Note: Descriptions are shown in the official language in which they were submitted.
lX75~47
FM COMMUNICATION SYSTEM WITH IMPROVED RESPONSE
TO RAYLEIG~-FADED RECEIVED SIGNALS
05
Background of the Invention
This invention generally relates to frequency
modulated (FM) communication systems. The present
invention is more specifically directed to improving the
lo audio sound ~uality by reducing the magnitude of audible
pops and bursts of noise that occur during received
signal minimums caused by Rayleigh-faded received
signals. The invention is especially effective for FM
systems which utilize peak deviations of less than 5.0
15 kilohertz (kHz).
In order to appreciate this invention, the
concept of a Rayleigh-faded signal should be understood.
Rayleigh fading refers to the rapid fluctuations in the
magnitude and/or phase of the received signal resulting
20 from multipath propagation. A Rayleigh-faded signal is
most objectionable to a listener when the magnitude of a
fade is large enough to cause a substantial momentary
decrease in the rec~ived signal amplitudeO This results
in the listener hearing an objectionable burst of noise.
25 A common example of Rayleigh fading occurs when a mobile
radio user travels down a highway while receiving a
signal having rapid and substantial magnitude
fluctuations in field strength. Such fluctuations may be
caused by the vehicle passing relatively nearby objects,
30 such as telephone poles or buildings, which result in
field strength variations at the vehicle's mobile
antenna.
If the reduction in signal amplitude due to
Rayleigh fading is sufficient to reduce the IF carrier-
35 to-noise (C/N) ratio to less than approximately lO dB,
s,~
7~L~47
- 2 - CM00415G
noise in the receiver input stages will dominate and
produce a burst of noise in the audio during the signal
null. On the oth~r hand, if the mean amplitude of the
signal is very larg~, such that reduction of the IF C/N
05 to less than 10 dB is unlikely, a noise llpopli will be
generated by the discriminator during each signal null.
This noise pop is due to the signal phase modulation
which occurs as the signal amplitude drops, and is
sometimes referred to as "random FM".
Thes~ undesired audio responses due to Rayleigh
fading become more objectionable as the peak transmitted
deviation decreases and the carrier frequency increases.
Deviation is a factor in determining the signal-to-noise
ratio. In order to achieva a given level of audio
15 output, more audio gain will be required in a system
having smaller peak deviation. Because this greater gain
also amplifies the undesired noise burst or pop, these
disturbances become louder relative to the desired audio
(having smaller deviation) and, hence, the disturbances
20 are rendered more objectionable. Since wavelength
decreases as frequency increases, a larger number of
signal strength nulls exist at higher frequencies. Thus,
these audio disturbances are more likely to be
encountered by a mobile radio operating at higher
25 frequencles.
Since the available frequency spectrum is
limited, and since there continues to be an increasing
demand for wireless communication channels, it is
apparent that better utilization of the current available
30 communication channels is desired. One way to increase
the available channels is to divide the current channel
bandwidths and provide more narrowband channels. For
example, if an existing 25 kHz channel is divided in
half, two 12.5 kHz channels would exist. Obviously,
35 narrower channels require that transmitted signals occupy
less bandwidth. Decreasing the peak deviation from 5 kHz
47
- 3 - CMo0415G
to 2.5 kHz in an FM system is one way to reduce the
bandwidth of the transmitted signal in order to create
more communication channels. New communication channels
are becoming available at higher frequencies. Thus, the
05 likelihood of having smaller deviation systems which
operate at higher frequencies makes the audio
disturbances associated with Rayleigh fading a
significant problem.
Various noise reduction techniques have been used
10 to control noise bursts and reduce the overall noise
level in FM radio systems. Syllabic companding, high
frequency equalization, automatic gain control (AGC)
circuits, noise blankers, space-diversity receivers, and
various other noise reduction techniques are known.
15 However, several unique problems arise in the
implementation of techniques compatible with the audio
disturbances associated with a Rayleigh-faded signal in a
FM receiver -- especially FM systems utilizing a peak
deviation of less than 5kHz. Moreover, there is often
20 the requirement that any noise reduction technique used
in commercial FM systems must be compatible with existing
radios. For example, conventional syllabic companding,
i.e., transmitter-compression and receiver-expansion,
cannot be retrofit into many existing systems. More
25 particularly, any noise reduction implementation
requiring an IF envelope or received signal strength
indicator (RSSI) signal cannot be added to many existing
xadios not providing these signals. As a further
complication, FM simulcast systems present the unique
30 requirement that the noise reduction system be
transparent to both high-speed and low-speed signalling
and/or coded squelch data which is sent simultaneously
with the voice.
A need, therefore, exists to provide an improved
35 audio noise reduction technique which is compatible with
existing FM communication systems.
~;~75~.47
- 4 - CM00415G
Summary of the Invention
Accordingly, it is a general object of the
present invention to provide an improved FM
communications system which produces a higher quality
05audio output under poor signal-to-noise and Rayleigh
fading conditions.
Another object of the present invention is to
provide a method and means for minimizing nois~ burst
problems associated with weak and Rayleigh-faded signals
lOin FM communication rec~ivers.
A further object of the present invention is to
provide an FM receiver with improved audio quality output
under adverse noise and fading conditions, which is
compatible with systems requiring simultaneous
15transmission of low-speed data.
This invention is especially, but not exclusively
suited ~or minimizing noise bursts under weak RF signal
conditions in an FM system utilizing a peak deviation of
less than 5 kHz, and for implementation in an FM receiver
20regardless of whether an RSSI signal is available.
These and other objects are achieved by the
present invention which, briefly described, is an FM
communications system incorporating various embodiments
of the following noise reduction techniques: an FM
25receiver having improved audio response to weak and
Rayleigh-faded received FM signals; an improved noisa
attenuation device having dual-thresholds to more
effectively attenuate low level noise; an improved
automatic gain control circuit which enhances the
30opera~ion of the noise attenuation device; and a noise
attenuation device adapted for use in simulcast FM
repeater systems transmitting low-speed data.
The FM receivex according to the present
invention comprises a receiver for receiving the FM
35signal, including a discriminator for converting the FM
into a baseband signal containing both in-band audio
5447
- 5 - CM00415G
components and above-band noise components; and a noise
reduction circuit for minimizing noise bursts in the in-
band audio components of the baseband signal in response
to the average value of the above band noise components
05from the discriminator, thereby producing noise-minimized
audio signals, where the noise reduction circuit operates
only when the signal-to-noise ratio (SNR) of the received
FM signal is less than a predetermined SNR. The noise
reduction circuit includes a circuit for generating a
lOnoise control signal which is proportional to the average
value of the high frequency discriminator noise, a
circuit for generating an attenuation control signal in
response to the noise control signal, and a variable
attenuator in series with the audio to attenuate the
15audio in response to the attenuation control signal. In
another embodiment, the noise control signal is
proportional to the strength of the received FM signal.
In still a third embodiment, the attenuation control
signal is proportional to both high frequency
20discriminator noise and RF signal strength.
The FM receiver also includes a noise attenuation
circuit for selectively attenuating received audio
signals having a low audio level. An improvement to the
noise attenuation circuit changes its attehuation
25characteristics below a first predetermined input level
to approximately 2:1 output/input amplitude response, and
to at least a 3:1 response below a second predetermined
input level, to enhance the noise suppression
characteristics of the device ~ithout significantly
30affecting the audio quality.
A further improvement to the noise reduction
techniques includes a "speech-controlled A~C" circuit to
control the amplitude of the audio signal applied to the
noise attenuator. The AGC must be able to distinguish
35between a noise burst and voice signals, and must not
produce erroneous detection under Rayleigh fading
conditions.
47
In the preferred embodiment of an 800 MHz FM
mobile radio simulcast repeater system, the combination
of the noise reduction circuit and noise attenuation
circuit according to the present invention provides a
su~stantial audio quality improvement. A further
improvement to the noise attenuation circuit permits its
use in a simulcast system in which continuous low speed
data is transmitted with the voice signal by adding a
complementary attenuation path to keep the low speed data
output at a constant value.
Brief Description of the Drawings
The features of the present invention which are
believed to be novel are set forth with particularity in
the appended claims. The invention, together with
further objects and advantages thereof, may best be
understood by reference to the following description
taken in conjunction with the accompanying drawings,
which are arranged in numerical order with the exception
of Figure 18 which follows Figure 15 and Figure 23 which
follows Figure 20, in the several figures of which like-
referenced numerals identify like elements, and in which:
Figure l is a block diagram illustrating an
improved FM receiver according to the present invention;
Figure 2 is a block diagram of an alternative
embodiment of an improved receiver according to the
present invention;
Figure 3 is a graph illustrating the variation
in average tone-to-noise ratio as a function of carrier
level for an FM receiver employing the concepts of the
present invention and for a conventional FM receiver;
Figure 4 is a graph which illustrates the
output of the clipper versus signal magnitude of the
embodiment of Figure l;
1~75447
- 6a -
Figure 5 is a graph illustrating the amount of
attenuation provided in response to the signal magnitude
embodiment of Figures 1 and 2;
~5~47
- 7 - CM00415G
Figure 6 is a graph illustrating thP output of
the clipper versus signal magnitude of the embodiment
shown in Figure 2:
Figure 7 is a hlock diagram of another embodiment
of the present invention which utilizes a received signal
strength indicator (RSSI) signal;
Figure 8 is a graph which illustrates waveforms
associated with the embodiment shown in Figure 7;
Figure 9 is a schematic diagram illustrating
selective portions of the invention as shown in Figure 7;
Figure lO illustrates a lo~-to~linear converter which is
used in a modified embodiment of the invention as shown
in Figure 7;
Figure 11 is a block diagram of an FM transmitter
in which companding is used;
Figure 12 illustrates an FM receiver for
receiving a companded signal;
Figure 13 is a partial block diagram of an FM
receiver for receiving a companded signal which
incorporates noise reduction in accordance with the
present invention;
Figure 14 is a simplified schematic diagram
illustrating the improved expandor circuit according to
the present invention;
Figure 15 is a graph illustrating the output of
~he improved expandor versus the input signal magnitude
of the embodiment of Figure 14;
Figure 16 is a partial block diagram of an FM
receiver incorporating the weak signal noise reduction
technique in accordance with the present invention;
Figure 17 is an improved noise attenuation device
for use in audio processing stage 192 of Figure 16;
Figure 18 is a graph illustrating the
input/output characteristics of the improved noise
attenuation device of Figure 17;
1~7t~447
Figure 19 is a circuit diagram of an
improvement to the automatic level control (ALC) system
of U.S. Patent No. 4,514,703;
Figure 20 is a graph illustrating the effect on
the noise attenuation device characteristics of Figure 18
resulting from the addition of the ALC audio processing
stage of Figure 19;
Figure 21 illustrates a simulcast FM repeater
system utilizing the noise reduction techniques in
accordance with the present invention;
Figure 22 is a block diagram of the simulcast
noise attenuation device used for blocks 256 & 274 of
Figure 21;
Figure 23 is a graph illustrating the low-speed
data amplitude level as a function of audio input level
for the two paths of the device shown in Figure 22; and
Figure 24 is a block diagram of another
embodiment of the simulcast noise attenuation device of
Figure 22 illustrating the implementation of two
threshold levels.
Detailed Description of t=he Preferred Embodiment
Figure 1 illustrates an improved FM receiver in
accordance with the present invention. Antenna 10
couples signals to bandpass filter 12 which couples the
filtered signals to mixer 14. The output of local
oscillator 16 provides the other input to mixer 14 which
has its output coupled via bandpass filter 18 to IF
amplifier 20. The gain of amplifier 20 is controlled by
means of an input 22. The output of amplifier 20 is
coupled through bandpass filter 24 to amplifier 25 which
further amplifies and limits the signal before it is
coupled to discriminator 26. The unprocessed baseband
audio output 28 from the discriminator can be used to
provide a squelch signal to a conventional squelch
circuit or an output to a digital signal demodulator, and
7 5 4 L~7
- 9 - CM00415G
also provides an input to attenuator 30. The output of
the attenuator provides a voice output which can be
coupled to a speaker 32. It will be apparent to those
skilied in the art that the output of the attenuator can
05also be amplified by an audio amplifier (not shown)
rather than directly driving speaker 32.
Tha output of bandpass filter 24 is also coupled
to an amplifier 34 which preferably has its gain
controlled means of an input 36. The output of amplifier
1034 is coupled to a linear envelope detector 38 which
det~cts the peak magnitude of its input signal. The
output 40 of the detector is coupled to a lowpass filter
42 which has its output coupled to amplifiers 44 and 46.
These amplifiers function as comparators and provide
15outputs when the input from lowpass filter 42 exceeds VRl
and VR2, respectively. The output from amplifier 46
serves as the input 36 to control the gain of amplifier
34. The output of amplifier 44 provides the input 22 to
control the gain of amplifier 20.
The output 40 from envelope detector 38 is also
coupled to a clipper 48 which has a selected voltage
level at which clipping begins so that its output, which
is coupled to summation network 50, is limited to a
predetermined value even if the input continues to
25 increase. The output 40 is also coupled to a highpass
filter 52 which highpass filters the signal and couples
same to amplifier 54 which amplifies the signal and
provides an input to envelope detector 56. The detector
detects the peak envelope of the magnitude of the signal
30 from amplifier 54. The detector's output is smoothed by
a lowpass filter 58 which may consist of a RC time
constant circuit which has its output coupled as an input
to summation network 50. The output of summation network
50 provides an input to attenuator stage 60 which
35 attenuates the magnitude of the audio signal in response
to the control signal provided from the network 50. The
~L~'7~44~7
-- 10 -- CM00415G
output of stage 60 is coupled to th~ input of attenuator
stage ~2 which also has its attenuation controlled in
response to an input providecl from clipper 48. The
output of stage 62 consists of the audio output.
05 I~ will be apparent to those having skill in the
art that elements 10-28 are a standard FM receiver
circuits with the exception that the IF gain associated
with amplifier 20 is controlled by an automatic gain
control (AGC) circuit. The AGC circuitry consists of
10 elements 34, 38, 42, 44, 46 and 20. The purpose of the
AGC circuitry is to control the gain of amplifiers 34 and
20 by means of the output of amplifiers 46 and 44 to keep
amplifiers 34 and 20 operating in a linear region for all
magnitudes of input signals.
Figure 4 illustrates the transfer characteristic
of clipper 48 by the solid line 66. The response curve
66 has a so-called "knee" 68 so that the clipper output
is a constant level for input signal levels greater than
that corresponding to the knee.
The clipping action provided by clipper 48 is
important to the present invention. The point indicated
by numeral 70 in Figure 4 represents the operational
point of the clipper for a preferred predetermined
limiting level such as at 15 dB. The time constant
25 associated with lowpass filter 42 is such that the AGC
control signals 22 and 36 do not respond rapidly enough
to follow a typical Rayleigh fade. The purpose o~ the
AGC is to follow the average input signal magnitude
(strength) and thereby control the gain for slowly
30 changing signal conditions which are much longer in
duration than Rayleigh fades. Thus, a time constant
associated with filter 42 could be on the order of one
second for UHF radio systems where Rayleigh fades have
time durations on the order of a few milliseconds. The
35 significance of the operational point 70 as shown in
Figure 4 is that a Rayleigh fade of significant magnitude
~L~t~4~7
~ CM00415G
will cause the clipper output to fall below the knee 68
for part of the duration of the fade and hence increase
the attenuation by attenuator 30 during this time.
