baud rate limit
Daniel Sadoc
In a typical phone line, it's possible to transmit on frequencies
aprox. between 300 Hz and 3300 Hz. Please, I would like to know what
is the relation between this bandwidth limitation and the maximum
signaling rate of 2400 baud.
Thanks in advance,
Daniel Sadoc
--
The Telecom Digest is currently mostly robomoderated. Please mail
messages to edi...@telecom-digest.org.
Alan Fowler
>Hi!
>
>In a typical phone line, it's possible to transmit on frequencies
>aprox. between 300 Hz and 3300 Hz. Please, I would like to know what
>is the relation between this bandwidth limitation and the maximum
>signaling rate of 2400 baud.
>
>Thanks in advance,
>Daniel Sadoc
Daniel,
Do you really mean 2400 baud or 2400 bit/s?
For instance a V.34 modem can transmit/receive at
bit rates from 2400 to 33,600 over a standard telephone
speech channel.
At 33,600 bit/s a V.34 modem is
transmitting/receiving at 3429 Baud.
The explanation is too long to type here. I suggest
you look at "Baud" in the 'Definitions' section of
http://www.emucities.com.au/member/whitethorn/hsmodem.html
regards, Alan
Daniel Sadoc
Thanks. The link is great.
But I would also like to know why the bandwidth limitation, in
general, restricts the maximum signaling rate. In the particular case
of a typical phone line, I would like to know why we frequently do not
reach more than 2400 bauds. Is there any mathematical relation between
the 3000 Hz bandwidth and the aproximately 2400 bauds limit? What?
Thanks in advance,
Daniel Sadoc
amfo...@melbpc.org.au (Alan Fowler) wrote in message news:<3b8cc65f...@news.melbpc.org.au>...
Scott Dorsey
>
>But I would also like to know why the bandwidth limitation, in
>general, restricts the maximum signaling rate. In the particular case
>of a typical phone line, I would like to know why we frequently do not
>reach more than 2400 bauds. Is there any mathematical relation between
>the 3000 Hz bandwidth and the aproximately 2400 bauds limit? What?
The theoretical bandwidth limitation of a channel is called Shannon's Limit
and it was worked out by Claude Shannon in the 1930s. It is a function
of both bandwidth and the noise floor of the channel.
The _actual_ bandwidth from a given modulation scheme is never very close
to the Shannon Limit for the channel, but some modulation schemes are
much better than others. This is why a Bell 103 modem using FSK modulation
can only get 300 to 600 bps on a dialup line, while you can get 9600 bps
easily with modern trellis encoding on the same line.
There are a number of good introductory books on information theory, including
one written by Shannon himself that used to be in the Bell Labs publications.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Reed
definition is at:
http://www.its.bldrdoc.gov/fs-1037/dir-033/_4816.htm
--reed
Alan Fowler
>Hi!
>
>Thanks. The link is great.
Thank you.
>
>But I would also like to know why the bandwidth limitation, in
>general, restricts the maximum signaling rate. In the particular case
>of a typical phone line, I would like to know why we frequently do not
>reach more than 2400 bauds.
Do you really mean 2400 bauds or 2400 bit/s (bits
per second) ? Because they are not the same thing, except
at 3400 bit/s.
If I have read the V.34 recommendation correctly,
the following bit rates are all available at 2400 Baud,
2400, 4800, 7200, 9600, 12000, 14400, 16800 and 19200 bit/s
What speed are actually getting, in bits per second?
regards, Alan.
Maurizio Codogno
: At 33,600 bit/s a V.34 modem is transmitting/receiving at 3429 Baud.
I am at a loss. If bandwidth in a telephne line is between 300 and 3400 Hz
(at least here in Italy) this means that I cannot possibly have more than
3400 symbols per second, that is 3400 *transitions* per second, i.e. 2400
baud.
Where is the fault in my reasoning?
ciao, .mau.
--
Tieni a mente il mio futuro indirizzo email! Get my new email address!
puntomaupunto(a)tin.it
Tra poco / Coming soon: http://xmau.com/
Wm. Randolph Franklin
In article <9mfptl$6...@beatles.cselt.it> on 28 Aug 2001 11:20:50 -0400,
m...@beatles.cselt.it (Maurizio Codogno) writes:
> I am at a loss. If bandwidth in a telephne line is between 300
> and 3400 Hz (at least here in Italy) this means that I cannot
> possibly have more than 3400 symbols per second, that is 3400
> *transitions* per second, i.e. 2400 baud.
