Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
METHOD AND APPARATUS FOR PERFORMING LEVEL COMPENSATION
FOR AN INPUT SIGNAL
BACKGROUND OF THE INVENTION
The present invention pertains to a method and apparatus for performing level
compensation for an input signal. More particularly, the present invention
pertains to a
method and apparatus that adjusts the output level of an input signal to
compensate for
level changes made to the input signal by a signal processor.
A category of devices referred to as 'signal processors' are known in the art.
These
processors can be used to modify an input signal, such as music (or other
audio sound)
that has been converted to electronic analog or digital signals. Common audio
processors
include tonal modifications (e.g., an equalizer or EQ, vocoders, distortion
effects, chorus
effects, flanger effects, ring modulators, wah-wah effects), dynamic
modifications
(compression, expansion, tremelo, vibrato,etc.), reverberation, delay, many
others, and
combinations of these. Other processors are known in the art such as common
video
processors (e.g., color filters) and other processors used in a variety of
fields. These
processors all have a potential "side effect" of changing the original input
signal's absolute
peak level in relation to other signals or devices present. This change in
peak level is
usually not desired, is often not readily apparent to an operator, and can
easily affect the
judgement of an operator of the device, often without the operator knowing it.
In the audio field, there are three common reasons to "equalize" a sound: 1)
to get
rid of a problem noise (e.g., cut out an annoying air-conditioner buzz), 2) to
tailor a
pleasant musical sound (e.g., make a singer's voice sound pleasantly 'husky'
by adding a
certain low range), and 3) to create an effect (e.g., make a singer sound as
if singing
through a 1920's megaphone, as in the Beatles' "Yellow Submarine").
When using an equalizer to boost or cut a portion of the sound for any
purpose, the
volume of the sound that is equalized actually gets louder (when boosted) or
quieter
(when cut). It is also understood in the art of sound engineering that it is
important to
allow a comparison between the original sound and the equalized sound as part
of the
-1-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
process of using an equalizer (or other sound processor). This is accomplished
with a
"Bypass" switch, which chooses whether the sound goes through the equalizer
circuit, or
bypasses it.
Psycho-acoustic Concerns and Results: The study of how a person hears is
commonly referred to as psychoacoustics. A well-known phenomenon in this field
was
first codified by Fletcher & Munson, and is most commonly displayed as a graph
called the
"Fletcher-Munson curves" or "equal loudness contours." The data shows that the
human
ear-brain system interprets loudness as a function of frequency with two
particular effects:
1 - A person hears unevenly at different frequencies (at the same volume
level,
low and high frequency sounds seem quieter than middle frequencies), and
2 - As the entire sound gets louder and louder, one will hear more and more
evenly. With very loud sounds, the above effect becomes small or disappears.
Thus, as one turns up the volume, a person seems to hear more bass and treble,
which is why some people listen to music so loudly.
A human being typically is capable of hearing sound in the frequency range
that is
made by the human voice (from about 100 Hz to 4 kHz) more easily than sounds
that are
at higher or lower frequencies. A low bass guitar note and a high violin note
both need to
be played with a lot more energy than a voice, if one wants these instruments
to sound as
loud as a voice.
The particular effects outlined above interact with one another. As the sound
gets
louder, frequency differences affect hearing less and less; at very loud
levels, it takes about
the same amount of energy for the bass, the voice and the violin to sound as
loud as each
other.
This is commonly experienced when using a stereo. With the volume at a normal
listening level, a compact disc will have an acceptable sound. If the volume
is turned down
a lot, the music will suddenly seem as if there is not enough bass, and also
not enough
cymbals or other high frequency sounds (the Fletcher-Munson curves describe
how much
less bass and high frequencies). Stereo systems are often supplied with a
"Loudness"
control (sometimes an on/off switch, sometimes a variable knob) to try to
compensate for
this (with very limited success). The Fletcher-Munson curves are averages of
data
-2-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
empirically derived from many people, thus they are approximations for any
given
individual's experience.
Audio engineers deal with this phenomenon on a regular basis. The following
are
two examples of how this issue is addressed in the art.