Figure 5 illustrates the attenuation provided by
05 attenuator 30 as a function of the output of clipper 48.
Each attenuation stage may consist of an operation
transconductance amplifier such as a CA3280E. The curve
72 illustrates that the attenuation characteristic
exhibits a knee 74 which corresponds to knee 68. For
10 clipper inputs greater than the level at the knee 74, the
attenuation is a constant. At clipper outputs below knee
74 the attenuation is inversely proportional to the
squar~ of the clipper output, that is, attenuation
increases at a greater than linear rate as the clipper
15 output decreases. Preferably the ratio X/Y is greater
than 1, where X is the increase in attenuation that
corresponds to a Y decrease in received signal strength.
This characteristic is due to the two cascaded attenuator
stages 60 and 62 each having its attenuation controlled
20in response to the clipper output. The operational point
70 in Figure 4 corresponds to operational po nt 76 in
Figure 5. Thus, during a fade as the operational point
moves below knee 68 on curve 66 and to the left of knee
74 on curve 72, it will he apparent that increasing
25 attenuation is provided which mutes or reduces the audio
gain thereby minimizing noise bursts or pops
corresponding to the Rayleigh fade.
If attenuator 30 had only a single attenuator
stage, the attenuation characteristic to the left of knee
30 74 would be linear. While such a single stage attenuator
would provide an audio reduction during Rayleigh fading,
attenuation increasing at a greater than a linear rate
provides a greater rate of muting. It will be apparent
to those skilled in the art that more than two
35 attenuation stages or attenuators having a controllable
transfer characteristic could be utilized to provide an
1~7~4~7
- 12 - CM00415G
even greater rate of attenuation below a given knee.
However, it will be appreciated that a complete or total
muting of the audio during the fade may be perceived by
the listener to be as objectionable as a noise burst.
05 Therefore, an empirical dete.rmination of the best rate of
attenuation can be selected based upon a subjective
evaluation of the audio output.
In narrow deviation FM systems, i.e., below 5 kHz
peak deviation, it is desirable at threshold or low
10 signal-to-noise conditions to provide additional muting.
~lements 52-58 provide an implementation to accomplish
this function. Highpass filter 52 passes the high
frequency or noise content above the audio range which is
amplified by amplifier 54 before being peak detected by
15 detector 56. It will be noted that the amount and
presence of the higher frequency noise increases rapidly
near threshold conditions. Time constant 58 is
preferably selected to be on the order of a few
milliseconds so that rapid fluctuations with respect to
20 the noise content can be followed. Attenuation is
provided by means of attenuation stage 60. This type of
attenuation functions for very low level received signals
such as approximately 12 dB SINAD or below because
moderate and strong signal levels have little high
25 frequency noise content. This enhances the ability of a
listener to attempt to understand low level signals by
providing audio attenuation which tracks rapidly changing
noise levels as opposed to an average signal level.
Figure 3 is a graph which visually illustrates an
30 improved audio response achieved by the present invention
in response to Rayleigh-faded received signals. The
graph shows the average tone to-noise ratio vPrsus
relative carrier level for a conventional full limiting
FM receiver by curve 78 and ~or an FM receiver embodying
35 the concepts of the present invention by curve 80. The
knee o~ curve 78 occurs at an average tone-to-noise ratio
5~7
- 13 - CM00415G
of approximately 20 and at a relative carrier level of
approximat21y 22 dB. The knee of curve 80 occurs at
slightly above an average tone-to-noise ratio of 36 and
at a relative carrier level of approximately 35 dB. For
05 relative carrier levels of approximately 15 dB and less,
a receiver according to the present invention has an
tone-to-noise ratio of approximately 8 dB better than the
conventional receiver. At a relative carrier level of
40 dB and above, a receiver accordiny to the present
10 invention has an average tone~to-noise ratio of
approximately 16 dB greater than that of a conventional
radio.
The curves in Figure 3 represent a ~.5 kHz peak
deviation receiver measured at 1.5 kHæ peak deviation
15 with a 1 kHz audio tone modulation. Simulated Rayleigh
fading having a maximum Doppler frequency of
approximately 54 Hz was utilized to simulate the fading
which might be encountered by a vehicle travelling at
approximately 40 miles per hour while receiving a signal
20 at 900 MHz.
The curves pre~ented in Figure 3 are intended to
merely provide 2 visual representation of the improved
audio response achieved by utilization of the present
invention. While Figure 3 does illustrate the benefits
25 o~ the present invention, it is difficult to fully
represent the subjective improvement in audio quality and
listening comfort achieved by the present invention by
purely visual means.
Figure 2 illustrates an FM receiver in which the
30 concept of t~e present invention is implemented by a
different embodiment. It should be noted that many of
the same elements referenced in Figure 1 are also ~resent
in the diagram shown in Figure 2.
The embodiment in Fiqur~ 2 illustrates that
35 additional circuitry 82 could be added to a conventional
FM receiver to achieve the improved results of the
~7~4~'~
- 14 - CM00415G
present invention. A single attenuation stage 84 is
utilized instead of the two stages 60 and 62 as shown in
Figure 1. A square law detector 86 is utilized instead
of the linear detector 38. Detector 86 has a square law
05 transfer characteristic instead of a linear transfer
function. For example, a properly biased diode could be
utilized as the detector 86. The general purpose of the
detector 86 is similar to that of detector 3~, that is,
it follows the peak magnitude envelope of the amplified
10 input signal except that it has a square law transfer
function.
Figure 6 illustrates the output of clipper 48 as
embodied in circuitry 82 versus the signal magnitude. It
should be noted that below the knee 88 on curve 90 the
15 response is a square law response representing the output
of d~tector 86. Above the clipper input voltage
corresponding to knee 88, the clipper output is a
constant. Point 92 as shown on curve 90 represents
operation at a preferred clipping level such as 15 dB of
20 clipping. During a Rayleigh fade in which the received
signal magnitude momentarily becomes less than the level
of clipping, the clipper's output will decrease. ~he
attenuator stage 84 responds to this change in level by
increasing the attenuation thereby decreasing the audio
25 gain during the Rayleigh fade to minimize the
objectionable noise burst. For moderate and strong
signal level conditions, the receiver shown in Figure 2
will perform substantially equivalent to the receiver as
shown in Figure 1. In this embodiment detector 86
30 provides the square law characteristic instead of the two
cascaded attenuation stages 60 and 62 of the embodiment
shown in Figure 1. Although the receiver in Figure 2
does not ~how additional audio muting due to low level
received signals near threshold, it will be apparent that
35 elements 50, 52, 54, 56 and 58 could also be utilized
with the embodiment as shown in Figure 2.
~7~ 7
- 15 ~ CM00415G
Lowpass filter 42 and amplifiers 46 and 44 derive
AGC signals that are coupled to inputs 36 and 22 of
amplifiers 34 and 20 to maintain the amplifiers in a
linear region for various average input signal levels.
05 The time constant associated with this AGC action is
substantially longer than the duration of a Rayleigh
fade, such as approximately one second.
Figure 7 illustrates another embodiment of the
present invention utilized with a conventional FM
10 receiver. Discriminator 26, attenuator 30 and speaker 32
operate in the same manner as previously described with
respect to the embodiment shown in Figure 1.
A received signal strength indicator (RSSI)
signal is provided at terminal 100 and is coupled through
15 capacitor 102 to a variable threshold clipper 104. The
generation of RSSI signals is well known. For example,
FM receivers used in cellular telephone mobile radios
generate an RSSI signal. The RSSI signal i5 preferably
proportional to the mathematical log of the magnitude of
20 the received signal in order to compress large signal
level variations into a smaller range of RSSI signals.