1 Hz involves 2 transitions per second, but that's insignificant.
More important, the bandwidth means that you can transmit signals
whose frequencies are between 300 and 3400 Hz. You can
simultaneously transmit many different signals, each at a
different frequency, and each signal can be transmitted at any one
of many different possible volumes. By detecting what frequencies
were transmitted at what volumes, the receiver decodes the
transmitted info.
This requires that the receiver be able to tell the various
transmitted signals apart. The higher S/N of the line, the closer
together two signals can be (in frequency and volume) and still be
distinguishable.
This gets more interesting because the signals are changing with
time. The shorter a signal is, the more uncertain its actual
frequency and volume are. This is superficially like Heisenberg
uncertainty.
All this gets summarized in the formula given in the previous
poster's citation:
http://www.its.bldrdoc.gov/fs-1037/dir-033/_4816.htm
C = W log2(1 + S /N ),
where C is the channel capacity in bits per second, W is the
bandwidth in hertz, and S /N is the signal-to-noise ratio.
----
(Wm. Randolph Franklin) <rfra...@altavista.net>
aes
m...@beatles.cselt.it (Maurizio Codogno) wrote:
> Alan Fowler scripsit:
>
> : At 33,600 bit/s a V.34 modem is transmitting/receiving at 3429 Baud.
>
> I am at a loss. If bandwidth in a telephne line is between 300 and 3400 Hz
> (at least here in Italy) this means that I cannot possibly have more than
> 3400 symbols per second, that is 3400 *transitions* per second, i.e. 2400
> baud.
One very elementary answer is: send pulses, at some pulse width and
pulse rate supported by that bandwidth (and recall that bandwidth of a
pulse does not depend on how strong it is). What's the weakest pulse
amplitude you can see above noise on your line? Call that amplitude
level E1. What's the strongest amplitude pulse that won't overload the
channel? Call that E2. Then you can send any one of roughly E2/E1
symbols in *each* pulse: a pulse level of E1 means symbol #1; a pulse
level of 2 times E1 means symbol 2; and so on.
Suppose for example you can distinguish 1024 pulse amplitude levels..
Each symbol can then be understood to be a different 10 bit number
(since 2^10 = 1024); so you can send 10 bits -- not just one bit -- per
pulse.
How do you calibrate the system, i.e. decide what level is E1, or 10 E1,
or 100 E1, especially if overall link gain changes? Answer: Every
once in a while send a string of amplitude calibration pulses at say,
100 E1?
The above isn't how it's actually done, but it illustrates the general
spirit of how it's done. Obviously low noise (small E1), large dynamic
range (large E2), and smart coding are the critical elements.
Daniel Sadoc
Thanks for the help.
So, please correct me if I'm wrong: "If bandwidth in a telephne line
is between 300 Hz and 3400 Hz, the chanel is nosyless, and I use just
one chanel of volume, I cannot possibly have more than 3400 symbols
per second, that is 3400 signalizations per second, i.e. 3400 baud -
that is the maximum imaginable signaling rate".
Why 1 Hz involves 2 transitions?
I haven't understood the phrase: "The shorter a signal is, the more
uncertain its actual frequency and volume are."
Is there a good reference about this subject, for beginers?
Thanks in advance,
Daniel Sadoc
(Wm. Randolph Franklin) <rfra...@altavista.net> wrote in message news:<9mg6i2$45f$1...@cp472970-a.alngtn1.va.home.com>...
> In article <9mfptl$6...@beatles.cselt.it> on 28 Aug 2001 11:20:50 -0400,
> m...@beatles.cselt.it (Maurizio Codogno) writes:
>
> > I am at a loss. If bandwidth in a telephne line is between 300
> > and 3400 Hz (at least here in Italy) this means that I cannot
> > possibly have more than 3400 symbols per second, that is 3400
> > *transitions* per second, i.e. 2400 baud.
>
> 1 Hz involves 2 transitions per second, but that's insignificant.
(...)
Alan Fowler
>sa...@rio.com.br (Daniel Sadoc) wrote:
>
>>Hi!
>>
>>Thanks. The link is great.
> Thank you.
