 |
Claims  |
|
|
We claim:
1. A self-contained, real-time spectrum analyzer, comprising:
(a) input means for accepting an externally applied electrical signal,
(b) analog-to-digital converter means for sampling said externally applied
electrical signal at said input means at periodic time intervals and for
providing a digital representation of said externally applied input signal
at said periodic time intervals,
(c) digital signal processor means having an input port and an output port,
said input port connected to receive said digital representation of said
externally applied input signal from said analog-to-digital converter
means, said digital signal processor means arranged to process said
digital representation and supply the results of said process to said
output port, said digital signal process or means comprising means for
providing real-time resolution of said digital representation of said
externally applied input signal into an array of coefficients in a
frequency domain, said coefficients representing a frequency spectrum of
said time-varying digital signal.
(d) a display interface connected to said output port of said digital
signal processor means for converting said results of said process into
electrical signals of magnitude and waveshape suitable for driving a
display, and
(e) display means connected to said display interface for providing a
graphic display of said frequency spectrum,
whereby the frequency spectrum of said externally applied electrical signal
will be computed and displayed in real-time on said display means.
2. The self-contained real-time spectrum analyzer described in claim 1,
further including a variable frequency digital clock generator wherein a
sampling frequency of said analog-to-digital converter is determined by a
digital clock generated by said variable frequency digital clock generator
and wherein a bandwidth of said frequency spectrum is proportional to the
sampling frequency of said analog-to-digital converter means, whereby the
bandwidth of said frequency spectrum is controlled by said variable
frequency digital clock generator.
3. The self-contained real-time spectrum analyzer of claim 1 wherein said
display means is an oscilloscope and said display interface is a
digital-to-analog converter.
4. The self-contained real-time spectrum analyzer of claim 1 wherein said
digital signal processor is arranged to provide a real-time resolution of
said time-varying digital signal into an array of coefficients in the
frequency domain by use of a fast Fourier transform.
5. The self-contained real-time spectrum analyzer of claim 1 wherein said
real-time resolution of a time-varying digital signal into an array of
coefficients in the frequency domain is computed at a speed equal to or
faster than the speed at which said output port contains all of the
information which is present in said externally applied electrical signal,
up to said cutoff frequency of said finite impulse response low-pass
digital filter.
6. The self-contained real-time spectrum analyzer of claim 1 wherein said
analog-to-digital converter means and said digital signal processor means
are included within a signal integrated circuit.
7. A self-contained, real-time spectrum analyzer, comprising:
(a) input means for accepting externally applied electrical signals,
(b) a periodic time interval generator for generation of a periodic time
interval output whose period is determined by a control,
(c) analog-to-digital converter means for sampling said externally applied
electrical signals at said input means at said periodic time intervals
determined by said periodic time interval generator, and for providing a
digital representation of said externally applied electrical signals at
said periodic time intervals.
(d) digital signal processor means having an input port, and an output
port, said input port arranged to receive said digital representation of
said externally applied electrical signals from said analog-to-digital
converter, said digital signal processor arranged to process in real-time
said digital representation and to supply the results of said process to
said output port, said digital signal processor means comprising means for
providing a resolution of time periodic digital signals at said input port
into an array of coefficients in a frequency domain, said coefficients
representing the Fourier transform of said time periodic digital signals,
(e) a window program with an input means and an output means, said input
means connected to receive said digital representation of said externally
applied electrical signals, said output means consisting of the
multiplication of said input by a stored table of numerical coefficients
in time sequence for providing a windowed data output,
(f) a fast-Fourier transform program for computing frequency spectra, with
an input and an output, whose input is said window program, and whose
output is a Fourier transform of said input, said output comprising said
array of coefficients in said frequency domain, each of said coefficients
having a real component and an imaginary component,
(g) a logarithm power spectral density program with an input and and
output, whose input is said real component and said imaginary component of
each of said coefficients, whose output comprises the computation of a sum
of a square of said real component and a square of said imaginary
component of each said coefficients, arranged to compute a logarithm of
said sum for each of said coefficients,
(h)a digital-to-analog converter having an input connected to said output
port of said digital signal processor means, and an output for providing
an analog output signal whose magnitude is proportional to the digital
signal at said output port of said digital signal processor means,
whereby the frequency spectra of said externally applied electrical signals
are computed continuously in real-time and provided in analog form at said
output of said digital-to-analog converter.
