grking@nunki.usc.edu (Greg King) (07/16/88)
I have seen some advertisements for 9600 baud modems that can operate over normal telephone lines. I've heard that the bandwidth of normal telephones is about 3 kHz (correct me if I'm wrong), so I'm wondering how do these modems work? Please e-mail your replies (I don't poke around in sci.electronics too often). Thanks.
grking@nunki.usc.edu (Greg King) (07/16/88)
My e-mail addresses are : grking@nunki.usc.edu
gking@hydra.usc.edu
...!oberon!hydra!gking
Sorry about that. I made my .sig file too long.
greg kinglogajan@ns.UUCP (John Logajan x3118) (07/19/88)
In article <1127@nunki.usc.edu>, grking@nunki.usc.edu (Greg King) writes: > 9600 baud modems can operate over normal telephone lines. > the bandwidth of normal telephones is about 3 kHz. > how do these modems work? With a bandwidth of 0-3000 hertz, you can simultaneously send tones at any arbitrarily spaced frequency intervals, say every 10 hertz. The presence or absence of a tone can have some information meaning. Thus you could, in the 10hz case, send some 300 bits of information simultaneously. The more tones you send, they lower in amplitude they have to be, because the aggragate volume could exceed that allowed. But by proper phase choice, the rate of volume increase to number of tones added, is much less than a linear relation! Anyhow, the basic idea is parallel information flow, rather than the serial, but much simpler, single frequency jobs such as the 110/300 baud modems of old. The 1200 and 2400 use a form of phase encodeing, which from another point of view is the same as multiple simultaeous frequencies. Just like the 9600 guy. - John M. Logajan @ Network Systems; 7600 Boone Ave; Brooklyn Park, MN 55428 - - {...rutgers!umn-cs, ...amdahl!bungia, ...uunet!rosevax!bungia} !ns!logajan -
dya@unccvax.UUCP (York David Anthony) (07/19/88)
In article <478@ns.UUCP>, logajan@ns.UUCP (John Logajan x3118) writes: > In article <1127@nunki.usc.edu>, grking@nunki.usc.edu (Greg King) writes: > > 9600 baud modems can operate over normal telephone lines. > > the bandwidth of normal telephones is about 3 kHz. > > how do these modems work? > > With a bandwidth of 0-3000 hertz, you can simultaneously send tones at any > arbitrarily spaced frequency intervals, say every 10 hertz. The presence or > absence of a tone can have some information meaning. Thus you could, in > the 10hz case, send some 300 bits of information simultaneously. The more > tones you send, they lower in amplitude they have to be, because the aggragate > volume could exceed that allowed. But by proper phase choice, the rate of > volume increase to number of tones added, is much less than a linear relation! This is how **some** 9600 baud modems work. Real 9600 baud modems work by quadrature amplitude modulation of a 1700 Hz (or so) carrier wave. Groups of 4 bits are encoded through the usual scrambler device to yield four scrambled bits (ensuring not having runs of ones and zeroes of significant length, so that the carrier clock may be recovered). Each four bits represents a unique amplitude and phase. The recovered clock then is multipled against the recovered QAM-encoded audio, yielding a magnitude and a phase. A trivial circuit decides which of the quadbits the recovered magnitude and phase corresponds. Now, much digital signal processing is required to keep the QAM signal error-free; but the technology in real, Honest-to-God 9600 bits/sec modems (such as the UDS 9600 A/B, or the Rockwell R96MD/R96FAX) is not frequency division multiplexing. It is QAM. Now, with trellis encoding of the original quadbit patterns, (forward error correction) you can obtain a better bit error rate for a given signalling rate, or pack in even more symbols per baud. This is the principle behind the UDS 14.4 Trellis (14.4 kbps) and the Rockwell 14.4 chipset to be introduced. They are also QAM modems. Note, that a baud is the number of symbols per second being transmitted. A 9600 bit/sec modem operates at 2400 baud. Each symbol represents a unique magnitude and phase for each group of 4 bits. York David Anthony dataSPan, Inc.
