Pending FCC ruling threat to modem users

[Alan Parker]

In article <403 at pixar.UUCP> aaa at pixar.UUCP (Tony Apodaca) writes:
>In article <3454 at curly.ucla-cs.UCLA.EDU> stiber at zeus (Michael D Stiber) writes:
>>In article <1572 at brl-adm.ARPA> OCONNORDM at ge-crd.arpa (OCONNOR DENNIS MICHAEL) writes:
>>>First: MODEM calls DO NOT cost the phone company the same amount as
>>>other calls. They tend to be longer, and don't tolerate noise as well.
>>              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>
>>1) You've obviously never had a sister (or daughter).  Do families with
>>teenage girls pay more for phone service?  Unlimited local calling
>>is just that _unlimited_.
>>
>>2) Modems use the same lines as voice.  The assertion that they tolerate
>>noise less well is irrelevant, since they do not get special
>>priveledges.
>
>I'm sure that this discussion is going hot and heavy in net.dcom.etc but
>I'll answer here anyway.  Flame off, Mike.  The point is true even if the
>rationale is messed up.  Modem calls DO cost the phone company more, for
>several reasons:
>	1) They are continuous.  The dual carrier never stops.  Therefore,
>the phone company must supply bandwidth to the call continuously even if
>there is no "valid" data.  They cannot time-multiplex their signals.
>There are small breathing and thinking pauses in all voice conversation,
>and 99% of voice is half-duplex, even a teenage girl's conversations.

This is wrong.   In almost all cases the phone company supplies
continuous bandwidth to calls.   Very few, if any, trunks care about
silence periods.   You get a channel.  That channel has a bandpass of
300 to 4000 Hz.   They have sharp filters to be sure that you stay in
this channel.    On analog carrier trunks (frequency-domain mux) you get
a piece of the carrier bandwidth (about 4Khz).  There is switching going
on during silence periods.   On digital trunks, the analog channel is
coded into a bit stream (usually 64Kbps for PCM, or 32K bps for APCM)
and muxed into a higer speed data stream.   Again, you get a channel of
about 4Khz.   They have sharp filters to make sure the input signal is
within this bandpass (you have to filter it anyway before going into the
ADC).

(There are schemes to exploits silence periods, but these are not widely
used for regular telephone service.)

>	2) They are high bandwidth.  The phone line was designed with human
>voices in mind, and they are pretty low bandwidth, as everyone knows.  Also,
>everyone knows that their modems strive to get the most out of it, so they
>use it all up (if they didn't you'd buy a new one!).  However, the phone
>company "counts on" the signals being voice-like, so they can cram as many
>signals into one wire as possible, and a modem transmission screws up their
>frequency-division multiplexing.

Wrong.   Doesn't matter how much bandwidth you ask for.  You get a
channel about 4Khz wide.   If you got what you wanted, it would be a lot
easier to design a 9600 bps modem (gee, I think I'll just demand a 10Khz
channel.)


>	3) The carrier on some modems just happens to overlap a critical
>region of part of the phone company's equipment's frequency allocation.
>The circuits known as "echo suppression" use it to monitor themselves, and
>kick in higher bandwidth and better circuits if there is too much echo on the
>line.  If they didn't have it, you'd bitterly complain about the quality of
>your long-distance connections.  Voice doesn't have much of these freqs,
>but the carrier flips it out, causing it to allocate too much signal to
>your call.
>

You don't understand why or how echo cancellers work.   Lines that run any
significant distance or 4 wire, that is there is a different path for
the signal going each way (as opposed to the 2 wire scheme used for
local loops and shorter trunks).   If you run a large distance (say,
greater than 1000 miles) if there is any mixing of the transmit and
receive signal (which will happen when they are converted back to two
wire at the other subscriber end) some of your voice comes back.  The
time delay of the return path causes you to hear it as echo.   You
always get some echo regardless of the length of the line, its just that
you don't notice it unless the time delay significant.   In traditional
(i.e. older) long distance systems they stuck analog echo suppressors
along the way.  They had to use this often, since they weren't great;
they suppressed echo, but didn't remove it all.   These days they have
digital echo suppressors that really clean things up well, so you can
have them just at each end.    These are *designed* to disable
themselves on a data call, since the echo doesn't bother the modems and
since both ends send continously the echo cancellers would tend to
suppress the signals.

This doesn't screw up the system.  Thats the way it was designed. 
By design modems send a tone to disable the echo cancellers.   The echo
canceller certainly doesn't not "switch in" a higher bandwidth or higher
quality line.   Bandwith has nothing to do with echo (echo the echo
would sound better on a wider bandwidth line!).