8. The self-contained real-time spectrum analyzer of claim 7 wherein said
analog-to-digital converter means is an oversampling analog-to digital
converter.
9. The self-contained real-time spectrum analyzer of claim 7 wherein said
digital-to-analog converter means is an oversampling digital-to-analog
converter.
10. The self-contained real-time spectrum analyzer of claim 7 wherein said
analog-to-digital converter means includes a finite impulse response
low-pass digital filter with a cutoff frequency less than or equal to
one-half the inverse of said periodic time interval.
11. The self-contained real-time spectrum analyzer of claim 7 further
including a self-contained power source with at least one output connected
to said digital signal processor means and with at least one power output
connected to said display interface for providng a self-contained source
of electrical operating power.
12. The self-contained real-time spectrum analyzer of claim 7 further
including at least one digital-to-analog converter having digital inputs
connected to said output port of said digital signal processor means, each
of said digital-to-analog converters providing an analog output signal
whose magnitude is proportional to a value present at said output port of
said digital signal processor means, whereby the frequency spectrum of
each of externally applied electrical signals is computed in real-time
provided in analog form by each of said analog output signals of said
digital-to-analog converters.
13. The self-contained real-time spectrum analyzer of claim 7, further
including a plurality of digital-to analog converters connected to said
output port of said digital signal processor means, and a program for
signal generation connected to said digital signal processor means,
whereby a digital signal may be generated in said digital signal
processor, outputted in analog form to an external system, so that an
output of said external system can be applied to said intput port for
analysis.
14. A self-contained, real-time spectrum analyzer, comprising:
(a) input means for accepting externally applied electrical signals,
(b) a periodic time interval generator arranged to supply a predetermined
output at periodic time intervals, said generator including a manual
control for determining the period of said output intervals.
(c) oversampling analog-to-digital converter means for sampling said
externally applied electrical signals at said input means at siad periodic
time intervals, and for providing a digital representation of said
externally applied input signals at said periodic time intervals, and for
frequency bandlimiting said digital representation of said externally
applied input signals,
(d) digital signal processor means having an input port and and output
port, said input port arranged to receive said digital representation of
said externally applied electrical signals from said analog-to-digital
converter, said digital signal processor arranged to process in real-time
said digital representation and to supply the results of said process to
said output port, said digital signal processor means comprising means for
providing a resolution of time periodic digital signals into an array of
coefficients in a frequency domain, representing the Fourier transform of
said time periodic digital signals,
(e) a digital-to-analog converter having an input connected to said output
port of said digital signal processor means, and an output for providing
an analog signal whose amplitude is proportional to a numerical value of a
digital signal at said output port of said digital signal processor means,
whereby frequency spectra of said externally applied electrical signals are
computed in real-time and provided in analog form at said analog output of
said digital-to-analog converter.
15. The self-contained real-time spectrum analyzer of claim 14 wherein said
analog-to-digital converter means is an integrated circuit including a
sample-and-hold function and a digital low-pass filter.
16. The self-contained real-time spectrum analyzer of claim 14 wherein said
analog-to-digital converter means includes a finite impulse response
low-pass digital filter with a cutoff frequency equal to or less than
one-half the inverse of said periodic time interval.
17. The self-contained real-time spectrum analyzer of claim 14, further
including at least a second input means connected to at least a second
analog-to-digital converter means, whereby real-time frequency spectra of
a plurality of externally applied electrical signals are continuously
computed.
18. The self-contained real-time spectrum analyzer of claim 14 wherein said
process or means for providing a resolution of time periodic digital
signals into an array of coeffiencts in the frequency domain includes a
constant percent bandwidth spectrum analysis program.