A DC bias voltage VR3 is gummed with the AC
coupled RSSI signal through resistor 106. The magnitude
of voltage VR3 sets the threshold level at which clipper
25 104 begins to clip. The RC time constant of resistor 106
and capacitor 102 is selected so that only the relatively
rapid Rayleigh fades reflected in the RSSI signal are
coupled to clipper 104. The functioning of clipper 104
will be further described below.
A circuit 108 receives a source of signal noise
at terminal 110 and provides an output signal voltage
VN which is proportional to the ampli~ude of noise
being received. The unprocessed audio output 28 from
discriminator 26 may be used as the noise source. The
35 circuit 108 includes a highpass filter 112, an envelope
detector 114, a lowpass filter 118, and an amplifier 116.
1~754~7
- 16 - CM00415G
The envelope detector and amplifier are biased by a
reference voltage VR4. The highpass filter 112 filters
out audio frequencies and passes higher frequencies.
These higher frequency signals are envelope or amplitude
05 detected by detector 114 whose output is lowpass filtered
by filter 118 and amplified by 116. The voltage VR4
provides a reference which corresponds to the magnitude
at which noise signals will be detected and amplified by
detector 114 and amplifier 11~. An output VN of
10 circuit 108 is provided only during low signal to noise
conditions. The purpose of output VN is to provide
additional attenuation of the audio output during weak
signal levels. This action results in a more pleasing
audio output as the audio message grows weaker relative
15 to the noise level.
Figure 8 is a composite graph showing three
waveforms: reference DC level VR3, waveform 120 which
corresponds to the magnitude of the received signal as
determined by the RSSI, and waveform Vc (shown in solid
20 line) from clipper 104 which controls the amount of
attenuation to be provided by attenuator 30, It should
be understood that waveform Vc is superimposed over
portions of waveform 120. The portions of waveform 120
which coincide with signal Vc illustrate Rayleigh
25 fading nulls which have a magnitude sufficient to cause
attenuation to be provided by attenuator 30. Bias
voltage VR3 is preferably selected so that Rayleigh fades
having a level 5-20 dB below the average magnitude of the
received signal will cause attenuation of the audio
30 output. The magnitude at which attenuation is provided
is illustrated graphically in Figure 8. The portion of
waveform 120 above waveform Vc represents signal levels
above the attenuation threshold; the threshold
corresponds to the straight horizontal line portion of
35 waveform Vc.
~.~75~4~
- 17 - CM00415G
The schematic diagram in Figure 9 illustrates a
particular embodiment of circuit 108 and clipper 104.
Amplifiers 122 and 124 and the associated capacitors and
resistors comprise highpass filter 112. Amplifier 126
05and diode 128 comprise envelope detector 114; resistor
130 and capacitor 132 comprise lowpass filter 118.
The variable clipper circuit 104 comprises an
amplifier 134 whose output is clipped by diode 136. The
control signal VN is summed through resistor 138 with
lOthe output Vc to provide additional attenuation during
low signal level conditions.
A conventional FM receiver can incorporate the
embodiment of the present invention as illustrated in
Figures 7-9 without modification of the other receiver
15circuitry. For example, there is no requirement for
automatic gain control of intermediate frequency
amplifiers. Thus, this embodiment of the present
invention can be easily added to existing FM receivers.
It is preferred that attenuator 30 in the
20embodiment illustrated in Figure 7 provide the same
attenuation characteristics as previously described.
That is, the attenuation provided when the received
signal falls below the predetermined threshold is
preferably greater than one dB of attenuation for one dB
25in reduced signal strength; a two dB increase in
attenuation for one dB loss in signal strength has proved
satisfactory. With the two attenuator stages 60 and 62
in series, a signal Vc corresponding to a linear change
of received signal strength would provide the 2:1
30attenuation change assuming each attenuation stage
contributes one dB of attenuation. However, if the RSSI
signal corresponds to a logarithmic variation of the
received signal strength, the rate at which the
attenuation will change varies with the signal strength.
35An attenuation change of approximately 2:1 can be
achieved despite the logarithmic response of the RSSI by
making an appropriate choicP of the bias voltage VR3.
1~754~7
- 18 - CM00415G
Figure 10 illustrates a log-to-linear converter
140 which can be utilized wit.h the embodiment of the
present invention as shown in Figure 7 by connecting the
output of th~ clipper 104 to its input and connecting its
05 output to attenuator 30. By providing a logarithmic to
linear conversion of signal Vc and by operating the
clipper 104 and attenuator 30 in a linear region, a
logarithmic RSSI signal is compensated for by the
converter 140 such that the control signal to attenuator
10 30 is substantially linear with respect to received
signal strength. Thus, an attenuation versus received
signal strength variation of 2:1 (during fades below the
predetermined threshold) can be precisely achieved. It
will also be apparent to those skilled in the art that
15 other ratios of attenuation versus received signal
strength can be achieved by varying the control signal
applied to attenuator 30. Alternatively, additional
series attenuator stages in addition to attenuator 60 and
62 could be utilized to achieve different ratios.
Figures 11 and 12 illustrate a typical
transmitter and receiver, respectively, in which
"syllabic companding" is utilized to improve the
effective signal-to-noise ratio. In a communications
system using syllahic companding, the amplitude of the
25 voice input signal is compressed by a compressor 142
before FM modulator and transmitter 143, and the receiver
141 utilizes an expandor 144 to expand the received audio
amplitude back to its original characteristics. Much of
~he improvement due to companding in FM communication
30 systems is due to the expansion provided at the receiver.
The expandor gain increases during audio envelope maxima
and decreases during audio envelope minima. This
produces an improvement in the average signal-to-noise
ratio of a signal which has a fluctuating amplitude, such
35 as speech, assuming that the amplitude of a noise signal
is below the amplitude of the desired received signal.
1~54~7
- 19 - CM00415G
During a Rayleigh fade which produces a noise burst with
an amplitude greater than the desired signal, the
expandor in the receiver tends to track the increasing
noise amplitude and increases the gain of the noise
05 signal. Thus, such fading will "talk up" the expandor in
the same manner as a voice signal so as to produce an
undesired audio output. This problem is especiallv
apparent when trying to receive a signal having a
relatively low signal strength. During such conditions,
10 a Rayleigh fade produces a noise signal output from the
FM receiver T~hich has a greater amplitude than the
desired signal. This noise signal causes the expandor to
increase gain, thereby amplifying the noise burst and
producing an audio output which sounds worse than a
15 system without companding.
The partial block diagram of the receiver shown
in Figure 13 illustrates that the prior embodiments of
the present invention represented by noise reduction
block 146 can be inserted between a conventional
20 discriminator 26 and circuitry 148 which is part of the
conventional circuitry utilized in a companding system.
It was not readily apparent that the combination of noise
reduction circuitry 146 in an FM receiver which utilizes
an expandor 144 would provide an improved result.
25 Although the noise reduction circuitry 146 will reduce
the noise output of the conventional receiver by reducing
the magnitude of the noise pops, it was thought that its
use with ~n expandor might produce an undesirable system
in which the expandor would ~ollow the reduced magnitude
30 level produced by the noise reduction circuits during
fading and introduce undesirable amplitude fluctuations
in the received signal.
Actual testing of the noise reduction circuitry
146 in a companding receiver showed that the overall
35 receiver system did not respond undesirably according to
the above prediction. The time duration during which
i4~7
- 20 - CM00415G
audio attenuation is provided by the noise reduction
circuitry for Rayleigh fades is shortar than the normal
time constant associated with the expandor 144. For
example, a typical time constant of approximately 20
milliseconds associated with an expandor is substantially
05 longer than the typical time intervals of audio muting
provided in response to the noise reduction circuitry 146
of the present invention. Thus, the expandor do~s not
follow the brief audio Level reductions caused by the
noise reduction circuit 146 and hence the expandor does
10 not introduce undesirable amplitude fluctuations in the
received signal.