>>
>>But I would also like to know why the bandwidth limitation, in
>>general, restricts the maximum signaling rate. In the particular case
>>of a typical phone line, I would like to know why we frequently do not
>>reach more than 2400 bauds.
>
> Do you really mean 2400 bauds or 2400 bit/s (bits
>per second) ? Because they are not the same thing, except
>at 3400 bit/s.
That should read "except at 2400 bit/s. My
apologies for typing mistake.
Alan Fowler
>Alan Fowler scripsit:
>
>: At 33,600 bit/s a V.34 modem is transmitting/receiving at 3429 Baud.
>
>I am at a loss. If bandwidth in a telephne line is between 300 and 3400 Hz
>(at least here in Italy) this means that I cannot possibly have more than
>3400 symbols per second, that is 3400 *transitions* per second, i.e. 2400
>baud.
>
>Where is the fault in my reasoning?
Part of the problem is the way the bandwidth of a
telephone line is specified. If I remember rightly the
response of the line is 3 dB down at 300 and 3400 Hz, but
still has a useful response outside that range. V.34 modems
(33,600 bit/s) need a bandwidth of 150 to 3750 Hz and can be
20 dB (?) down at these frequencies.
At 33,600 bit/s they use a carrier frequency of 1959
Hz and a modulation rate of 3429 symbol/s which should fit
into that 3600 Hz bandwidth (3750 - 150)
V.34 uses QAM (Quaternary Amplitude Modulation .)
which allows you to pack a lot more bits into a symbol. As
I understand it, each combination of say 10 bits is
represented by a particular value of phase and amplitude of
about a bit more than half a cycle of the 1959 carrier.
I don't understand it fully myself, I'm still trying
to understand the V.34 document.
Perhaps someone else in this newsgroup can help both
of us with a better explanation.
regards, Alan
Peter_...@ne.3com.com
(although my company affiliation appears in my email address,
this is not an official company statement, purely my own)
For a real lesson in how to cram the most through a
telephone pipe, check out DSL!
I'm currently working on an ADSL product. It uses
"Discrete Multi Tone" modulation. Basically,
it's a bunch of carriers (255 of them, going out to
1 MHz).
When a connection is initiated, these carriers are
used to measure the frequency response of the line.
Initial carrier amplitude and number of bits transmitted
on each carrier are assigned based on this measurement.
Here's the neat part. Using Quadrature Amplitude Modulation
on each carrier, they send a varying number of bits per
symbol. How many? Well, for each carrier, the modulator
measures the error rate and reduces (or increases) the
number of bits per carrier to maintain a given error rate.
So the whole modulation scheme is dynamic, with carrier
bit loading varing as the channel varies; guaranteed to
get the most information possible through the channel.
Regards,
Peter Simpson
Wm. Randolph Franklin
In article <d016349.01082...@posting.google.com> on 28 Aug 2001 21:53:39 -0400,
sa...@rio.com.br (Daniel Sadoc) writes:
> So, please correct me if I'm wrong: "If bandwidth in a
> telephne line is between 300 Hz and 3400 Hz, the chanel is
> nosyless, and I use just one chanel of volume, I cannot
> possibly have more than 3400 symbols per second, that is 3400
> signalizations per second, i.e. 3400 baud - that is the
> maximum imaginable signaling rate".
Every channel has noise. Otherwise you could send a lot more than
3100 bps by sending or not sending a signal at each of the
following frequencies simultaneously:
300
300.00001
300.00002
300.00003
...
3399.99998
3399.99999
3400.
The problem is that the noise prevents the receiver from
separating, say, the possible 300 Hz signal from the possible
300.00001 Hz signal.
It's a major error to equate 1 Hz with one symbol per second. The
stated bandwith just means that signals at frequencies between 300
and 3400 Hz will be transmitted. You can however transmit many
simultaneous signals at different frequencies, each between 300
and 3400 Hz.
> Why 1 Hz involves 2 transitions?
Up and down. This is getting irrelevant and confusing to this
particular discussion. However, this is why you have to sample a
signal at twice the highest frequency in order to reconstruct it.
The relevant keyword is probably "Nyquist theorem".
> I haven't understood the phrase: "The shorter a signal is, the
> more uncertain its actual frequency and volume are."
