PE> Why don't they just ask the other modem to put out something
PE> constant then, to eliminate any error introduced by that factor?

Thats essentially what happens.

BL> Because they are testing the comms link, not just the line.

Thats bullshit. Thats how you test the comms link, tell the other modem
to put out a known set of stuff on the line, see what comes out at the
other end, that tests the comms line. There aint no other way to do it.

That long table of readings like this

Freq     Level
 150     34
 300     24
 450     23
 600     23
 750     23
 900     23
1050     23
1200     24
1350     24
1500     24
1650     25
1800     25
1950     25
2100     26
2250     26
2400     26
2550     27
2700     28
2850     28
3000     28
3150     29
3300     30
3450     31
3600     36
3750     44

Is produced by telling the modem at the other end to put out a FIXED level
at each of those frequencys, and then measuring the level of that frequency
at the other end, with the other modem. The loss is that column labelled
Level. You are basically plotting how the loss varys with frequency.

In practice its a bit more complicated than that, but thats the basic idea.

PE> I had no idea it was an inherent part of all protocols,
PE> even ones designed 10 years ago, to honour a "transmit
PE> multiple frequencies at fixed level" command.

It isnt, only V34 and VFC.

BL> I don't think it does that either. I assume it sends
BL> certain code to be echoed, and that the code itself
BL> contains those frequencies. I assume that this is what
BL> happens upfront, with the 4 seconds of squawks and stuff.

Thats crap, he is confusing the V8 tones that just specify
the basic stuff on the capabilitys of the modem, like whether
it can do stuff like V34 at all, with the line probing, what
produces that graph of levels, amoungst other things.

PE> One thing that doesn't sound reasonable, is your previous bit about
PE> them testing the comms link.  How is the other modem meant to know
PE> how loud to yell unless they test the line at a fixed level FIST,

Yes, thats the flaw in his argument.

PE> and THEN try to test different levels of loudness?

You dont necessarily need to actually test different transmit levels.
In practice the transmit level is ONLY ONE of quite a few parameters
that can be adjusted to achieve maximum thruput. Those include the
carrier freq, the transmit level, the shaping, the symbol rate and
some other more subtle stuff, and you can largely CALCULATE those
ONCE you have MEASURED how the line varys with frequency and some
of the other line parameters like the noise level.

PE> I think your theory isn't complete.

Its hopeless, as usual with Bob waffling from ignorance.

PE> As to a THEORY about how I would attempt to measure line quality if I
PE> were designing a modem, I will first ask what the test is that Telecom
PE> do when they come out to your place after you have reported a bad line.

These tests for voice lines are MUCH cruder. Nothing like
the comprehensive test of a line that a V34 modem does.

BL> They call the exchange which sends test signals.

Thats garbling the story very comprehensively indeed. Each telephone
exchange has a bit of fancy test equipment called a test desk which
you can sit at and measure particular lines. But thats mostly quite
different tests to what a V34 modem does, stuff like what they call
'foreign battery' which is essentially a voltage on the line which
isnt put there by the phone exchange, whether both wires in the pair
of wires which form your phone line back to the exchange are balanced
with respect to ground and other stuff like the line resistance etc.

At YOUR place, its also possible to dial 199 and then hang the phone up.
This essentially tells a bit of test equipment called a SALT to test your
line. It then rings you back and communicates the result to you in the
tones it sends to you. You can also with the dial on the phone tell it
to do some other tests like ring you with half the usual ring voltage etc.

It is possible to put other test equipment on the line, but they dont
do that sort of testing with a range of tones on a normal voice line.

If you have a data grade line, they can do more fancy tests, but they
wont be doing that on a line you are paying for as a voice grade line.
And that essentially is just a manual test using test equipment like
the V34 modems do. The test equipment aint cheap, tho you can rent it.

PE> Ok, that sounds reasonable.  I bet the exchange sends
PE> a fixed test signal too, not varying before you've even
PE> started.  In an effort to test the LINES, NOT the exchange.

Voice lines arent ever measure extensively like that. There isnt any
need to because the normal faults you get dont produce just a distortion
of the frequency response curve that affect voice communications.

In fact there are specific conditions on your line which rather
grossly distort the frequency response like a system call bridge
taps, which have no effect on the use of the line for voice, but
stuff up modem comms unless the modem allows for it. V34 can do so.
@EOT:

---