[comp.dcom.modems] baud vs. bit/sec.

WITLICKI@WILLIAMS.BITNET (04/22/87)

In reference to the article Keith Petersen posted about Baud vs. Bits/sec.

> .....  This is nothing new, texts on the subject generally avoid
>        the term "baud"

  Yes, but computer companies use the word continually.

  ...Rather than do a point by point response to what I feel is a very
muddy, although perhaps technically correct article, let me give
you a few of my explanations... (hopefully a little more clearer then
the recursive definition of 'bit rate' in the glossary).

  We first have to separate our wires.  The RS232 wire between your
modem and your terminal or computer talks in binary (0 or 1) - the
data rate here is _Bits Per Second_.
  The Phone line which is between your modems talks in _Baud_.
  Now comes in the common confusion.  Computer and terminal makers
typically give the speed of their RS232 interfaces in terms of
'Baud' - this is wrong - it should be Bits Per Second.
  At modem speeds of less than 2400 baud the modulation techniques used
usually result in 1 baud = 1 bit per second, hence the casual and
incorrect use of terms.
  Definition: A Sine Wave is a signal at one frequency and one frequency
only.  (no harmonics if you think in terms of music).  On a telephone
channel which is bandwidth limited (by the phone company) to 3000 Hertz
- this is highest frequency (== fastest) which you can push through the
wire.  Because of signal degradation and the chaining of telephone
company equipment, the maximum signal frequency used in practice is
a few hundred Hertz lower.  If you are still confused - think
in terms of music - a higher frequency == a higher pitch.
  The Nyquist theorem says that for a sine wave of 2400 cycles per
second (I never did like renaming cycles per second into Hertz, next
thing you know, phonograph records will go at 33 1/3 Edisons, but I
date myself and I digress...) ... for a a sine wave of 2400 hz you
can have 1200 bits per second.
   Shannon, on the other hand, said hey fellows, we got us a Noiseless
Channel.  We don't have to push sine waves through it, we'll push
voltage levels through it which you can sample at each interval.
Heck, with no Noise at all, say we have at least 65,536 voltage levels
in each sample of a 600 baud channel and we can do 9600 bits per second...
(to use the example in the original piece).  Life is not so simple.
We do have noise.  The Nyquist Theorem does apply (...  but...).
   In the case of 9600 BPS over a voice grade channel, Phase Modulation
uses each sampling interval (baud rate time interval) to push several
bits of information over the line.

   I hope this helps some.

   Quite frankly, I think it's time to drop the term Baud from
the data communications consumer's terminology (who's the consumer?,
who's the engineer? who's the user? why do I have to go through
this to set a few a dip switches and use the thing?).  Computer systems
people have serial interfaces which talk at Bits Per Second.
I don't give a hoot what sideways modulation method and baud rate
you use over the phone line.  I want to talk in terms of 'I have a 9600
Bits Per Second Asynch. Serial Interface' and 'Does it talk to a Hayes
or a Bell 212A' and 'Can I use a voice grade line or do I order a
Type 3002 unconditioned line or what?'.

- randy

P.S. I am in the process of writing some introductory material
on data communications and would appreciate any pointers to
published stuff anyone has seen which explains bits per second,
parity, etc. for the user/system manager who shouldn't be subjected
to the Nyquist/Shannon intimidation.

steckel@alliant.UUCP (04/22/87)

BAUD = signalling elements per second
BITS per SECOND = data bits per second

Any ASYNCHRONOUS communications device on RS232 or RS422/423 has start and
stop bits attached to each and every character transmitted; these are required
for the devices on each end of the line to detect where the data characters
live.  Typical amounts are:
  ancient technology @ 110 baud: 1 start bit, 8 data/parity bits, 2 stop bits
  wrong-headed technology @ 134.5 baud: 1 start bit, 5 data bits, 1.5 stop bits
  newer technology @ 300 baud & up: 1 start bit, 8 data/parity bits, 1 stop bit

These start & stop bits are transmitted (for instance) from terminal to host,
over low speed asynchronous modems, etc.  In these cases BAUD is the correct
term to use to determine signalling speed.  Bits per second will always be
lower, but it is format and user dependent.

SYNCHRONOUS communications does not use start and stop bits; it uses sync
patterns at the beginning of a block to co-ordinate the sender and receiver.
Again, the BAUD rate is the signalling rate over the channel, while the
data rate depends on the size of sync pattern, block header and trailer,
etc., etc., and is format and user dependent.

I hope this clears up the confusion.
	geoff steckel (steckel@alliant.uucp, gwes@wjh12.uucp)

WITLICKI@WILLIAMS.BITNET (04/22/87)

Phil Ngia writes:
>
>You are wrong. Ever heard of Bell 212? Try looking it up.
>Also, you surely mean less than 2400 bps, not baud.
>--

  ooops, sorry you're right.
  i guess that shows i can just as easily fall into the trap of equating
bps and baud.  (for the $ set terminal  command on the VMS vax which
I use every so often, the Help command says you are setting the
speed in baud...)
  am i being uptight about people using 'baud' to describe bps,
especially when it is RS232 (yes i know when you are doing binary
signaling that baud does == bps, but i think bps is more proper).
  old habits die hard.

- randy

pozar@hoptoad.uucp (Tim Pozar) (04/23/87)

    If you want a very good definintion of how modems work (>300 b/s), get
the technical Manual for the Advanced Micro Devices Am7910 WORLD-CHIP FSK
modem.  Section 2 of the manual has a tutorial for modems.  
    One analogy they draw is:
    "Think of the information being transmitted as a stream of automobiles
with people in them moving down a highway (the phone line).  Consider 
each care a baud (symbol).  Consider each person a bit. 
    "Bit Rate = Baud Rate * No. of Bits/Baud 
  or people/sec = cars/sec * people/car.
    "Bit Rate >= Baud Rate.

