Note: Descriptions are shown in the official language in which they were submitted.
20 983 19
- 1 - -
TITLE OF THE INVENTION:
SIGNAL PROCESSOR FOR RECREATING ORIGINAL AUDIO SIGNALS
BACKGROUND OF THE INVENTION
The present invention relates generally to electronic
circuits for audio signals, and is specifically concerned with
circuits for restoring realism to stereophonic and monaural
audio signals that has been lost during transduction and
recording.
In 1969, R.C. Heyser in "Loudspeaker Phase Characteristics
and Time Delay Distortion: Part 1", J. Audio Eng. Soc., Vol.
17, p. 30, demonstrated that it was possible to obtain not only
the amplitude spectrum of sound from a loudspeaker, but also
the phase spectrum by an in-place measurement using a
microphone. This provided an answer to the question of how
well a speaker recreates the original sound field recorded by
a microphone. In "Loudspeaker Phase Characteristics and Time
Delay Distortion: Part 2", J. Audio Eng. Soc., Vol. 17, p. 130,
Heyser showed that all loudspeakers that are absorptive and
dispersive possess time-delay distortion, due to the fact that
acoustic pressure waves do not effectively emerge from the
transducer immediately when excited. The wave emerges with a
time delay that is a multiple valued function of frequency.
As a result, the sonic image is "smeared" in space behind the
physical loudspeaker, the positions being a function of
frequency. The time delays are measured in
P.
~O 92/10918 2 0 9 8 3 19 PCT/LJS91/09375,...
-2-
terms of the number of milliseconds required for the sound
pressure wave to emerge from the position in space
occupied by the loudspeaker.
Theory indicates that acoustic events can be
described in at least two ways: in a time domain or a
frequency domain, each convertible into the other via a
Fourier transformation. The mathematical formulation for
this process is well known. The time-domain
characterization of an acoustical event is a scalar, while
the frequency-domain representation is a complex vector
quantity containing amplitude and phase information. The
time domain representation can also be expressed as a
complex quantity. The scalar portion of the time domain
vector represents performance based on impulse excitation;
the imaginary part of the vector is the Hilbert transform
of the scalar.
Loudspeakers and electrical networks which transfer
energy from one form to another can be characterized by
response to an impulse function, because the impulse
response can be manipulated to predict the behavior of the
system in response to any arbitrary signal. Fourier
transforms work for predictive systems as well as causal
systems. However, the group velocity of a set of aduio
signals is not related to time delay for all possible
systems, and uniform group delay does not insure a
distortionless system.
Because the time and frequency domains are two ways
of describing the same event, accurate time domain
representation cannot be obtained from limited frequency
domain information. For example, the time delay of a
frequency component passing through a system with
nonuniform response cannot be detenained with accuracy.
However, a joint time-frequency characterization can be
made using first and second order all-pass networks.
3S This is consistent with ordinary human experience. At any
frequency there are multiple arrivals of the audio signal
at the listener's location as a function of time. These
:1~'O 92/10918 PCT/US91/09375
-3- 20 983 19
can be viewed as a collection of perfect loudspeakers at
various positions in space, each producing single-valued
signal arrival times as function of the propagation
velocity of each frequency. With conventional
loudspeakers and electronics, therefore, a burst of energy
from a loudspeaker will create what may be described as a
"spatial" smear.
In 1971, R.C. Heyser in "Determination of Loudspeaker
Signal Arrival Times: Part I", ~. Audio Eng. Soc., Vol.
19, No. 9, pp. 734-742 (October 1971) , showed that it is
not possible to derive a unique time behavior based on
incomplete knowledge from a limited portion of the
frequency response. It is necessary to know the phase
spectrum of the loudspeaker response in order to uniquely
determine its impulse response. The general time domain
vector contains information relating to the magnitude and
partitioning (exchange) of kinetic and potential energy
densities of the loudspeaker signals and is given by the
equation:
h ( t) =f ( t) +ig( t)
where f(t) is a time-dependent disturbance and g(t) is its
Hilbert transform. The exchange ratio of kinetic to
potential energy determines the upper bound for the local
speed of propagation of the signal from a loudspeaker.
Heyser showed that the actual versus virtual distances
(i.e., the spatial smear) can be on the order of 12
inches. For recordings made with more than one microphone,
the time delay distortion inherent in loudspeakers becomes
audible.
Live sound is a transient phenomenon. The only
available way to compare two non-repetitive audio signals
is to record them. For example, this can be done by
monitoring transduction of a loudspeaker output signal via
a microphone, and then recording the information and
comparing it with the electrical signals which drove the
loudspeaker. However, since these latter signals were
n
WO 92/10918 PGT/US91/09375_
20 98 3 19
also recorded from the output of microphones, the
comparison is in reality between two recordings rather
than between a recording and a live sound.
A purely mathematical analysis of a single
loudspeaker reveals that there is a direct tie between the
frequency response and time smear of information (signal)
received by an observer. If the loudspeaker were isolated
in an anechoic chamber, the acoustic response of the
speaker could be replaced by a number of perfect response
]0 loudspeakers, each positioned in its own frequency-
dependent location in space behind the physical position
of the original imperfect loudspeaker. Time delay
distortion is caused by this multiplicity of delayed
equivalent sources. Experimental evidence has provided
verification of this analysis.
The individual time-frequency components of an audio
signal, predicted mathematically, overlap in the time and
frequency domains. Therefore, a graphical presentation is
not possible, because it is impossible to separate
2o simultaneous arrival times in a single time domain plot.
Potential energy (i.e., pressure expressed in dB) and
comparisons of input to output signals directly (a measure
of distortion) do not completely describe the performance
of audio equipment quality such as loudspeakers,
microphone, and electrical networks. Total sound energy
provides phase distortion information and, although phase
is not detectable consciously for simple signals, there
are indications that the human hearing mechanism is
capable of processing complex functions and perceiving
phase information as part of total sound perception.
The square root of the total energy density vector,
E_, is equal to the sum of the square root of the potential
energy vector and the imaginary component of the square
root of the kinetic energy vector:
WO 92/10918 PCT/US91/09375
-5-
20 983 19
=~+i~
Attempts to measure the total energy density at a
microphone responding to a remote sound source will only
yield part of the total energy density of the source.
Thus, at any given moment, a microphone will not directly
measure E. Essentially, a microphone compresses a complex
spatial, multi-dimensional acoustic signals into a single
point in time and space, effectively making the signal
two-dimensional as a function of time. However, the
1o information necessary to unravel the entire original
signal is contained in the compressed signal and can be
retrieved if processed property.
Although the threshold of hearing has been
established in terms of vector orientation and frequency
~5 of pure tones (see, e.g., L. Sivian and S. White, "On
Minimum Audible Sound Fields, ~ Acoust Soc Am , Vol. 4,
pp. 288-321 (1933)), pure tones have no Fourier
transforms. The human hearing mechanism processes total
energy density, not just the "minimum audible pressure"
20 associated with a pure audio tone.
The ability to localize direction and distance from
a sound source has something to do with the orientation of
the ear with respect to the vector components of sound.
For pure tones, simply the phase differences between
25 arrival of the signal at the two ears provides a clue to
the direction of the source. See Kinsler and Frey,
Fundamentals of Acoustics (New York: John Wiley and Sons,
1950), pp. 370-392. Thus, the minimum audible field for
binaural hearing varies with amplitude, frequency, and
30 azimuth relative to the source signal.
J. Zwislocki and R. Feldman (1956) "Just Noticeable
Differences in Dichotic Phase", J Acoust Soc Am , Vol.
28, No. 5, p. 860 (Spetember 1956) pointed out that the
ears may not be able to detect phase or time differences
35 above 1300 Hertz and the only directional clues above 1300
Hz are contained in relative intensity differences at the
n
WO 92/10918 PCT/US91/09375
2 0 9 8 3 19 _6_
ears. Because Zwislocki and Feldman also state that the
same intensity difference can occur at different azimuths,
there is little directional information in the higher
frequencies that provide accurate positioning. However,
as noted above, the hearing mechanism is not limited to
pressure stimulation alone.
In reality, the human auditory system binaurally
localizes sounds in complex, spherical, three dimensional
space using two sensors (ears) that are unlike
l0 microphones, a computer (brain) that is unlike any
computer constructed by man, and, at a live performance,
the eyes. The eyes allow us to "hear" direction by
providing a sensory adjunct to the ears for localization
of sound in azimuth, distance and height. During
reconstruction of a familiar sound, such as a symphony
orchestra, the brain remembers instrument placement and
correlates this information with auditory clues to provide
a more complete sense of the individual orchestra sections
and sometimes of the locations of individual instruments.
Techniques for localizing sound direction by the ears,
neural pathways, and the brain have been termed
"psychoacoustics". U.S. Patent 4,817,149, to Myers,
points out that the brains of all humans look for the same
set of clues (normally provided by the ear structure)
concerning the direction of sound even if ear structures
differ from person to person.