Taken to the ridiculous limit, if you look at a 1000 Hz signal for
1 microsecond, it's going to be hard to know if it's not really
1001 Hz.
Another way to look at this is to think of statistics. If you
want to find the mean of a population, the larger the sample you
use, the more accurate your estimate is. Simularly the longer a
signal you look at, the more accurately you can determine its
properties, like frequency and volume.
Getting a little far afield, here's another instance of this sort
of idea. If you look at a small object, you can't tell its color
as accurately as you can when the object is larger.
> Is there a good reference about this subject, for beginers?
I'd have to think awhile. I have trouble thinking of some of this
myself, and I teach computer engineering (tho not communications
engineering).
Maurizio Codogno
: >: At 33,600 bit/s a V.34 modem is transmitting/receiving at 3429 Baud.
: >
: >I am at a loss. If bandwidth in a telephne line is between 300 and 3400 Hz
: >(at least here in Italy) this means that I cannot possibly have more than
: >3400 symbols per second, that is 3400 *transitions* per second, i.e. 2400
: >baud.
:
: Part of the problem is the way the bandwidth of a
: telephone line is specified. If I remember rightly the
: response of the line is 3 dB down at 300 and 3400 Hz, but
: still has a useful response outside that range. V.34 modems
: (33,600 bit/s) need a bandwidth of 150 to 3750 Hz and can be
: 20 dB (?) down at these frequencies.
Thank you very much! I thought that filtering was very sharp above
3400Hz, so it would have been impossible to use that frequencies.
Of course, I knew that since the days of V.32 more symbols were packed
in a single transition to enhance bitrate - that was no problem :-)
Thanks again, .mau.
Denis Mcmahon
> The explanation is too long to type here. I suggest
>you look at "Baud" in the 'Definitions' section of
>http://www.emucities.com.au/member/whitethorn/hsmodem.html
Dead site!!
Rgds
Denis
--
Denis McMahon
Mobile: +44 7802 468949
Email: de...@pickaxe.demon.co.uk
I always trim ng when posting!
Alan Fowler
>amfo...@melbpc.org.au (Alan Fowler) wrote:
>
>> The explanation is too long to type here. I suggest
>>you look at "Baud" in the 'Definitions' section of
>>http://www.emucities.com.au/member/whitethorn/hsmodem.html
>
>Dead site!!
>
Denis,
Thank you for letting me know. It was OK about a
week ago, but something was changed on the server on 29-8-01
and I cannot get access to any part of it.
regards. Alan.
The Old Bear
>From: m...@beatles.cselt.it (Maurizio Codogno)
>Newsgroups: comp.dcom.telecom
>Subject: Re: baud rate limit
>Date: 28 Aug 2001 11:20:50 -0400
>Organization: Telecom Italia Lab, Torino, Italy
>Lines: 19
>
>Alan Fowler scripsit:
>:
>: At 33,600 bit/s a V.34 modem is transmitting/receiving at 3429 Baud.
>
>I am at a loss. If bandwidth in a telephne line is between 300 and 3400 Hz
>(at least here in Italy) this means that I cannot possibly have more than
>3400 symbols per second, that is 3400 *transitions* per second, i.e. 2400
>baud.
>
>Where is the fault in my reasoning?
>
>ciao, .mau.
Because the definition of bandwidth "300 to 3400 Hz" is that frequencies
in this range are passed with no greater than about 3db of attentuation.
Frequencies outside of this range are not blocked -- they're just not
passed as strongly and the amount of attenuation increases as one moves
farther outside of the 300-3400Hz range. A certain amount of this
"roll-off" can and is tolerated.
While this may not answer all of your questions, below is something which
I wrote some time ago in response to various questions about transmission
schemes over the public switched telephone network.
Cheers,
The Old Bear
------------------------ begin included text -----------------------------
"Mike" <no...@none.com> writes:
>From: "Mike" <no...@none.com>
>Newsgroups: comp.dcom.modems.cable
>Subject: Re: Modem question
>Lines: 29
>
>Yes. I am talking about just two regular phone lines. They are analog
>right? Also, does this slowdown happen when I connect to my ISP?
>
>Thanks
Your v.90 modem, which plugs into a regular analog Plain Old Telephone
Service (POTS) line, sends data using essentially the same technology
as the earlier v.34+ modems which max out at 33.6kbps.
