a special case. These modes all run about 50 or so wpm. See Figure 7 at: http://ecjones.org/pactor.html Note that the test claims that Pactor 2 runs at around 1000 wpm down to about -10 db. Pactor 3 is only slightly better at that noise level at around 1500 wpm. Of course Pactor 3 scales up so that it can run at around 3000 wpm at 0 db and over 6000 wpm at very high signal levels ( 30 db). Those are magnitudes of difference. The weak signal ability of Pactor 2 and 3 converge around -12 db, when the other modes were inoperative, but are still close to 1000 wpm. Note also in Figure 6, the real world test by using distance on 80 meters daytime. The worst performance was by Amtor, followed by Pactor 1 and closely by PSK31. The best performer was RTTY at these slow speeds and he gives his explanation as why he believes this occurs. It sounds reasonable to me. And also note that the non-ARQ modes always had some errors and the ARQ modes were error free. But the P2 and P3 modes are so far ahead of anything else, even with weak signals. And others who have compared the modes in real life also make similar claims. Some of this is due to tweaking techniques such as compression, but it is only part of the explanation. 73, Rick, KV9U I am not sure I quite follow you Rick, I am not sure that the Pactor modes have any higher throughput vs. s/n ratio than the sound card modes. I have not performed the testing myself, but my gut tells me that the PACTOR modes require a fairly high S/N ratio to move data at their highest rates. I think the only limitation in implementing these modes in the Windows environment is timing. The other limiting factor is licensing. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
To be onest, Walt, I don't see Rick's claim of such a good performance level for MT-63. If you look at his presentation on comparing several modes with Pactor, at: http://winlink.org/Presentations/RFfootprints.PDF he seems to suggest that all the non-ARQ sound card modes (e.g, PSK-31, MT-63) will only work down to about 0 db S/N. He also claims that P1, 2, and 3 all go to about -5 before they shut down. I am not as pessimistic as his data shows, but real world (not necessarily simulator) tests seem to put P modes well into the noise. I wish the RSGB would have another series of tests with some additional new modes and compare with P modes. I believe that if you check the baud rates of MT-63, you will find them to be even lower than 31 baud. MT-63 1K at only 10 baud and MT-63 2K at 20 baud. So I agree that it should be able to work quite well at 40 baud and 80 baud and even 160 baud under good conditions. I asked about this in the past and got the impression that this would be hard to do. But then again, Patrick was able to take PSK31 and increase the baud rate a great deal for a faster mode. A pipelined ARQ MT-63 mode (or something along those lines) running at several speeds that can auto switch on the fly would be a major coup for amateur radio. When word throughput is used, I consider the following: cps (characters per second) is about equal to wpm (words per minute) times 10 so 10 cps ~ 100 wpm This is because one uses an average of 5 character words and a space which comes to six total characters per word and it also makes it easy to calculate in your head. 73, Rick, KV9U DuBose Walt Civ AETC CONS/LGCA wrote: Rick, You asked...Why do the Pactor modes work so well? They have the same bandwidth, power, and fairly similar coding to sound card modes. Is their coding something that can not be implemented on current sound cards in terms of the modulation? Here may be part of the answer... Measurements made by Rick, KN6KB, when working on SCAMP measured PI/II/III with the KC7WW HF Channel Simulator found that PIII at a -10 dB SNR had a slightly better throughput than MT63. MT63-2K has been measured by KC7WW using the same simulator that he sold KN6KB and found that MT63 needed a about a 5 dB better SNR than Pactor III to have the same throughput. The problem with MT63 is that it does not change its modulation dynamically as the SNR changes but Pactor III does. So when conditions are good, Pactor III screams. But when conditions are very poor, Pactor III is not that much better than MT63. Another thing is that MT63 doesn't use ARQ and Pactor does. Also, the modulation rate is lower than optimum for all of the HF bands...31 Hz. Research for the past 30 years has reveiled that a 45-50 baud modulation rate works very well on HF. Thus if MT63 kicked up its modulation rate and added ARQ, it might very well outperform Pactor III and low SNRs. If you added dymanic modulation changes to MT63, you might very well have a throughput of 400-800 WPM. A typed page is about 720 words. Copy some E-Mail into your word processor some time and do a word count...you might be surprised. 73, Walt/K5YFW Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
RE: [digitalradio] The digital throughput challenge on HF