In addition to direction, the brain will interpret
distance as a function of intensity and time of arrival
differences. These clues can be provided by reflected
3o sound in a closed environment such as a concert hall, or
by other means for sound originating in environments where
no reflections occur, such as in a large open field. In
a closed environment, there is a damping effect as a
function of frequency due to reverberations. When
3f acoustic energy is reflected from a surface, a portion of
the energy is lost in the form of heat. Low frequencies
tend to lose less energy and are transmitted more readily,
_... _ . T
~0 92/10918 PCT/US91/09375
2098319
whereas high frequencies tend to be absorbed more quickly.
This makes the decay time of high frequencies shorter than
that of low frequencies. The air itself absorbs all
frequencies, with greater absorbtion occurring at high
frequencies.
In "Biophysical Basis of Sound Communication" by A.
Michelsen (in B. Lewis (ed.), Bioacoustics. A Comparative
Approach (London: Academic Press, 1983)), at pages 21-22,
the absorption of sound in air is described as a
combination of dissapation due to heat and other factors
not well understood. In air, the absorbtion coefficient
in dB/100 meters is 1 at about 2 I~iz. At about 9 IQiz, the
signal is down by 10 dB: at 20 HI~iz it is down by 100 dB;
and at 100 KHz (the upper harmonics of a cymbal crash), it
is down by about 1000 dB. Thus, higher harmonics
generated by musical instruments are drastically
attenuated (in a logarithmic fashion) by even a distance
of a few feet when traveling to microphones, and then even
more when traveling from speakers to the listener's ears.
With conventional stereophonic sound reproduction
systems, it is necessary to be equidistant from the
speakers in order to experience the proper stereo effect.
With earphones, standard stereo provides a strange ping-
pong effect coupled with an elevated "center stage" in the
middle and slightly above the head. At best, ordinary
stereo is an attempt to spread sound out for increased
realism, but it is still basically two-dimensional.
In the 1920s Sir Oliver Lodge tested human hearing
range out to 100 I~iz . It has been suggested that the true
range of human hearing is not completely known. However,
the outer ear, inner ear (cochlea), auditory nerve, and
human brain are capable of detecting, routing, and
processing frequencies in excess of 100 IQiz, and possibly
to 300 IQiz and beyond. However, conscious hearing is
limited by the brain to roughly 20 Hz to 20 IQ~iz.
There is no currently accepted theory of how humans
actually hear outside the voice range of acoustic signals.
2 0 9 8 3 19 ~~/U~9 ~ /fl9 3 75
-8- , IPFAIUS 2 2 JAN 1993
Below about 200 Hz, the wavelength of an acoustic pressure
wave is too large to enter the ear canal. Experience with
low frequency standing waves suggests an interaction with
the cochlea or auditory nerve directly. Indeed, standing
wave acoustic emitters produce the perception of
distortion-free sound throughout the hearing range. Above
about 6 Hz, the "volley" theory and active cochlear
processes could account for an increase in hearing range
beyond 20 KHz. The volley theory is derived from the fact
that there is not a single stimulus-response event per
nerve: rather, higher frequency stimulation results in a
multiplicity of neural firings. The process is one of
bifurcation wherein the higher frequencies cause a greater
number of neurons to fire. This suggests the possibility
of fractal pattern generation. How the brain interprets
the volley of information presented to it is unknown,
however.
In Auditory Function, edited by G. Edleman, W. Gall,
and W. Cowan, (New York: John Wiley & Sons, 1986), a class
20 of experiments is described which demonstrate acoustic
emissions from animal and human ears. The cochlea can
function as a generator of acoustic signals which can
combine with incoming signals to produce higher
frequencies. Both empirical and theoretical studies
25 (indicating that active cochlea processes are necessary
for basilar membrane tuning properties) support the
concept.
P. Zurek, in "Acoustic Emissions from the Ear - A
Summary of Results from Humans and Animals", J. Acoust.
30 Soc. Am., Vol. 78, No. 1, pp. 340-344 (July 1985),
indicates that frequency selectivity results from active
cochlear processes. When the ear is presented with a non-
linear pulse, in addition to the stimulus response
mechanism, another response with an 8 millisecond (or
35 longer) delay is produced. This phase-shifted signal,
generated by the,ear, may play a role in the actual way in
which we hear music and other high frequency sounds. When
~U&~Tk~IfTE SHcET
1PEA/US
2 0 9 8 3 19 ' p~~ICJ~ 9 i / 0 9 3 l5
_g_
~PEAiUS 2 2 JAN X993
musical instruments produce sound, the various Fourier
waveforms are not simply produced independently of each
other, but exist in a phase space wherein there are phase
interactions among all of the sounds. Even a single
string plucked on a harp.or struck on a piano will produce
phase-related signals and harmonics, not simply
frequencies and amplitudes. Thus, the ear must be capable
of decoding phase information in order to properly
transduce complex sounds such as music.
The evoked time-delayed response in the ear is not
simply a phase-shifted replica of the original sound,
because the higher frequency components are time delayed
less (about 8-10 milliseconds) than the lower frequency
components of the emission (about 10-15 milliseconds).
Also, the amplitude of the evoked response is non-linear
with respect to the stimulus for high stimulus levels,
amounting to about 1 dB for every 3 dB increase in the
stimulus. The interaction of the stimulus and acoustic
emission occurs increasingly with lower and lower levels
Zp of input, suggesting that the ear may have a compensation
mechanism for low level signals. People with certain
types of hearing loss do not product acoustic cmissions.
At low levels of auditory stimulus, the emissions are
almost equal in amplitude to the incoming signal itself,
and they occur even for pure tones. The ear can generate
continuous signals, and generated signals as high as 8 KHz
have been observed.
As noted earlier, the conscious hearing range is
roughly between 20 Hz and 20 KHz. Audio equipment has
3p been designed to be optimal within that range. Also, most
equipment has been designed to accurately reproduce that
which has been transduced and recorded. However, live
sound' is a transient phenomenon. It is not possible to
compare a live sound with anything, because in order to do
so, it must be transduced and recorded in some way. It is
this fact that forms the motivation for the present
invention, and discussion of the prior art that follows.
iPcA~'JS
~° WO 92/10918 2 0 9 8 3 19 pCT/LJS91/09375
-10-
The previous discussion of auditory perception can be
summarized as follows:
(a) There are at least six mechanisms involved in
the interpretation of the intensity and direction of
sound, including:
(1) Relative phase (for low frequencies).
(2) Relative intensity (for midrange
frequencies).
(3) Relative time of arrival (for high
frequencies) .
(4) Angle of the sound vector relative to the
ears.
(5) Total energy density of the sound signal.
(6) Active and passive cochlear processes.
~ 5 (b) Arrival time at a listener for a signal emanated
by a loudspeaker is a function of the total energy vector
and can cause certain frequencies to appear to be located
behind the speaker in a spatial smear.
(c) It is not possible to fully describe the time
behavior of an audio reproduction system from limited
frequency (bandwidth) information.
(d) The quality of audio recording and playback
systems is not completely described by input versus output
distortion and potential energy variations.
(e) The human sound detection, transmission, and
analysis system is capable of sensing phase distortion, of
sensing frequencies far beyond 20 IQiz, and of determining
relative distance and direction of sounds.
(g) Live sound is a transient phenomenon which
cannot be directly compared with transduced and recorded
sound.
The prior art has addressed only some of the problems
that are encountered in seeking to reproduce sound that
more nearly resembles what the sound was like when it was
live. The following patents and commercial audio systems
are illustrative:
WO 92/10918 PCT/US91/09375
20 983 19 ~-
-11-
U.S. Patent No. 4,100,371 to Bayliff, describes a
two-loudspeaker system with phase difference compensation.
Two sources of phase distortion were addressed. One is
inherent in crossover networks and provides 90-degree
shifts. The other is due to the fact that high and low
frequencies emanate from different parts of the speaker
cones and can cause 180-degree shifts at 3,000 Hertz. The
proposed solution consisted of an active filter and
speaker cabinet construction techniques.
U.S. Patent No. 4,875,017 to Sakazaki, describes a
digital technique for shifting phase of a video signal.
U.S. Patent 4,626,796 to Elder, describes a digital
technique for shifting audio tones in a cellular
telephone.
U.S. Patent No. 4,727,586 to Johnson, describes a
technique for correcting phase distortion due to
differences in frequency propagation times by placing
speakers at different locations in their an enclosure.
This is similar to the commercial BSR dbx "Soundfield V"
system which uses biamplification instead of crossover
networks and positions the high-frequency speakers behind
the low-frequency driver and at an angle in the enclosure.
U.S. Patent No. 4,866,774 to Klayman, describes a
servomechanism to drive potentiometers in a device that
provides sums and differences of right and left channels
of stereo for directivity enhancement of the stereo image.
U.S. Patent No. 4,218,585 to Carver, describes a
three-dimensional sound processor that improves stereo by
adding and subtracting inverted portions of the right and
3p left channels via equalizers, passing them through a time
delay, and mixing portions of the left channel to the
right and vice-versa. Switches control the amount of
dimensional effect.
U.S. Patent No. 4,769,848 to Eberbach, describes a
complex design for a passive delay network inserted into
the high frequency signal path of a crossover network in
a loudspeaker.
i
WO 92/10918 PGT/US91/09375
2 0 9 8 3 19 -~ -12-
Correction of phase distortion caused by air path
differences has been addressed in U.S. Patent Nos.
