Re: [digitalradio] Re: RFSM 8000
Hi Dmitry, Although I may not completely understand S/N ratio, if we can haves some kind of relative comparison between two or more modes, and done consistently, should give us a way to compare in a practical manner. Examples of claims that PSK31 can work down to perhaps-10 dB S/N and Pactor 2 and 3 stop working by -15 dB as do most modes including CW, depending upon the operator. A practical ham measurement is the hroughput in characters per unit of time at a given S/N and BW,necessary to help determine practical the mode is ... or is not. And the robustness from difficult conditions with ISI, doppler, QRM, QRN, etc. The standard CCIR Channel parameters for Doppler and Multipath are very modest with a poor channel being considered to have 1.0 Hz of Doppler and 2.0 ms of delay. In fact, it is not that uncommon to have over 5 ms of multipath delay spread and Doppler spread of 5 Hz. Indeed, under the really difficult conditions of polar paths, the delay can be 10 ms and 50 Hz of Doppler. (Source: Iterative Equalization and Decoding of Stanag 4539 9600 PBS HF Waveforms, John W. Nieto paper). While I don't have any practical way to measure this, do the STANAG modems detect the ISI and doppler spreads? Do they display these parameters? Many of us have used different modes (including PSK and RTTY and the throughput went to zero, even with signals present that one might think should be working. This is where some specialty modes such as DEX (Domino EX) might be able to work well for keyboard speeds due to the very low baud rates possible with multitone modems. Under those really difficult conditions, I suspect that PSK modes, including Pactor, will not work at all. For amateur use, the high baud rate modes are rarely possible to use based upon comments I have received from the digital SSTV operators. They find that 16-QAM is usually the highest constellation that works with the OFDM modems that are commonly used. I found it surprising how rapidly the DSSTV operators moved away from the short lived RDFT waveform, toward OFDM. And this was primarily because of performance of only a few dB improvement. When I look at the required (and even measured computer simulations ... but not real world measurements) S/N for low BER (10 -5) and in a 3 kHz bandwidth with say, STANAG 4539, (functionally the same as MIL-STD 188-110 B), the government expectation is +9 dB for 3200 bps. And much higher of course for faster data rates. Thus, I don't see these data rates as anything very useful for the kinds of things that we hams are involved in on HF. My experience seems to be that we are lucky to have a solid +10 dB S/N, but usually there is still some QSB. Then we can expect signals to deteriorate to -10 dB or even worse. My thinking is that any practical amateur digital modes need to focus on the +5 dB and downward and not be too concerned about faster speeds that are possible with +10 dB or greater signals. Many of the military designs require at least 1000 watts or even much higher power levels plus good antennas on each end of the circuit, something we hams typically don't have available to us. One thing that disappointed me about the STANAG standards is what I view as poor performance at the low bit rates. Even at the slowest data rate of 75 bps, the main standard only has to meet +2 dB S/N. Simulations show some competing products to work close to -5 dB. The STANAG 4415 75 bps very robust mode, is designed to work down to around -9 dB and apparently can work down to -11 dB on computer simulations with some products. Also, it is supposed to be able to work down to at least zero dB under extreme 50 Hz Doppler spread and 10 ms Multipath and that may not be possible for most of the amateur modes. I would like to try the slower and more robust STANAG modes when they become available for amateur use. Don't you think that the the most robust modes, often running at the slower speeds will be used on HF amateur frequencies, if they can compete with existing modes? 