A modem, standing for modulator-demodulator, converts outgoing digital data into analogue signals over the telephone line, and converts incoming analogue signals back into digital data.

Over the last 20 years, the progress in the functionality, size and cost of modems has been amazing (though not so amazing as the progress in microcomputers). In the mid 1970s when I started with dial-up computing, modems were large wooden boxes (nicely made) rented from the Post Office which transmitted data at 110 bit/s, or 300 bit/s for an advanced model. Within a few years, the advent of viewdata (the earliest online service, when most developers of the Internet were still in nappies) had pushed the speed up to 1200 bit/s - although (and it was seen as an aberration at the time, not a strategy) the speed in the reverse direction was a miserly 75 bit/s. It took a long time for the speed to rise to 2400 bit/s in both directions. That was a plateau, then speeds rose to 9600, 14400 and 28800. That was by some people called the "final generation" of modems. Many people firmly believed that ISDN would sweep all this away. Speeds rose a little in the last two years to 33600 bit/s but one could see Shannon's theorem biting.

What few people thought of was to take advantage of the networks which combined analogue and digital communications. Rockwell and US Robotics did, and have come out with 2-way 56000 bit/s modem (with a deal of fudging about the actual speed of the reverse channel).

There are a number of modem standards in common use today. They are known as "V" standards, which is the letter used for data communications over analogue lines in the nomenclature of the ITU-T (formerly CCITT) standards body for telecommunications. We shall not in this Unit consider the Bell standards previously widely used in North America for lower-speed modems, since they are rapidly being overtaken by the ITU-T standards.

In addition to these ITU-T standards, from time to time modem vendors, normally very keen on standards, "break ranks" and produce a new "industry standard" just to leapfrog the competition. Usually after a year or two things settle down and the international standards bodies catch up.

This happened with V.34, there was a pre-V.34 industry standard called V.32terbo for modems running at around 28800 bit/s; and minor extensions to V.34 were introduced to produce modems running at 33600 bit/s.

Many commentators described V.34 as the "final generation" of modems, believing that (a) vendors were getting close to the theoretical limits imposed by Shannon's theorem, and (b) the success of ISDN would mean that modem vendors would no longer have the incentive (or cash) to invest in the high levels of R&D to produce a new generation of modems. Note that the modem industry is an unusual sector of the computing industry - normally in the computing industry advances (for example in processor speed or memory size) are doublings or order of magnitude (ten-fold) increases; but in the modem industry speed increases of around 15% (28800 to 33400) are seen as giving a competitive edge.

Well, as so often, the final generation turned out not to be final, as a generation of "56K" modems has been developed.

Various developers were trying to push the speed of modems up beyond 33600 bit/s, and they have based their work on one key insight: the signal/noise assumptions in the telephone network (see Shannon's theorem) no longer apply for the typical modem application. This is for two linked reasons:

The area of 56 kbit/s modem technology was first proposed to the public by Rockwell in a series of announcements around August 1996, under the codename "K56".

This was followed in October 1996 by announcements from Texas Instruments and US Robotics, concerning the 56K modem technology called "x2".

Rockwell continued the escalation by using the codename "K56Plus" and "K56flex" - no doubt assuming that developers will read into it a promise of perhaps 64 kbit/s modems or even higher speeds.

By amending the FCC regulations in the USA so that higher signal levels are allowed - speeds of 56 kbit/s "and potentially higher rates" could be achieved. It is possible that higher signal levels are already allowed in some countries.

The standard adopted for 56k modems is V.90.

There are several difficulties that may arise during the field trials of modems. The industry has already had two bad experiences with fast modem technology which appeared to work well in the lab but not in the real world:

If the technical difficulties are overcome then this technology could, in some European countries where ISDN tariffs are still high, cut deeply into the home-based ISDN market or postpone its very existence in some countries. However, it is true that industry commentators see less pressure from the 56K technology on the office-based ISDN market, which is a little less cost-sensitive and where other features of ISDN (like fast call set-up) play a larger role in selection of technologies.

There are a mass of modem vendors but only a few industry leaders. Based on a casual survey in the Usenet Newsgroup comp.dcom.modems, here are the top names that crop up (not all the top names are the same every year):

Although there are hundreds of modem manufacturers, there are very few manufacturers of the digital signal processing chips which make up the "heart" of a modern modem. Two of the big names are:

Texas Instruments make the chip at the heart of the 56k technology from US Robotics, it is the signal processing chip which when driven by the software developed by US Robotics delivers the 56k technology.

Rockwell make the chip set which is used by Motorola, Hayes and others.

In addition to the "raw" speed of modems, further speed increases have been obtained over the years by the techniques of data compression.

The basic mechanism of data compression is to recognise patterns in the byte stream presented to a modem and find more compact ways of representing it. An interesting early example of data compression was Morse Code, because in Morse Code, the more frequently used symbols had shorter representations. Here are some more modern examples:

The main uses of modems were covered in Section One. Why not re-read Section 1.3 there? As said earlier, the main use now is to link microcomputers (at homes or small offices) into central services. These central services are mainly Internet-based.

ISDN was covered briefly in Section One. You may want to re-read that to refresh your memory.

ISDN stands for Integrated Services Digital Network. It is a digital enhancement of the usual telephone service which allows data to be sent over telephone lines at speeds rather faster than the fastest modems in common use. The call set-up time for ISDN is also fast - much faster than for a modem to make a connection.

In most countries, an ISDN call costs the same as a telephone call.

ISDN was specified by the CCITT in the early 1980s but it was many years until it became deployed. Even now, it is only in perhaps France and Germany that its use could be called widespread. There are various things that have caused problems for ISDN:

4 The connection charge for ISDN is quite high in some countries such as the UK.

Vendors seem mainly interested in two main areas:

Many commentators in the past have thought that ISDN is a technology in search of an application. They argued that often-proposed ISDN applications like high-quality fax do not seem very useful on the mass scale, compared with modern electronic mail.

However, some sectors have taken up ISDN in quite a large way.

There is a great deal of use of ISDN for videoconferencing in education.

ISDN is having a significant effect on teleworking. This is because ISDN allows one's favourite applications on the LAN to be available in one's own living room.

There are also several interesting applications in the publishing sector - these include:

2 Supplying images to clients from colour scanner bureaux.

3 Access to photo libraries.

5 Advertising agencies sending roughs of artwork to clients.

7 Remote technical support of users, by using screen-sharing software.

These are likely to break down into three main kinds:

2 Doing new things. The obvious example is video communications.

3 Doing things which did not work well on analogue networks.

Some applications of ISDN have been covered in outline above. Case studies may be found on the WWW.

There are many vendors of ISDN equipment. In particular the US vendors have discovered ISDN as a product, even if the deployment is patchy within the continental US. However, there are many vendors with specific national or European orientations - how long this will last is not clear.

Two main service providers are:-

There are many kinds of products associated with ISDN. The most relevant are: