NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/dcom/modems-part1/249

Internet Message Format | 1996-08-05 | 5.3 KB

Path: polarnet.com!floyd
From: [e-mail] (Floyd Davidson)
Newsgroups: comp.dcom.modems
Subject: Re: high speed modems ?????
Date: 3 Jan 1996 07:13:55 GMT
Organization: __________
Message-ID: <4cdabj$[e-mail]>
References: <[e-mail]> <4c4kgu$[e-mail]> <4c674e$[e-mail]>
Reply-To: [e-mail]
NNTP-Posting-Host: tanana.polarnet.com

In article <4c674e$[e-mail]>,
Carl Renneberg  <[e-mail]> wrote:
>[e-mail] wrote:
>> >  Estimates for the speed limit of the public switched telephone network vary
>>>  based on assumptions about bandwidth and signal to noise ratio... and of
...
>>The false claims are on both sides. The theoretical limits based on several
>>analog line characteristics have not changed and stay the same since
>>they had been proved 50 years ago.

I agree that that is basically a true statement.  Our ability to
engineer and manufacture a modem that can use the PSTN to maximum
effect has certainly changed though.  V.32 specs were around for
relatively a long time before it was practical to manufacture such
a modem (never mind getting the price down where anyone would buy
one!).

>Those old "theoretical" limits are not so much due to the line's
>characteristics as to the equipment along the call path.

I think most people realize it isn't just the cable loop, but the
entire circuit including switching and transmission systems...

>In the days of FDM and analog switching, a 28-30 dB SNR was a
> reasonable figure ...

I wouldn't agree that those figures are any more or less reasonable
for practical circuits over TDM than over FDM.

>With PCM, the major source of noise is the A-D conversion at the
>local exchange (or local digital pair-gain system).  The
>signal-to-quantization noise introduced by the A-D conversion
>itself is about 38 dB (Reference:  Introduction to Digital
>Communications Switching, John Ronayne, ISBN 0-273-02178-8).  The
>relevent ITU-T spec., which specifies the conversion rules (A-law
>and u-law), is G.711.

That is correct, but also realize that 38 dB is the absolute
_best_ S/N available at some given level and frequency.  I've
never measured a channel that had that good a S/N (maybe 35 dB or
so is about the best with TDM).

>The PCM equipment, like FDM equipment, bandpass-filters your
>signal, to prevent aliasing.  The relevent ITU-T specs on the
>bandwidth characteristics are G.712 and Q.552.  The 3 dB points
>are 300 Hz and 3400 Hz (I believe, from memory).  These are
>minimum figures; some PCM equipment will give you a bit more
>bandwidth.

That is an excellent take on bandwidth.  The low end is commonly
considerably better than 300Hz, but 3400Hz is about the end of it
on the upper end.

>To operate at full speed, V.34 modems require an upper 3 dB
>frequency of 3200 Hz, and an SNR of around 32-34 dB SNR (source:
>"Issues and Responses - V.  34 and V.FAST Class, Rockwell
>Telecommunications,
>http://www.nb.rockwell.com/ref/v34-vfc-comparison.html).

The bandwidth is commonly that good, the S/N rarely gets higher
than that.

>The old FDM equipment wouldn't allow a full-speed V.34 connection,
>but properly set up PCM equipment will.  The limits *have*
>changed.

I can't agree with that.  For example, Lenkurt 46A3 FDM carrier
was probably the most commonly used analog FDM carrier before
digital carrier became popular.  It happens that I know of one
Lenkurt 78A microwave system using 46A3 carrier today!  And even
now, at 20+ years of age, it isn't uncommon to measure S/N at
better than 40 dB on it!  It is much less likely to have a channel
bandwidth of greater than 300-3400 Hz, but commonly 300-3200 Hz is
found, and 400-3000 Hz normal.  The one difference is that it also
isn't really rare to find a channel that is only barely able to
meet 400Hz to 2800 Hz specs too, and it will not be replaced
either.

The variations in FDM channels was common, whereas digital
carrer channels tend to be uniformly good.

>>.. Nowadays, with DSP processors you can make a few
>>improvements with some algorithms and crank analog kilobytes from your
>>garage research lab.

And that is why the "theoretical" limits are higher today than
15 years ago.  The channel is about the same, but a DSP can use
it and that just wasn't possible in 1980.

>According to Rockwell (ibid.), the most common reason for not getting a
>full-speed V.34 connection is that the overall signal path just doesn't have
>enough bandwith.

I find that surprizing.  Commonly the bandwidth seems to be there,
but each time the number of A-D conversions doubles the S/N is reduced
by 3dB, and if robbed bit signaling is used and enough DS1 interface
points are in the circuit (ie. it gets reframed) to cause all 8th
bits to zapped, the S/N is reduced another 6 dB.

That adds up pretty fast...

>(I suspect that you also need 64 kbits/s in the digital path; 56
>kbits/s may be just too low.)

That accounts for the 6 dB mentioned above.


In essence I think the average channel today is probably just
slightly better than the average channel was 20 years ago.  The
best is considerably less, and the worse is much less common.
The channel that Shannon was concerned with is almost the same
now as it was in the 1940's!  What has really improved is the
modem itself.

Floyd
--
Floyd L. Davidson          Salcha, Alaska         [e-mail]