A first example involves an instrument that produces only one sound, a tom-tom
on
a drum set. Given a recording of the tom-tom, the audio engineer is looking to
make it
sound better using an equalizer. One may hear a pleasant sound near the drum's
fundamental resonance, for example between 400-800 Hz. There is also a very
unpleasant
sound just above this frequency range. An engineer can use a low pass
equalizer (which
cuts out high frequency signals) that has a control to select frequency (this
lets one choose
the point above which sound is cut). This should allow the engineer to cut out
the
unpleasant area and then fine tune the lower frequencies to find a spot where
the drum
sounds the best.
First, an engineer may cut out the unpleasant frequencies above the nice
range. But
now, the sound is much too quiet, so the volume is turned up significantly.
Next, the
engineer fine tunes for the nicest sound. The drum itself has varying degrees
of loudness at
each frequency, and the nicest place to set the control may be in a frequency
range where
the drum is relatively quiet. If the nicest part of the drum's sound is very
quiet, it may go
unnoticed (again because of the Fletcher-Munson effect). From experience or
training, the
engineer will (again) turn the volume up significantly, to help hear the best
sound. Next, a
bypass switch is used to see if the equalized signal is an improvement over
the original
(unequalized) sound. The much louder unequalized sound makes it impossible to
tell what
the engineer has accomplished (the louder one almost always sounds fuller and
better just
because it is louder, as described by the Fletcher-Munson curves) and can hurt
ones ears
because the sound is so much louder (the equalized sound had to be turned up a
lot so that
one could hear it). This increased volume could also damage the speakers
because the
sound is so much louder.
In a second example, the recorded track of a singer is mixed with the rest of
the
band (drums, bass, guitar, piano, etc.). As stated above, the music sounds
"bassier" when
the volume is turned up (or less "bassy" when the volume is turned down), as
described by
-3-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
the Fletcher-Munson curves. As more bass is added to a sound with an
equalizer, the total
volume is being increased.
If the recording of the voice sounds "thin" because the microphone was poorly
placed, the engineer may try to compensate by turning up the bass equalizer.
The voice
sound improves, but the extra bass has made the voice louder than the rest of
the band, so
it is turned down to an appropriate level. The result is that, the voice
doesn't have enough
bass anymore (due to the Fletcher-Munson effect). So the process is repeated a
few times,
until a satisfactory result is achieved.
A known technique for solving the problems addressed above is set forth in the
following complex procedure.
1 - Select a channel to be equalized ("EQ Ch.").
2 - Set up a separate channel that duplicates the unequalized original input
("UNeq
Ch").
3 - Listen only to the EQ Ch. (i.e. 'SOLO' the EQ Ch. - listening only to that
channel, turning off all other channels).
4 - Adjust the equalizer settings on the EQ Ch. until satisfactory.
5 - Solo the EQ Ch. and the UNeq Ch. back and forth to:
a - match the EQ Ch.'s output level to the UNeq Ch.'s output.
b - adjust the equalizer settings on the EQ Ch. if necessary.
6 - In the mix (with the other sound channels turned back on),
a - Set (one at a time) both the EQ Ch. and UNeq Ch. levels
b - compare the EQ Ch. and UNeq Ch. to judge and adjust the equalizer
settings.
7 - Adjust as necessary, repeating steps 3-6 as needed.
Not only is the foregoing solution complex, but many musicians and many sound
technicians do not understand or know of this problem nor how to compensate
for it. This
is evidenced by the amplified sound in many clubs, concerts, weddings, etc.
which is often
poor (e.g., harsh, piercing, boomy, too loud, etc.).
In the art, there are two very common devices that actually change the source
material volume (other than equalizers, reverberators, and other processors).
These devices
make no moment-to-moment changes that affect the signal.
-4-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
1 - The Loudness Control (a switch or knob found on many consumer listening
devices) is an attempt to compensate for the Fletcher-Munson effect at low
listening levels.
It is a special tone control that boosts both the bass and treble with one
control (but the
bass and treble it turns up are somewhat different from typical bass and
treble controls).