The device at your ISP is connected digitally to the telephone network,
usually by a channelized T1 or an ISDN PRI circuit. These digital
connections are capable of 64kbps but v.90 throttles them down to about
53kbps max because at some point, that nice clean digital signal has to be
converted to an old fashioned analog signal to send over the phone line
from your telephone company's local central office to your home.
Hence, v.90 will always be 33.6kbps max from you to your ISP and 53kbps
max from your ISP to you. (Actual connection speeds will vary depending
on line quality and conditions; see the lengthy discussion below.)
When two v.90 modems connect to each other, they both have the transmission
limitation of their respective analog phone lines, and hence they establish
at best a v.34+ 33.6kbps modem-to-modem link.
Interestingly, 3COM makes an ISDN "modem" which, as far as I know, is the
only desktop device that can accept incoming v.90 calls. This is because
it is a digital device and must be connected to a digital ISDN line.
For some more background on the what and why of all of this, I've pieced
this response together from several other postings I've made to
comp.dcom.telecom and comp.dcom.isdn over the past couple of years.
Pardon any roughness in the prose.
--------------------------------------------------------------------
You need to understand that a twisted pair copper telephone line using
analog transmission technology where a 48-volt DC voltage is simply
amplitude modulated is a 1200 baud device. Notice that I use the term
"baud" and not "bits-per-second."
Think of baud as the equivalent of symbols per second. If you are using
a number system which only two symbols, like binary with its 1 and 0, then
symbols per second is the same as bits (binary digits) per second.
But if you have a number system that is base 8, a symbol can be 0, 1, 2,
3, 4, 5, 6, or 7. These symbols have the binary equivalent of 000, 001,
010, 011, 100, 101, 110, and 111. Therefore, each symbol is equivalent
to three bits (binary digits).
Early modems using the Bell 103 standard made use of two audible tones
to represent 0 and 1. (There was a second pair of tones for data
coming in the other direction.) By having only two states representing
0 and 1, the Bell 103 standard was used at data rates of 50, 110 and
300 bits per second. Later Bell 212 standards raised these limits to
1200 bits per second and, ultimately, 2400 bits per second.
Subsequently, modem standards evolved which used multiple frequencies,
phase shifts, and relative amplitudes to create "symbol tables" which
were much more robust -- culminating in v.34+ with a data rate of
33,600 bits per second under optimal conditions. These modems actually
negotiate a symbol table at the start of the connection by sending
various combinations of frequency, amplitude and phase to one another
and determining what the specific connection can support. They also
are smart enough to retrain themselves if line conditions change
during the duration of the connection.
v.90 takes things a step farther by exploiting the fact that almost
all telephone connections are now totally digital except for the
"last mile" to the customer's end. By having the host end totally
digital (v.90 systems require ISDN or T-carrier providing a digital
connection to the host end terminal equipment), v.90 is able to
exploit the 56kbps digital path of modern telephone networks. But
the customer, who has a v.90 modem on an analog phone line, is still
limited to 33.6kbps from his end to the host end -- because of the
baud rate limitations of his "last mile" analog local loop and
central office hardware.
Although the inherent 1200 baud limitations are a serious impediment
to high speed datacom over conventional voice-grade analog phone lines,
one must pause in amazement that a telephone that great-grandmother
used in 1935 still can be plugged into a local loop from a modern
digital central office and that there will be dial tone, acceptance of
the dialing pulses from the rotary dial, and a fully functional talk
path established. What other electronic system, with the exception
of AM broadcast radio, can claim a backward compatibility with
hardware manufactured over a half-century ago?
But why is there this limiting factor is the "bandwidth" of an analog
telephone line... and why has it remained something which pretty much
unchanged since Alexander Graham Bell's first telephone in 1876.
Picture the twisted-pair copper telephone line as if it were a garden
hose. You have the hose out on the lawn one hot summer day and it is
flowing water from the faucet to the far end of the hose where your
father is watering the lawn. He takes the hose nozzle end around the
corner of the house while you stay at the faucet valve so you that you
can't see each other.
You decide to try to send him messages by turning the water pressure
up and down. If you do it slowly enough, he can cound the pulses of water
and you can agree on a code: 1 pulse, come for lunch; 2 pluses, the
mailman is here; 3 pulses, the ballgame is starting on TV; etc.
