When I first started poking around, I found a simple comparison paper at http://ieeexplore.ieee.org/iel2/672/6329/00247152.pdf . It appears to require an IEEE membership, or a university internet connection(I'm VPNed into my uni right now). The bottom of each page does say "U. S. Government work not protected by U. S. copyright." I've hated not being able to access academic papers in the past, so I'm half tempted to just observe the copyright notice and post it here.
The problem with that comparison, and in fact many real world tests is the single tone modem(STM) ends up using adaptive equalization(the STM wouldn't work at such a high rate without it), while the parallel tone modem(PTM) doesn't. Adaptive equalization is the "secret sauce" of high speed STMs. It's what gives them their multipath handling ability. They basically model the transmission channel, find the inverse transformation, and de- spread the multipath laden signal to form a clean one. In such tests, the STM seems to come out on top for anything but very local (<50km) ground paths. PTMs inherently have multipath tolerance due to their MUCH lower baud rates. But there's nothing that says you can't strap an adaptive equalizer onto a PTM. You'd get natural multipath tolerance, along with whatever multipath correction needed to be performed. Most current PTM modems ignore multipath all together, and late signal is lost at best and causes intersymbol interference at worst. Adding an adaptive equalizer would allow PTMs to recover this late signal energy along with further reducing ISI. So we'd have the secret sauce adaptive equalizer, along with natural multipath tolerance, instead of just one or the other. It appears that the ham OFDM modes DO include adaptive equalizers(at least, WINDRM does). I'm not that familiar with DRM, so I'm not sure how much of the transmitted signal is dedicated to training data(if any at all). Most STMs dedicate a large portion of the transmitted signal to keeping the receiver's equalizer up to date. Other modes that weren't designed for adaptive equalization simply use the FEC of the transmitted signal to guess at the actual transmitted signal, which is then used to train the equalizer. Here in the US, the ATSC digital television standard uses extremely simple modulation(8VSB)...if you sent an analog TV transmitter an 8-level voltage signal with a little filtering, you'd have an 8VSB transmitter. The first generation receivers were horrible with multipath handling, but as time went on, adaptive equalizer chipsets got better and better and were "strapped on" to the standard. I have a little HD receiver for my laptop with a fairly recent chipset/equalizer. I can receive a perfect copy of the signal in a metal-walled narrow dorm room(structurally similar to a prison cell, although far more pleasant). Yet if someone walks around the room, the signal will drop completely. For all I can tell, the signal got stronger as the person moved, but the equalizer has to sit there for 2-3 seconds trying to retrain itself. And sure enough, 2-3 seconds later, I'm back to perfect copy again. So I'm not sure what we could achieve if we willy nilly started to strap adaptive equalizers to PTMs. The PTMs may have to be modified to transmit a training signal, so the receivers have something definite(and quick!) to model with, rather than a slow adaptive equalizer that simply guesses at the received signal. On the other hand, PTMs tend to naturally be far more multipath tolerant, so it's likely we'd have to dedicate less of the transmitted signal energy to the training signal. What's ideal needs some simulation and real over the air testing. Heck, we should let a computer search for the right combination of baud, tones, and power dedicated to training data over a wide variety of conditions. Pick a small set of winners for varying band conditions, and some synchronous(but slow enough for PC s) ARQ, and bam, you got yourself a data mode :). Mike KF6EYU --- In digitalradio@yahoogroups.com, Rick <[EMAIL PROTECTED]> wrote: > The question becomes: if you had two modems, one using single tone high > baud rate vs. one using multi tone OFDM, which one would perform the > best in varying conditions. > > Various documents on the internet suggest that there is not much > difference, but there is at least one that does show a difference with > computer simulations in favor of the multi tone modems. I tend to > discount computer simulations as not adequate and prefer the real world > under many different conditions that gives you a more accurate practical > feel for what can and can not be done. That same document, done as a PhD > paper, admitted that some waveforms that worked well on computer > simulation, actually did not work at all in an actual real world test.