2 - AGC (Automatic Gain Control) amplifier circuits are used to compress (make
smaller) a signal's dynamic range. This means that, no matter how loud or soft
the original
sound (the input) is, the output is always at the same volume. It is usually
found on
consumer items with low quality built in microphones (VCR's and inexpensive
cassette
tape recorders) and on devices that intentionally sacrifice sound quality for
speech clarity
(CB and Ham radios, some telephones). The output is not dependent on either
the sound
source or any particular setting. Audio fidelity is intentionally sacrificed
to overcome the
limitations of noisy environments, or to allow the use of cheap devices. In
fact, these
usually don't even have an input volume control, although some higher quality
products
allow a choice between using the AGC circuit and a real input level control
circuit. These
have an on/off switch for the AGC circuit; AGC ON bypasses the real input
circuit, AGC
OFF bypasses the AGC circuit.
Related to the AGC are dynamic processors in general. These are professional
devices that allow the user to control how the dynamic range reduction
happens. As with
the AGC, these devices make actual changes to the dynamic range of the input
signal.
In view of the above, there is a need for a method and apparatus that can make
intelligent corrections in level to compensate for changes in level caused by
a processor
and allows the user to make better choices about the use of the processors
without having
to do, or even understand the need for, the compensation.
SUMMARY OF THE INVENTION
These and other needs are met by a method and apparatus of the present
invention.
According to an embodiment of the present invention, intelligent corrections
in the level of
an output signal can be made to compensate for changes in peak level.
In a first embodiment of the present invention, the unprocessed source signal
is
first compared to the final processed signal to determine an amount of change,
and then
-5-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
uses that comparison to make a level adjustment. Standard components can be
used to
build a stand-alone device that:
1 - compares the pre-processor signal (original unprocessed signal) with the
post-
processor signal (after it has been processed),
2 - notes the difference in level between them, and then
3 - compensates the processed signal, for example, by matching the post-
processor
signal level to the pre-processor signal level.
In this embodiment, one objective is to effect a single overall level change,
and
since any number of processors in combination result in only one final level
change, this
circuit will work for any combination of changes. Therefore, only one circuit
is needed for
a given signal path, regardless of how many processors are between the "pre-"
and "post-"
reading points.
Alternatively, the processed signal is provided to a level compensation
device, and
the output of the level compensation device and the unprocessed signal are
provided as
inputs to a comparison device. The output of the comparison device then
provides a
feedback signal to the level compensation device. Accordingly, the level
compensation
device modifies the level of the processed signal, until the output of the
level
compensation device has a level equal to the level of the unprocessed signal.
In a second embodiment of the present invention, the compensation that is
provided to the processed signal is empirically derived rather than by direct
measurement.
Actual level changes that result from an operator's use of a given processor
controls are
measured. Each control on the processor which `causes' an undesired level
change can
also be used to control a circuit that provides level compensation by
producing a desired
opposing level change. This embodiment is well suited to a situation where one
knows the
range of variation in the character of the signal to be processed, and knows
the range of
variation in the settings of the processor's control. According to this
embodiment of the
present invention, development of compensation for a single control on a
processor
includes the following steps:
1 Adjusting the processor's control in adequately small steps (e.g., 30
degrees of rotation
for a rotary knob) while listening normally and observing a representative
signal
source.
-6-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
2 For each position of the control, the resulting change in overall signal
level is noted
(with objective measurement and/or with an experienced observer matching
perceived
levels).
3 Repeat with an adequate variety of source materials.
4 Compile the results, and design a circuit that counteracts the level changes
caused by
the changes of the processor control. The circuit is to be adjusted by the
same control
that adjusts the processor parameter. Two implementations of this, using a
standard
potentiometer as the control, are:
a. Have a separate gain circuit in series with the processor, whose gain is
adjusted,
e.g., by a separate potentiometer element on the same shaft as the processor
control.
b. Modify the processor to incorporate the desired changes in gain.
The ability to specifically tailor the compensation for a specific situation
and/or use is a
particular advantage of this method. Another is that it can require a minimum
of circuitry
to accomplish (thus, can be inexpensive).
In a third embodiment of the present invention, a method is provided to
compensate for level changes, for example, where the processor itself makes
dynamic
changes in level, as do compressors, expanders, and the like. While it is
desirable for these
to change the dynamic range of the source material, it is usually not
desirable for these
devices to change the peak level (loudest spots) relative to the rest of the
acoustic
environment.