19. The selft-contained real-time spectrum analyzer of claim 1, further
including a variable frequency digital clock generator in which said
periodic time interval is determined by said variable frequency digital
clock generator and wherein the frequency bandwidth of said real-time
frequency spectra is inversely proportional to said periodic time
interval, whereby siad frequency bandwidth of said real-time frequency
spectra is controlled by said variable frequency digital clock generator.
20. The self-contained real-time spectrum analyzer of claim 1 wherein said
analog-to-digital converter means is an oversampling analog-to-digital
converter. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND
1. Description of Prior Art
The usefulness of frequency spectrum analysis has long been known to the
engineering community.
Electrical signals, such as speech, music, vibration, etc. coming from
transducers, such as microphones and pick-ups, may be analyzed and
examined by various techniques. Displaying these signals on an
oscilloscope provides a represenation of amplitude versus time. It is very
useful to obtain information about the frequency content of a signal. This
may be done by using fixed frequency analyzers, such as filters attached
to voltmeters and frequency selective voltmeters. However, these provide
only a single point of frequecy information. The mathematics of spectrum
analysis was worked out by the French mathematician Fourier, who proved
that a periodic signal may be represented by a sum of sine waves, which
include the fundamental and various harmonics. The amplitude of these
harmonics represents a spectrum and may be graphically displayed to
provide a visual indication of the frequency content of the signal,
frequecy on one axis and amplitude on the other axis. The technique of
spectrum analysis has been evolving over the years, becoming faster and
more accurate. The earliest techniques took a sample of data, for example,
a one second tape recording, and repeatedly analyzed it by mechanically or
electrically moving a filter. The result would take many times the length
of the sample to provide a complete set of data. For example, to analyze a
100 harmonic spectrum of data lasting one second, a moving filter analyzer
would require 100 seconds. Obviously, if data is coming into the analyzer
at a continuous rate, the operation of analysis takes so long that input
data accumulates much faster than it can be analyzed. This is called a non
real-time situation.
In most applications, it is critical that analysis be performed in
real-time, that is, data is analzed as rapidly as it enters the system. A
delay may be experienced between the input data and the output spectra,
however, every piece of input data is being analyzed and outputted at the
same rate. The techniques of real-time analysis have greatly benefited
from computer technology and mathematical advances, such as the fast
Fourier transform. The invention addresses itself to certain improvements
in this field.
Early analog spectrum analysis techniques that generate a graph of
amplitude versus frequency included moving filters, swept local
oscillators, and multiple filter banks. All of these techniques suffer
from a limited resolution on the frequency axis and poor dynamic range on
the amplitude axis. Another limitation was their substantial size and
weight. Over the years, such analyzers, which have occupied up to a cubic
meter of volume, have been shrunk down in size but still are as big as a
computer.
Some filter bank analyzers provide a constant percent bandwidth "third
octave" analysis, which is very limited in frequency resolution. For the
band from 20 Hz-20 kHz there are only 29 one-third octave bands.
In may applications, such as rotating machinery, vibration analysis,
underwater sound, speech research, etc., much higher resolution is
required. The fast-Fourier transform (FFT) technique has allowed computers
to take over the function of analog filters and analog technique. The
computing power required to do this is substantial and has, therefore,
limited application of real-time spectrum analysis to bulky computers
which are not portable and require substantial power. The prior art
available high resolution, real-time analyzers required either a great
sacrifice in dynamic range, bandwidth, resolution, etc., and are large,
expensive, and difficult to use, oftentimes occupying an entire rack of
equipment and costing from $10,000 to $30,000.
The prior art devices has a few fixed ranges for analysis bandwidth in a
power of 2 sequence or a 1,2,5,10 progression. Each range needs additional
filters and other hardware, increasing size and weight, as well as cost.
A spectrum analyzer using digital signal processing has a frequency range
that is determined by the sample rate of the A/D converter. Changing this
frequency range requires a corresponding change in the input anti-aliasing
filter, which is used to limit the bandwidth of the input signal,
according to the Nyquist criteria, as described in Mischa Schwartz,
"Information Transmission, Modulation, and Noise," MaGraw Hill, 1959, p.
