BG> In fact, we don't KNOW that at all, and it's likely that if the
BG> USR-measured line levels vary with different remote modems, it's
BG> including the other modem's impedance, or loading on the line.

RS> Thats rather implausible, a MUCH more likely explanation is that
RS> the USR stats are stuffed unless there is a Courier on BOTH ends.

BG> And you have the cheek to accuse ME of jumping to
BG> conclusions in the absence of absolute proof!  So
BG> far, BOTH explanations are only guesses, nothing more.

Thats crap, its nothing like a guess. Its in fact putting your brain
into gear and thinking about how modems work, particularly hybrids.

PE> Yeah, in the absence of some scientific evidence
PE> from Rod about why modem's impedence is constant,
PE> the guesses seem equally as valid, yours more so.

Welp they aint, tho in your case you can be excused because you
dont know that much about the finer detail of electronic hardware.

The short story is that the 'modem impedance' cant be anything, and is
in fact under the control of the modem with a modem echo cancelling modem.
Its not that easy to summarise the detail in a para or two, but there is
a thing called a hybrid which combines the receive and transmit onto the
single pair of wires that are the phone line. How well the modem can cancel
out its OWN outgoing data so it can here the data from the other end at all,
and remember they are on the SAME frequency with modern modems, depends on
how close the modems impedance is to the lines impedance for complicated
reasons. And thats why you see the near end and far end echo figures in
the stats, they are a measure of that impedance imbalance.

So Bills waffle about modem impedance is just that, waffle.

The 'loading on the line' is even worse, we are actually talking
about the design of the hybrid, and its capacity to handle varying line
impedances, and its got sweet fuck all to do with 'loading on the line'.

In fact the variation in the loss between the modem and the
exchange line card is MUCH more important, and the whole point
of a modern protocol like V34 is to be able to adjust to all that
stuff as it probes the line and sets the considerable number of
variables it has that can be optimised for a particular line.

PE> I find it hard to believe that the USR Courier can conform to
PE> the V.34 spec AND simultaneously decide that the other end is
PE> a non-Courier so it should start faking away at the numbers.

Thats just playing silly buggers. No one is suggesting that it does
that on purpose. Its clear from the MASSIVE difference of 6db, which
is a HELL of a fucking lot, that it claims to see with and M34F on
the other end, its having a massive brain fart. Just why is currently
harder to say. One obvious possibility is just a bug in the USR code
in that particular situation. It would be interesting to see what
happens with a non rockwell non USR on the other end, like a Motorola.

PE> Although maybe the variable transmit rates or something have a factor.

It shouldnt, that line probe data, which is what that graph is supposed
to be, is supposed to be BEFORE the modems work out whats best to use
for the stuff they can vary. In fact the DATA they use to do that.

PE> I guess if we're guessing, we should ask the question, if YOU
PE> were designing a modem, how would YOU measure the line quality.

The proper way is to use proper test equipment, and then compare
those results with what the Courier/Courier session claims. Ditto
for the other configs. Its easy enough to MEASURE what the M34F
is like impedance wise etc with the right equipment.

PE> Except you've already agreed that you can't measure the
PE> line quality without a 600 ohm resistor on the other end.

Well, its a bit more complicated than that when the modems can vary the
impedance they present to the line to balance the hybrid. Part of what
modern modems do is recognise that real life lines are NEVER actually 600
ohm at the wall socket. And in fact they are a complex impedance anyway,
not that simple figure. So even if you do say use a 600 ohm resistor, it
doesnt prove a damned thing if the line is nothing like that, and the modem
adjusts itself to the actual line impedance. And they do, they HAVE to do
that with echo cancellation protocols otherwise they cant hear the other
end at all over the echo of their own outgoing carrier.

PE> So if YOU were manufacturing a modem, and you couldn't figure
PE> out a way of getting the other end to emulate a 600-ohm resistor

Thats precisely what the line probe is supposed to be doing. It really
does attempt to do a full test of the line characteristics, then it
adjusts the various stuff it can adjust to do the best with what its
just measured the line to be.

PE> (after a connect even), then what would YOU measure INSTEAD of line quality

That question is meaningless. There are a variety of parameters like
noise, impedance, the variation in loss with frequency, and other more
complex stuff like group delay which are in total the line quality.

There is nothing BUT the line quality that matters
when attempting to achieve maximum thruput.

And thats all a completely separate issue to the gross handshaking
failures that Bill is seeing on Courier/M34F sessions on SOME calls.
@EOT:

---