Incorporation of the noise reduction circuitry
146 according to the present invention in a narrow
deviation FM system which utilizes companding provides a
15 significant improvement in the quality of received audio
signals at all signal levels. The improvement provided
by this combination is most apparent at relatively low
signal levels where there is a tendency of conventional
FM companding receivers to be talked up in response to
20 noise bursts which occur during Rayleigh fades.
In companding systems, the equal gain point --
defined as the point of unity gain for the compressor or
expandor - should be the same for the compressor as for
the expandor when referred to peak deviation. This
25 standardization of gain levels will give the same
loudness when changing between linear and companded modes
of operation. It is also desirable to set the
transmitter compressor equal gain point as high as
possible to keep the modulation index high. However,
30 because the compressor is syllabic with a finite attack
time, setting the equal gain point too high will result
in increased distortion due to clipping of the initial
overshoot in the transmitter audio processing.
Furthermore, if too much of the initial overshoot is lost
35 in clipping, the expandor will be slow to build up gain
at the beginning o~ syllables.
~5~47
- 21 - CM00415G
The receiver expandor also benefits from having a
equal gain point set as high as possible. A high
expandor equal gain polnt prevents the expansion of noise
at the receiver threshold. A high equal gain point in
05 the expandor also minimizes the upward expansion of audio
signals which would require increased dynamic range in
the audio power amplifier to prevent clipping.
Nevertheless, many portable radios, and some
mobile radios, have limited audio power available in
10 either the receiver audio power amplifier stage or the
receiver expandor stage, or both. This lack of dynamic
range results in clipping at the audio amplifier output.
Moreover, since the louder companded si~nals are usually
in the lcwer frequency voice range, the limited dynamic
15 range exhibits itself as a "hoomy" characteristic, which
significantly degrades the audio intelligibility. The
intelligibility is degraded even further when pre-
emphasis and de-empha6is audio shaping is usadr since the
higher frequency audio is rolled off at -6 dB/octave in
20 the receiver. Tharefore, what is needed is a
distortionless means to limit the expandor gain above a
certain audio power level.
Figure 14 is a simplified schematic diagram
representing an improvement to compandor expandor 144 of
25 Figure 13. The improvement is the clamp circuit, shown
as block 170, which causes the 2:1 Pxpandor gain to
change to a constant 1:1 value above a certain value of
audio output level. An input signal is applied to node
151, buffered by amplifier 152, and routed to the inputs
30 of both the variable gain stage 154 and the rectifier
stage. The output of the variable gain stage is
amplified by op amp 156, and the expanded output siynal
is available at node 157.
The full wave averaging rectifier stage is
35 comprised of op amp 158, current sources 159, 161, and
163, diode 160, resistor 164, and capacitor 162. The
- 22 - CM00415G
input current to the rectifier stage is supplied by the
output of op amp 158. Diode 160 performs the full wave
rectifier averaging function. The output current is
averaged by resistor 164 and capacitor 162, which sets
05 the averaging time constant. The output current is then
mirrored with a gain of two in current source 163 to
become the control current for variable gain stage 154.
When the input signal drops by 6 dB, the gain stage
control current will drop by a factor of 2, such that the
10 gain will also drop by 6 dB. The output signal at 157
will thus drop 12 ds, producing the desired 2:1
expansion.
Clamp circuit 170 serves to change gain of the
expandor so as to produce a linear 1:1 response for input
15 signals above a predetermined level. Since the path for
the gain stage controlling current is through R164, the
voltage drop across R164 is the gain term being
controlled by limiting circuit 170. The voltage drop
across R176 is the same as the voltage drop across R164,
20 since the base-emitter voltage of transistors 172 and 174
are appro~imately the same as those in current source
163. For small signals, the voltage across R164 is
small, such that transistors 172 and 174 are not forward
biased. Hence, the conventional 2:1 expansion occurs.
25 However, when the voltage drop on resistor 164 exceeds a
predetermined level (2.37 volts in the preferred
embodiment), then transistors 172 and 174 turn on to
shunt additional current supplied by current source 161.
This clamping action prevents any further increase in
30 current through R164, and hence, limits the expandor gain
at that point.
Figure 15 illustrates the gain characteristics of
the improved expandor circuit of Figure 14. The
horizontal axis represents the input voltage VIN to the
35 expandor, i.e., the voltage available at node 151. The
vertical axis represents the output voltage ~OUT f the
~75447
- 23 - CM00415G
expandor, i.e., at node 157. For reference, a constant
1:1 output/input slope line 186 has been included to more
clearly illustrate the gain altering characteristics of
the improved expandor. Additionally, a constant 1:2
05 slope line 184 has also been included to illustrate the
compression characteristics of a compandor compressor,
such as block 142 of Figure 11.
A co~ventional expandor has a constant 2:1 slope,
illustrated as line 182 in Figure 15. Point 183 is the
10 e~ual gain point, i.e., the point where a 0 dB input
produces a o ds output. A typical equal gain point would
correspond to 50~-60~ of full sys~em modulation. Solid
line 1~0 represents the gain characteristic of the
improved expandor circuit of the present invention. Gain
15 characteristic 180 follows the conventional 2:1 expandor
slope below equal gain point 183, then changes the
expandor gain to 1~1 at point 181 having an input level
above the equal gain point. In the preferred embodiment,
point 181 would correspond to at least 50% of full system
20 modulation, preferably 60% - 65%. This change in slope
is advantageous to control the upward expansion of audio
signals above the equal gain point. As an example of the
resulting improvement, a representative audio clip level
188 has been included in Figure 15 at +25 dB output to
~5 illustrate that approximately 7 dB of additional input
range can be obtained in this case. Not only does the
improved expandor circuit increase the dynamic range as
illustrated, but it also does so without adversely
affecting the audio quality of the companding system.
As previously mentioned in conjunction with
Figure 13, noise reduction circuitry 146 reduces the
magnitude of the noise pops and bursts without degrading
the operation of expandor 144. The use of the noise
reduction circuit of the present invention provides a
35 significant improvement in the received audio quality at
all signal levels. Noise reduction block 146 is
1~75~7
- 24 - CM00415G
-
preferabl~ constructed as shown in Figure 7 utilizing
both the RSSI-derived attenuation signal and the
discriminator noise-derived attenuation signal. Due to
practical considerations involving the design of RSSI
0~ circuits, it is common for the IF-~SSI signal level to
diminish under weak signal conditions. However, the
discriminator output noise increases very rapidly for IF
C/N less than 10 dB, so it is particularly important to
provide some attenuation of the audio when the IF signal
10 level is weak~ There may also be some circumstances in
which the RSSI circuits are not incorporated into the
radio design, or are unavailable for some other reason.
Thus, it is often necessary under these conditions to
provide an attenuation control signal which is not
15 derived from the RSSI.
Figure 16 illustrates a partial block diagram of
an FM receiver incorporating only the discriminator
noise-derived noise reduction elements. Discriminator
26, variable attenuator 30, weak signal noise circuit
20 108, and sp~aker 32 operate in the same manner as
previously described with respect to the embodiment shown
in Figure 7. Audio shaping block 190 may include
filtering circuitry such as a receiver de-emphasis
network. Audio processing block 192 would include
25 compandor expandor 144 in a companding system, or may
comprise additional audio processing circuitry such as
audio noise suppression or AGC.