logajan@ns.UUCP (John Logajan x3118) (07/20/88)
In article <1044@unccvax.UUCP>, dya@unccvax.UUCP (York David Anthony) writes: > This is how **some** 9600 baud modems work. Real 9600 baud > modems work by quadrature amplitude modulation > Each four bits represents a unique amplitude and phase. This is point of view dependent. You can say QAM is built your way or you can say that QAM represents the resultant waveform of multiple simultaneous frequency synthesis. After all, any arbitrary waveform can be constructed from the proper combination of various sinewaves. My point is that high baud rates can be seen as a form of parallelism. - John M. Logajan @ Network Systems; 7600 Boone Ave; Brooklyn Park, MN 55428 - - {...rutgers!umn-cs, ...amdahl!bungia, ...uunet!rosevax!bungia} !ns!logajan -
henry@utzoo.uucp (Henry Spencer) (07/24/88)
> This is how **some** 9600 baud modems work. Real 9600 baud > modems work by quadrature amplitude modulation... Real 14000 baud modems, mind you, work by the multi-carrier approach originally described. ("Real" meaning "Telebit Trailblazer", of course.) Why buy 9600-baud QAM when you can have 14000-baud PEP? :-) -- Anyone who buys Wisconsin cheese is| Henry Spencer at U of Toronto Zoology a traitor to mankind. --Pournelle |uunet!mnetor!utzoo! henry @zoo.toronto.edu
dya@unccvax.UUCP (York David Anthony) (07/25/88)
In article <1988Jul23.203605.20716@utzoo.uucp>, henry@utzoo.uucp (Henry Spencer) writes: > > This is how **some** 9600 baud modems work. Real 9600 baud > > modems work by quadrature amplitude modulation... > > Real 14000 baud modems, mind you, work by the multi-carrier approach > originally described. ("Real" meaning "Telebit Trailblazer", of course.) > Why buy 9600-baud QAM when you can have 14000-baud PEP? :-) Not true at all. CCITT specification V.33, which applies to trellis-encoded 14,400 bit/sec and 12,000 bit/sec (corresponding to 6-bit per symbol and 5-bit per symbol encoding, respectively) specifies only QAM of the carrier. A V.33 modem, when operating over a suitable dial-through circuit, **guarantees** a throughput of 14,400 bits per second, period. Unless and until Telebit Trailblazer, or any other of these "tier-two" high speed modems can demonstrate a throughput of 14,400 bits/sec using random synchronous data, these modems will be rightfully considered to be "second class" compared to the UDS 14.4 Trellis and the Rockwell R144HD. Having transmitted medical images (essentially random data) through every kind of modem known to God, I can with certainty state that the Telebit Trailblazer does not even achive achieve 9600 bits/sec on random data. The Trailblazer does deal with certain line impairments better than the earlier QAM modems. However, the two technologies are **not the same**, despite what you can do with Fourier transforms and their ilk. Of course, most people do not send random data (or nearly random data), the adaptive compression and equalization of the Trailblazer (plus many manufactuers' nice big "fool UUCP" buffers) can make burst rates of 14,400 bits/sec possible. As far as I am concerned, the only measure of a modem's performance is the worst case continuous speed at a specified bit error rate. Why not run crypt on a file like /vmunix and send that through the Trailblazer and a UDS. Hmmmmmmmmmmmm? Yours for truth in modem advertising; York David Anthony
dave@onfcanim.UUCP (Dave Martindale) (07/29/88)
Actually, the Telebit Trailblazer's modulation technique allows it to transmit at about 18000 bps on a relatively clean line. With protocol and error control overhead, plus occasionally turning around the line for ACKS (the modem is really half-duplex) the actual throughput in one direction is about 14000 bps, provided there is no traffic in the reverse direction. I have personally seen throughput of above 1300 characters per second of approximately-random data (compress output) with the modem's compression algorithm turned off. And the limiting factor was that my poor 780 was spending 100% of its time pushing characters around; the modem was probably capable of more.