3,824,343, 3,927,261, and 4,015,089, wherein various
schemes are used to physically provide phase compensation.
U.S. Patent No. 4,151,369 discloses an active network
for four speakers in a "surround sound" configuration.
Japanese Patent Publication No. 54-13321 discloses
the combination of a phase delay circuit in the low
frequency filter and phase-correcting speaker placement in
the enclosure, while Japanese Patent Publication No. 52-
33517 discloses a phase delay circuit in the high
frequency filter circuit.
U.S. Patent No. 4,567,607 to Bruney et al., discloses
a method and apparatus for improving the accuracy of
psychoacoustic images by cross-feeding signals from one
channel to the other in an out-of-phase relationship in
the 1-5 KHz frequency range, and by increasing the gain
for frequencies in the 100-1000 Hertz range (except for
signals in the 200-900 Hertz range that are applied to
2o both channels). The circuit also inverts the phase of low
frequencies.
U.S. Patent No. 4,817,149 to Myers, discloses a very
complex three-dimensional sound apparatus that imparts
front-to-back and elevation cues by selectively boosting
and attenuating certain frequencies, azimuth cues by time-
delaying part of the signal, and various other cues by
adding reverberation.
U.S. Patent No. 4,748,669 to Klayman, discloses a
system that is extremely complex for creating three
dimensional sound by generating sum and difference signals
from the left and right stereo channels, selectively
altering the amplitude of signal components, and
recombining the altered signals with the left and right
channels to provide direction information for the
listener.
Quadraphonic and so-called "surround sound" systems
involve four or more speakers. These and other attempts
~'O 92/10918 PGT/US91/09375
2098319
to correct the problems involved in transducing,
recording, and playing back sound have created
environmental ambience, but have fallen short in re-
creating the original sound.
The "QSound" system developed by Archer
Communications of Calgary, Alberta, provides three-
dimensional sound but does not improve the basic quality
of the sound. Also, it is a relatively complex system
that uses computers for sound mixing.
The "Natural Sound Digital Sound Field Processor"
developed by Yamaha Corporation of Hamamatso, Japan,
attempts to recreate the ambience of several types of
spaces, such as concert halls, movie theaters, and so on,
by adding reverberation, echo, and presence. The signal
processing is digital, rather than analog, and hence the
processor is relatively complex.
The "Dolby Surround Pro Logic" system developed by
Dolby Laboratories claims to recreate a live performance
by using four speakers to provide "reflected" sound.
2p The listener is placed in the center of the sound to
enhance the dimensionality of video tape sound tracks.
Although the prior art discussed above has attempted
to correct some of the problems associated with distortion
in audio systems due to phase shifts as a function of
frequency, and spatial distortion due to the inherent
inaccuracies in standard stereo, these attempts have not
completely succeeded in restoring lost realism to recorded
sound. At best, some prior art processors create the
illusion of ambience.
3o The prior art provides single and, in some cases,
multiple corrections to recorded signals. The object of
the prior art is, in general, to control the location of
sound cues and provide phase correction, not to increase
the quality of the sound by putting back in to the signal
what was removed by the transduction, recording, and
playback systems.
20 983 19 -=--
~e~IUS ~ ~ i 0 9 3 7 5
I P~~/~IS 2 2 ~ A I~ 193
As previously pointed out, microphones compress
signals that can consist of many fundamental frequencies
from different instruments at different spatial locations.
These signals also contain complex interactions of the
fundamentals and harmonics produced by the same
instruments. When cymbals crash, for example, the
harmonics produced reach above 100,000 Hertz. As the
complex signal develops from these interactions, it can
become non-linear and sub-harmonics will be present.
l0 At first, it would appear impossible to retrieve or
reconstruct a complex signal whose spectral content has
been compressed by microphones in both the time and
spatial domains. The digital sampling rate of information
that is recorded on compact discs and digital audio tapes,
for example, results not only in a loss of information,
but also in an absolute frequency cutoff that is lower
than the upper harmonics produced by some musical
instruments. The present invention arises from the
recognition that, if the harmonics and subharmonics of
recorded sound are allowed to develop from the fundamental
frequencies, and if the spectral content of the signal is
spatially separated, the original live sound can be
recreated.
SUMMARY OF THE INVENTION:
The present invention overcomes the shortcomings of
the prior art by causing harmonics and sub-harmonics to
develop for all frequencies, by continuously correcting
the phase of the signal logarithmically as a function of
frequency, by spatially separating the spectral content of
the signal, and by dramatically increasing the audio
bandwidth of the signal. Preferably, the audio bandwidth
is expanded beyond the normal range of human hearing (20-
20,000 Hz) to a range of 0 Hz to more than 2 MHZ. The
invention is based, in part, on the recognition that the
human hearing mechanism for sensing audio signals (as
SU~S'I11U7E SHEET
IPEA/US
PC~'/~s~ 9 ~ ~ p q 3 75
20 983 19
15 _ If~F.~II,~S 2 2 Jq N
193
opposed to electromagnetic and tactile signals) is
different from the electronic circuits used to construct
amplifiers, microphones, tape recorders, and other types
of audio equipment. Thus, when humans hear or sense an
audio signal, it is processed differently than standard
apparatus attempting to transduce, record, and playback
the original sound.
In accordance with one aspect of the present
invention, an audio signal processor comprises an input
terminal for receiving an audio signal, first and second
processing stages for processing the audio signal, and an
output terminal for coupling the processed audio signal to
an output device. The first and second signal processing
stages are arranged in a series or cascade configuration,
and each stage functions to generate harmonic frequencies
related to fundamental frequencies in the audio signal and
to phase shift the fundamental and harmonic frequencies as
a function of frequency. The phase shift increases in a
negative direction with increasing frequency, so that
higher frequency signals lag the lower frequency
signals.
In accordance with another aspect of the present
invention, a method for carrying out two-stage processing
of an input audio signal comprises the steps of generating
harmonic frequencies related to fundamental frequencies in
the input audio signal and adding the harmonic frequencies
to the input audio signal: phase shifting the fundamental
and harmonic frequencies as a function of frequency, with
the phase shift increasing in a negative direction with
3o increasing frequency so as to cause higher frequency
signals to lag lower frequency signals; generating
harmonic frequencies related to fundamental frequencies in
the partially processed audio signal resulting from the
preceding method steps, and adding the harmonic
frequencies to the partially processed audio signal; and
phase shifting the fundamental and harmonic frequencies of
the partially processed audio signal as a function of
S~BsrrttJTE Si-iEET
y~us
WO 92/10918 2 ~ 9 8 3 1 9 PGT/iJS91/09375
-16-
frequency, with the phase shift increasing in a negative
direction with increasing frequency so as to cause higher
frequency signals to lag lower frequency signals.
The present invention can be implemented by means of
a relatively simple electrical circuit that can be
manufactured and sold at very low cost. The principal
components of the circuit can, if desired, be reduced to
a single dual inline package (DIP) which can be
incorporated into existing types of audio equipment. The
to invention can be utilized with nearly all existing types
of power amplifiers, stereo tuners, and phonographs with
preamplifiers, as well as with compact disk (CD) players,
digital audio tape (DAT) players, and conventional analog
tape recorders and players. All recorded media can be
reproduced with a sound that is very close to that of a
live performance, and the spatial smear that is perceived
with standard loudspeakers is reduced. The invention can
be used with any number of audio channels or speakers; the
resulting sound will be dimensionalized, to some extent,
with even a single speaker. The signal processing that is
carried out by the present invention transfers to tape and
to virtually any other type of recording medium. Thus,
for example, a digital CD output can be processed using
the present invention, and the result can be recorded on
ordinary stereo audio tape. The present invention
restores information that has been lost during digital or
analog processing, as well as during the transduction of
the original sound, and may be employed at a radio or
television broadcasting station to improve the quality of
the audio signal received by the listeners.
Further objectives, advantages and novel features of
the invention will become apparent from the detailed
description which follows.
CVO 92/10918 PCT/US91/09375
_ 20 983 19
BRIEF DESCRIPTION OF THE DRAWINGS:
The details of several preferred embodiments of the
present invention are illustrated in the appended
drawings, in which:
Fig. 1 is a functional block diagram illustrating the
operation of an audio signal processor constructed in
accordance with the principles of the present invention;
Fig. 2 is a graph illustrating the manner in which
the audio signal processor carries out phase shifting and
spatial separation of an audio signal as a function of
frequency:
Fig. 3 is a detailed schematic diagram of a first
embodiment of an audio signal processor constructed in
accordance with the present invention:
Fig. 4 is a detailed schematic diagram of one of the
operational amplifiers used in the signal processor
circuit of Fig. 3:
Fig. 5A is a graph illustrating the phase shift
characteristics of the operational amplifier shown in Fig.