In audio, these devices generally work by comparing the level of the input
signal to
a threshold, which is a constant reference voltage whose level is set by the
user with a
control. When the signal source crosses the set threshold level, the signal is
processed,
most often via a single control that sets (e.g.) the ratio of processing
applied. At any given
setting of these two controls, the total effect on the signal's peak (loudest
moment) is a
simple function of its difference from the threshold times the ratio setting,
in decibels. This
function is often represented as a Control Voltage, which is then commonly
applied to a
device such as a VCA (voltage controlled amplifier) which continuously enacts
the desired
changes in level.
An embodiment of this method uses the same control settings for the
processor's
effect to manipulate a second control signal path, which uses a separate
constant reference
-7-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
voltage as an input. The first control signal responds to the audio input
signal. The second
control voltage is dependant on the processor's control knobs, but not the
input signal,
creating the desired fixed level compensation. This compensation control
voltage may be
applied to the same device (e.g. VCA) that is used by the processor. Also,
because the
amount of level compensation is determined by the same controls which set the
changes in
peak level output, the amount of level compensation will automatically track
(and
accurately compensate for) changes in the processor's settings. The result is
an integrated
processor and level compensation system.
to BRIEF DESCRIPTION OF THE DRAWINGS
Fig.1 is a general block diagram of an apparatus constructed according to an
embodiment of the present invention.
Fig. 2 is a general block diagram of a second embodiment of the present
invention.
Fig. 3 is a detailed circuit diagram of the processor of Fig. 2.
Fig. 4 is a detailed circuit diagram of the amplifier of Fig. 2.
Fig. 5 is a graph showing the relationship between frequency and level
compensation in an example of the operation of the circuits of Figs. 2-4.
Fig. 5A is shows a notch filter with interactive depth and frequency level
compensation constructed according to an embodiment of the present invention.
Fig. 6 is a block diagram of a level compensation apparatus where a single
element
of a potentiometer can change both frequency and gain constructed according to
an
embodiment of the present invention.
Fig. 7 is a detailed circuit diagram for the apparatus of Fig. 6.
Fig. 8 is a block diagram of a digital level compensation system constructed
according to an embodiment of the present invention.
Fig. 9 is a block diagram of an apparatus constructed according to a first
embodiment of the present invention.
Fig. 10 is a detailed example of the apparatus of Fig. 9.
Fig. 11 is a further detailed circuit diagram of several elements of Fig. 10.
Fig. 12 is a further detailed circuit diagram of several elements of Fig. 11.
Fig. 13 is a block diagram of a third embodiment of the present invention.
-8-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCT/US99/17321
Fig. 14 is a graph showing the effects of compression as is known in the art.
Fig. 15 is a graph showing the effects of compression with level-compensation
according to an embodiment of the present invention.
Fig. 15A is a graph similar to Fig. 15 normalized to a different level.
Fig. 16 is a detailed circuit diagram of a dynamic processor for implementing
the
third embodiment of the present invention.
DETAILED DESCRIPTION
Referring to Fig. 1, a block diagram of an apparatus for performing level
compensation according to an embodiment of the present invention is shown. The
level
compensation can be referred to as performing one of the following steps:
1. When the processor reduces the overall level of the input signal, the
apparatus of the present invention provides an increase in gain to match the
original signal
level.
2. When the processor increases the overall level of the input signal, the
apparatus of the present invention provides a decrease in gain to match the
original signal
level.
This allows a user to make judgments about a processor's effect without being
"fooled" by the differences in signal level. It can do so without involving
the user at all
(which can be a tedious process as described above). A switch can be made
available to
disable the circuit if desired. As described in the embodiments below, the
present
invention can be applied to compensate volume for audio signals. One skilled
in the art
will appreciate that the present invention (e.g., the circuit of Fig. 1) can
be applied to
compensate for level changes caused by a processor of other types of signals
(e.g., video
signals, MRI's, CAT scans).
Referring to Fig. 1, an input signal 10 is provided to a processor 11 (e.g., a
digital
signal processor) where the signal is modified. For example, the signal can be
modified by
the processor by decreasing the level of a particular frequency band of the
signal (e.g., as
in an audio filter). The original, unprocessed signal and the modified signal
from the
processor can be supplied to a comparison device 13 (i.e., the modified signal
may be
supplied either before or after level compensation device 14). The comparison
device 13
-9-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
outputs a difference signal indicating the difference in level between the
original signal
and the modified signal. The level compensation circuit 14 receives the
modified signal
and the difference signal and level compensates the modified signal (e.g.,
corrects the
increase or decrease in gain caused by the processor 11). If desired, the
level compensated
signal can be provided to an output device 12, such as a recording device, a
speaker, etc.