At 04:29 PM 8/23/2006, you wrote: It in deed would. That is the reason Pactor and Amtor work so well. It's the AQR - even with the hi S/N needed. There is some value to ARQ, I wonder how we would quantify the advantage? In a point to point link I think it would be easy, but in a point to multipoint network, I think the value is significantly diminished. From an efficiency standpoint, broadcast modes like soundcard modes are very efficient. Point to point modes can be very reliable and very accurate, but very inefficient. I am not sure how one quantifies these differences. When it comes to speed and or throughput, we have the bandwidth, power, and coding barrier with which we much deal. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
Note also in Figure 6, the real world test by using distance on 80 meters daytime. The worst performance was by Amtor, followed by Pactor 1 and closely by PSK31. The best performer was RTTY at these slow speeds and he gives his explanation as why he believes this occurs. It sounds reasonable to me. And also note that the non-ARQ modes always had some errors and the ARQ modes were error free. Rick, If I boil your argument to 2 points it would be that the advantages of the Pactor modes are ARQ and low crest factors? 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
Hi Mark, The ARQ is really important. You really should have this for serious messaging via RF and must have it if you want to interface with a mailbox system or internet. Even one bad character trashes everything when negotiating a menu. Those who are OT's with Amtor know what I mean. I used to be so frustrating trying to use Amtor even though signals were still fairly good but you could not get anywhere with a message storage bbs as there were slight errors at the edges of what Amtor could still operate. Amtor was not a very good ARQ mode when condx got a bit bad. Pactor I did not seem to have this problem. It would just fail, even with signals still observable by ear. With ARQ, you can work deeper into the noise and still get some traffic through. My preference is to have modest FEC and then the ARQ. It amazes me that hams and especially developers are not truly excited about this. Rick, KN6KB has shown the way to get a pipelined ARQ method and eliminate the bottleneck that everyone claimed made ARQ not possible on PC's. I realize that there are only a few dozen hams who have the knowlege and can actually do this, but it only takes one to make it happen. I can not describe the thrill I got when using the beta SCAMP software. It is just so cool to watch something work so well from a soundcard at modest signal strengths. All we need is a bit lower (-5?) S/N mode that can scale upwards when condx are good enough and yet scale downwards when you have to in order to get some throughput. I am not very knowledgeable on CRF (Crest Factors). Can you give us an idea of converting peak power/average power into CRF? My understanding is that Pactor 2 has a CRF of 1.45 and if I understand things correctly, many of the raised cosine modulation schemes are similar. Am I correct that the rectangular waveforms have a CRF of 1 since square waves have a crest factor of 1? I understand that Pactor 3 has quite a variable CRF depending upon the number of tones: 1DBPSK = 1.9 3 DBPSK = 3.1 6 DBPSK = 5.7 Meanwhile, MT-63 has a large ratio between peak and average power. So that means it has a very high crest factor? 73, Rick, KV9U Mark Miller wrote: Note also in Figure 6, the real world test by using distance on 80 meters daytime. The worst performance was by Amtor, followed by Pactor 1 and closely by PSK31. The best performer was RTTY at these slow speeds and he gives his explanation as why he believes this occurs. It sounds reasonable to me. And also note that the non-ARQ modes always had some errors and the ARQ modes were error free. Rick, If I boil your argument to 2 points it would be that the advantages of the Pactor modes are ARQ and low crest factors? 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/
Re: [digitalradio] The digital throughput challenge on HF
At 10:33 PM 8/23/2006, you wrote: I am not very knowledgeable on CRF (Crest Factors). Can you give us an idea of converting peak power/average power into CRF? Using powers, crest factor = Peak Instantaneous Power / Average Power. A more piratical way of measuring crest factor is (PEP/Average Power) + 3dB. I agree that ARQ has its benefits, but we still have to rely on the modem scheme. This was my point earlier, that we reach a limit because we are power and bandwidth limited. Because we are using HF frequencies, we pay a coding penalty. Also if we look at the broadcast nature of non-ARQ modes, it is apparent that they are much more efficient than ARQ modes. This does not mean that ARQ does not have its place, it certainly does. The more tools in the tool box the better. BTW I am an AMTOR OT myself. I remember when APLINK was used before unattended operation was allowed on HF. I miss keyboarding with AMTOR/PACTOR and CLOVER. 73, Mark N5RFX Need a Digital mode QSO? Connect to Telnet://cluster.dynalias.org Other areas of interest: The MixW Reflector : http://groups.yahoo.com/group/themixwgroup/ DigiPol: http://groups.yahoo.com/group/Digipol (band plan policy discussion) Yahoo! Groups Links * To visit your group on the web, go to: http://groups.yahoo.com/group/digitalradio/ * To unsubscribe from this group, send an email to: [EMAIL PROTECTED] * Your use of Yahoo! Groups is subject to: http://docs.yahoo.com/info/terms/