    A "baud" is the speed at which the carrier is Frequency Shifted.  If
it is moved from, say 1000cps to 2000cps 100 times a second, it would
be delivering 100 "baud".  If this shift shows is representive of a change
in state of data (1 to 0, or hight to low, or the other way around), then
it would be sending 100 bits per second or 100 b/s.
    In order to put more that one person per car for FSK modems you can
phase shift the frequency.   A phase shift of 90 degrees can have 4 people
in a car, or 2 bits per baud.  Since 00b = 0, 01b = 1, 10b = 2, and 11b = 3.
    This is how 2400b/s modems work.  By spliting the phase "cycle" into
eigths, we can send 3 bits per baud, thus sending 4800b/s.  
    This is about the max we can send.  What we are talking about is 45
degree phase angle shifts.  We start getting into distortion with the
phone line that makes getting smaller incraments pretty rough.  So,
there is one other way to get more bang for our buck.  Quadrature Amplitude 
Modulation or QAM.  QAM phase shifts the frequency and amplitude shifts is
as well.  Also the pattern of the bits are layed out differntly.  Instead
of a circle, with each bit being a different point on the edge (the size
of the circle would represent the amplitude of the signal), the pattern
is broken up into squares:

                 0
     1000   1100 | 0100   0000
                 |
     1001   1101 | 0101   0001
  270------------+------------90
     1011   1111 | 0111   0011
                 |
     1010   1110 | 0110   0010
                180 

    Now we can send 4 bit/baud or 9600 b/s.
    According to the tech manual:
    "QAM is even more sensitive to line distortions than PSK (Phase Shift
Keying).  Automatic equalizers are required to compensate for these 
distortions.  This type of modem is the most complex and expensive of
all the modems to implement.

-- 
        Tim Pozar
UUCP    pozar@hoptoad.UUCP
Fido    125/406
USNail  KLOK-FM
	77 Maiden Lane
	San Francisco CA 94108

howard@COS.COM (Howard C. Berkowitz) (04/23/87)

The previous article made some excellent points on the confusion 
between baud and bits per second.  While I agree the "consumer"
should exclusively use BPS/K[ilo]BPS/M[ega]BPS (and always spell
out bytes per second), there is at least one place that knowing
the true analog baud rate can be useful:  generating "eye"
patterns.

An eye pattern is an oscilloscope display which gives an idea
of total system transmission quality.  Sales strikes again here,
for most commercial "eye pattern generators" actually are
constellation or phasor displays.  For my purposes, eye patterns
are NOT the neat geometric displays of dots (probably more
useful than eye patterns) which the diagnostic sections of 
good high-speed modems can produce.  Eye patterns look like
they came out of a science fiction movie, and are good
1) for impressing management and visitors amd 2) getting
a visual go-no/go check if you have no other equipment.

Eye patterns are generated with an oscilloscope, using the signal
on the line and the baud rate as the sweep rate.  Baud, in this
context, is defined as the number of times per second the line
changes state.  With this setup, you will see a number of overlapping
more-or-less sine waves, which in their crossing produce a central
empty "eye," something like the eye of a hurricane.  The larger
the eye opening, the better the transmission performane.

This is basically for entertainment, although it is useful in
some engineering applications.

One point about the previous article -- I never thought before to
use the rule that it's bps on the digital side and baud on the
analog, and borrow it with gratitude as a teaching device.
To be technical, though, baud=bps on the digital side, because
there is one signal state change per bit on a digital channel.

Randy's article was excellent for the consumer of modems.

......

"Consumers" of modems remind me of a note rumored to have generated
the largest letter-writing campaign in the history of one IEEE
magazine.  We have, for a long time, tacitly understood that a
Consumer of Single Socks lived in washing machines.  The pubished
note suggested this belief was wrong.  It proposed that Socks are
the immature but sexual form of the Wire Coat Hanger.  Have you
ever noticed how wire coat hangers multiply?  This is due to
one member each of two sock pairs (i.e., of different sock sexes)
mate to form a wire coat hanger.  The divorced partners of these
socks are too shaken to mate (or afraid of incest), and thus become
the odd socks in the drawer.

jdf@pbhyc.UUCP (Jack Fine) (04/28/87)

Ummm... I think I agree with most of the explanations I have seen here on
what a bit versus a baud is.  Though I am not sure if what I think I read
is what I think I understand.  Understand?  Well anyway I have been
advocate for sometime (teachine datacomm) of trying to get people to drop
the baud term when referring to data communications.  Yes a 300 bps modem's
signial is modulated at 300 baud per second times one bit per second
giving you 300 bps.  However, 1200 bps modems divide each second into 600
parts which is done by a method called four-level phase shift keying.  Each
part then can then hold two bits of data.  This is at 600 baud and if you
multiply 600 times the 2 bits you get the 1200 bps transfer rate.  With
2400 bps modems you use a 16-level phase shift keying (still at 600 baud)
which allows 4 bits of data per baud.  So I second the vote for dropping
the term baud when used to refer to electronic data transmission.

What may be better is the characters per second.  With most transmissions
it takes 10 bits for a character of data.  So 1200 or 2400 divided by 10
would give 120 or 240 characters per second which has more meaning to me as
a user.  Now if we can just get the modem companies to advertise 240
character modems.....

Jack