169-180, par. 4-5, 4-6. Prior art spectrum analyzers included a series of
analog, switched capacitor, or other types of anti-aliasing filters for
this purpose.
OBJECTS
It is accordingly an object of the invention to provide a self-contained,
real-time spectrum analyzer capable of providing high-resolution frequency
spectra of audio bandwidth analog signals.
Other objects are to provide such an analyzer in a self-contained,
battery-operated, hand-held form.
Another object is to provide an analyzer whose analysis bandwidth and
resolution are continuously variable over a range of frequencies through
the manipulation of a single control.
Yet a further object is to provide such a variable frequency capability,
which may be controlled manually or by an external input, such as that
derived from a tachometer.
Another object is to provide a spectrum analyzer that includes a flat
screen visual display in a hand-held package.
Yet another object is to provide such a spectrum analyzer with a digital
output suitable for interfacing to a computer.
A further object is to provide a spectrum analyzer with wide dynamic range
and high resolution.
Another object is to provide a spectrum analyzer whose power source and
power supply are self-contained within its hand-held package.
A further object is to provide a spectrum analyzer whose analog-to-digital
conversion process eliminates the need for a separate input anti-aliasing
low-pass filter and a separate sample-and-hold circuit.
A further object is to provide analysis of a plurality of channels
simultaneously.
Other objects and advantages are to provide an improved spectrum analyzer.
Further objects and advantages will become apparent from a consideration
of the ensuing description and accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a general block diagram of a self-contained real-time spectrum
analyzer, in accordance with the invention.
FIG. 2 is a block diagram of the invention, including additional components
for continuously variable resolution and bandwidth.
FIG. 3 is a block diagram of the invention, including additional components
for a visual display and additional computer programs for fast-Fourier
transform, windowing, and log power spectral density.
SUMMARY OF THE INVENTION
A miniature spectrum analyzer with real-time audio frequency capability is
pocket size, self-contained, and battery operated. This device provides
the capability of a hand held signal processing system for analyzing
signals in the audio band. This function is extremely valuable in field
use, for industrial, and military applications. The analyzer uses an
oversampling analog-to-digital converter and a digital signal processing
integrated circuit, in combination with fast Fourier transform, and other
algorithms to compute frequency spectra in real-time. The output may be a
display such as a liquid crystal or electroluminescent matrix display, or
an analog electrical output may be provided for display on an external
oscilloscope.
BACKGROUND OF THE INVENTION
The modern techniques for spectrum analysis use digital signal processing
implement a fast Fourier transform algorithm. In this technique, the
analog input signal to be analyzed is first sampled periodically and then
converted into a sequence of digital words, which represent the analog
signal at the sampled instants in time. These digital words are applied to
a computer, which has the capability of performing mathematical operations
on the digital words under control of a fast Fourier transform algorithm.
This algorithm computes the Fourier coefficients of a set of sampled data
and provides them as a sequence of digital words, which can be further
processed or displayed.
The fast Fourier transform is a technique that was developed to increase
the processing speed of the computations required to generate the Fourier
transform of a series of time data. These techniques are well known to
those in the art and have been employed for 20 years in the signal
processing industry. See, for instance, "Digital Processing of Signals,"
by Bernard Gold and Charles M. Rader, McGraw Hill, 1969, Chapter 1 and
Chapter 6.4, pp. 173-186. The advance of the integrated circuit art has
allowed the entire digital signal processing computer to be inplemented on
a single integrated circuit. The computational algorithms may be stored in
a read only memory and the entire system can operate on signals in
real-time. The input part of the system is the conversion from the analog
to the digital domain.
The are many techniques for implementing the conversion from an analog
input signal to the digital form. The simplest technique is the use of
successive approximation or dual slope converters, which are techniques
well known to those of skill in the data conversion art. These methods
work for low speeds (below 3 kHz), low resolution (12 bits or less) but
will be too noisy or slow to covert wide dynamic range audio signals
(80-100 dB) at full (20 kHz) audio bandwidth.
In addition, these converter methods have excessive distortion, and power
consumption. In addition, the input signal must be band limited through a
low-pass filter and held constant during the conversion time by a separate
sample-and-hold circuit prior to entering the A/D converter.
