[Below are comments on] two of the documents which Sharon Weinberger 
received as a result of a FOIA request for documents
related to the patent:

    O'Loughlin JP, Loree DL: Method and device for
    implementing the radio frequency hearing effect.
    US patent #6470214, Oct 22, 2002

One document is an Aug. 2001 letter from O'Loughlin to the Judge
Advocate handling the patent case, and the other is a draft of the
patent from 1994 which contains some different information than the
final patent.

The scientific principle underlying the invention, and assumed by the
inventors, is that thermal expansion can convert microwave pulses to
acoustic signals.  As everyone knows from basic science, when things
heat up they tend to expand, and when they cool down they tend to
contract.  Microwave pulses carry energy, and when a pulse is absorbed
by biological tissue this energy is converted to heat.  This causes
the tissue to expand very slightly, and then to contract back when it
cools.  This is basically the standard model for describing and
analyzing the microwave hearing effect (and has been for several
decades).  A single microwave pulse causes a click to be heard.  More
complex signals can be broken down into strings of "clicks" by various
modulation methods.  The amplitude of the induced acoustic wave is (at
least roughly) proportional to the power level of the microwave pulse.

The basic modulation method used in the O'Loughlin patent is an
amplitude modulation.  In a simple amplitude modulation, the signal is
broken up into discrete samples in time and then the power level of
the microwave pulse train is modulated accordingly.  This is "an AM
modulated envelope over a pulsed RF carrier train."  This method
unfortunately introduces nonlinear distortion which cannot be
compensated for by preprocessing.  It works for tones, but not for
complex signals such as speech.  This is theoretically worked out in
the draft patent, which also describes experimental verification that
the simple AM modulation produces barely intelligible speech.  The
experiments were conducted at "the Air Force Phillips Laboratory
during the week of 24 Oct 94, using the AM sampled data modulation
process."  Subjects could recognize the encoded messages, but only if
they were told what the message was.

The patent actually makes use of a different form of AM modulation,
the "AM double sideband suppressed carrier" method, which is linear in
the amplitude and allows for preprocessing of the audio signal.  They
state (and show) that "conventional AM modulation ... is not useful
for the implementation of this invention."

O'Loughlin points out in his letter that since the power of a
microwave signal is proportional to the square its the amplitude, and
since the acoustic signal at the head is roughly proportional to the
microwave signal's power level, the microwave signal's amplitude
should be modulated according to the square root of the original audio
signal's amplitude.  According to O'Loughlin, "This is the basic
essence of the invention."  It is essentially a preprocessing of the
input audio signal.  Other preprocessing is also applied, for example
to decrease the higher frequencies according to a spherical model of a
human head (which shows "a 40dB per decade slope in favor of the
higher frequencies").

In the Washington Post article "Mind Games," Sharon Weinberger writes
that she obtained "records that note that the patent was based on
human experimentation in October 1994 at the Air Force lab, where
scientists were able to transmit phrases into the heads of human
subjects, albeit with marginal intelligibility."  But these were
initial experiments, using the simple AM modulation (which would be
expected to give marginal intelligibility).  Using the actual process
in the patent "will produce an undistorted subjective sound; which is
the invention," according to the draft of the patent.  O'Loughlin's
letter to the Judge Advocate states, "... the fact that when the
signal is processed by the teachings of the invention the signal is
intelligible has also been experimentally demonstrated."

In a calculation in the draft patent, the sound level is calculated
for a single tone sent via the AM balanced modulator method.  A 1GHz
RF carrier is used, with a power level of 100mW/sqcm.  The calculation
yields a sound level of ~50dB, which is said to be at the level of "a
normal male voice at one meter."  The calculations do not take several
loss sources into account, though, and the actual sound level would be
somewhat lower.  It is calculated that the 100mW/sqcm signal could be
applied for ten seconds and stay within the ANSI exposure level at
1GHz.

The suppressed-carrier method only outputs power while a signal is
present, so power levels would be much less of a problem unless
someone were really being blasted by a constant auditory stream.  (As
if any experimenters on nonconsensual subjects would necessarily
follow the ANSI exposure levels while psychologically manipulating
involuntary human subjects.)

Justesen reported in "Microwaves and Behavior," American Psychologist,
March 1975, that Sharp and Grove had successfully encoded speech (the
spoken digits from one to ten) in a pulsed microwave signal.  The
method reported in that article is essentially an FM modulation,
rather than an AM modulation.  That paper states that "the electrical
sine-wave analogs of each word were then processed so that each time a
sine wave crossed zero reference in the negative direction, a brief
pulse of microwave energy was triggered."  That sounds like they are
taking a Fourier decomposition of the speech signal and then using a
frequency modulation on all of the Fourier components, simultaneously.
The results reported in the O'Loughlin patent for AM modulation do not
necessarily apply to the FM modulation methods.

Of course, neither the Justesen article nor the O'Loughlin patent
discuss the psychological sequelae that would result from testing
these devices on nonconsensual subjects and/or applying them against
unwitting citizens.