73, Rick, KV9U dmitry_d2d wrote: Hello Rick. Let's turn our attention to the astonishing robustness of Pactor-2/3. Surprisingly low signal to noise ratio can be reached due to using low-speed rate of the low band signal ONLY. How does it work? First of all low signal to noise ratio declared by the producer are not measured in the small band, used by Pactor signal, but in the whole band of the standard telephone channel (0,3-3,4 kHz). So the noise is assigned evenly to the whole band. Then the process of treatment is as follows: as the useful signal is concentrated in a certain definitely appointed frequency area the preliminary digital filtration is used. As a result the noise that does not get to the useful signals would be filtered consequently the signal to noise ratio will be much better in fact. So the question arises it still works where is dirty trick? In fact it
Re: [digitalradio] Re: RFSM 8000
Thank you for that information, Dimitry, When I read over the MIL-STD/FED-STD/STANAG standards, I could not understand how these slower modes would work all that well by the multi-repeating approach. They repeat multiple times, even if you do not need the repeat (sort of a non automatic ARQ) which seemed very wasteful in terms of throughput. Especially because this was happening at the slower speeds which made them even slower than they might otherwise need to be. When I saw the charts that compare the throughputs with the S/N ratios, there was some improvement in terms of weak signal capability over the non repeating speeds, but none of the modes seemed to be able to compete with Pactor2 speeds at the low S/N levels. When I have asked those who work daily with STANAG type modems how successful they are, they tell me that they often have difficulty with certain paths. Also, because they use inefficient document types (MS Word .doc files) just to send the shortest possible message can take a very long time. They even have some technologies that have duplex voice and that does not always work either. The reason that they use these technologies is because they have to use them due to the need for interoperability. In fact, this is now mandated more and more with regulations. I often wonder if they may eventually rethink some of the modulation schemes or have a switch over to something that can tolerate severe ISI/doppler/ under weak signal conditions that are common with amateur signals? The STANAG modem standards do allow for OFDM modems as well, but I don't think that anyone switches over to OFDM for part of the time when they are normally using a single tone modem. But perhaps the standard allows it? After seeing which modes seem to work the best under the most difficult conditions, (well below zero dB S/N) which are the conditions we often have with amateur communications, it does seem like a few (two under worst conditions) spaced a moderate distance apart, may have the best compromise on throughput vs. robustness. That is why Pactor 2 (and Pactor 3 when operating in the Level 1 speed) work so well. At this time, it does not appear that there is any other mode that can compete with that modulation technique based on all the comparisons that I have been able to find. Again, thanks for your help on this. 73, Rick, KV9U dmitry_d2d wrote: Hello Rick. As regards the speed that is slower 600 in MIL-STD 188-110A/B. There are 300, 150, 75. In my opinion speed reduction has been made nonoptimal, using dumb repetition of data in 300 and 150 is not needed. The theory of coding says that repetition is the worst way to improve noise immunity. Speed 75 based on the method of spectrum spread by orthogonal consecution by Walsh. It's rater good but this speed uses repetition as well. We consider that the speed 300, 150, 75 allows reaching better characteristics of noise immunity that the standard MIL-STD 188- 110A/B allows. Frankly speaking the standard MIL-STD 188-110A/B has been used our product to be noticed by customers. But true to say it contains nonoptimal solutions. Turning to the point of RFSM we should admit that we have mistaken making the minimal speed - 600. I hope we improve it in the near future. Dmitry.
