(Continued from previous message)


RS> The proper way is to use proper test equipment, and then compare
RS> those results with what the Courier/Courier session claims. Ditto

PE> The question I was asking was how to design a modem.

And I was saying thats the wrong question in this situation.
What should really be done is to measure the line with some
real test equipment and see how the results compare with what
the Courier/Courier stats FOR THAT LINE say about that line.

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

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

PE> Ok, so how does it perform the line probe?

It essentially tells the other modem to put a known test on
the line in terms of the tones and levels, and sees what comes
out the other end. Thats how the test equipment does it too.

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

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

PE> And which ones can be measured?

All of them. Some are less important than others tho when
trying to get the best thruput thru a particular line.

PE> Or can you only measure the sum, or what?

No, you can measure most of them individually. Most are listed individually
on the stats, there are more presented than the graph of level and freq.

PE> And is your measurement method dependant in any way on the other modem?

Well, clearly if you ask the other modem to put a standard set of
tests on the line, so you can see what comes out the other end of the
line, if it doesnt do what its supposed to do, that possibility arises.

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

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

PE> It could be related.

Well, its certainly possible that the Courier gets rather confused about
what it thinks of the line with a non Courier on the other end of the line
and so adjusts its params to not very satisfactory values, and thats what
causes the handshaking that happens LATER to fail. But it clearly thinks
the line is a lot WORSE than it really is in that situation, so it should
be deliberately using stuff like a lower line rate that can be achieved,
so you would expect that to be LESS likely to fail the handshaking when
it uses a lower rate on on what it thinks is a pretty bad line.

Tho that would certainly explain why sessions which do handshake fine
achieve a lower thruput than say an M34F/M34F pair or an Courier/Courier
pair would get on that particular line. The Courier/M34F gets confused
about the state of the line and gets a lower thruput because it thinks
the line is worse than it actually is. Tho you would really need to do
some M34F/M34F pair tests from Bill to you to be sure on that.

PE> It might be that the failures only happen
PE> on noisy lines, never on the good lines.

Again, its NOT just 'noisy lines' V34 is about a LOT more than
noise and the truly magnificent thruput Russ gets to the bulldog
on lines which have very audible noise on them shows that.

PE> BTW, my line to the exchange is presumably constant, ditto for Bill's,

Well, the do vary somewhat with weather etc, particularly with
high moisture levels, particularly if you have some not very good
joints. Bills line isnt that terrific, and may well vary a bit.

PE> and the STD link should be perfect,

Yes, once its digital you can forget about variation.

PE> which means that for us to have noisy lines,

It AINT about 'noisy lines' Paul. The world has moved on with V34.

PE> would mean that my wire to the exchange would need
PE> to have some crosstalk from someone else's wire.

Thats one possible effect, but since the begley gets perfect calls
every time with an M34F/M34F pair, its clearly not happening unless
there is some quirk of it only happening at the times Bill calls and
not when the begley calls.

I bet that if Bill had an M34F he would get no handshake failures either,
tho since his line isnt terrific, he may not get 28800/28800 every time.

The problem appears to just be Sportster/M34F interoperability which
manifests itself at a high rate with Bills rather grotty line.

Which explains why he ALONE saw handshake
problems with Sportster V32bis to your Spirit.

PE> And presumably at 3am there is less
PE> chance of interference from others?

You really cant say, it depends on whats causing the crosstalk.
If its some dork whose device is pumping out much to much signal,
what counts is when he uses it. If its a fax, it could be any time.

Its most unlikely its a crosstalk problem tho. And even if you do
have crosstalk, V34 is supposed to be able to handle that without
handshake failures anyway. It should only degrade the thruput.
@EOT:

---