These problems have led to the search for better and faster conversion
techniques, which obviate the need for these additional components in
front of the converter. The preferred technique is to use an oversampling
converter in which the input signal is applied to a low resolution (1-6
bits) very high speed A/D converter, sampling the input at say, 64-128
times the audio sampling frequency. This technique is described in: "A
Stereo 16 Bit Delta-Sigma A/D Converter for Digital Audio", by D. R.
Welland, B. P. Del Signore, B. J. Swanson, T. Tanaka, K. Hamashita, S.
Hara, K. Takasuka; paper presented at 85th Convention of the Audio
Engineering Society, Nov. 1988 and "Design and Implementation of an Audio
18-Bit Analog-to-Digital Converter Using Oversampling Techniques", by
Robert W. Adams, Journal of the Audio Engineering Society, Vol. 34, No. 3,
Mar. 1986.
It can be shown that all of the information required for an N bit
conversion at a sampling frequency fs is obtained by oversampling at M
times fs with a resolution of N minus log.sub.2 M. The decimation of the
oversampled data to the audio frequency sampling rate is done in the
digital domain by digital filtering, decimation, etc., as described in the
references. A single integrated circuit can contain the high speed A/D,
latch, decimator, digital filter, needed for a complete A/D converter.
There are several different topologies for this technique, including:
feedback differential converters, Delta modulation, differential pulse
code modulation, feedback pulse code modulation, and noise shaping. All of
these techniques are similar and differ only in the position of input and
feedback components, and in the number of bits in the high speed
oversampling converter. The output of these types of converters is linear
pulse code modulation, and practical implementations can yield 16 or more
bits for resolution.
The successive approximation and dual slope A/D techniques require an input
anti-aliasing low-pass filter to bandlimit the input signal prior to
sampling, and a sample-and-hold circuit, which samples the input signal
periodically and provides a held output constant between sample times, to
the A/D converter. These components are is necessary because these
converters produce erroneous results if the input voltage changes during
the course of a conversion cycle.
The oversampling A/D converters do not require either of these components
and, thus, provide substantial savings in space, power consumption, and
cost, which is particularly important for the present invention. The
function of the input low-pass filter is replaced by a digital filter,
within the oversampling converter, which operates on the oversampled data
prior to exiting the A/D converter. The effect of this digital low-pass
filter is to provide the same function of anti-aliasing as the external
low-pass filter serves in the sucessive approximation system.
The oversampling converter input stage, which is a high speed comparator or
a "flash" converter, which is generally 3-6 bits in resolution. Because
these types of converter circuits are composed of high speed comparators,
they operate continuously and do not require a sampled-and-held input. The
instant of sampling is determined within the oversampling converter by a
digital latch on the data coming out of the compartor or flash A/D. This
internal latch replaces the function of the external sample-and-hold.
By using an oversampling A/D converter, the overall volume, complexity,
cost, and power consumption of the system is greatly reduced, to the point
where a hand-held, lightweight conversion system becomes practical.
Additional savings are realized in the elimination of the low-pass filter
and sample-and-hold circuits.
DESCRIPTION OF FIG. 1
An audio bandwidth analog input signal 100 is applied to an input 102 of an
analog-to-digital (A/D) converter, 104, which changes the analog
information into digital form. A/D converter 104 takes a "snapshot" or
sample of the input voltage at a particular time instant and changes the
analog voltage into a digital representation using techniques well known
to those of skill in the art. The instant of sampling is defined by an
edge of the digital signal applied to line 105, called the sample command.
This can be generated periodically by an oscillator, such as a quartz
crystal, with a stable period and a fast rising pulse or squarewave
output. By sampling at uniform time instants, it an by shown that a
complete representation of the signal will be obtained if the sampling
frequency is at least twice the signal bandwidth. See, eg., op. cit.,
Mischa Schwartz, "Information Transmission, Modulation and Noise." For
example, with an input signal of bandwidth B Hertz, the Nyquist sampling
rate is 2 B Hertz, and this would by the minimum sampling frequency for
the A/D conversion process.