Re: [digitalradio] Re: RFSM 8000
dmitry_d2d wrote: 1. A few words about OFDM and serial tone modem. Let's find out how the fight between ISI and Doppler shift takes place in these systems. OFDM uses the great number of low speed channels so the symbol duration increases. While the duration of ISI is much smaller than symbol duration everything goes well. Consequently there is an aim to increase the number of channels ad infinitum BUT at the same time natural limitation takes place. It's just a Doppler shift effect. Hence there is always a compromise between ISI and Doppler shift. Moreover we should take into consideration a big peak factor which results in non-effective usage of power of transceiver. There are methods directed at improvement of peak-factor, but the most part of them makes the system characteristics worse. In case of serial tone modulation the fight ISI with Doppler is provided with adaptive algorithms. The more effective and faster they are the larger number of Doppler and ISI the modem can manage. As for RFSM it should be mentioned that now it includes rather efficient adaptive algorithms that work properly at a speed of 600(500) up to 4800(4000) bps (wide/narrow mode). To work at a speed 6400(5333) - 8000() much more compound algorithms are needed. In particular using turbo-equalization will improve noise proof feature at all rates. Therefore OFDM and serial tone modem can be more efficient in dependence on channel statement. In my opinion serial tone modem with effective adaptive algorithms is the most effective. We'd like to mention that under certain circumstances either serial tone or OFDM modem can fail to provide connection, for example, when the Doppler shift is extremely high (polar communications). In that case one should use the methods of spectrum spread that extending the symbol in time and frequency. Unfortunately the speed would not be high in this case. So the best way out is to measure the channel characteristics and choose the speed of transmission and modulation method according to them. The full adaptation of the all characteristics is required. 2. About our users. The project RFSM-2400/8000 was initially aimed at organizations (not for HAMs)! (First version had no 0,3-2,7 band, which is adapted for HAMs). Its prime value is that high-performance algorithm is used in it. Consequently only technical specialists of organizations where data (files, mail etc.) transmission through HF is needed can estimate the program at its true worth. They need the following: high speed of connection and data transmission. They are the FIRS GROUP OF OUR USERS. For example there are organizations (our users at the moment) who even haven't looked upon HAM -modems (little speed, instability, absence of files transmission in spite of excellent chat-exchange). If you are interested in RFSM as in a program for chat- exchange (or even for file transmitting but you do not need a high speed) and runner is not important for you:. You are the SECOND GROUP OF OUR USERS. $60 may be a pretty penny for this product for you. There is also not numerous GROUP OF USERS - THE THIRD ONE The representatives of this group are specialists in HF- radiocommunications and radioamateurs at the same time who is interested in algorithms of a high efficiency - the runner of the program. May be $60 is rather expensive for them but they can trial versions for free. They communicate with us suggesting interesting and moreover useful ideas. We really appreciate their advices and suggestions. Due to the THIRD GROUP the first version of RFSM has transformed in the product adopted for HAM. 3 . There are several remarks on the open source codes. a) RFSM-2400 (and all the more RFSM-8000) is not just a dumb modem though such a rate is also possible (it was used in PSKMail). Our product is an accomplished system of communication that provides different types of services including receiving/transmitting e-mail on Internet. b) Speaking about OFDM it should be pointed out that we have got experience in such a kind of modulation and can remark that to construct this modem is incommensurably easier than Serial Tone Modem. But the modem of this kind doesn't compare with RFSM characteristics. If we were not be able to realize Mil-STD correctly and use OFDM in RFSM, we would not be sorry to distribute source codes. c) Philosophy. Professional free software is possible because qualified developer has been grown up by certain company. The buyers have already paid for software and progressive developer as well. Then at the same time free software appears (like RFSM-2400) - like an ad, to create an image or ease consumers' tasks. The fact that software is free is a result of successful sales of developer. However free software is not possible in fact. The bigger the quantity of it the poorer it's quality. So said Write on C++ for food ;) There is also rather INTERESTING free
Re: [digitalradio] Re: RFSM 8000
