Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!mips!pacbell.com!tandem!netcom!gandrews
From: gandrews@netcom.COM (Greg Andrews)
Newsgroups: comp.dcom.modems
Subject: Re: PEP vs. v.32
Summary: The line isn't the same across the whole bandwidth
Message-ID: <26916@netcom.COM>
Date: 5 Mar 91 07:37:16 GMT
References: <6927.27D24513@zswamp.fidonet.org>
Followup-To: comp.dcom.modems
Organization: Netcom - Online Communication Services  UNIX System {408 241-9760 guest}
Lines: 57

In article <6927.27D24513@zswamp.fidonet.org> root@zswamp.fidonet.org (Geoffrey Welsh) writes:
>In a letter to All, Greg Andrews (gandrews@netcom.COM ) wrote:
>
> >PEP isn't limited to a 16 point constellation.  In the 
> >cleanest parts
> >of the line you'll see carriers using a 64 point (6 bit) 
> >constellation.
>
> >Unless the modem decides the line won't support that kind of 
> >signal resolution...
>
>   My disappointment has doubled.  At first I believed that Telebit & co. were 
>simply not pushing the technology.  Now I that I have reason to believe that 
>they are, I can't avoid the conclusion that they are doing so poorly!
>
>   If there are 511 carriers operating at a maximum of 8 baud with 4 bit QAM, 
>the maximum possible raw throughput is in the vicinity of 16000 bps, a figure 
>which I have seen Telebit modems approach on clean lines.  If, as you say, 
>they're using a 64 point constellation, the maximum raw throughput should 
>be around 24000 bps, and I've never seen a TB come even close to that.
> 

Geoffrey, I'm sorry you're disappointed, but I think your disappointment
stems from a misunderstanding.  You say "If there are 511 carriers...
with 4 bit QAM...", but there *aren't*.

First of all, the modem typically uses only about 400 carriers.  Those
at the top and the bottom extremes of the frequency range aren't often
used because the line isn't good enough.

Second, the signal quality of the phone line looks something like a
bell curve.  The best reponse is found in the middle range and the
quality falls off as you go higher and lower.

This results in the best transmission through the middle of the
frequency range, and poorer transmission as you go farther into
the fringes.  PEP follows the line quality, using the most complex
modulation in the middle ranges where quality is best (the 6-bit,
64 pint QAM), dropping to a simpler modulation where quality is
a little worse (the 4-bit 16 point QAM), dropping further down
to the simplest modulation where the line is bad (the 2-bit 4 point
QPSK), and finally abandoning the carrier frequency where the line
is really awful.

The modem doesn't automatically use the same modulation across the
whole frequency range.  It uses the fastest modulation on each
carrier that it can.  On clean lines I typically see a mass of
6-bit carriers in the middle of the range, with 4-bit ones on the
outside, and a few 2-bit ones toward the fringes.  (this is shown
by querying the S71 and S73 registers)  The upper and lower extremes 
aren't used.  Just as you'd expect, an adaptive modulation uses the 
fastest method where the line is cleanest, in the midrange.

-- 
.-------------------------------------------.
| Greg Andrews      |   gandrews@netcom.COM |
`-------------------------------------------'