Now, this code works up to a point. But if you start pulsing the water
up and down too quickly, the combination of the compressibility of the
water and the internal friction of the hose will smooth out the pulses
so that by the time they get to the nozzle end, the water will appear
to be coming in a more or less contiuous stream.
Visualize:
- _____ _____ _____ _____
- | | | | | | | |
faucet end - ___| |___| |___| |___| |___
- - - - - - - - - - - - - - - - - - - -
- ___ ___ ___ ___
- ____/ \___/ \___/ \___/ \__
mid-way -
- - - - - - - - - - - - - - - - - - - -
-
- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
nozzle end -
---------------------------------------
So, as you can see, there is a limit to the rate at which the
flow can "change state" before the the changes all smooth out into
a blur at the far end.
For a telephone line, this limit is about 1200 times per second -- what
we call 1200 "baud" -- the unit baud being named after the French
engineer Emile Baudot who developed one of the first teletype machines.
Please note that "baud" is not the same as "bits per second" -- it is
a measure of changes of state per second. If our bits are represented
by the water being either on or off (two states), then baud and
bits per second are the same.
But we can get crafty and introduce some new states other than just on
or off. For example, with some clever mechanics, we might be able to
switch from cold water to hot water. Now we have another 'state' of
the flow which we can use to encode more than just one bit per change
of state. We can build a 'symbol table':
cold water high pressure 00
cold water low pressure 01
hot water high pressure 10
hot water low pressure 11
Here, each change of state represents a unique two bit symbol -- and
our garden hose can be made carry twice as many bits per unit of time.
Well, various modem schemes above 1200 bits per second use things like
frequency, pitch, phase, volume and other audio signal characteristics
to build a symbol table and encode a vocabulary of bit clusters. (v.32
and v.34 modem attempt various symbol tables as part of their
initial connection handshaking and then agree at the beginning of the
connection on what rate of change the particular phone line circuit
will accommodate.)
However, there is a mathematical limit to how many symbols you can
encode using a limited vocabulary of 'states' and a maximun baud rate
of 1200 changes of state per second.
Modem engineers do some pretty fancy stuff with "trelis coding" and
other mathematical gamesmanship but they are always running up against
Alexander Graham Bell's analog phone line's limits.
Now, digital transmission is an entirely different story and that is
why things like ISDN, ASDL, XSDL, etc. provide greater bandwidth than
plain old telephone service. And it is also why hybrid technologies,
like x2 or v.90, which use some of the digital signal path attributes
available in modern telephone exchanges, achieve a somewhat greater
transmission speed.
Today, virtually all of the public telephone network is digital
except for the last mile or so to residential subscribers' homes.
That last mile, over twisted pair copper, is pretty much the
same as it has been for the last hundred years.
Hybrid technologies like v.90 take advantage of the clean 64kb/sec
signal path which is established over the digital portion of the
telephone network. The "host" end of a v.90 connection -- that
is, the end usually at your ISP -- must be connected directly
by a digital circuit to the telephone network. That could be
a standard ISDN line, but more likely it is a 23-circuit ISDN
"PRI" or a 24-circuit T1 between the ISP's v.90 host device and
the telco central office switch. This provides a nice, clean
64kb/sec all-digital signal path to within that last mile of
the customer's home. At that point, the telco uses a device
called a line card to take that digital signal and make it look
like an analog signal from grandmother's day.
Your v.90 modem takes advantage of the best signal it can get
from that line card over your last mile "local loop" to achieve
data rates of as much as 53kb/sec. And, because this v.90 also
has to send data, it uses something similar to the older v.34
technology to send no more than 33.6kb/sec back in the other
direction. (Yes, v.90 is asymmetrical.)
DSL technologies sidestep the 64kb/sec channels of the switched
telephone network by doing something very different. DSL
providers locate their own equipment in space rented from
the telephone company inside the telephone company central
office. There, they connect their equipment directly to the
incoming copper wires from subscribers' homes so that the DSL
signal never reaches the public switched telephone network.
Instead, the circuits from many DSL subscribers are terminated
into a Digital Subscriber Loop Access Module (DSLAM) and
aggregated onto a private very broadband circuit for carriage
back to the DSL provider's network.