As indicated above, an example of a processor is one that changes tone
quality,
such as an equalizer. Tone quality, also called timbre, is a result of the
balance of different
frequencies, measured in Hertz (cycles per second). A human's hearing is
limited to a
range between 20 Hz and 20KHz (20,000 Hz), sometimes called the sound
spectrum.
Equalizers, tone controls, filters, etc. are devices which change tone
quality. They all work
by allowing either an increase or a decrease in the volume of a selected
portion of the
hearable frequency range. A device that raises or lowers the entire frequency
range evenly
is simply a volume control. An equalizer works as a volume control for a
limited portion
of the hearing range.
A first embodiment of the present invention is shown with respect to Figs. 9-
12.
Referring to Fig. 9, a block diagram of this embodiment is shown. In Fig. 10,
a more
detailed example of the circuit is shown (i.e., a Root-Mean-Square (RMS) Level
Compensation System). In Fig. 9, three signals are input to an RMS level
compensator 62:
a post processed signal ("Audio Signal") from signal processor 61, a pre-
equalized
"control" signal 10, and a post-compensator "control" input from said RMS
level
compensator 62. The RMS level of the input signal and the control signal are
measured by
RMS converter 72A and RMS converter 72B (Fig. 10), respectively, and compared
(e.g.,
in a window comparator 73). If the control signal is greater than the output
signal, then the
signal processor 61 is reducing the level of the signal at that moment, so the
compensator's
gain is increased by adjusting a digitally controlled resistor 78. If the
control signal is less,
the gain is reduced. The rate of change in gain is determined by the frequency
of the
oscillator 75 and by the resolution of the resistor. The gain will change
until the output
level matches the control signal level, or until the gain limits are reached.
If there is no
output signal detected as sensed by signal detector 76 or if the operation
mode is on
"Hold," (as determined by Match/Hold Switch 77), the resistor will not change
value, and
the gain remains the same.
-10-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCT/US99/17321
Referring to Fig. 11, a more detailed schematic of the circuit of Fig. 10 is
shown
where the block designated DS 1666 is functioning as a digitally controlled
attenuator,
range limited by resistor R2, with output gain stage at op-amp AR2 to allow
boosting of
the audio signal. The control inputs to the DS 1666, UP/down, increment, and
chip select,
are operated by "Window comparator," "oscillator," and "disable" circuits,
respectively.
"RMS_conv" converts the "CTRL" signal and the "OUT" signal into RMS log-
equivalent
signals. "Hysteresis" circuits determine whether "CTRL" is significantly
greater or less
than "OUT." If there is a significant difference, "oscil" increments DS 1666
up or down.
The direction is determined by the top "hysteresis" circuit output. XOR gate
U3 provides
necessary timing of the chip select input with the increment input.
Referring to Fig. 12, a schematic of the subcircuits of Fig. 11 is shown. The
RMS
conversion is accomplished with a THAT2252 log converter and rectification (by
diode
Dl and capacitor C3). Resistor R16 provides independent release time by
discharging
capacitor C3. The oscillator is a standard 555 Timer circuit activated by the
Reset pin.
Note that the comparators VR1 in the hysteresis blocks have a degree of
hysteresis to
allow approximate level match, for instance, to within 2 dB. The signal
detection VR2
responds to a negative going AC signal crossing below a preset noise
threshold. This
functions as a safety feature that prevents the compensation circuit from
responding to
inappropriate signals accidentally, especially during the release time of the
RMS
converters.
A potentially more accurate version of this embodiment of the present
invention
uses the same steps as above, but the comparison includes an analysis of the
frequency
spectrum of the pre-processed and post processed signals. The applied level
compensation
is based upon the signal level differences and the frequencies at which the
level differences
occur. For example, a Fast Fourier Transform (FFT) could be used, but other
frequency
band examination techniques are possible. An example of this procedure is
given below:
1 - FFT's are taken of the pre-processed signal and the post processed signal.
2 - The RMS value of both for selected bandwidths is derived. 1/3 octave is
common in
general audio work, which results in about 30 bands of calculation for each of
the 2 FFT's.