The unprocessed baseband signal output of
discriminator 26 contains both in-band speech signals,
30 i.e., 300 Hz to 3 kHz, as well as above-band noise, i.e.,
frequencies above 3 kHz. The output signal VN of weak
signal noise circuit 108 corresponds to the magnitude of
the noise envelope above the audio band. This average
value of noise begins to increase rapidly in an IF C/N of
35 approximately 10 dB, and continues to change until the IF
C/N is below 0 dB. The rate at which the magnitude of
~754~7
- 25 - CM00415G
noise changes and the particular IF C/N at which it
begins to change depends somewhat on the particular noise
frequency band that is selected. In general, the noise
behavior is determined by natural laws which govern the
05 process of frequency detection. The rate at which noise
output signal VN changes can be varied by selecting the
gain of circuit 108, but this will also affect the
particular IF C/N at which the attenuation begins.
If the gain of circuit 108 is too low, the noise
10 output signal vN will change very little at weak signal
levels, such that the discriminator signal will not be
attenuated significantly. A burst of noise which is
produced by the discriminator when a weak, Rayleigh-faded
signal is received will then be heard by the listener.
15 If the gain of circuit 108 is too high, ~N will cut off
the attenuator as soon as any noise is produced by the
discriminator. Since this occurs when the signal is
still intelligihle, the attenuator will "chop up" the
voice signal and suppress significant message information
20 in an undesirable and annoying manner. Accordingly,
there exists an optimum value for the gain of circuit
108, which has been experimentally found to be that which
produces a reduction of 5 to 15 dB (8 to 10 dB is
preferred) in the magnitude of the audio signal level
25 when no FM signal is being received.
The operation of weak signal noise reduction
circuit 108 may bP improved by clipping the noise output
signal VN such that the attenuation control signal Vc
will not change until the IF C/N falls below some preset
30 level, i.e., 4 - 10 dB C/N. Clipper 48 performs this
attenuation threshold function. In this manner, audio
signal levels are unaffected by attenuator 30 until they
have reached the threshold of intelligibility. In the
preferred embodiment, such a clipping threshold is set
35 near the FM noise threshold of the receiver, which, in
th~ preferred embodiment, corresponds to an in-band audio
;'54~
- 26 - CM00415G
signal-~o-nolse ratio (SNR~ of approximately 12 dB SINAD
(S+N+D/N+D).
As we have seen, attenuator control signal Vc
for the noise reduction circuits of the present inYention
05 may be derived from the IF envelope, from discriminator
noise at frequencies abov~ 3 kXz, or from a combination
of both signals. If the noise reduction is to be used
with existing receivers which do not incorporate a means
of sensing the IF envelope, then the control signal Vc
10 must be derived solely from discriminator noiseO The
additio~ of only the weak signal noise reduction block to
a companding system still results in a significant and
unexpected improvement.
However, there is often the requirement in
15 commercial FM systems that any noise reduction technique
be compatible with existing radios. For example, it may
not be feasible to retrofit transmitter compression and
receiver expansion as part of the noise reduction
technigue into an existing system. However, it is still
20 possible to obtain some of the noise reduction properties
of a full companding system without compression at the
transmit end by using alternative noise reduction
techniques of the present invention.
Without full companding, and without the
25 availability of an RSSI signal, the weak signal noise
reduction technique shown in Figure 16 may still be
employed as a feedforward gain control to attenuate the
speech signal momentarily when a burst of noise occurs in
the output of the discriminator. Moreover, it has been
30 found that a significant improvement in audio quality may
be obtained through the use of an additional feedforward
gain control stagP located in audio processing block 192.
An appropriate circuit would be a speech operated noise
attenuation device (SONAD) as known in the art. SONAD
35 does not alter the amplitude of speech having high signal
levels, but significantly diminishes the amplitude of
low-level noise and weak speech signals.
~ ;~7S4~7
- 27 - CM00~15G
A SONAD is a modified syllabic expandor in which
the control signal is clipped so as to prevent the
attenuator from acting on high amplitude signals.
Moreover, unlike an expandor, the speech lev~l detector
05 of a SONAD has a faster response time to signals which
increase in level than it does to signals which decrease
in level. An improvement in the average signal-to-signal
noise of speech signals is possible with a SONAD, because
its attenuator will diminish the amplitude of noise which
10 occurs during the frequent pauses between syllables and
words in speech, without altering the level of the higher
amplitude portions of the speech. Distortion i5
minimized by choosing an appropriate level at which the
attenuator begins to act, and by choosing an appropriate
15 time constant associated with the control signal.
Known SONAD circuits produce a very unnatural
sounding suppression of the background noise when
operated under weak signal conditions. This is due to
the fast change in attenuation, which produces a dramatic
20 contrast between no attenuation and full attenuation.
Such behavior is inevitable, since a single threshold
point is used in order to achieve the desired degree of
noise suppression.
Figure 17 illustrates an improved SONAD circuit
25 for use in audio processing stage 192 of Figure 16. The
circuit of Figure 17 avoids the aforementioned problems
by employing two attenuators in tandem, each having a
different threshold point. In operation, the audio
signal from audio shaping network 190 is applied to input
30 node 195. Bandpass filter 196 receives the unprocessed
audio, and restricts the audio frequencies to those of
speech, i.e., 300 Hz - 3 kHz. If audio shaping network
190 includes a de-emphasis network, then bandpass filter
196 may be omitted. The in-band audio fxequency signals
35 are full-wave rectified and amplitude detected by
envelope detector 114, whose output is low pass filtered
- 28 - CM00415G
by filter 118, and amplified by amplifier 116. The
detected speech control signal VB is applied to
threshold 1 clipper 197 as well as threshold 2 clipper
198. The two separate threshold points produce two
05 individual attenuation control signals Vcl and Vc2,
respectively, for use by the two attenuators 60. The
noise attenuated speech is then provided at output node
199 .
The graph of Figure 18 illustrates the
10 input/output characteristics of the improved noise
attenuation device of Figure 17. The 1:2 compression
line 184, the 1:1 non-companding line 186, and the 2:1
expansion line 182 have been shown for reference, as well
as equal gain point 183. The attenuation characteristic
15 200 of the improved noise attenuation device follows the
1:1 non-companding response for strong signals. At
weaker input signals corresponding to 50% full system
modulation or less, clipper threshold 1, represented by
"kn_P" 201, is reached and the device then follows a 2:1
20 expandor curve. As the input level is lowered further,
i.e., to less than 15% full system modulation, clipper
threshold 2 is reached at knee 202, wherein the device
changes to a 3:1 attenuation characteristic. Hence, the
attenuation changes more rapidly as the input signal
25 level is reduced. This improved attenuator
characteristic minimizes speech distortion, since only
the lower level speech sounds are significantly
attenuated without affecting the higher level speech
sounds. Furthermore, the use of separate threshold
30 points more effectively reduces the noise between
syllables to produce more natural sounding speech.
Although the combination of weak signal noise
reduction and dual-threshold SONAD lacks the noise
reduction benefits of transmitter compression and the
35 high-dynamic range advantage of strong signal noise
reduction, it still provides a dramatic signal-to-noise
~75447
- 29 - CM00415G
ratio improvement. This improvement is especially
important in 800 MHz FM systems where compatibility and
retrofit capability are significant issues. Moreover,
this combination is also capable of reducing background
05 noise at the microphone, provided the deviation of the
noise is less than the first threshold (knee 201).
Still further, since a SONAD is controlled by
audio signal amplitude alone, it cannot distinguish
between a speech signal and a noise burst caused by
10 multipath propagation. The noise burst causes the SONAD
to stay "talked up" -- operating in the high VIN region
of the graph -- such that the SONAD attenuator will
seldom function. This anomaly greatly xeduces the
effectiveness with which conventional SONAD circuits
15 eliminate noise. Accordingly, the use of the instant
noise reduction circuitry preceding the SONAD attenuates
these noise bursts, and hence, greatly enhances the
effectiveness with which the SONAD attenuates noise.