henry@utzoo.uucp (Henry Spencer) (07/30/88)
In article <1047@unccvax.UUCP> dya@unccvax.UUCP (York David Anthony) writes: > Unless and until Telebit Trailblazer, or any other of these >"tier-two" high speed modems can demonstrate a throughput of >14,400 bits/sec using random synchronous data, these modems >will be rightfully considered to be "second class"... My recollection is that the link-level bit rate of the Trailblazer on a good phone line (it adapts much more gracefully to poor ones than the V.xx modems) is 18k, and the delivered error-free rate is 14k. This is raw random-data rate, mind you, WITHOUT DATA COMPRESSION. > Having transmitted medical images (essentially random >data) through every kind of modem known to God, I can with >certainty state that the Telebit Trailblazer does not even achive >achieve 9600 bits/sec on random data. You're transmitting with uucp, I assume? Have you forgotten protocol overhead? (For that matter, have you considered the possibility that the CPU is the bottleneck?) -- MSDOS is not dead, it just | Henry Spencer at U of Toronto Zoology smells that way. | uunet!mnetor!utzoo!henry henry@zoo.toronto.edu
dya@unccvax.UUCP (York David Anthony) (08/01/88)
In article <1988Jul29.194813.27599@utzoo.uucp>, henry@utzoo.uucp (Henry Spencer) writes: > In article <1047@unccvax.UUCP> dya@unccvax.UUCP (York David Anthony) writes: > > Having transmitted medical images (essentially random > >data) through every kind of modem known to God, I can with > >certainty state that the Telebit Trailblazer does not even achive > >achieve 9600 bits/sec on random data. > > You're transmitting with uucp, I assume? Have you forgotten protocol > overhead? (For that matter, have you considered the possibility that > the CPU is the bottleneck?) Bad assumption. I'm transmitting with proprietary hardware which can keep up with 2.048 mbps serial synchronous data, and has DMA access to zero wait-state RAM with 4 mb capacity. I never said that the Trailblazer wasn't theoretically capable of handling X mb/sec. What I did say (and using various methods of very high performance equipment to back it up) is that when the criteria for a plain old telephone line is satisfied, the V.xx modems **guarantee** a **specifed** bit error rate and a **specified** bit stream rate. Period. Our customers demand a specified BER at a specified data rate using random serial synchronous data. Period. If nothing else, all someone needs out in Bumfuk, Iowa is a modem which, upon adapting to a marginal line, transmits at significantly less than 9600 bps. In my application (where I have several lines located in some major, major, MAJOR markets, say Boston, NY, which can be used to dial back around to myself, as well as other test setups in some tiny, tiny, TINY markets of similar ilk) UDS and Rockwell beats Trailblazer. The Rockwell R96MD operates error free on 98% of our installations, (in the hundreds) on unconditioned residential type circuits, even where the loop loss is very high (such as Chillicothe, Missouri, or Potomac, Md.). This is essentially the same OEM modem used by every fax machine manufacturer on the face of the earth. In general, the guaranteed BER is 10 e-07. The observed BER is 0. This, with files typically four meg or more. My only transmission overhead is the time it takes to execute two move.l's to turn on the timer, and turn off the timer and dump variables containing error counts. Microseconds, at worst. I think my testing protocol is an order of magnitude more robust than using a Unix box. I'm glad that the Trailblazer works for everyone concerned. All I wanted to call attention to is the fact that 1) QAM does NOT equal Trailblazer and 2) QAM modems get a bad rap from people who have installed and operated about a few hundred less than I have and 3) the Trailblazer statement about bit error rate and transmission speed is not particularly honest advertising. Now, if Telebit publishes charts of various line impairments and the bit error rates (or speeds) which then ensue, I'm all wet. But, all I hear is "well, the Trailblazer does 1300 char/sec (lets assume 8 bit synchronous characters where all the data is valid and not asynchronous ASCII, or a rate of about 10400 bit/sec) assming that it is on a good line, or it isn't transmitting anything in the reverse channel, blah, blah, blah. I know people have to do things like this. There are probably step-by-step metallic exchanges somewhere where the phone lines are essentially transparent where the Trailblazer might hit 18,000 bit/sec. My whole point is, is, that the Trailblazer does not guarantee a specified data rate at a specified bit error rate. I can buy a UDS ECC-1 box, too, and get some error rate like one error every 7 years, too! (This is a packet error correction box which handles buffering and other junk between the V.xx modem and the DTE.) We don't use it, BTW. Out of the SIO and into the modem. What could be simpler? Frankly, if you need 100% link availability, you have no business using dial-through modems to begin with. However, until you have used V.29/V.33 modems, and have an installed data base (remember, 98 % link availability at 10e-07 BER and the guaranteed data rate of 9600, 12000 or 14400 bps) of a few hundred (I lost count at about 275 or so of our customers) please don't bash 'em. They work quite well. My challenge still stands. Yours for truth-in-modem-advertising York David Anthony DataSpan, Inc.