Fewer bands may be satisfactory, especially if carefully determined.
3 - "Weigh" the results of step 2 with an "average" Fletcher-Munson curve.
-11-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
4 - Sum the results of step 3 into pre-processed and post-processed subtotals.
- Take the difference between the step 4 subtotals (subtract one from the
other).
6 - Adjust the final output level according to the results in step 5.
A goal of this embodiment of the invention is a single overall level change,
and
5 since any number of processors in combination result in only one final level
change, this
circuit will work for any combination of changes. Therefore, only one circuit
is needed for
a given signal path, regardless of how many processors are between the
"pre-" and "post-" reading points.
A second embodiment of the present invention is shown in Figs. 2-8 where the
level compensation is empirically derived. Deriving an integrated level
compensation for
an audio signal processor depends on an assumption of the character of the
signal to be
processed and level compensated. That is, one must assume a change in setting
of the
processor will effect the level similarly because all expected input signals
would be
similar. For instance, a notch filter can be used to modify the sound of a
cymbal, and
setting the notch frequency at 400 Hz may increase the auditory level by 6 dB
above a
setting of 1.25 kHz for most cymbals. To maintain equal level, the gain must
be
compensated by -6 dB for a change in frequency setting of 400 Hz to 1.25 kHz.
Referring
to Fig. 2, a broad view of how frequency and gain may be tracked via a dual
potentiometer
24 is shown.
An example of the method of this embodiment is given below:
1 Select the processor desired. In this example it is a notch filter (as shown
in Fig. 3)
with variable frequency (400 to 1250 Hz) and fixed depth.
2 Select a typical sound source, e.g., a cymbal being hit. Set a starting
parameter and
output level, for example with the notch frequency at 1.25 kHz and the output
gain
referenced to 0 dB.
-12-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
3 Change a parameter by a chosen increment and adjust gain to match previous
level,
if necessary. The new parameter setting and new gain setting are recorded.
Repeat
over entire useful range of the variable parameter. For instance, as shown in
Table
1, row 1 shows the relative gain changes needed to maintain equal level of
cymbal
sound at one-third octave intervals from 400 Hz to 1.25 kHz.
4 Repeat steps 2 and 3 for many similar sound sources. Table 1 is data from 10
Table 1
H Gain change for cymbal # 1 r -8 -6.5 -5 -4 -3 0
cymbal#2 -7 -6.5 -5 -4 -3 0
#3 -6.5 -5 -4 -3 -2.5 0
# 4 8 8 6.5 5 3 0 E- These gain settings in dB,
relative to gain at
#5 -6 -6 -5 -4 -3 0 1250 Hz=0dB.
#6 -6.5 -6 -5 -4 -2.5 0
#7 -5 -5 -4.5 -4 -3 0
#8 -7 -6 -5 -4.5 -3 0
#9 -6 -5.5 -5 -4 -3 0
#10 I_ -7 -6 -5 -4 -3 0_
F in Hz: 400 500 630 800 1k 1.25k (F = Notch frequency setting)
H = Relative dB gain needed based on subjective comparison at six one-third
octave notch frequency
settings (F) using ten similar sound sources through Fig. 3 Notch circuit
(e.g., 10 cymbals).
similar cymbals.
5 Determine an average relative gain desired at each increment of the
parameter.
Using Equationl, the mean of each column of Table 1 can be computed. It also
may be helpful to calculate the range of gain at each frequency. Equation 2
first
adjusts the relative gain data to center around a total average point,
optimizing the
deviation. Equations 3 and 4 calculate the maxima and minima of the adjusted
columns. The results are shown in Fig. 5.
A, = mean (H<x>) Eq. 1
H<X> = H<X> + Q where Q = [0.3, 0.133, -0.617, 0.967, -0.117, -0.117, -0.533,
0.133, -0.2, 0.05 Eq. 2
LX = max (H<">) Eq. 3
S,, = min (H<") Eq. 4
-13-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
6 In the analog circuit of this example, the frequency settings must correlate
to gain
settings via the potentiometer. Using a dual potentiometer, for instance,
allows the
addition of a gain circuit (Fig. 4) so that gain and frequency are controlled
simultaneously from a single knob. Equations 5 and 6 describe the
relationships
between notch frequency, gain, and resistance (f(x), G(x), and R(x)), in this
example. Results are shown in Table 2.
