Meng-Shi, at 20:57 on May 22 1996, you wrote to All...

ML> I haven't used a 28.8K (or > 14,4k) modem prior to this, so is it always
ML> that difficult to set up for a 28.8K connection? 

Apart from locking your port rate to 57600bps or higher, not at all.

ML> I don't remember it being such a problem when I started using 14.4k. My 
ML> problem is difficulty with a consistent high speed connection. I have 
ML> tried gradually going up the scale and the most consistent connection I 
ML> can get is 21600 LAP-M V42bis.

There's something wrong with your modem's implementation of V.34 if it's
incapable of analysing line conditions during the negotiation phase, then
connecting at the highest link rate possible for the current conditions. 
V.34 is an asymmetrical protocol (you might receive data at say 26400, yet
send at only 24000, for example), and should also automatically and
transparently speed-shift (in either direction) as line conditions vary
during the call.

ML> I think the problem is probably due to the phone line but twice already
ML> Telstra has said the line checked out fine. 

That means absolutely nothing, I'm afraid.  It could still be a line
problem, although V.34 is supposed to compensate under those circumstances.

ML> I once got a beautiful 26400 without error correction or compression (is 
ML> this possible ?)

Yes, very.  Some Rockwell chipsets have poor implementations of both V.42
and  the 3429 symbol rate, and are both fixable in firmware.  For example,
I quite often see 28800/NONE connects when calling one particular Rockwell
with my Courier, but disabling the V.42 detect phase (and forcing LAPM)
solves this problem.  If calling Hayes Optimas, I have to disable the 3429
symbol rate in order to connect at 28800 as well.  Not all V.34 modems are
created equal.

ML> that lasted for hours but most times I get a 26400 LAP-M v42bis that would 
ML> drop off after minutes.

Does the speed simply drop off, or do you actually lose carrier?  If the
latter, one of the modems has likely had its retraining ability disabled.

ML> 1) What is the major difference between v32bis and v34 at 14.4k? Is v34
ML> more reliable?

Yes.  Its instantaneous speed-shifts and protocol asymmetry see to that.

ML> 2) Should I enable v32/v32bis/v34 automatic retrain?

Absolutely, or the modems will disconnect if one or the other can't respond
to a retrain request.  Some Rockwells actually default to retrains
disabled!  :(

ML> 3) Should I enable MNP5?

Up to you.  I have MNP5 disabled, and use V.42bis only. 

ML> 4) Should I enable auto-fallback?

Definitely.

ML> -BTW how do I know if auto falback is in action? Does the modem make
ML> a handshake noise?

Only if they retrain.  A fallback (or speed-shift) should occur instantly.

ML> 5) What is the difference between reliable and auto-reliable v42?

Not sure in your modem's context, but auto-reliable will normally stay
connected if no EC is negotiated.  Reliable requires EC to stay on line.

ML> 6) Should I enable Trellis encoding? (What does this do anyway)

It's a form of signal encoding, and if you disable it, your modem will
likely not connect at high-speed (if at all).

ML> 7) I use this mainly for Internet so should I turn this on: Conditional
ML> data compression. A LAP-M connection negotiates v42bis. An MNP
ML> connection does not negotiate MNP5 data compression.

I would (and have).  Almost nobody uses MNP4 these days anyway.

ML> 8) When a high speed modem is handshaking and finds that the line is not
ML> up to scratch, in what sequence does it drop back? eg does it go
ML> 28.8k+EC+Data compression -> 28.8k without EC/compression or straight to
ML> 26400+EC+compression? 

The link rate will drop, but ALWAYS with EC and compression (if enabled). 

ML> I'm curious because I have EC+compression set all the time but when I check 
ML> the statistics I sometimes get a connection without EC or compression. I 
ML> thought that the modem would have preferentially chosen a slower speed.

It should always connect with EC.  If it doesn't, you have a problem.

Regards, Bill

--- Msgedsq/2 3.20