4 ;
Fig. 5B is a graph illustrating the variation in
phase shift with frequency that is provided by the phase
shift stages used in the audio signal processor of Fig. 3:
Fig. 6 is a graph illustrating the bandwidth and
frequency response characteristics of the signal processor
circuit shown in Fig. 3, based on actual measurements
thereof;
Fig. 7 illustrates a second embodiment of an audio
signal processor in accordance with the present invention,
intended for use with an audio system of the type
typically found in an automobile:
Fig. 8 illustrates a third embodiment of an audio
signal processor in accordance with the present invention,
intended for use in conjuction with a home audio. system:
i
WO 92/10918 PCT/US91/09375
~ 0 9 8 3 19 ~ -18-
Fig. 9 illustrates a third embodiment of an audio
signal processor in accordance with the present invention,
also for use with a home audio system:
Fig. 10 illustrates a fourth embodiment of an audio
signal processor in accordance with the present invention,
intended for use in recording studios, broadcast stations
and other professional environments;
Figs. 11A-11D illustrate various methods for
connecting the audio signal processor of the present
invention to an existing audio system in a home or
automobile:
Figs. 12A and 12B depict oscilloscope traces
generated by the audio signals at the input and output,
respectively, of the audio signal processor of the present
invention;
Fig. 13A and 13B depict x-y oscilloscope traces
comparing the input and output of the audio signal
processor of the present invention; and
Figs. 14A and 14B depict spectrum analyzer traces
illustrating the harmonic content restored to an audio
signal by the signal processor of the present invention.
Throughout the drawings, like reference numerals
should be understood to refer to like parts and
components, unless otherwise indicated.
DESCRIPTION OF THE PREFERRED EMBODIMENTS:
The basic functions carried out by the audio signal
processor of the present invention are illustrated in the
functional block diagram of Fig. 1. At the input 16, an
audio input signal is received from an audio source which
will typically consist of a playback device such as a
phonograph, tape player or compact disk player. However,
the audio source may also consist of a receiving device
such as an audio or video tuner, or a live output from a
microphone or musical instrument transducer. The audio
input on line 16 is applied first to a harmonic generator
1,y0 92/10918
PC'T/US91/09375
_19_ 2 0 g 8 3 19 J
10, which enhances the signal by adding harmonics and
subharmonics of the fundamental frequencies found in the
original audio performance. Typically, these harmonics
and subharmonics have been lost during the recording
and/or transducing process. As will be described in some
detail hereinafter, the harmonic generator 10 is
preferably implemented by means of one or more operational
amplifiers. The output of the harmonic generator 10 is
applied to the input of a phase shift and spatial
separation network 12. The function of the network 12 is
to automatically shift all input frequencies in phase as
a logarithmic function of frequency. Thus, all of the
audio frequencies appearing at the output of the harmonic
generator 10 are phase-shifted, including the newly
generated harmonic and subharmonic frequencies. The
output of the network 12 is applied to the input of a left
and right channel comparator and mixer 14. The function
of this portion of the audio signal processor is to mix
portions of the right and left channels of a stereophonic
audio program, in order to product a dimensional expansion
of the stereo image which includes, for example, depth and
height cues. Therefore, the output signal which appears
on line 18 differs from the input signal at 16 by virtue
of having additional harmonic and subharmonic content, by
virtue of having phase shifts among various frequency
components of the composite audio signal as a logarithmic
function of frequency, and by virtue of having the right
and left channel information mixed to some extent to
produce dimensional expansion of the sound.
It should be understood, of course, that the input
and output lines 16, 18 each represent two distinct lines
(i.e., left and right channel inputs and outputs) in
stereophonic embodiments. In monaural applications, the
same input signal may be applied to the two input lines
3S~ 16. It should also be understood that, in actual hardware
implementations of the present invention, the functions
represented by the blocks 10-14 in Fig. 1 may, at least to
a i
WO 92/10918 PGT/US91/09375_
2 p g g 3 1 g . -2 °-
some extent, be carried out simultaneously by the same
components or groups of components, rather than occurring
separately and sequentially as Fig. 1 might suggest.
Fig. 2 is a graph illustrating in a general way the
manner in which the phase shift and spatial separation
network 12 of Fig. 1 operates. As harmonics are generated
by the harmonic generator 10, they are automatically
shifted in phase by the network 12 as a function of
frequency. This has the effect of creating an increasing
time delay for the higher frequencies, as illustrated in
Fig. 2. Because higher frequency sounds move faster
through the air than lower frequency sounds, the function
of the network 12 in Fig. 1 is to correct for the "spatial
smearing" created by microphones and loudspeakers. At the
same time, the harmonic generator 10 adds back into the
signal harmonic information that was lost due to the
frequency limitations of the recording equipment, the time
and frequency domain compression caused by the microphones
used during the recording process, and the high frequency
cut-off inherent in any digital equipment used during
recording and playback.
Without limiting the invention to any particular
theory of operation, it is possible that the spatial
separation of different frequency components of the
incoming two-dimensional compressed audio signal into
time-delayed harmonics and fundamentals allows the human
brain more time to process the received information, and
to interpret the resulting audio signal as more closely
resembling a live performance (in which the fundamentals
and harmonics are, in fact, separated spatially). Also,
the possibility exists that the outer ear, inner ear,
neural pathway, and brain process signals well beyond
100 KHz, possibly into the gigahertz range. It is well
known that microwave frequencies and ultraviolet and
infrared frequencies are radiated from living :cells at
times, and that a small percentage of the population can
._...~..~____...._ .__._. ___..__._ .._r~_
WO 92/10918 PCT/US91/09375
-21- 20 983 19 =~.
"hear" microwave frequencies through auditory or cerebral
processes that are not completely understood.
Fig. 3 is a detailed schematic diagram of a first
embodiment of an audio signal processor constructed in
accordance with the present invention. The audio signal
processor comprises a total of six operational amplifiers,
three for processing the left channel of the audio signal
input and three for processing the right channel. Various
resistors, capacitors and switches are also provided, as
will be described. Referring first to the left channel
processing circuit, an input signal from the audio source
is applied to the non-inverting input of an operational
amplifier 20. The output of the operational amplifier 20
is fed back to its inverting input, so that the
operational amplifier 20 functions as a unity gain stage.
As such, the operational amplifier 20 serves primarily as
a buffer stage to reduce input noise, although in practice
some harmonic content may be added to the audio signal in
this stage.
The output of the operational amplifier 20 is
connected through a first resistor 22 to the non-inverting
input of a second operational amplifier 30, and through a
second resistor 24 to the inverting input of the
operational amplifier 30. The node between the resistor
22 and the inverting input of the operational amplifier 30
is connected to ground through a capacitor 26. The output
of the operational amplifier 30 is connected to its
inverting input through a feedback resistor 28. The value
of the feedback resistor 28 is adjusted so that the
overall gain of the operational amplifier stage 30 is
approximately unity.
The operational amplifier 30, resistors 22, 24 and
28, and capacitor 26 constitute a phase shift circuit 21
which carries out the functions represented by block 12 in
Fig. 1. In other words, the individual frequency
components of the audio signal at the output of the
operational amplifier 20 are phase-shifted as a
WO 92/10918 PCT/US91/09375
2 0 9 8 3 19 ' -22-
logarithmic function of frequency, with the phase shift
occurring in a negative direction and being greater for
higher frequency signals. In effect, therefore, the
higher-frequency signals are time-delayed with respect to
the lower frequency signals, as illustrated in Fig. 2, and
this serves to reduce the spatial smearing effect
discussed previously.
In addition to the phase-shifting function, the
operational amplifier stage 30 also restores harmonic
content to the audio signal. This is believed to result
from two separate effects. First, because there is a
phase shift between the signals applied to the inverting
and non-inverting inputs of the operational amplifier 30,
the amplifier is effectively comparing a phase-shifted
version of an audio signal to the original signal. This
will inherently result in the generation of harmonic
frequency components. The second effect has to do with
the interconnections between the left and right channels
of the audio signal processor, as will be described
shortly.
With continued reference to Fig. 3, the output of the
operational amplifier 30 is connected to the non-inverting
input of a third operational amplifier 32. The output of
the operational amplifier 32 is connected directly to its
inverting input, as shown. Thus, the operational
amplifier 32 serves as a unity gain stage, similar to the
initial operational amplifier stage 20. As such, it
functions primarily to reduce noise in the audio signal.
The output of the operational amplifier 32 constitutes the
left channel output of the audio signal processor, as
shown.
The right channel of the audio signal processor of
Fig. 3 is essentially identical to the left channel that
has already been described. The right channel input from
the audio source is applied to the non-inverting input of
an operational amplifier 36, which is connected in a
unity-gain configuration as shown. The output of the
._.~ T _ ~__._.
:~O 92/10918 PGT/US91/09375
2098319 :;
-23-
operational amplifier 36 at node 37 is connected to the
non-inverting input of a second operational amplifier 46
through a resistor 38, and to the inverting input of the
amplifier 46 through a resistor 42. A feedback resistor
44 connects the output of the operational amplifier 46 to
its inverting input, and a capacitor 40 is connected
between the non-inverting input of the amplifier 46 and
ground. The circuit 31 comprising the operational
amplifier 46, resistors 38, 42 and 44, and capacitor 40
l0 performs phase shifting and harmonic generating functions
similar to those performed by the circuit 21 of the right
channel, as described above. The output of the
operational amplifier 46 is connected to a third
operational amplifier 48, which is connected in a unity-
gain configuration as shown. The operational amplifier 48
is provided primarily for noise reduction. The output of
the operational amplifier 48 constitutes the right channel
output of the audio signal processor of Fig.3.
