The history of private datacommunication 4283

This is really a shift between technologies that is obfuscated by the speeds. The term "Baud" is probably misused here without the poster realising it.

Baud is the rate (per second) of state changes on the line, or the "encoded symbol rate".

The original V.21, Bell103 etc modems are simple modems that translate logic levels to frequencies and back. They use 2.5-3 times as much bandwidth as the original symbols, and use a frequency range in each direction. They really have no concept of "baud" inside the modem.

A V.21 modem will work just fine for bit rates from DC to around 500 baud. Some BBS'es supported 450 baud connections just for this reason. The same modem works just fine for telex (45.5, 50, 75, 110 , 134.5 or 150 baud; depending on standard).

The is a simplex (one-way) version of V.21 called V.23. This standard reserves almost the entire band for one direction, and a small back channel the other way (leading to 75-1200). The 75 channel is pretty tight; but the 1200 channel can handle around 1800 baud. Tested that, it works.

This technology grew incrementally from pre-WW2 until the sixties. The technology with plain FSK coding of logic levels is present for a lot of simple applications. All kinds of simple digital transmission systems use it. Your garage door opener, your house alarm, the radio for the remote toy car you gave your kid probably all use V.23 with a dirt cheap chip to be modem. Because it is robust, cheap, and fast enough.

I think the poster refers to the phase shift keying persent in V.22 and similar bell standards. uses a frequency band in each direction. They laid the frequencies out a little better, so (as we saw with v.23) they get 600 baud in each direction. They don't encode logic levels anymore. They do synchronise with the bit stream, and don't encode tones anymore, they ancode phase shifts. With 0, +--90 and 180 degreees they can encode two bits per baud; and you get two-way 1200 bps modems. These are now real, digital modems.

There also exists a similar "shadow version" like V.23 shadows V.21 for single direction. ISTR it is V.27 that does plain PSK in a single direction. These modems need to synchronise, and start the dialogue with a handshake of tones.

This technology arrived around 1970, but it took a few years to get it standardised; and commercial modems were available only around 6-8 years later. They were horrendously big and expensive at the start, and only saw much use in the early eighties.

WHoa. First we must go through the PASK modems. After PSK the intelligent signalling designers though of including even more bits per symbol. They started encoding bits also on amplitude, but that was unreliable; so they ended with a common encoding of phase and amplitude which can be plotted as a "star" diagtram,

Otherwise it uses the framework from V.22; as in V.22 bis. There are also one-way versions; most notably V.29 that gives single direction 9600 bps with the same 75 baud return channels as in V.23. This is what the standard, G3 fax uses.

These modems followed quite fast onto the bandwagon the V.22 modems created when they finally went commercially viable. The fax contributed a lot to bring down the cost of these. We are still in the very early eighties.

The next step was to apply some signal processing, and detect the received signal as differential from the received signal. This led to some changes in the symbol encoding "star" to make it more reliable. This "new star" is called "trellis encoding", and looks more like a diamond than a star.

Now modems could transmit full blast in both directions, and use signal processing hardware to pick the bits out of the return stream. The standards went to 14.400 baud pretty rapidly, but these modems often fell back to removing symbls from the encoding to make it more robust, and had fallback to 12800, 9600, 7200, 4800 and 2400 bps. They often degraded, and had no good way of climbing back up without redialling.

This happened in the late 1980's. Technology stopped at 14400 for a while because of limitations in the signal procesing chips of the time. In the mean time the modems got a link layer; with MNP as a pioneer. The standards rapidly included a LAPB-like layer that did retransmissions on errors. Thereby the chips could be pushed even harder.

While phone lines got better and better the concept of error recovery was in place the encodings could be pushed even further. The symbol rate also went up at some stage; but in principle it is just "more of the same" as the 14400 error correcting modem.

Wrong order, really.

The 56k modem uses a whole different encoding. Signal processing and phone lines had become so good that they could just push more or less raw bits out the ISDN interface in one end, and let the remote end deduct from the audio levels what these bits were.

By this time (late 1990s) all core phone networks were digital, and ISDN really reached almost out to the customer. The last bit was still analog, but if there were only one digital-to-analog encoder you could do such a wild hack as the 56k "mo"dem.

This requires the sendin end to be connected with ISDN, and the conversion to analog signals only done once on the entire connection; probably pretty close to the customer end.

It really does not modulate anymore. It pushes raw bits, and lets the other end rebuttemble it. This requires processing power that 10 years earlier would be a supercomputer task. A 56k modem probably has the processing power of a Cray-1 for it's dedicated task.

But it does work, sometimes. And the 33.8 modem can take over, and wit error correction it can fall back; all the way to 2400 bps and back up.

IBM also had systems that used V.23 modems intensively, and controlled transmit from the stations carefully. This was really what Bisync was all about. It had a good fit with the block transmission concepts of the systems at the time.

Travel agencies are still full of these modems, or later, faster modems that still do transmission control. You can detect them from the special blinkenlight pattern on the modem lights as they transmit, listen etc.

-- mrr