home *** CD-ROM | disk | FTP | other *** search
- Path: zeno.ibd.nrc.ca!roberson
- From: roberson@zeno.ibd.nrc.ca (Walter Roberson)
- Newsgroups: comp.compression,comp.dcom.modems
- Subject: Re: Use a full-duplex modem as a asymmetric one?
- Date: 3 Mar 1996 19:41:47 GMT
- Organization: National Research Council of Canada
- Message-ID: <4hcslr$hon@canopus.cc.umanitoba.ca>
- References: <4h7290$3ql@bcarh8ab.bnr.ca>
- NNTP-Posting-Host: zeno.ibd.nrc.ca
-
- In article <4h7290$3ql@bcarh8ab.bnr.ca>, Zaihua Luan <hualuan@bnr.ca> wrote:
- :I wonder if it is possible to use a full-duplex modem as a asymmetric
- :modem. For example, suppose I have a 28.8kps full-duplex modem. Can I
- :set the modem to a channel of 43.2kbps and a channel of 14.4kbps. If
- :it is possible, please let me know how to do it.
-
- comp.dcom.modems would be a better newsgroup than comp.compression for
- modem questions. I've added it to the Newsgroups list.
-
-
- The answer to the question is "No." Modern V.32/ V.42 modems
- essentially work by using a fixed carrier frequency and modifying the
- signal phase at several well-defined points along each cycle to
- indicate a 1 or 0 bit. The maximum transmission rate is determined by
- the number of phase changes per wave that can be robustly detected,
- times the carrier frequency.
-
- The math works out for using a 2400 Hz carrier frequency and detecting
- at 12 points per cycle (every 30 degrees) for 28800. 14400 is 2400 Hz
- and detect 6 points per cycle (every 60 degress).
-
-
- In order to go for 43200 bps in one channel, one would have to either
- alter the carrier frequency, or else sample at 18 points per cycle
- (every 20 degrees). Those are not yet feasible -- all the more so when
- you consider that if you go beyond 32000 bps then by the pigeon-hole
- principle at least one datapoint is going to have to be resolved in a
- single bit when transmitted via a digital line {as those sample at 8000
- Hz, 8 bits per sample.} Below 32000 bps you have 2 bits per datapoint,
- which makes accurate detection much easier.
-
-
- Recall that 28800 modems transmit information by -audio- techniques,
- not digital techniques and not by bit-streams. What you are asking
- would require packing more -audio- information per second into one
- direction. Sampling 18 points per cycle is not necessarily impossible,
- but it is not yet practical.
-
-
- Since you posted to comp.compression, I will link to a topic in
- compression theory. People have suggested from time to time
- that perhaps higher data rates might be achieved by having the
- remote end send back 'guesses' about what was about to be sent, and
- the transmitting end ack or nak these 'guesses'. Transmission time
- and latency issues make this scheme problematic at the best of times,
- but it just doesn't work anyhow. The transmitting end has at least as
- much information as the remote end does, and so could predict what
- the remote end would predict and issue the ack or nak without ever
- receiving anything from the remote end. Such a scheme is therefor
- useless unless the transmission time and latency are *very* low,
- and the remote end is able to model the data better than the local
- end can (perhaps the local end has very little memory or a very slow
- processor, or is required to hand off the data sooner than it would be
- able to complete the better compression.)
-
- Walter Roberson roberson@Ibd.nrc.ca
-