PW>> So the scrambler turns "0000000011111111110000000" into
PW>> "0101010101010101010" (or something like that) ???
DH>> You got it.  That's the desired outcome.  And the practical outcome,
DH>> unless your gambling luck is GASTLY bad.

PW> Hmmm...  And that 0101010101010101010 is then turned into a mish-mash of
PW> different frequencies (as detailed below by you) ???

Yup.  That's the idea.

PW>> Oh?  I thought the "spectrum" was just one frequency
-- the carrier?
DH>> The act of modulating splatters the carrier into a smear of
DH>> frequencies.  To make matters more interesting, in the presence of the
DH>> more modern modulation techniques, it's almost a case of "will the
DH>> real carrier please stand up".

PW> So, while the specs say xxxx Hz is the carrier, after modulation, there is 
PW> no real carrier as such?

There is - mathematically.  And the "errors" from that pure
sinewave carrier are the carriers of the actual information.   Some of the
earlier modems even actually worked that way.

DH>> In theory, ideal modulation will place signal energy evenly from one
DH>> end of the available bandwidth to the other. 
("Evenly" is a rather

PW> What is the "available bandwidth" on your average
average-quality line?  
PW> I'm
PW> not talking about the sort of string-quality lines of the
"outback" here...

A cruddy one is usually useable from 300hz to 3000.  A good one will
stretch that to 250 to 3600.

DH>> esoteric bit of math in and of itself.)  This maximizes the use of the
DH>> channel capacity.

PW> So that polynomial is just one of many equations necessary for modulation?

Yup.

PW>> what does QAM stand for AND what does it mean?  Ditto for PCM and TAM and
PW>> line probing and all that sort of thing!

DH>> QAM "Quadrature Amplitude Modulation".  Sneaky pete in
operation -

PW> Who? <%-\

DH>> mathematically consider several different signals, encode each as AM
DH>> at a particular phase angle.  (Uh.. yeah.  What he said..:-)   Turns

PW> Wait a second...  Several different signals?  So you're splitting up the
PW> "carrier" into parts and encoding those in Amplitude
Modulation with 
PW> various
PW> "phaser angles"???  What is a phase angle -- frequency of the wave, 
PW> gradient of the wave at amplitude 0, etc.???

No - try it like this.   Two identical sine waves, phase shifted.  One the
carrier, the other the quadrature component.  Now AM both of them.  Then
matrix multiply the pair to get a single value, which is the real
instantaneous voltage sent to line.   Believe it or not, the process is
reversible at the other end - and is even moderately resistant to noise in
the middle.

BTW - don't depend on this process actually being physically done.  It's
more likely done at the math level, in a digital signal processor nowdays.

DH>> out the math is not only doable, but it can be physically implemented
DH>> as well.  Conceptually - well, it talks a good line of math...:-)   It
DH>> also makes pretty pictures on the right kind of test set...:-)

PW> So you just plop your probes down on the line somewhere and watch them on
PW> the test-set?  I guess you then use Fourier decomposition or something to
PW> break down the signal into simple sine waves or something?

No - what it actually does is to encode the data into a differential on a
circle, which represents the original carrier plus its quadrature
component.
Modulation yanks a point away from the quiescent circle.

The live output looks like a bunch of bright star points, hence the name of
"constellation".

DH>> PCM - "Pulse Code Modulation".  Modern digital data
transmission for
DH>> phone lines.  Sample the signal, digitally encode, send the digital
DH>> info.

PW> Is that like for CD-DA -- sample at so many Hz at this many bits, etc.???

Exactly.  A CD could be, and actually sometimes is, sent over a studio
quality phone line, using this same technology.

DH>> Line probing - the attempt to quantify the particular line impairments

PW> I.e. delay, noise, etc.?  What sorts of "impairments" are
we talking about
PW> here?

Shudder.   Noise.  Delay variances.  Gain slope.  Echo.  Return ratio
error.  Return ratio changes.  "Hits" or transients, on all of
those.  Stolen pulses used for signalling in PCM systems.  Cheating PCM
systems that change from 64k to 32k while you're using 'em.  And on and
on...

DH>> of this exact connection, prior to making efforts to at least roughly
DH>> compensate for some of 'em.  Send known signals each way, then listen
DH>> to the other end and compare what you actually get vis what you know

PW> So noise in a simple sine wave would result in a queer wave, the modem 
PW> would
PW> then work out what the "waveform" of the noise is and then
compensate???

That's the idea.  It's more like the chirp transform of a pulse for the
more interesting bits - but what it's measuring is amplitude loss variation
with frequency, and delay variation with frequency, and seeking to
linearize both.

DH>> was sent.  Basic drill for setting up a high speed modem link - the
DH>> horrendous and weird noises during modem initial handshake nowdays are
DH>> doing just that.

PW> So the "boing boing" is a "set" signal for line probing???

Yep.  So is the back-and-forth blat-blat that happens during a retrain.

PW> BTW sorry to keep bothering you, but what does TCM mean?

Trellis Code Modulation.  That's a description of one of the options in the
QUAM we were discussing above.  The idea is that you use the possible
modulation points sparingly, so a single noise pulse must push the signal
through _two_ wrong states instead of one to successfully register an
error.

Direct application of s/n vis data rate signal theory to increase the noise
resistance of a link at the expense of the data rate.

Regards,
Dave Hatch.

--- Msgedsq 3.20