As noted previously in connection with block 14 of
Fig. 1, some degree of mixing between the left and right
channels is desirable in order to increase the
dimensionality of the sound produced by the audio signal
processor. In Fig. 3, the mixing is achieved by
connecting the non-inverting input of the operational
amplifier 30 in the left channel to the inverting input of
the operational amplifier 36 in the right channel through
a switch 34. In addition, the output of the operational
amplifier 20 of the left channel is connected to the non-
inverting input of the operational amplifier 46 of the
right channel through a second switch 35. The switches 34
and 35 may be combined into a double-pole, single-throw
switch as indicated in Fig. 3. When the switches 34, 35
are in the open position, no mixing occurs between the
left and right channels of the audio signal processor.
When the switch 35 is in the closed position, however, the
output of the operational amplifier 20 is applied to the
non-inverting input of the operational amplifier 40. This
WO 92/10918 PGT/US91/0937~.
2 4 9 8 3 19 - -24-
effectively feeds the left channel signal into the right
channel. Thus, the circuit 31 also serves as a comparator
network for comparing the left and right channel signals
and for removing part of the right channel information
which is also present in the left channel. In a similar
manner, the closing of the switch 34 causes the output of
the operational amplifier 36 in the right channel to be
applied to the non-inverting input of the operational
amplifier 30. As a result, the right channel signal is
fed into the left channel, and the circuit 21 functions as
a comparator for removing part of the audio information
from the left channel that is also present in the right
channel. Accordingly, when the switches 34 and 35 are
closed, the spatial separation between the right and left
15 channel outputs is enhanced. Monaural signals are
unaffected by the comparators 21 and 31, since the audio
signals in both channels are identical.
As noted briefly above, the interconnections between
the left and right channels in the audio signal processor
20 of Fig. 3 enhance the generation of harmonic and
subharmonic frequencies by the operational amplifier
stages 30 and 46. For example, by adding the left channel
signal to, the right channel signal after the right channel
signal has been phase-shifted by the circuit 31 and
25 attenuated by the resistor 38, a new left channel signal
is created that is rich in harmonics. The harmonics
result from the addition, at the non-inverting input of
the operational amplifier 46, of the signals present in
the non-phase-shifted left channel with the signals
30 present in the phase-shifted right channel. The new
signal is no longer coherent in phase with respect to the
left or right channels, unless both channels contain the
same information. In a similar manner, a new right
channel signal with added harmonics is created by adding
35 the right channel signal to the left channel signal at the
non-inverting input of the operational amplifier 30.
_.~~___ .
~,VO 92/10918
PCT/US91/09375
-25- _ 2 0 9 8 3 1 9
Preferred values for the components used in the audio
signal processor of Fig. 3 are provided in Table 1 below.
These values are given merely by way of example and are
not intended to limit the scope of the present invention.
TABLE 1
Component Value or TvDe
Resistors 22, 38 100 K11, ~ W
Resistors 24, 28, 42, 44 10 Kfl, ~ W
Capacitors 26, 40 0.1 ~,F Mylar
Operational amplifiers National Semiconductor
20, 30, 32, 36, 46, 48 LF147/LF347 or
equivalent
Switches 34, 35 (2) SPST or (1) DPST
Fig. 4 is a detailed schematic diagram of one of the
operational amplifiers 20, 30, 32, 36, 46 and 48. All of
the operational amplifiers are identical, with a frequency
bandwidth of 0 Hz to more than 4 MHz and a fast slew rate
(greater than 12 volts per microsecond). Each operational
amplifier is provided with the required plus and minus DC
supply voltages by a suitable power supply circuit (not
shown in Figs. 3 and 4). The specific type of operational
amplifier illustrated in Fig. 4 is a National
Semiconductor LF147/LF347 wide-bandwidth JFET input
operational amplifier, although other commercially
available types of operational amplifiers can be used if
desired. Examples include the Analog Devices AD713JN quad
operational amplifier, and similar types of operational
amplifiers that are manufactured by Motorola and Texas
Instruments. Low noise, JFET input operational amplifiers
are particularly preferred for use in connection with the
n
WO 92/10918 PCT/US91/09375
2 0 9 8 3 19 -26-
present invention, although any operational amplifier with
a high slew rate can be used.
Fig. 5A is a graph illustrating the phase shift
characteristic of the National Semiconductor LF147/LF347
operational amplifier depicted in Fig. 4. The graph is
generic, inasmuch as various resistive and capacitive
loads will alter the degree of phase shift as a function
of frequency. In the case of the phase shift stages 21,
31 shown in Fig. 3, the variation in phase shift with
l0 frequency is illustrated in Fig. 5B. In the general case,
the resistors 22 and 38 may be referred to as R1, resistors
24 and 42 as RZ, and feedback resistors 28 and 44 as R3.
The capacitors 26 and 40 may be referred to as C.
Assuming that the values indicated in Table 1 are used,
the gain of each of the phase shift circuits 21 and 31,
where R3 = R2 = Rl/10 and R1C = r = 10, is unity. The
transfer function G(s) for the circuits 21 and 31 is:
G(s) - 1 - '~ ~ s -_ 1 - lOs
1 + z ~ s 1 + 10s
where s is the Laplacian operator. The phase for
frequencies higher than 100f in Fig. 5B is retarded or
delayed by 180 relative to frequencies lower than f/100.
Comparing this to Fig. 5A, it will be noted that a 180
phase shift will occur at 100 l~iz . This corresponds to
the 100f point in Fig. 5B, and hence f is about 1 I~iz. As
will be described shortly in connection with Figs. 7-10,
it may be preferable to provide cascaded phase shift
stages in order to enhance the performance of the audio
signal processor of Fig. 3. When this is done, the
transfer function becomes:
G(S) 1 T ' S 2 _ 1 - 1~S 2
( 1 + i ~ S) ( 1 + lOs)
where r is equal for each phase shifter. If the time
constant r is not equal in the two stages, a composite
non-linear phase shift can occur, creating harmonics and
_____.. . .._..___
PCT/U59 ~ ~093~5
~PE~~u~ 2 ~ JA Pu ~99~
2o g83 ~ s ~-2~-
changing the characteristics of the phase shift to tailor
the performance of the audio signal processor. Monaural
sound processing is enhanced when the two channel time
constants rl and r2 are different, since the resulting
differences in the phase shifts and harmonic generation in
the left and right channels can simulate or replace the
left and right channel separation which characterizes
stereophonic recordings. The net
change in phase shift (i.e., delay time) for identical
0 cascaded stages is about 10 milliseconds at 50 Hz and
decreases with increasing frequency.
Fig. 6 is a graph showing the frequency response and
bandwidth of the output versus input signals for each
channel of the audio signal processor of Fig. 3. At 0 Hz,
~S there is a 0.5 dB drop, increasing to about 0 dB from
20 Hz to 400 KHz. This is followed by a roll-off to -3 dB
from 400 KHz to about 2 MHz, and then by a sharp drop to
about -40 dB at 8 MHz. Substitution of a higher frequency
operational amplifier and adjustment of appropriate
20 biasing components will produce a circuit capable of
processing higher frequency signals.
Having now described the audio signal processor of
Fig. 3 and its individual components in detail, it will be
seen that the functions broadly outlined in the block
25 diagram of Fig. 1 have been accomplished. The circuit of
Fig. 3 adds hanaonic and subharmonic frequencies to the
input audio signal to compensate for information lost
during transduction and recording. Continuous phase
shifting of the fundamental and harmonic frequencies is
30 carried out as a function of frequency in order to reduce
the spatial smear that is associated with most types of
loudspeakers. In addition, while preserving the original
stereo~image, the signal processor mixes the right and
left channel information in such a way as to restore a
35 sense of dimensionality to the sound. When the audio
signal processor of Fig. 3 is used to reproduce recorded
music, the realism of the processed sound is commensurate
SUBSIITUTE SHEET
lPEA/US
~'VO 92/10918 PGT/US91/0937~"
2098319
-28-
with that of a live performance. Each instrument
maintains its individuality and the feeling and emotion of
the original performance are recreated.
Although the audio signal processor of Fig. 3
accomplishes the objectives of the present invention, it
has been found that cascaded versions of that circuit
(i.e., versions in which the outputs shown in Fig. 3 are
applied to the inputs of an identical circuit) yield even
better performance. In these embodiments, the already
processed audio signal is provided with additional
harmonic content and phase shifting, and the resulting
output signal is markedly improved over the output of the
non-cascaded circuit shown in Fig. 3. While circuits
containing three or more cascaded stages are also
possible, the improvement that is obtained with more than
two cascaded stages does not outweigh the cost of
providing additional circuit components. Accordingly,
audio signal processors comprising two cascaded processing
stages are regarded as the most preferred embodiment of
the present invention, and four different examples of such
a circuit are illustrated in Figs. 7-10, respectively.
