We think that this paper describes a simple and elegant
method to use multiple antennas to transmit data in an unknown
changing wireless channel. The method is relatively easy to 
understand and implement, especially for workers in the field
who are familiar with single-antenna methods that use some of
the same principles. Although we described the general method,
we left many research sub-areas open, and perhaps this has led
to the follow-up work we are seeing by others.
Does
it describe a new discovery or new methodology that's useful to
others?
The paper describes a new methodology that is itself built on
earlier work within the Math Center. We believe that the method
provides a practical way to achieve high data-rates with many
antennas on a time-varying unknown wireless channel.
Can
you give us some background on this research?
The basic idea of using multiple antennas to increase
wireless system capacity and data rates dates to the mid-1980s
with some work by AT&T on how capacity increases when the
number of transmit and receive antennas grow simultaneously. By
the mid-1990s, work within the Math Center and in other parts of
Lucent showed that the capacity growth was substantial and that
a system could be designed to realize a large fraction of this
capacity (this system was called BLAST).
However, all of this work relied on the fact that the
wireless channel was relatively stable and known to the
receiver. Our group initiated a study of what happens when the
channel changes too rapidly to be learned reliably. In the
process, we designed some transmission techniques that could be
used in a completely unknown channel. One of these techniques is
the differential method described in this paper.
Could
you summarize the significance of your paper in layman's terms?
The basic differential method works like this: Suppose we
wish to transmit data on an unknown channel. Since the signal
that we transmit is multiplied by the unknown channel, the
receiver would have difficulty recovering what was transmitted
because of the distortion introduced by the channel. If we
instead first transmit a "training signal" that the
transmitter and receiver have agreed upon, the receiver can use
the training signal to estimate the channel and then look at the
received data signal to recover what was transmitted. The
differential technique combines the training and data
transmission into one operation. That is, information is
transmitted not on any one signal, but on the relative
"difference" between two consecutive signals (hence
the name "differential"). While it was known how to
accomplish such a differential transmission with one antenna, we
provided a framework for how to accomplish it with any number of
antennas.
Bertrand M. Hochwald
Member of Technical Staff
Lucent Technologies
Bell Laboratories
600 Mountain Avenue, Rm. 2C-363
Murray Hill, NJ 07974
