VICTOR@YKTVMZ.BITNET ("Victor S. Miller") (11/28/85)
I recently reposted some of the discussion about 3400 versus 212A on a local bulletin board, and got the following information which may be of interest: ----- MODEM FORUM appended at 21:33:07 on 85/11/27 GMT (by LWEST at AUSVM1) ---- Subject: Filtering of Modem Noise Reference: Question from Henry Schaffer of NCSU via VICTOR Henry, you are correct that no analog technique can filter all noise from a signal, if the noise is non-deterministic and within the frequency band of the signal. Neither can any digital technique. The comments about "feedback filter" versus "harmonic technique" relate to methods of demodulation of a signal. For DPSK or QAM signals, as well as for the other common modulation techniques, any of several demodulation techniques can be used. I suspect (but do not know) that the RV3400 used a "coherent detector", which is a form of phase locked loop, for demodulation. I do know that most (maybe all) of the DPSK demodulators prior to the IBM entries did not. They use an inferior technique known as "differential detection". It is simpler. This is one reason that RV3400 works better than most 212 types. However, the new IBM modem uses (I think) both coherent detection and a relatively new process called "adaptive equalization" which should make it better than the older RV3400. Another reason RV works better than most 212's is the nature of QAM versus DPSK. There are many forms of both, but, in general, DPSK techniques spread the signal over a broader bandwidth than does QAM. In simplistic analysis, this should be better. However, on phone lines, the bandwidth is extremely limited, and two bands are utilized simultaneously - one in each direction. The upper portion of the upper frequency band (sent by the answerer to the caller) is often severly altered by the phone line. Distortion results, and more errors occur. The QAM techniques in general "crowd" more energy into a smaller band, which makes the transmitted symbol for one state look a lot like the symbol for another, so one might expect more errors. However, the bandwidth is smaller, and the receiver filter can be designed to reject all noise outside this smaller band. The upper portion of the upper band does not extend as far up into the "bad" frequency portion of most phone lines, so the signal is less distorted (has less envelope delay and frequency distortion) than DPSK. Incidentally, we also ran across the frequency tolerance problems mentioned above: the problem can certainly originate in the transmissions from a bad modulator. Lynn West