The testing that was done on DRM showed that the worst multipath spread is on NVIS paths and is about 8 ms. They originally could accomodate only about 5 ms of multipath but had to create a new mode with a longer guard interval to support South American stations using the 41, 49, 60, 75, 90 and 120 meter broadcast bands.
73, John KD6OZH ----- Original Message ----- From: Rick To: digitalradio@yahoogroups.com Sent: Wednesday, October 10, 2007 14:05 UTC Subject: Re: [digitalradio] Dev: Multipath SCS decided on 100 and 200 baud rates when they developed Pactor 2 around 1995. They also used constellations as large as 16PSK. At the time, Dr. Rink explained why they selected certain design parameters and it was his view that with strong FEC, the high baud rates would be workable and 100 baud would be a good compromise lower speed. After many years of experience they must have proven to themselves that 100 baud was the sweet spot. There was no shifting back and forth between baud rates which can really slow down a link. G-TOR had 100/200/300 baud rates and from what I have heard had a lot of problems with taking too much time trying to hunt to find the right baud rate for conditions at that moment. They also must have found that the complex constellations greater than 4PSK were not very productive compared with a much wider bandwidth. The tones are spaced slightly wider than normal orthogonal spacing and, if my understanding is correct, normal spacing would have been one Hz for each baud. Perhaps this reduces the tones from overlapping if there is increasing ISI? The MIL-STD/FED-STD/STANAG modems use the 2400 baud 8PSK waveform which is many, many times higher in baud rate than anything we normall use on HF here in the ham bands. They supposedly use very strong DSP to cope with ISI, but when you look at the actual specifications of these modes, they consider 2 ms and particularly 5 ms of ISI as worst case. This might be possible close to the MUF, but on the lower bands in the evening, this seems unrealistic. We have only heard a few comments from hams outside the U.S. who have used these modes and while the throughput can be quite good under good conditions, as you would expect, they don't do well when conditions are more difficult. I am betting that even if 8PSK2400 modes were permitted here in the U.S., these modes would not work as well as some think they will. People who design, install, and maintain military ALE equipment have told me that the modes are not that great in many conditions and can be frustrating to use. Especially when you consider the astronomical cost of this equipment. The one problem that SCS noted with too slow a baud rate is that short term ISI can be a problem, such as fast doppler. It was his position that we will not normally expect anything worse than a 5 ms ISI and a few Hz dispersion. Of course we do run into much worse ISI at times and I suspect that Pactor modes will not work at all under those conditions. Then a mode like Domino EX might perform better. But from what I have been able to gather, PSK modes are the best route to go at this time. One of these days, some enterprising ham is going to build a multi-tone design modem similar to Pactor 3 and we will find some very impressive throughputs. This fresh start would be easier to program than to make it backward compatible to previous modes such as had to be done with Pactor 2 and 3. Perhaps it could follow the FAE asynchronous ARQ design? It could be manually adjustable at first so that you can switch between 2PSK and 4PSK, but forget higher constellations. And it could be manually adjustable so that it could drop tones to get fewer tones with wider spacing as needed for more robustness. Eventually, further progression of the design could lead to automatically changing numbers of tones and/or constellations, but at least it would be usable for manual control. In other words, Pactor 3 has proven what works and tells you where to focus for higher throughput modes. 73, Rick, KV9U Rud Merriam wrote: > > Okay, the MFSK decoding via FFT gives me some things to work on the > next couple days. > > We know from Shannon-Hartley the limit for throughput based on > background Gaussian noise. Another major problem with HF > communications is multipath. This is where a signal takes multiple > paths through the atmosphere. Some of the paths may delay the signal > by many milliseconds, even 10s of milliseconds. This cause > self-interference where a symbol is distorted because of the late > arrival of a previous signal. Also known as Inter-symbol Interference > or ISI. > > It seems the best way to handle this is by a slow symbol or baud rate. > I find it interesting that Pactor III maintains a constant 100 baud > while adjusting the waveform in other ways to adjust for adverse > conditions. At a 100 baud the timing is 10 ms per symbol. Does anyone > know if they arrived at this because 100 is a round number? Or is > their research behind this choice? > > >