Field evaluation of the weak signal noise
20 reduction/improved SONAD system revealed yet another
complication: the modulation level variation in typical
FM radio systems is sufficiently great that the SONAD
would sometimes operate at lower input levels, often
causing significant attenuation ~o the high level speech
25 signal. This undesirable condition creates objectionable
fluctuations in the speech level during parts of words or
syllables as the SONAD expands the dynamic range of the
input signal. Furthermore, in repeater systems, the
reduced average modulation level may degrade the signal-
30 to-noise ratio, and necessitate frequent adjustment of
the receiver volume control.
Since one of the principal advantages of the
noise reduction system according to the present invention
is the ability to retrofit into existing systems, the
35 level variation problem must be handled at the fixed end
as part of audio processing stage 192. Some type of
~ ~75~47
- 30 -
automatic gain control (AGC) prior to the SONA~ is
required in order to control these speech level
variations. If the AGC is located in the repeater
transceiver transmit audio processing, then the noise
reduction technique of Figure 16 may be e~ployed in
existing systems with transceiver units having a
conventional design.
A conventional AGC would bring up the level of
noise between words or syllables, which would degrade the
operation of a SONAD. Consequently, a "speech-controlled
AGC" is required which would allow the gain to be
adjusted only on voice peaks, and hold its gain fixed at
all other times. One example of this type of AGC is the
automatic level control (ALC) system described in U.S.
Patent No. 4,514,703 by Maher et al. The Maher ALC
generates an AGC control signal only when an audio signal
peak having a frequency between 300 and 1200 Hz is
detected. Since the energy content of voice is
predominantly lower than 1200 Hz, a detected spectral
energy above 1200 Hz is indicative of noise. Hence, the
desired "speech-controlled AGC" behavior is achieved.
The Maher ALC will accommodate an input signal range of
approximately 35 dB with negligible change in output
level.
The Maher ALC, as originally designed, will
operate continuously when subject to the noise produced
by strong signal multipath at fading rates corresponding
to vehicle speeds in excess of 45 miles per hour at 900
MHz. Under these particular conditions, the ALC will act
like a conventional AGC and bring up noise between words
or syllables. The problem is produced by random FM pops
which excite the audio bandpass filter and produce a
damped sinusoid transient with high peak-to-average
amplitude ratio. This ringing frequency, which roughly
corresponds to the peak of the audio shaping network
~754~7
- 31 - CM00415G
response at approximately 300 to 450 Hz, confuses the ALC
with its voice-like characteristics. A weak signal noise
reduction circuit controlled only by discriminator noise
will not remove this strong signal multipath noise. This
05 problem however would be easily obviated if the complete
strong signal/weak signal noise reduction technique of
the present invention were employed. Another benefit of
using the complete technique would be a substantial
reduction in noise when the modulation level is low and
10 the AGC is operating with high gain. However, as
previously discussed, it is not always possible to use
the complete strong signal/weak signal noise reduction
configuration of Figure 7.
Referring now to Figure 19, an improvement to the
15 Maher ALC is illustrated, which permits the weak signal
noise reduction/SONAD implementation to properly
function. The general circuit blocks of Figure 19
correspond to those of the Maher patent. (Refer to U.S.
Patent NQ. 4, 514, 703 for their description). The
20 additional circuit components 206 through ~26 represent
the improvement, and will be described herein. The ALC
improvement prevents erroneous AGC operation by
generating a variable width pulse for each audio cycle at
frequencies between 300-500 Hz. The pulse width in
~5 seconds for any audio frequency less than 500 Hz is given
by the expression: [l/f - lJ500~. The variable width
pulse is used to discharge the activity checker
integrator capacitor, which helps to prevent the activity
checker from giving a false detecticn of voice activity
30 when strong, rapidly fading signals are being received.
Some pulses will be generated when normal voice activity
is detected, but these are few enough in number that the
sensitivity of the activity checker is not significantly
degraded.
~.~7~447
- 32 - CM00415G
As illustrated in Figure 19, the scaled audio
output of the peak detector is applied to NOR gate 210
via capacitor 206. NOR gate 210, having resistor 208
connected from its input to its output, simply acts as an
05 amplifier for the audio signal~ NOR gate 212 converts
the amplified signal into a squarewave, while capacitor
214 and resistor 216 differentiates the squarewave into
pulses. TAese pulses are applied to reset the S-R flip-
flop 220/222, as well as to reset counter 224. The
10 15 XHz clock is divided by the Q6 (26 = 64) output of
the counter, such that a time interval of approximately
2 milliseconds exists between the occurrence of a reset
pulse and the positive-going transition of output Q6.
If the time between reset pulses is less than 2 msec,
15 then transistor 226 will be biased off and the normal
operation of the activity checker w-ll not be affected.
If the time between reset pulses is greater than 2 msec,
then transistor 226 will conduct during the time that the
reset pulse interval exceeds 2 msec. When transistor 226
20 turns on, the activity checker integrator capacitor
becomes clamped to VB, which prevents a voltage build
up on the detector from successive audio transients.
The improvement in operation of the SONAD with
the use of the "speech-controlled AGC" may be seen from
25 Figure 20. Without the use of the speech-controlled AGC,
the input level VIN will be allowed to vary. At normal
speech input levels corresponding to 0 dB, a o dB output
is produced. If the noise input level is -20 dB with
respect to this normal speech input level, the noise
30 output level would correspond to -40 dB. Hence, the
SONAD will improve the 20 dB average signal-to-noise
ratio of unprocessed normal level speech to 40 dB. At a
reduced speech input level of approximately -10 dB, the
unprocessed signal-to-noise ratio of 10 dB will be
35 improved to 25 dB. However, the output amplitude of the
reduced level speech has been diminished an additional
5 dB in passing through the SONAD.
1~7S44~
- 33 - CM00415G
If, however, the speech-controlled AGC is used in
combination with the SONAD, the AGC will act to maintain
a constant input level such that the effective noise
level varies with the speech level. In terms of the
05 graph of Figure 20, the speech-controlled AGC will keep
the input levels of either normal speech or reduced
speech at 0 d~. Therefore, at the SONAD input noise
level associated with reduced speech would be -10 dB, and
the noi~e level associated with normal speech would be
10 -20 dB. The SONAD will improve the 10 dB input signal-
to-noise ratio of reduced speech to 15 dB (as compared to
the previous result of 25 dB), while the SNR improvement
of normal speech remains unchanged. Although the s gnal-
to-noise ratio improvement is not as great for reduced
15 speech with the speech controlled AGC/SONAD combination,
the output level of reduced speech is maintained instead
of being further diminished. This output level
improvement is noted with vertical distance arrows in
Figure 20, comparing the constant 1:1 slope line 186
20 (dashed) to the improved SONAD response curve 230 at two
different reduced speech input levels. When preceded by
the AGC, the SONAD output is approximately 5 d~ higher
for -10 dB reduced speech input levels at point 231, and
approximately 20 dB higher for -20 dB reduced speech
25 input levels at point 232.
In review, the addition of the speech-controlled
AGC prevents the SONAD from having to operate at low
levels of input modulation. Hence, the aforementioned
problems of objectionable fluctuations in the speech
30 level and degraded signal-to-noise ratios are alleviated
with this aspect of the present invention.
Referring now to Figure 21, a simulcast FM
repeater system is illustrated utilizing the noise
reduction techniques of the present invention. The
35 repeater system incorporates multiple transmitters
operating in a "simulcast" or "quasi-synchronous"
~75447
- 34 - CM00415G
configuration in order to ensure signal penetration over
wide geographic areas. In a simulcast FM repeater
system, at least two FM signals are simultaneously
transmitted from separate base sites. The remote base
05 sites/ which are not at the same geographic location as
the repeater, are coupled to the primary base site by
means of microwave links. In Figure 21, mobile
transceiver unit A is transmitting inbound to the
repeater, while mobile transceiver unit B is receiving
10 the outbound repeated signal simultaneously broadcast
from the primary and remote base transmitters.