Ironically, the tricks that telephone engineers have employed
over the years to improve the carrying capacity of twisted pair
copper phone lines are detrimental to the use of those lines
for high speed digital signals. In addition to just better
manufacturing techniques for making insulated cable which is
uniform over its length, telephone company engineers have added
things like loading coils to smooth out electrical bumps in the
signal path caused by things like splices, cables of different
ages or quality, bridge taps (like rail sidings and spurs
leading off from the cable to other locations), or other things
which cause changes in impedance.
These tricks of the trade have been in use for quite a while -- long
before digital signal transmission schemes such as IDSN and xDSL
were ever thought of. And while they may smooth the path for
analog signals, they have the effect of favoring the passage of
frequencies within the voice spectrum (30-3,000 hz) and attenuating
frequencies outside of that range.
Every day DSL engineers keep coming up with more ingenious methods of
running megabit-per-second signals over the remnants of a network
which was originally designed for simple telephony. One can only
wonder at what point it becomes more cost effective to install
coaxial cable or fiber optics rather than make that last tweak to
wring more life out of existing copper pairs.
And keep in mind the irony that the Public Switched Telephone
Network (PSTN) in this country is designed to be totally backward
compatible to every terminating instrument made since the
introduction of the rotary dial in the 1930s. Imagine a computer
system that would still run a peripheral made almost 70 years ago!
So the PSTN uses digital signalling and digital transmission, but
adds "line cards" just before the last mile of twisted copper
wire to the subscriber's premises so that the subscriber can
plug-in grandma's telephone with the rotary pulse signalling and
90-volt AC magnetic ringer and it will still work.
That line card at the central office accepts all the bits of the
modern digital PSTN and uses them to activate analog DC talk
voltage, AC signally, pulse (or tone) conversion, dialtone buzz,
ringing tone brrrrrng, slow and fast busy noises, etc. It's all
a piece of electronic theater designed to make the system appear
to the end user like the familiar noises and protocols which we
all recognize.
In reality, when you place a phone call, you are opening up a
64kb/sec digital channel point to point (called a DS-0) and it is
this fact that the x2, 56kflex and v.90 technologies exploit.
Take away that line card and put real digital hardware on your
end and you could pass a 56kb/sec bit stream (reserving the
balance of the 64kb/sec channel for digital signalling -- i.e.,
call set-up, mangement and tear-down). But if you really want
to do that, the phone company will charge you for an ISDN line
which consists of two 64 kb/sec channels over which you can
move digitized anything -- voice, data, video, etc. -- plus a
16 kb/sec signalling channel.
Humorously, this is like so many things today: if you want it
without sugar, without fat, without preservatives, without
lead, without caffeine, or without alcohol, you'll just have
to pay a little more for what you're not getting.
Similarly, if you want the phone company to just deliver a
bitstream to your line (without dialtone, without ringing
voltage, without audio busy signals, without pulse/tone
conversion to digital signalling, and without doing a
digital-to-analog conversion, etc.), you'll just have to pay
a little more for all those things you're not getting. :)
Cheers,
The Old Bear
Leonard Erickson
wrote:
>Of course, I knew that since the days of V.32 more symbols were packed
>in a single transition to enhance bitrate - that was no problem :-)
Actually, that's been true since the days of v.22 and Bell 212 modems.
They got 1200 bits per second, by using four different "symbols" sent at
*300* baud.
V.22bis (2400 bps) modems were still 300 baud as I recall. Though they
might have been 600. And V.32 and later ran at 2400.
*Only* Bell 103, Bell 202, V.21 and V.23 modems run at bit rates that
match their baud rates.
Leonard Erickson (aka shadow{G})
sha...@krypton.rain.com <--preferred
leo...@qiclab.scn.rain.com <--last resort
Raymond Mereniuk
> V.22bis (2400 bps) modems were still 300 baud as I recall. Though they
> might have been 600. And V.32 and later ran at 2400.
V.22 is a 1,200 bps standard similar to Bel 212A. V.22bis is a
2,400 bps standard and the baud rate is 2,400 baud. V.32 is a
9,600 bps standard and the baud rate is 1,800 baud. V.32bis is a
14,400 bps standard with a data signaling rate of 2,400
symbols/sec or baud.
Virtually
Raymond D. Mereniuk
Ray...@fbntech.com
FBN - Offering LAST, Large Array of Stale Technology
http://www.fbntech.com/product.html