Ax)= 1 Eq. 5
2=7c =,\'C3=C4=(R(x) + R13)=(R12+ R26- R(x))
R2 R21
x = Eq.6
g(X)
(R(x)+R13)+R41
F, G(x) A,,
400 -6.354 -6.7
TABLE 2 500 -5.758 -6.05
630 -4.999 -5
800 -3.951 -4.05
1000 -2.521 -2.9
1250 -0.651 0
In Formula 6, the value of resistor R4 determines the shape of the curve shown
in
Fig. 5.
7 Find the closest match of the gain settings with the frequency settings.
Since the
frequency circuit has been predetermined as the functionally best arrangement,
the
gain circuit must be adjusted to match it. By choosing resistance value R4
appropriately, the resultant change in gain as seen by the line in Fig. 5 can
approximate the desired change in gain with respect to frequency as shown by
the
diamonds in Fig. 5. Resistor R2 sets the overall gain.
8 Repeat for other parameters.
Alternatives to the example method in step 6 include integrated gain change
with
frequency change by manipulating the filter circuit (see Fig. 5A, also Fig. 6
and 7) and a
digital version (see Fig. 8). One variation of a digital version has the
average gain data
-14-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCT/US99/17321
stored in memory, interpolating gain at frequencies between measurement
points; another
variation calculates gain using the relationship between f(x) and G(x).
Fig. 5A shows a notch filter with interactive depth and frequency level
compensation. The gain transfer function for audio frequencies is given by
Equation 7.
R14=(R24+ R19)=(R25+ R18+ R16) + R14
(R16=(R24+ R19) + R17.(R16+ R24+ R19)).(R25+ R18) + RI7.R16.(R24+ R19) R15 Eq.
7
A filter can be arranged so that a single element of a potentiometer can
change both
frequency and gain. Fig. 6 shows a functional diagram with a standard filter
41, namely a
Sallen-Key second-order high pass filter, that has been manipulated to yield
frequency-
gain tracking of a dual potentiometer depicted by R21 and R22 in Fig.7,
element 41.
Normally Output #2 (see Fig. 6, element 43) would be connected to ground.
Instead,
connecting R22 to current amplifier AR6 in Fig. 7, element 44 allows the same
transfer
function at Output #1 (see Fig. 6, element 43) and a second output whose
current is
inversely proportional to R3+R22. Note that the transfer functions of the two
outputs are
both high passes with poles at f=1/(2rrRC) and Q=0.707, where R=R4+R21=R3+R22
and
C=C 1=C2. The high frequency output gain of the mixer is -R7/(R8+R).
Accordingly, the
frequency of the high pass filter and the gain can be controlled with a single
control
element of said potentiometer. Referring to Fig. 8, a digital version of level
compensation is shown next to a known system. In the art, it is known to
provide a user
interface to select a function (e.g., set a high pass filter) and provide an
input as to
frequency (e.g., the cut-off frequency for the high pass filter). The filter
coefficients that
are necessary to implement the desired function are selected from memory based
on the
input frequency. The output then is the resulting adjustment of the filter
frequency.
According to an embodiment of the present invention, the memory is further
used to store
gain values that are selected based on the frequency input from the user.
Accordingly,
level compensation is performed based on the selected frequency as described
in more
detail above.
Aside from needing to separately compensate each control that causes a change
in
level, there are circumstances under which this solution will not be adequate.
Under
-15-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
predictable circumstances (where combinations of sound source and equalizing
type are
consistent), the above process produces very good results. The more limited
the type of
source material, and the more limited the range/effect of the equalizing
element, the easier
it is to develop an empirically derived compensated processor. But the more
varied the
source, the more tests are needed, and eventually, test results will be
contradicting.
The most complicated situation would be to design for all possible human
listeners
(acknowledging that each one is different as seen by the Fletcher -Munson
curves, which
are averages), for all possible sound sources (speech, music, sound effects,
etc.,), with a
knob that has a wide range (e.g., a frequency control with a range of 20-
20khz). This is
actually impossible, because of how variations in the source interact with a
given
equalizer. For example, assume one designs a bass boost frequency knob that
works well
for a broad source like an orchestra, and assume that in the low range, the
compensating
circuit amplification level increases (this is likely because of the Fletcher-
Munson effect).