It is necessary that the modulation phase and
amplitude of each of the base transmitters be as nearly
identical as possible in order to avoid poor reception in
15 overlapping transmission areas. Any modulation phase or
amplitude imbalance will produce noise in the receiver
output, even when the transceiver is motionless. This
noise makes the regions in which overlapping signals are
received appear to have poor system coverage. It has
20 recently been discovered that this poor performance in
the overlap regions is often caused by bac~ground noise
in the simulcast signal distribution system which may
originate in the inter-site transmission between the
primary and xemote transmitter stations. The background
25 noise produces uncorrelated modulation of the simulcast
transmitters when received by a mobile unit located in
the overlap region.
This appears to be an ideal application for the
noise reduction technique of the present invention.
30 However, a unique reguirement is placed upon the noise
reduction system in that it be transparent to both the
3600 bit-per-second (bps) data used as signalling, and to
the 150 bps data used for coded squelch which is sent
simultaneously with the voice. The circuit which
35 satisfies these objectives represents th~ next aspect of
the present invention.
~75~47
- 35 - CM00415G
Again referring to Figure 21, mobile transceiver
unit A 240 transmits an FM signal to base receiver 242.
The transceiver units need not incoxporate any special
transmit audio processing beyond that which is normally
05 employed, i.e., pre-emphasis, clipping, and splatter
filtering. However, the mobile receiver audio processing
should include the weak signal noise reduction circuit in
combinatlon with the improved noise attenuation device
implemented as shown in Figure 16. The received signal
10 at the base is then applied to noise reduction block 244,
which, in order to be retrofit into an existing system,
may be comprisad of only the weak signal noise reduction
circuit of Figure 16. Block ~46 performs audio shaping
to meet the requirements of the particular system.
15 Speech-controlled AGC 248 is then used to maintain a
constant modulation level for noise attenuation device
250 and for transmit audio processing block 252. A
representative speech-controlled A&C would be the Maher
ALC circuit with the improvement shown in Figure 19.
20 Noise attenuation device 250 performs the audio
proc2ssing for the control console audio. In the
preferred embodiment, block 250 comprises the improved
SONAD circuit illustrated in Figure 17. Also note that
noise attenuation device 250 may alternatively be
25 interposed between AGC block 248 and the transmit audio
processing block 252 for non-simulcast repeater systems.
Data modulation is applied to transmit audio
processing block 252 for signalling and squelch. The
transmit audio is then applied to audio delay 254 which
30 may be necessary in order to equalize the modulation
phases among the base transmitters. Simulcast noise
attenuation device 256, which will be described in detail
in accordance with Figure 22, is a novel improvement to
SONAD for use in systems in which sub-audible signalling
35 must be sent together with speech on a single channel.
Primary base transmitter 258 then transmits one of the
simulcast FM signals to mobile transceiver unit B 260.
1~54~7
- 36 - CM00415G
The transmit audio from audio processing block
252 is also sent through modem 262 and multiplexer 264 to
microwave transmitter 266 for transmission to microwave
receiver 268. The received microwave signal is then
05 demultiplexed in block 270, delayed in block 272, and
audio processed by simulcast noise attenuation device 274
as was done for the primary base transmitter. The remote
base transmitter 276 then sends a simulcast FM signal to
mobile transceiver B. Similarly, if a third simulcast
10 base transmitter is required, microwave transmitter 27
would forward the signal to an appropriate microwave
receiver at another remote site.
Figure 22 illustrates simulcast noise attenuation
device 256 and 274 of Figure 21, in accordance with the
15 present invention. The audio input signal at 280 is
applied to three paths: the upper path, which will pass
any frequency input siqnal; the lower path, which passes
only low speed data; and the center path, which
determines the control signal proportional to the
20 amplitude of either the voice or high-speed data.
Current-controlled attenuators 290 and 310 are arranged
in a complementary manner such that the subaudible
signalling is always passed through the simulcast noise
attenuation device, even when the speech signal is being
25 attenuated. This constant low-speed data output level is
illustrated in the graph of Figure 23.
In the upper (voice) path, the audio is applied
through capacitor 282 to charge coupled delay line 284.
~n conjunction with clock 286, the delay line performs
30 the function of csmpensating for the delay of the noise
rejection low pass filter 308 in the lower (data) path.
The amount of loss introduced by the first current-
controlled attenuator 290 is constant for input signal
amplitudes greater than a given threshold, but increases
35 for input signals less than the threshold. As a result,
the attenuation characteristics for the upper voice path,
5~47
- 37 CM00415G
whi~h are shown in Figure 23 as response 320, has a 1:1
slope above the threshold, and a 2:1 slope below the
threshold. The 2:1 attenuation for low input l~vels
provides the noise reduction function.
05 However, the 150 bps low speed data cannot pass
through the upper path when attenuator 1 is in the high
attenuation state. Consequently, the lower path is
provided for the low speed data under these conditions.
Noise rejection filter 308 prevents input noise from
10 reaching the outpu~. This filter should have a cutoff
~requency less than 300 HZ, and have reasonahly good
phase response. The attenuation of the second current-
controlled attenuator 310 is reduced as the attenuation
if the first current-controlled attenuator 290 is
15 increased. Figure 23 illustrates this complementary
attenuation of the data path as response 322. In the
preferred embodiment, the attenuation paths cross at
approximately -1~ dB input level. The upper and lower
path outputs are summed in summation network 292, and
20 applied to the output at 294, thereby maintaining a
constant transmission path for the low speed data. This
combined path is represented in Figure 23 as constant
amplitude line 324 for the low speed data.
Voice detection prefilter 296 in the center path
25 serves to filter the noise and low speed data from the
input signal. Full wave rectifier 298 and envelope
detector 302 measure the magnitude of the voice as was
done in previous noise attenuation device embodiments.
Clipper 304 presents the aforementioned threshold such
30 that khe attenuation is constant for higher input signal
amplitudes. Complementary voltage-to-current converter
306 applies the attenuation control signal from the
clipper tu the two attenuators 290 and 310 in such a way
that the resultant amplitude of the low speed data at the
35 output of the circuit does not change as the path loss
~witches between the upper and lower paths. A DC balance
~L275~
- 38 - CM00415G
signal may be applied to current-controlled attenuator 1
to ensure that the DC level at output node 294 is
independent of the path selected. Monostable 300 is
employed to pre-charge the envelope detector ~ilter for a
05 predetermined length of time when the transmitter PTT
(push-to-talk) is activated. This prevents the loss of
initial bits of data due to the syllabic response time of
the envelope detector.
Figure 24 illustrates another embodiment of the
10 simulcast noise attenuation device having two thresholds.
The second amplitude threshold was added to the two
transmission paths in the same manner as was done in
Figure 17. A second clipper 312, a second complementary
voltage-to-current converter 316, and a pair of current-
15 controlled attenuators 314 and 318 have been added inthis embodiment. The attenuation response
characteristics of the dual threshold embodiment of
Figure 24 would appear similar to the graph of Figure 23,
except that a greater amount of attenuation would be
20 evident for output levels corresponding to low voice path
(and high data path) input levels.
In review, several embodiments of noise reduction
techniques have been illustrated in accordance with the
present invention. The noise reduction technigues are
25 especially effective in reducing the noise generated in
an FM receiver which occurs when a vehicle moves through
a region of multipath propagation. The technique may
also alleviate acoustical background noise generated at a
microphone input, as well as electrical noise generated
30 in a speech transmis ion path within a simulcast FM
repeater system. Various embodiments of the noise
reduction techniques may be incorporated into existing
systems and transceivers without creating compatibility
problems.
~X7544~7
- 39 - CM00415G
Although embodiments of the invention have been
described and shown on the drawings, the scope of the
present invention is defined in the appended claims.
What is claimed is:
05
