AT&T PARADYNE MODEM OPTICAL LINE INTERFACE
 ------------------------------------------

We were intrigued by AT&T Paradyne's announcement in last month's issue
regarding their sysop discount program for the DataPort 14.4 modem they are
making available to qualified BBS operators at a price of $206. They alluded
to a new feature called Optical Line Interface. We guessed this was some form
of impedance matching interface, but it was pretty much a guess. We've since
received further information from their engineering staff in Largo Florida
and think it's worth a mention. Technophobes will probably want to move onto
the next story, and kids, don't try this at home.

The concept is not entirely new, but it is interesting in this modem.
Basically, modems are not really connected to telephone lines. For that
matter, telephones aren't either. The two copper wires dangling from the wall
carry a DC voltage that is "modulated" by audio tones. Devices connected to
these wires must comply with certain FCC regulations that basically define
how devices appear to the telephone company equipment and the line. Within
modems, telephones, answering machines, fax machines etc., some provision
must be made to "isolate" the device from the line. Typically, this is an
impedance matching transformer.

Transformers operate on the principle that if you run an alternating current
waveform through a coil of wire, a similar waveform of opposite polarity will
be "induced" in a second coil of wire held in near proximity to it. The two
coils have no physical or electrical connection beyond the inductance "field"
setup by the waveform in the primary coil. In this way, the primary coil with
the dc carrier will "couple" the ac waveform only to the secondary coil -
there is no connection for the dc component and so the device is said to be
"isolated" from the line.

There are two problems with this almost universal method. First, for the
maximum amount of signal "coupling" between the two coils, the vectored sum
of inductance, capacitance, and resistance (impedance) in the transformer
should match exactly the inductance, reactance, and resistance (impedance) of
the telco line circuit. Further, any mismatch not only reduces the amount of
signal that is coupled to the device, it also causes signals to be
"reflected" back up the line. This is often termed near-end echo.

The second problem has to do with the tendency for transformers to introduce
harmonic distortion. More expensive and physically larger transformers
exhibit less of this, but the trend in modems is of course toward very small
devices, and cheaper ones.

At today's higher modem speeds of 14.4 kbps and higher, one of the critical
elements in passing data is echo cancellation. Broadly, this is the ability
of a modem to eliminate its own transmitted signal from interfering with the
signal received from the other modem. Impedance mismatch that causes near-end
and far-end echoes causes some problems here. And harmonic distortion appears
simply as noise that the signal must rise above to be heard.

In a perfect world, transformers actually don't do a bad job of impedance
matching and isolation. The problem is that we all live varying distances
from the local telephone central office switching station. The line impedance
of a loop between the CO and your house varies dramatically with distance. If
you are 1000 feet away, it is much different than if you are 6000 feet
distant. If your modem was designed and tested on a short local loop, it will
suffer a fairly noticeable drop in performance on a long loop. Oddly, the
reverse is also true. If it was designed on a long loop, it can actually
perform worse on a short loop because of the importance of this impedance
matching.

Optical line interface takes a little different approach to isolating the
modem from the line while still coupling the ac waveform. Basically, small
light emitting diodes are modulated by the waveform to produce a light beam
of varying intensity and photosensitive detectors convert this varying light
beam back into an electrical signal. Like the transformer, there is no
physical or electrical connection to the telephone line across the path of
the light beam. Light itself is used to couple the audio waveform to the
modem.

The advantage here is that the components used to create this light bridge
have extremely low levels of capacitance and inductance. It presents an
almost purely resistive load to the line. This basically causes a much less
critical impedance match between the device and the line. Further, there is
virtually no harmonic distortion. The result is a lower noise level and
ostensibly a much better impedance match over a wider range of line
impedances.

Consequently, the AT&T modem will not likely show you much improvement on a
good line. But on noisy lines and particularly in situations where you are
either very close or very far away from the CO, it can perform noticeably
better. And this improvement is not at all dependent on what kind of modem is
on the other end of the line. Modems calling your BBS are just as subject to
the reflected images and noise caused by impedance match on your end as they
are on their own end.

We've fallen seriously out of romance with line simulator test sets. They do
not appear to accurately simulate the cumulative variety of damage done on a
long distance telephone connection. So we've adopted a bit of our test set
consisting of bulletin boards around the world and a known 9dB drop we can
induce through our telephone system at will. A real life opportunity
presented itself the other day when we were unable to link up with Merlin
Systems BBS in Ottawa using three different modems. We would get sort of a
connect at 7200 or 4800 bps, but the noise was so bad we couldn't actually
get characters to make the trip with any reliability. And we wanted some
largish files from this system. "Why not connect the AT&T DataPort and see if
this Optical Line Interface means anything," suggested Earstwhile Assistant.

The DataPort got a 14.4 connect on the second ring with a very brief
handshake. While the difference was dramatic, these WERE bad lines. It
dropped to 9600, 7200, and eventually 4800 bps over the course of the next 10
minutes. But there were no lengthy retrain pauses, even the downshift was
almost unnoticeable. When we actually began downloading, we were down to
about 450 characters per second which held quite well at that level.
Anecdotally, this modem looks like a champ on poor lines. And while it does
use AT&T's own chipset instead of the nearly ubiquitous Rockwell chipset, we
suspect that the Optical Line Interface plays a significant role. For more
information, contact Scott Frazee, AT&T Paradyne, Mailstop LG219, 8545 126th
Ave North, Largo, FL 34649; (800)554-4996 voice; (813)530-8276 international;
(813)530-2398 fax; Internet: s.frazee@ pdnis.paradyne.com