Wireline modems were for use over dial-up or leased telephone-type lines and are divided
into two classes of specifications: Bell (the name Bell is generally no longer used, just the
specification number to describe a modem) and the International Telegraph and Telephone
Consultative Committee (CCITT), now ITU-ITS, which issues the "V." and "X." standards.
Table 7-1 lists various standard modems with their speed and modulation types. Most
modern modems are synchronous between the modems, using LAP-M packets, and may
have a synchronous or asynchronous interface. On PC modems the interface will be asynchronous
as the PC bus for I/O is asynchronous. Depending on price and application,
stand-alone modems may be configurable for interface through either programming or the
use of a dip switch.

Typical of many of the low-speed legacy wireline modems is the Bell 212 type, which contains
a Bell 103 set of carrier frequencies. A plug-in card for many PCs, this once was the
most widely used modem on dial-up lines. The Bell 212 is capable of identifying what type
of modem it is talking to and adjusting itself to become that type. Considering the great
differences in modulation schemes, and so on, this is no small feat.

Faster Modems

The Bell 212 modem had to use data encoding to overcome the bandwidth restrictions of a
typical dial-up wireline. An older specification, Bell 202, operated at 1,200 bps using FSK,
with one carrier at 1,200 Hz and the other at 2,200 Hz. This scheme used almost all of the
bandwidth available to the typical two-wire wireline and had to operate half duplex. This
caused throughput problems because the predominant method of error recovery is ARQ
(Automatic Retransmission Query), and after a number of transmitted blocks an acknowledgment
must be returned to the transmitter. To avoid having to perform handshaking
every time the line was turned around (in order to send an acknowledgment of message
receipt), the 202 used a secondary channel. A small portion of the spectrum near the 300
Hz lower limit was used for a 5 bps channel. Though this could be called duplex, it is only
technically so. The problem was the use of FSK as the scheme of modulation and the
limited wireline bandwidth. The Bell 202 could actually be called the originator of the asymmetrical
data line, in which data in one direction proceeded at a much higher line rate than
data in the other direction.

Enter the 212 modem. It used the di-bit-encoded, quaternary phase shift keying with
coherent detection. It is a synchronous modulation/demodulation scheme used because it
has noise advantages over a four-level AM signal. The 212 has a unique answer tone that
allows for the convenient identification of modem speed when the called end answers. This
modem could be used in asynchronous or synchronous systems, offering the choice of
recovered receive clock (or not). From the 212, modem types data speed progresses until
you reach the V.90 modem. V.92 is not considered a legacy modem yet.

We have not mentioned error-detection and -correction schemes. Operating above 2,400
bps over a wireline presents opportunities for many more errors than at the slower speeds.
Modem manufacturers therefore offer different data compression as well as error-detection
and -correction schemes. Unless one is operating with identical schemes used in both
modems, the error-detection and -correction process will generate errors. Two sets of standardized
schemes are in general use. One scheme is the V.42 international standard, which
in the original and bis (second modification) allows for data compression at two to one
(original) and four to one (bis). The others would be the Microcom Networking Protocols
(MNP), which come in various flavors, the first five having been released to the public
domain. MNP-1 and MNP-2 are generally concerned with the packetizing of data, but MNP-
3 through -5 get serious about a whole range of parameters, including losing the start and
stop bits (as does V.42) for a 20 percent gain in throughput and packetizing the data. V.42
employs the LAP-M frame and uses code lists to represent strings of characters, replacing
the strings with the shorter code lists. MNP-6 through -10 are still in the proprietary
domain, meaning that you have to have modems at both ends that have the appropriate
level of software.

How does a modem know what to do? Part of the modern modem technique is something
called "training" or "negotiation." In this process, the two modems negotiate which
speeds, compression, and error-detection schemes they will use.

Modems are used for frequency translation, and wireline types are only a small segment of
the devices called modems. Others are used in local area networks (LANs) and cable television
(CATV, which stands for Community Antenna Television, where cable has its roots). Still
others are found in satellite and ground communications, just to name a few. Depending
on the medium, modems use varying techniques to transmit information. There is a continual
push to increase transmission speeds. Especially in the wireline arena, the speeds
keep increasing. The highest-speed "standard" modems for wireline at the time this is
written are the V.90 and V.92 at 56 Kbps. However, because of the need for increased
throughput over the wireline medium exists, efforts are being made to go past these speeds
when economically possible. The V.92 standard improves slightly on the V.90 specifications by
adding a method for disconnecting the modem long enough to let you know that someone is
trying to call you without losing the connection, a feature referred to as "Internet call
waiting." Also, the maximum upload speed has been increased from 33.6K to 48K.

Higher-speed wireline modems must use a variety of techniques to coax high speeds from a
3 KHz wireline. One of the techniques used prior to the V.32 modem series (and still used
on hard-to-transmit cases) is multi-tone. In multi-tone, the data stream is divided into eight
(or more) parallel streams. Each stream is di-bit encoded, after which each performs phase
shift modulation on its own carrier (one of the eight or more tones separated by about
twice the tone bandwidth in Hertz). All the tones are used as a simultaneous baseband to a
vestigial sideband AM modulator. At the receive end, the reverse process takes place until
the signal is reassembled. This scheme is the basis for DMT (discrete multi-tone), which is
used in DSL lines.

Most modern high-speed modems now incorporate a "fallback" system in which, if the
error-detection rate rises above a preset level, the speed falls back to a lesser data rate.
Some drop in integrals of standard data rates, some by smaller increments. As the data
error rate improves, the speed is adjusted back up to the desired rate. None of these newer
modems would have been possible without the large advances in microcircuitry. What
would have been literally tons of equipment is now routinely packaged in one integrated
circuit (IC). It's interesting to note that high-speed modems typically start up at a low speed
(1,200 bps). If transmission is not possible at 1,200, there is no use in trying higher speeds.
If transmission is possible, they then negotiate the speed, data compression, and errordetection
scheme the modem at each end will use, arriving at the highest common
denominator of operation. This procedure is required because of all the different types and
speeds of modems and the various error-detection and compression schemes available.