Fig. 7 is a detailed schematic diagram of a cascaded
audio signal processing circuit which is intended for use
in connection with an audio system of the type that is
typically found in an automobile. The right channel
operational amplifiers 60, 62 and 64 correspond to the
operational amplifiers 36, 46 and 48 in Fig. 3, and the
resistors 66, 68, 70 and capacitor 71 correspond to the
resistors 38, 42, 44 and capacitor 40 in Fig. 3.
Similarly, the left channel operational amplifiers 72, 74
and 76 of Fig. 7 correspond to the operational amplifiers
20, 30 and 32 of Fig. 3, and the resistors 78, 80, 82 and
capacitor 83 of Fig. 7 correspond to the resistors 22, 24,
28 and capacitor 26 of Fig. 3. In the circuit of Fig. 7,
however, the first stage of the right channel which
comprises the operational amplifiers 60-64, resistors 66-
70 and capacitor 71 is cascaded with an identical second
1~'O 92/10918
PCT/US91/09375
2oss3~s " -29-
processing stage which comprises operational amplifiers
84-88, resistors 90-96 and capacitor 97. Similarly, the
first processing stage of the left channel which comprises
the operational amplifiers 72-76, resistors 78-82 and
capacitor 83 is cascaded with an identical second
processing stage which comprises operational amplifiers
98-102, resistors 104-108 and capacitor 109.
In the cascaded signal processing circuit of Fig. 7,
crossover or mixing between the right and left channels is
not provided in the first stage and is permanently hard
wired in the second stage. To this end, the output of the
operational amplifier 84 in the second stage of the right
channel is hard-wired to the non-inverting input of the
operational amplifier 100 in the second stage of the left
channel. Similarly, the output of the operational
amplifier 98 in the second stage of the left channel is
hard-wired to the non-inverting input of the operational
amplifier 86 in the second stage of the right channel.
Although it is possible to provide the mixing or cross-
over connections in either stage of the signal processor
(or in both stages), it has been found that the provision
of channel mixing in the second stage has the greatest
effect on the perceived dimensionality of the resulting
audio signal.
Since the audio signal processor of Fig. 7 is
intended for use in an automobile, the number of switches
and indicators has been kept to a minimum. A power switch
110 operates a power supply which provides +8 volts DC and
-8 volts DC to each of the operational amplifiers 60-64,
72-76, 84-88 and 98-102. A power-on LED indicates that
power is being supplied to the audio signal processor. A
second switch 114 serves as a bypass switch for allowing
the audio signal processor to be selectively switched into
or out of the existing audio system of the automobile.
3: The bypass switch 114, which may be implemented as three
ganged single-pole, single-throw switches or as a single
3PST switch, connects the right and left channel inputs to
n
WO 92/10918 PGT/US91/09375
-30-
the corresponding right and left channel outputs via the
signal processing stages 60-64, 84-88, 72-76 and 98-102
when the switch is in the position shown in Fig. 7. When
the switch is in the opposite position, the right and left
channel inputs are connected directly to the right and
left channel outputs without passing through any of the
signal processing stages. In this switch position, the
bypass LED 116 is illuminated by current drawn from the +8
volt DC power supply through a current limiting resistor
118 to indicate to the user that the bypass mode has been
selected.
The power supply for the audio signal processor of
Fig. 7 is specifically designed for use with a +12 volt DC
battery of the type that is found in most automobiles.
The +12 volt DC terminal 120 is connected via the power
switch 110 to a choke coil 122 which serves to reduce
ignition noise. The opposite side of the choke coil 122
is connected to two circuit branches, one producing +8
volts DC and the other producing -8 volts DC. The
positive circuit branch includes a current limiting
resistor 124, a series of smoothing capacitors 126-130,
and a positive voltage regulator circuit 132. The output
of the voltage regulator 132 is connected to a filter
capacitor 134 and provides a +8 volt DC output for the
operational amplifiers 60-64, 72-76, 84-88 and 98-102. A
current limiting resistor 136 is connected between the
voltage regulator output and the power-on LED 112 referred
to previously, so that the LED is illuminated whenever the
power switch 110 is in a closed position.
Referring now to the negative branch of the power
supply circuit, the output side of the choke coil 122 is
connected through a current limiting resistor 138 to the
input of a polarity converting circuit 140. A Zener diode
142 and a smoothing capacitor 144 are connected between
the input of the polarity converter 140 and ground, in
order to reduce spikes and ripple in the input voltage.
The polarity converter 140 has an internal oscillator
_._. _ _._.._ _ . _.~ _. _
WO 92/10918 PCT/US91/09375
3 i- 2 0 9 8 3 19
which requires an external capacitor 146, and also
requires a connection, 148 to ground. The output voltage
from the polarity converter on line 150 has a negative
polarity and is applied to the input of a negative voltage
regulator 152. Two smoothing capacitors 154, 156 are
connected between the line 150 and ground to eliminate
voltage transients of either polarity. The output of the
negative voltage regulator 152 on line 158 provides -8
volts DC to the operational amplifiers 60-64, 72-76, 84-88
20 and 98-102. A filter capacitor 160 is connected between
line 158 and ground in order to stabilize the output of
the regulator 152.
Preferred values for the circuit components of Fig.
7 are provided in Table 2 below. The values of the
resistors and capacitors used in the signal processor
stages are the same as those of Fig. 3, but the
operational amplifiers are designed to operate with +8 and
-8 DC supply voltages. As in the case of Fig. 3, the
specification of particular component values or types is
intended merely by way of example and not by way of
limitation.
TABLE 2
Component Value or Type
Resistors 66, 78, 90, 100 Kft, ~ W
104
Resistors 68, 70, 80, 10 Kft, ~ W
82, 92, 96, 106, 108
Capacitors 71, 83, 97, 0.1 uF, Mylar
109
Operationa l amplifiers Texas Instruments
72-76, 84-88, 98-102 TL084BCN or
equivalent
Switch 114 3PST
LEDs 112, 116 2 volt nominal
II '
WO 92/10918 ~~ PCT/US91/09375
2 0 9 8 3 19 - , -32-
TABLE 2 (Continued)
ComDOnent Value or Type
Resistor 118 1 Kfl, ~ W
Switch 110 SPST
Choke coil 122 350 ~H
Resistor 138 50 ft, ~ W
Capacitor 126 100 ~CF, electrolytic
Resistor 124 5 n, ~ W
Capacitor 128 470 ~CF, electrolytic
Capacitors 130, 156 0.1 ~F, disk
Voltage Regulator 132 Texas Instruments
LM78o8 or
equivalent
Capacitors 134, 146, 10 ~F, electrolytic
154,
160
Resistor 136 1 Kil, ~; W
Polarity Converter 140 Harris 7660/76605 or
equivalent
Voltage Regulator 152 Texas Instruments
LM7908 or
equivalent
A modified version of the cascaded audio signal
processor of Fig. 7, intended for use in connection with
home audio systems, is illustrated in Fig. 8. As far as
the signal processing stages are concerned, this version
is essentially the same as the previous version of Fig. 7,
except that the operational amplifiers 60'-64', 72'-76',
84'-88' and 98'-102' operate on +15 and -15 volts DC
rather than +8 and -8 volts DC. Also, in order to allow
the user to control the overall gain of the audio signal
processor, as well as the amount of the processing (i.e.,
._ ~.__._.~._._.____._ .._____
CVO 92/10918 PCT/US91/09375
-33- 20 983 19
phase shift and harmonic generation) that is carried out
on the incoming audio signal, feedback resistors 96' and
108' in the second cascaded signal processing stage of
each channel are provided in the form of potentiometers
rather than fixed resistors. Preferably, the
potentiometers 96' and 108' of the right and left channels
are ganged, as shown, so that they can be adjusted
simultaneously by the user. The other major difference
between the circuit of Fig. 8 and that of Fig. 7 has to do
with the nature of the power supply, as will now be
described.
Since the audio signal processor of Fig. 8 is
intended for use with a home audio system, the power
supply is designed to operate from a 120 volt AC power
source 162, which will typically consist of a wall outlet.
The 120 volt input is applied across the primary side of
a step-down transformer 164, and the secondary of the
transformer is connected across a filter capacitor 166
through a power switch 168. The node between the power
switch 168 and the capacitor 166 is connected to a
positive circuit branch through a first diode 170 and to
a negative circuit branch through a second diode 172.
Referring first to the positive circuit branch, the
cathode of the diode 170 is connected to the input of a
positive voltage regulator 174. Smoothing capacitors 176
and 178 are connected between the input to the voltage
regulator 174 and ground. The output of the voltage
regulator 174 is a regulated +15 volt DC level which is
connected to the +15 volt power supply input of each of
the operational amplifiers 60'-64', 72'-76', 84'-88' and
98'-102'. Two capacitors 180 and 181 are connected
between the output of the regulator 174 and ground in
order to stabilize the output voltage level and polarity.
In addition, a current limiting resistor 182 and an LED
3; 184 are connected in series between the regulator output
and ground. When the power switch 168 is in a closed
WO 92/10918 PCT/US91/09375
2 4 9 8 3 19 -34-
position, the LED 184 is illuminated to provide the user
with a power-on indication.