Thanks for your comments Dimitry as they do help with understanding your approach. One thing that I wondered about when using the program for our test was the relatively slow speed at the ~ +5 dB S/N. The 600 showed on the screen which now if I understand your comments below, the slowest bps rate for your version of the STANAG modem is 600 bps. For around a +5 dB S/N is that about right for the 600 bps rate? If the program was running at the 600 bps rate, and had a throughput of around 300 wpm, is that about what we should expect? Did you try using the other STANAG modems down to the more robust 75 bps too or did you find that impractically slow? 73, Rick, KV9U dmitry_d2d wrote: 1. A few words about OFDM and serial tone modem. Let's find out how the fight between ISI and Doppler shift takes place in these systems. OFDM uses the great number of low speed channels so the symbol duration increases. While the duration of ISI is much smaller than symbol duration everything goes well. Consequently there is an aim to increase the number of channels ad infinitum BUT at the same time natural limitation takes place. It's just a Doppler shift effect. Hence there is always a compromise between ISI and Doppler shift. Moreover we should take into consideration a big peak factor which results in non-effective usage of power of transceiver. There are methods directed at improvement of peak-factor, but the most part of them makes the system characteristics worse. In case of serial tone modulation the fight ISI with Doppler is provided with adaptive algorithms. The more effective and faster they are the larger number of Doppler and ISI the modem can manage. As for RFSM it should be mentioned that now it includes rather efficient adaptive algorithms that work properly at a speed of 600(500) up to 4800(4000) bps (wide/narrow mode). To work at a speed 6400(5333) - 8000() much more compound algorithms are needed. In particular using turbo-equalization will improve noise proof feature at all rates. Therefore OFDM and serial tone modem can be more efficient in dependence on channel statement. In my opinion serial tone modem with effective adaptive algorithms is the most effective. We'd like to mention that under certain circumstances either serial tone or OFDM modem can fail to provide connection, for example, when the Doppler shift is extremely high (polar communications). In that case one should use the methods of spectrum spread that extending the symbol in time and frequency. Unfortunately the speed would not be high in this case. So the best way out is to measure the channel characteristics and choose the speed of transmission and modulation method according to them. The full adaptation of the all characteristics is required. 2. About our users. The project RFSM-2400/8000 was initially aimed at organizations (not for HAMs)! (First version had no 0,3-2,7 band, which is adapted for HAMs). Its prime value is that high-performance algorithm is used in it. Consequently only technical specialists of organizations where data (files, mail etc.) transmission through HF is needed can estimate the program at its true worth. They need the following: high speed of connection and data transmission. They are the FIRS GROUP OF OUR USERS. For example there are organizations (our users at the moment) who even haven't looked upon HAM -modems (little speed, instability, absence of files transmission in spite of excellent chat-exchange). If you are interested in RFSM as in a program for chat- exchange (or even for file transmitting but you do not need a high speed) and runner is not important for you:. You are the SECOND GROUP OF OUR USERS. $60 may be a pretty penny for this product for you. There is also not numerous GROUP OF USERS - THE THIRD ONE The representatives of this group are specialists in HF- radiocommunications and radioamateurs at the same time who is interested in algorithms of a high efficiency - the runner of the program. May be $60 is rather expensive for them but they can trial versions for free. They communicate with us suggesting interesting and moreover useful ideas. We really appreciate their advices and suggestions. Due to the THIRD GROUP the first version of RFSM has transformed in the product adopted for HAM. 3 . There are several remarks on the open source codes. a) RFSM-2400 (and all the more RFSM-8000) is not just a dumb modem though such a rate is also possible (it was used in PSKMail). Our product is an accomplished system of communication thatprovides different types of services including receiving/transmitting e-mail on Internet. b) Speaking about
Re: [digitalradio] Re: RFSM 8000
Dmitry, Can you tell us why you chose a non-standard 2000 Baud symbol rate? I could understand a much lower baud rate. How much of an effect do you think a half rate 1200 baud speed have on the ability to use the mode under more difficult conditions? Did you try other speeds, including even relatively slow speeds, such as 300 Baud and, if so, can you tell us if that made much difference? I am still very skeptical that these high baud rate modems will be able to compete with multi tone modems when ISI and doppler get severe, while other modes will be able to continue to work. What is your view on comparing the multitone vs single tone modems? 73, Rick, KV9U dmitry_d2d wrote: Hi, All. About bit-rates and Baud-rates (in RFSM-8000). RFSM-8000 uses vary bit-rates: 1) 600, 1200, 2400, 3200, 4800, 6400, 8000 - in wide (standard) mode; 2) 500, 1000, 2000, 2666, 4000, 5333, - in narrow (non-standard) mode. But, Baud-rate is not changing: 1) for standard mode - 2400 Baud (symbols per seconds) - for any bitrates; 2) for non-standard mode - 2000 Baud (symbols per second) - for any bitrates also. Dmitry. Announce your digital presence via our Interactive Sked Page at http://www.obriensweb.com/sked DRCC contest info : http://www.obriensweb.com/drcc.htm Yahoo! Groups Links