home *** CD-ROM | disk | FTP | other *** search
- Path: news.gate.net!not-for-mail
- From: dhaire@gate.net (doug haire)
- Newsgroups: comp.dcom.modems
- Subject: Re: Supra will offer upgrade to 33.6!
- Date: 26 Feb 1996 15:36:00 -0500
- Organization: CyberGate, Inc.
- Message-ID: <4gt5jg$tte@hopi.gate.net>
- References: <312172b0.302209@uchinews.uchicago.edu> <4ftaot$rpc@shellx.best.com> <31226a7d.63749347@uchinews.uchicago.edu> <4g182k$1mu2@navajo.gate.net> <31243778.561299@uchinews.uchicago.edu> <4g3t66$26um@hopi.gate.net> <dan.878.312C8AD9@supra.com> <dan.
- NNTP-Posting-Host: hopi.gate.net
- X-Newsreader: TIN [UNIX 1.3 950824BETA PL0]
-
- 890.3131E1C6@supra.com>:
- Distribution:
-
- Dan Moore (dan@supra.com) wrote:
- : In article <4gl4d7$1nss@seminole.gate.net> dhaire@gate.net (doug haire) writes:
- : >Dan Moore (dan@supra.com) wrote:
- : >: The default setting is %E0 (do not initiate retrains based on line
- : >: conditions) and %G1 (do initiate rate renegotiations based on line
- : >: conditions). This is exactly what you want the modem to do, quickly adjust
- : >: the data rate to match the current line conditions. Retrains are slow (eight
- : >: or more seconds), rate renegotiations are relatively fast (hundreds of
- : >: milliseconds with a max of around two seconds). So the default setup
- : >: provides rate renegotiations instead of retrains.
- : >However, is there not a time when a retrain is important to do? And why
- : >would you restrict your own modem from doing this, leaving it dependent
- : >upon the remote modem to be one that does NOT restrict this?
- : >If there is no time that this would be useful, why is it an option at all?
- :
- : In V.34 a retrain causes a full line reprobe and can choose a new
- : symbol rate, carrier frequency, pre-emphases filter, etc. based on the current
- : line bandwidth, SNR, near and far echo levels and echo delays. This is the
- : same analysis that is done when a connection is first made, and it can take as
- : long (or longer) than the initial training sequence. A retrain is the only
- : way to change the symbol rate of a connection.
- :
- : In V.34 a rate renegotiation is simply a change in the bits per symbol
- : without changing the symbol rate. This is a very quick change since no
- : probing, etc. is done. When a modem initiates a rate renegotiation there is
- : no requirement that the remote modem accept the rate change, during the
- : parameter exchange it can refuse to change the rate. Typically modems always
- : accept rate renegotiations downward and may refuse rate renegotiations upwards
- : if the current EQM is too high. If a rate renegotiation fails (ie. the remote
- : modem doesn't respond at all) a retrain will occur instead.
-
- Not if both modems are defaulted to *not* initiating a retrain.
-
- : The protocols that support rate renegotiation (V.32bis, VFC and V.34)
- : all require a modem to follow a remote rate renegotiation request. So no
- : matter how the local modem is configured to do rate changes if the remote
- : modem initiates a rate renegotiation the local modem must respond to it. If
- : the remote modem uses a retrain to change data rates the local modem must
- : respond to that also.
-
- Yes, but if *both* modems are Supras, no retrain will be initiated so
- neither one will have to respond since the request will never be given.
- Instead, a few rate changes will be tried (allegedly) and carrier will be
- lost.
-
- : During a call it is fairly common for the SNR to change. As it
- : changes the modems will initiate rate renegotiations up or down as needed. It
- : is fairly rare for the bandwidth or echo level to change during a call, so
- : there typically is no need to initiate a retrain. (NOTE: Without reprobing
- : the line there is no way to tell if the bandwidh, echo levels, echo delay,
- : etc. have changed. So the only way to tell for sure that a retrain is needed
- : is by reprobing the line, which only happens during a retrain...)
- :
- : During a connection the modem monitors the EQM, if the EQM stays below
- : a "good" threshold for a period of time (set by S190) the modem will attempt
- : to increase the data rate via a retrain or rate renegotiation (based on %E and
- : %G settings). If the EQM exceeds a "bad" threshold the modem will attempt to
- : lower the data rate via a retrain or rate renegotiation (again based on %E and
- : %G). If the EQM exceeds a "retrain" threshold the modem will initiate a
- : retrain no matter how %E and %G are set. The "good", "bad" and "retrain"
- : thresholds vary with the connection protocol (V.32/V.32bis/VFC/V.34) and with
- : the data rate.
- :
- : As far as I know all commercial V.34 modems default to using rate
- : renegotiations to change data rates instead of using retrains.
- :
- : The option to do a retrain instead of a rate renegotiation is there
- : for historic reasons. V.32 does not support rate renegotiations to change
- : data rate, only retrains. This lead to the %E command to enable/disable
- : rate changes via retrains. V.32bis added rate renegotations as a way to
- : change the data rate, so the %G command was added to enable/disable rate
- : changes via rate renegotiations. (NOTE: In a V.32 connection rate changes
- : are done with retrains when %G1 is set to enable rate renegotions based on
- : line quality.)
- :
- : The %E and %G commands are really of limited use since the best
- : preformance is achieved by simply taking the factory default settings (&F
- : sets %E0, %G1). The commands are still present since many programs have
- : really old init strings that contain them. These programs issue an AT&F2 and
- : then proceed to set all the settings that the &F2 had set. The additional
- : commands achieve nothing except making the init string longer.
-
- While I agree that many programs have really old init strings, that is an
- incredibly stupid reason to leave otions in. In addition, there is
- nothing consistent about extended commands for high speed modems. Even
- the Rockwell chipset based modems have some differences in these. I
- really don't believe that Supra used this rational or that even Rockwell
- (who really is the owner of the command set) used that rationale when
- designing the command set to be used.
-
- I appreciate your attempt to explain but I think you have consistently
- missed the point.
-
-
-
-