Referring now to the negative branch of the power
supply circuit, the anode of the diode 172 is connected to
the input of a negative voltage regulator 186. Smoothing
capacitors 188 and 190 are connected between the input of
the voltage regulator 186 and ground, as shown. The
negative voltage regulator 186 converts the negative half-
cycles from the diode 172 to a stable -15 volt DC output
which is applied to the -15 volt DC inputs of the
operational amplifiers 60'-64', 72'-76', 84'-88' and 98'-
102'. Two capacitors 192 and 194 are connected between
the regulator output and ground in order to stabilize the
output voltage level and polarity.
Preferred values for the components used in the audio
signal processor of Fig. 8 are provided in Table 3 below.
As in the case of the previous Tables, these values are
provided simply by way of example and not by way of
limitation.
TABLE 3
Component Value or Type
Resistors 66, 78, 90, 100 Kfl, ; W
104
Resistors 68, 70, 80, 10 Kfl, ~ W
82, 92, 106
Potentiome ter 96', 108' 0-100 Kft, ~ W
Capacitors 71, 83, 97, 0.1 ~F, Mylar
109
Operationa l amplifiers National Semiconductor
60'-64', 72-76', LF147/LF347 or
84'-88', 98'-102' equivalent
Switch 114 3PST
LEDs 116, 184 2 volt nominal
~ __ . __. ~_.._,. _.....H.. ..___~. ~._~. _ .. T.._._..~.__~_w
WO 92/10918 PGT/US91/09375
-35- 20 983 19
TABLE 3 (Continued)
Component Value or Tube
Resistor 118 1 Kfl, ~ W
Transformer 164 120 volt input, 22.5
volt output, 0.5
ampere secondary
Switch 168 SPST
Diodes 170, 172 1N4002
Capacitors 176, 188 10 ~,F, electrolytic
Capacitors 178, 181, 0.1 ~F, disk
190, 192
Voltage regulator 174 National Semiconductor
LM7815 or equivalent
Capacitors 180, 194 470 ~F, electrolytic
LED 184 2 volt nominal
Voltage regulator 186 National Semiconductor
LM7915 or equivalent
Fig 9 illustrates a modified version of the cascaded
audio signal processing circuit of Fig. 8. The embodiment
of Fig. 9 is also intended for use in connection with home
audio systems, but provides the user with the additional
option of employing crossover or channel mixing in the
first cascaded signal processing stage while maintaining
the hard-wired crossover in the second signal processing
stage. To that end, a two-position mode switch 196 is
provided between the respective first stages of the right
and left channels, as illustrated in Fig. 9. When the
mode switch 196 is in the position shown, the output of
the operational amplifier 60' in the right channel is tied
to the non-inverting input of the operational amplifier 74
in the left channel. Similarly, the output of the
WO 92/10918 PCT/US91/09375
36
operational amplifier 72' in the left channel is tied to
the non-inverting input of the operational amplifier 62'
in the right channel. In this condition, crossover or
mixing between the right and left channels takes place in
the first cascaded stage of each channel, and this
supplements the mixing that takes place in the second
stage of each channel as a result of the hard-wired
connections. In order to indicate to the user that the
circuit is operating in this mode, the mode switch 196
l0 includes an additional pole which connects an LED 198 to
the +15 volt DC power supply through a current limiting
resistor 200. For convenience, this mode is referred to
as "Mode 2", although the designation of modes is purely
arbitrary.
When the mode switch 196 is in the position opposite
from that shown in Fig. 9, the crossover connections
between the first stage operational amplifiers of the
right and left channels are broken. Accordingly, no
channel mixing takes place in the first stage of the audio
signal processor, although channel mixing continues to
occur in the second stage as a result of the hard-wired
crossover connections. This results in a reduced amount
of processing of the audio signal, which may be desired by
the user for certain types of audio programs. When the
mode switch is in this position, the +15 volt DC supply is
disconnected from the LED 198 and is connected to a second
LED 202. This causes the LED 198 to be extinguished and
the LED 202 to be illuminated, thereby indicating to the
user that the audio signal processor is operating in "Mode
1". The mode switch 196 may take a variety of forms, such
as a two-position toggle or rocker switch, a two-position
rotary switch, or a pair of pushbuttons. The bypass
switch 114 may be implemented in a similar manner and may
be combined with the mode switch 196 into a single six-
pole, three-position switch.
Preferred values for the circuit components in Fig.
9 are provided in Table 4 below. As before, these values
~w_._____..._....___.~._. _ ..~ ...~..._... ....__. ___.~~...~_ _. . .r _. w.
xV0 92/10918
PCT/US91/09375
-37- 2 0 9 8 3 19
are provided merely for the purpose of example, and are
not intended to limit the scope of the present invention.
TABLE 4
ComDOnent Value or Tune
Resistors 66, 78, 90, 100 Ktt, ~ W
104
Resistors 68, 70, 80, 10 Kfl, ~ W
82, 92, 106
Potentiometer 96', 108' 0-100 Kfl, ~ W
Capacitors 71, 83, 97, 0.1 ~F Mylar
109
Operational amplifiers National Semiconductor
60'-64', 72'-76', LF147/LF347 or
84'-88', 98'-102' equivalent
Switch 114 3PST
LEDs 116, 184, 198, 202 2 volt nominal
Resistors 118, 200 1 KiZ, ~ W
Transformer 164 120 volt input, 25.5
volt ouptut, 0.5
ampere secondary
Switch 168 SPST
Diodes 170, 172 1N4002
Capacitors 176, 188 10 ~F, electrolytic
Capacitors 178, 181, 0.1 ~F, disk
190, 192
Voltage regulator 174 National Semiconductor
LM7815 or equivalent
Capacitors) 180, 194 470 ~F, electrolytic
Voltage regulator 186 National Semiconductor
3p LM7915 or equivalent
Switch 196 3PDT
WO 92/ 10918 PCT/US91 /09375
~2 0 8 8 319
A further embodiment of an audio signal processor in
accordance with the present invention is illustrated in
Fig. 10. This version is intended for use in recording
studios, broadcast stations and other professional
environments. The embodiment of Fig. 10 is similar to
that of Fig. 9, except that low-noise operational
amplifiers 60"-64", 72"-76", 84"-88" and 98"-102" are used
and the bypass switch 114, mode switch 196 and LEDs 116,
198 and 202 have been deleted. In addition, a pair of
potentiometers 204 and 206 has been interposed in the
crossover connections between the respective first stages
of the right and left channels, and a similar pair of
potentiometers 208, 210 has been interposed between the
respective second stages of the right and left channels.
The potentiometers 204, 206 and 208, 210 allow the amount
of crossover or mixing between the right and left channels
to be controlled in a continuous manner, rather than
discretely as is the case when a switch is used.
Moreover, by providing separate sets of potentiometers
204, 206 and 208, 210 as shown in Fig. 10, the amount of
crossover or channel mixing can be controlled separately
for each stage of the cascaded signal processing circuit.
With appropriate selection of potentiometer values, the
amount of crossover can be varied between essentially zero
and 100%. If desired, the crossover potentiometers can be
ganged by stage (i.e., 204 with 206 and 208 with 210)
and/or by channel (i.e., 204 with 208 and 206 with 210).
The audio signal processing circuit of Fig. 10 also
differs from that of Fig. 9 in that the feedback resistors
70', 82' associated with the operational amplifiers 62",
74" are implemented as potentiometers, rather than as
fixed resistors. This provides the user with additional
control over the amount of gain and signal processing that
occurs in the first stage of each channel. The feedback
3f resistors 70' and 82' may be ganged, if desired, or
controlled individually as shown. The feedback resistors
96', 108' associated with the operational amplifiers 86",
__ __~..._ __._....r.. _~ _ _....
WO 92/10918 PCT/US91/09375
-gg- 2 0 9 8 3 1 9
100" are shown as being individually controlled in the
circuit of Fig. 10, but may be ganged in the manner shown
in Fig. 9 if desired. The final difference between the
audio signal processor of Fig. 10 and that of Fig. 9 is
that the input resistors 92, 106 to the inverting inputs
of the operational amplifiers 86", 100" in Fig. 9 have
been replaced by potentiometers 212, 214 in Fig. 10. Each
of the potentiometers 212, 214 is placed in series with a
fixed resistor 216, 218 in order to insure that some input
resistance remains when the potentiometer is adjusted to
its zero-resistance value. The use of the potentiometers
212, 214 provides the user with additional control over
the amount of signal processing that occurs in the second
stage of each channel of the audio signal processor in
Fig. 10. By approximately varying the potentiometers 212
and 214 as well as the feedback potentiometers 70', 82',
96' and 108', the audio sound field can be manipulated in
order to move the sound image in space. This is
accomplished by phase additions and subtractions of the
2o spatially separated harmonic and subharmonic frequencies.
Preferred values for the components used in the
embodiment of Fig. 10 are provided in Table 5 below.
These values are provided merely by way of example and are
not intended to limit the scope of the present invention.
TABLE 5
Component Value or Tvt~e
Resistors 66, 78, 90, 100 Kfl, ~ W
104
Resistors 68, 80 10 Kil, ~ W
Potentiometers 70', 82' 0-10 Ktl
~ W
,
Potentiometers 96', 108' 0-100 Ktl, ~ W
n
WO 92/10918 PCT/US91/09375
-40-
2098319 .