It is assumed that there is a place in the music where only a musical triangle
is playing. As
the equalizer knob is adjusted in the low region, it has no effect on the tone
of the triangle,
because the triangle has no low frequencies that can be boosted. We expect to
hear no
change at all, but the gain compensator is still adding level, so the triangle
changes volume
as we turn the knob back and forth. For a triangle (or any sound with no low
frequencies),
this knob has turned into a volume control, and is confusing to a user. Thus,
a circuit that
compensates based only on where the equalizer knob is set (as determined by
any method,
including the one above) won't work well in broadly different circumstances.
For those
circumstance, the first embodiment described above is preferred.
A third embodiment of the present invention is shown in Figs. 13-16 as a
dynamic
processing compensation circuit. A dynamic processor changes the gain of a
signal to
achieve any number of effects: limiting peaks to prevent overload, compressing
the
dynamic range of a signal, expanding the range, sustaining a note longer, etc.
Since the
total gain changes based on the input signal level, it is very difficult to
compensate for
adjustments in parameters using the source-adjusted compensation method
described
above. Such methods are likely to undo the work of the processor. Instead,
according to
an embodiment of the present invention, one can compensate for expected gain
changes
with the following procedure:
-16-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
1. Assume a normalized listening level desired after processing. The simplest
normal _
point is 0 dBV.
2. Since without compression at unity gain an input signal of 0 dBV yields an
output
of 0 dBV, any change in gain from compression of a 0 dBV input signal should
be
matched by an equal and opposite change of gain to maintain a 0 dBV output.
For
instance, if a reduction in dynamic range is desired from 90 dB to 60 dB,
starting at
-90 dBV, a compressor is used to steadily reduce gain as an input signal rises
up to
0 dBV. But the output of a 0 dBV input would be -30 dBV. It is more likely
that a
range of -60 to 0 dBV is desired, not -90 to -30 dBV. So the gain should be
increased by 30 dB.
3. The gain adjustment is shown in Fig. 13. Using a Voltage-Controlled
Amplifier
(VCA) with positive and negative control voltage inputs, a constant control
voltage
is supplied to one terminal that is equal to the control voltage expected at
the other
terminal with an input of 0 dBV. For instance, a 0 dBV input signal at certain
parameter settings may produce +100 mV at the negative control port, reducing
the
gain. A +100 mV voltage at the positive control port cancels this reduction.
Alternatively, equal and opposite control voltages may be combined at a single
terminal for the same result.
To implement this method, identical circuitry is used to modify both the RMS
output signal, and a constant voltage source that is equal to the RMS output
at the
normalized input level (OdBV in this case). In this embodiment, dual
potentiometers for
threshold and ratio knobs can be used to simultaneously adjust both control-
voltage
circuits, so that the two control voltages change in tandem. Other circuits
can be used to
achieve the same functionality.
Referring to Fig. 14, a graph of a typical compressor's transfer function Vin
versus
Vout is shown. Note how much the peak level is reduced for low thresholds and
large
ratios. In Fig. 15 a graph of a level-compensated compressor at various
settings is shown,
normalizing the level to 0 dBV (in Fig. 15A, the same effect is shown
normalized to 14
dBV). For instance, a 2:1 ratio, -16 dBV threshold compression reduces the
gain -8 dB at
0 dBV input. The gain is therefore compensated +8 dB. At o0: 1, the gain must
be
-17-
SUBSTITUTE SHEET (Rule 26)
CA 02380658 2002-01-29
WO 00/07175 PCTIUS99/17321
compensated +16 dB. Note that normalizing at a level high in the dynamic range
increases
the volume of a bulk of the signal, which generally lies below 0 dBV. But
unless the aim
is to merely "shave off the top," this gain adjustment is usually desired. In
Fig. 16, a
detailed schematic of a dynamic processor using a THAT4301 dynamic processor
chip is
shown integrating an RMS converter and a log-based VCA.
The embodiments of the present invention compensate for level differences
caused
by changes in the signal processor. The compensator should only adjust level
based on the
processor's modification of the signal. To avoid false triggers of level
compensation, a
detection circuit or switch should freeze the level compensator from changing
its
modification when the signal processor is not being adjusted.
-18-
SUBSTITUTE SHEET (Rule 26)