TABLE 5 (Continued)
Component Value or Tune
Capacitors 71, 83, 97, 0.1 ~CF, Mylar
109
Operational amplifiers Analog Devices AD713
60"-64", 72"-76" or equivalent
84"-88", 98"-102"
Transformer 164 120 volt input, 25.5
volt output, 0.5
p ampere secondary
Switch 168 SPST
Diodes 170, 172 1N4002
Capacitors 176, 188 10 ~F, electrolytic
Capacitors 178, 181, 0.1 ~F, disk
~ 190, 192
5
Voltage regulator 174 National Semiconductor
LM7815 or equivalent
Capacitors 180, 194 470 ~F, electrolytic
LED 184 2 volt nominal
2p Voltage regulator 186 National Semiconductor
LM7915 or equivalent
Potentiometers 204, 206, 0-100 Kit, ~ W
208, 210
Potentiometers 212, 214 0-10 Kil, ~ W
Resistors 216, 218 4 Kfl, ~ W
Figs. 11A-11D illustrate several ways in which an
audio signal processor of the type contemplated in the
present invention can be used in connection with an
existing audio system. In Fig. 11A, the preamplified line
30 outputs of an audio source 220 such as a compact disk
WO 92/10918 PCT/US91/09375
-41- 2 0 9 8 3 19
player, a digital or analog tape deck, a tuner, or a video
cassette recorder with audio outputs, are connected to the
left and right inputs of the signal processor 222. The
signal processor 222 may comprise any one of the
S embodiments shown in Figs. 3 or 7-10, described
previously. The left and right channel outputs of the
signal processor 222 are connected to the corresponding
inputs of an audio amplifier or receiver 224. The left
and right channel outputs of the audio amplifier or
receiver 224 are connected in a conventional manner to a
pair of high-fidelity loudspeakers (not shown). Fig. 11B
illustrates an alternative arrangement in which the signal
processor 222 is connected in the tape monitor loop of the
audio amplifier or receiver 224 to allow for processing of
the audio signal from any source that is connected to the
amplifier or receiver 224. In a similar manner, the
signal processor 222 can be installed in the processor
loop of a recording studio mixer. The signal processor
can also be installed on the input side of a broadband
limiter at an FM stereo radio station.
Figs. 11C and 11D illustrate two possible ways in
which the signal processor 222 may be used in connection
with an automobile audio system. In Fig. 11C, the
preamplified line outputs from an automobile radio 226 are
applied to the inputs of the signal processor 222, which
is preferably of the type described previously in
connection with Fig. 7. The outputs of the signal
processor 222 are applied to the line inputs of an audio
booster 228. The outputs of the audio booster 228 are
connected to the left and right speakers (not shown)
installed in the automobile. In Fig. 11D, the amplified
speaker outputs of the automobile radio 226 are applied to
the audio signal processor 222 through attenuating
resistors 230 and 232. The outputs of the signal
3f processor 222 are connected to the line inputs of the
audio booster 228, which drives the left and right
n
WO 92/10918 PCT/US91/09375
2 0 ~ s ~ ~ 9 -42-
speakers of the automobile in the same manner as in Fig.
11C.
Figs. 12A and 12B depict oscilloscope traces
generated by audio signals at the input and output of the
audio signal processor illustrated in Fig. 9. These
traces were obtained with the mode switch 196 in the
position opposite to that shown in Fig. 9 (i.e., with no
crossover between the first signal processing stages of
the left and right channels), and with the potentiometers
96' and 108' set at 10 Ktl. The input audio signal was a
musical program reproduced on a stereo compact disk (CD)
player, and only one output channel of the CD player was
used. Fig. 12A illustrates the unprocessed voltage output
of the CD player versus time at the input of the signal
processor of Fig. 9, and Fig. 12B illustrates the
processed output from the signal processor of Fig. 9
during the same time interval. It can be seen that the
output signal in Fig. 12B is a more complex waveform than
the input signal shown in Fig. 12A. This is a result of
the harmonic generation and phase shifting that is carried
out by the audio signal processor of Fig. 9.
Figs. 13A and 13B depict x-y oscilloscope traces
comparing, the input and output of the audio signal
processor of Fig. 9. The processor settings were the same
as those used in Figs. 13A and 13B, and the same CD player
and musical program were employed. In Fig. 13A, the
switch 114 of the audio signal processor was set to the
bypass position, eliminating any signal processing. The
resulting trace on the oscilloscope is a straight line,
indicating that there is essentially no phase difference
between the audio signals at the input and output of the
signal processor. In Fig. 13B, the bypass switch 114 was
moved to the opposite position to allow processing of the
audio signal. The resulting waveform is a complex
Lissajous pattern whose shape continually varies in
accordance with the changing phase relationship between
the input and output signals. When the musical program
..:fVO 92/10918 PGT/US91/09375
-43- ,20 983 19
consists of a horn or other instrument whose output
approximates a pure tone, more conventional types of
Lissajous figures are observed.
Figs. 14A and 14B depict spectrum analyzer traces
illustrating the harmonic content of the signals present
at the input and output of the audio signal processor of
Fig. 9. The same processor settings, CD player and audio
program were used as in Figs. 12 and 13. In Fig. 14A,
which shows the unprocessed audio spectrum at the input of
the signal processor, a normal audio spectrum exists
between 0 Hz and 20 KHz, and there is relatively little
activity above 20 I~iz. In the processed spectrum shown in
Fig. 14B, however, the audio spectrum between 0 Hz and
I~iz is noticeably richer in harmonics, as indicated by
15 the greater number of peaks and valleys. In addition, the
audio bandwidth has been expanded well beyond 20 IQiz, as
indicated by the peaks and valleys in the area of the
power spectrum between 2 0 I~iz and 2 5 IQiz .
Further testing has suggested that the audio signal
20 processor of the present invention may expand the audio
bandwidth far beyond the normal range of human hearing.
Harmonics have been observed out to 50 I~iz and beyond, and
there are indications that the signal processor may
generate frequencies as high as 4 MHz. Although a 4 MHz
signal is far outside the audible range, the possibility
exists that a loudspeaker voice coil excited by a 4 MHz
signal may generate electromagnetic waves that affect
human auditory perception in a manner that is not yet
understood.
In view of the preceding description, it can be seen
that the present invention produces a significant
improvement in sound quality by recreating and spatially
separating harmonic frequencies that would otherwise be
lost during the transduction, recording and playback
processes. In addition, the dimensions of depth and
height are added to standard stereo signals, and the
II
WO 92/10918 PCT/US91/09375
2 0 9 8 3 1 9 -44-
spatial smearing that is inherent in conventional
loudspeaker designs is reduced.
The audio signal processing circuits shown in Figs.
3 and 7-10 can be manufactured as stand-alone devices that
may be used in conjunction with existing types of audio
systems. Alternatively, the audio signal processing
circuits of Figs. 3 and 7-10 may be incorporated into
stereophonic amplifiers, receivers or other types of audio
components as an integral part thereof. Equalizers and
tone controls are generally not needed for two-speaker
systems when the present invention is employed, since the
dimensional enhancement of the audio signal has been found
to be substantially independent of the acoustics of the
space.
Although the present invention is of particular
utility in enhancing the quality of recorded music and
other types of recorded audio programs, the invention can
also be used during live performances to enhance the sound
of the audio signal after it has been transduced by
microphones, and to correct for phase shifts caused by the
distance of the listener from the stage or loudspeaker.
In other words, the listener can effectively be brought
closer to, the performance, regardless of the listener's
actual location. When the present invention is used to
make an original recording, whether on a record, compact
disk, audio tape, film, video sound track, laser disk, or
computer disk, the recording can later be reproduced on
conventional sound reproduction equipment while preserving
much of the enhancement provided by the audio signal
processor.
The present invention can be used in virtually any
type of device or application in which audio recording,
reproduction, or transmission is involved. Examples
include stereophonic and monaural amplifiers and
preamplifiers, automobile radios (including FM, AM and AM
stereo), boat radios, portable tape and compact disk
players, portable stereo radios, electronic organs,
_._. . ~ _ _. _.___ _. _.._r.,..,._.__.. _
,;CVO 92/10918 PGT/US91/09375
-45- ~ p g 8 3 19
synthesizers, theater sound systems, citizens band radios,
walkie talkies, sound effects generators, electronic
keyboards, electrically amplified orchestras and bands,
radio and television stations, recording studios,
television sets, cellular and other types of telephones,
video cassette recorders, radio receivers of all kinds,
airline entertainment systems, military and non-military
communication systems, public address systems, night clubs
and discotheques, and background music for elevators,
]o skating rinks, shopping malls, stores and so on.
Although the present invention has been described
with reference to a number of preferred embodiments
thereof, the invention is not limited to the details
thereof. For example, although the operational amplifiers
]5 are preferably of the integrated solid state type,
discrete solid state devices or even vacuum tubes can be
used in higher power applications. In addition, large
scale integration (LSI) techniques can be used to reduce
most of the circuitry of the audio signal processor to a
20 single chip, or surface mount technology can be used to
produce hybrid chips. These and other modifications are
intended to fall within the spirit and scope of the
present invention as defined in the appended claims.