The preferred embodiment of the real-time analyzer uses an oversamping A/D
converter 104. This provides low power consumption, high speed, and high
resolution in a single integrated circuit chip. These converters also
provide an inherent anti-aliasing filter and sample-and-hold capability,
thus eliminating the need for these external discrete circuits. Note that
no separate sample-and-hold or low-pass filter are indicated in the
figures, although they would be neccessary if a non-oversampling A/D
converter were used for 104.
Commerically available integrated circuits, such as the Crystal
Semiconductor part number CS-5326 or the Motorola DSP56ADC16, include all
of these functions within a single package in the dimensions of an
industry standard, 28-pin or 16-pin, dual-in-line package.
The self-contained analyzer is made realizable by the use of these
oversampling integrated circuits because other A/D converter techniques
providing high resolution (16 bit) conversion at audio frequencies consume
so much power and space that they cannot fit into a hand-held portable
instrument.
The digital output 106 of A/D converter 104 is applied to a digital signal
processor integrated circuit 108. The so called "real-time" operation on
the analyzer is critical in many applications. The speed of the digital
signal processing (DSP) program will be a trade off between the processing
power of the DSP integrated circuit, the bandwidth of the input signal,
and the resolution of the frequency spectrum, which is the result. The
prior art systems were incapable of providing high resolution real-time
analysis in a hand-held package. So called 1/3 octave analyzers provide a
constant bandwidth analysis useful mainly for room equalization. However,
these devices are incapable of providing a high resolution constant
resolution analysis.
The capabilities of the system described herein are limited only by the
analysis speed of the digital signal processor integrated circuit. For
example, commercially available units such as the Texas Instruments
TMS320C25 or Motorola DSP56001 operate at such a high rate of computation
speed that they are capable of computing 1024 point, 20 kHz spectra in
real-time.
This digital signal processor is controlled by a stored program, unit 112,
which is, in turn, controlled by a digital signal processor program with a
set of spectrum analysis processing programs, unit 116. Unit 112 contains
the stored programs. Digital signal processor 108 selects and controls a
program in unit 116 and unit 112, under control of a manual control panel
118. Processor 108 receives data 106 in real-time from A/D converter 104
and performs various computations, continuously outputting results 122 in
real-time.
Processor 108 is controlled by the user through manual controls 118 via
lines 120. The user can change the type of analysis, frequency range and
resolution, amplitude scale and various other functions by manipulating
manual controls 118.
The output of processor 108 is a serial or parallel stream of digital data
122. This may be directly utilized by an external computer or other
digital device via output connection 140. The preferred embodiment of the
invention uses one or more digital-to-analog (D/A) converters 124
connected ot output 122 of processor 108. D/A converters 124 change
digital information 122 to analog form at outputs 126 using techniques
that are well known to those skilled in the data conversion art. Outputs
126 are applied to connectors 128. The user may then observe the results
of the analysis on an oscilloscope or other external indicating device.
Processor 108 will generally provide several channels of output 122. The
data may be simultaneously available in parallel form, or may be
alternately outputted via time multiplexing techniques on a single serial
channel.
One use of the several outputs is to use one channel to represent the
frequency axis and the other channel to represent the amplitude axis of
the analysis. By connecting these channels to an external X/Y
oscilloscope, a frequency spectrum is displayed. Alternatively, one
channel could provide a synchronization signal and the external
oscilloscope could then operate in a time base mode, triggering from one
channel and observing the amplitude on the other channel.
Another application of these multiple D/A channels is to provide dual
output functions, which have two simultaneous output channels of
amplitude. For example, one channel could represent phase and the other
one magnitude. Alternatively, more than one input channel 100 and A/D
converter 104 could be connected to processor 108, thereby providing a
dual-channel input function. In this embodiment, two simultaneous analysis
could be performed by processor 108 and outputted on two analog channels
via D/A converters 124 and output connectors 128. Cross spectra and
coherence functions are examples of dual channel functions which can by
implemented in software, using the same hardware. These are well known
techniques.
The real-time spectrum analyzer is completed by adding a power source 132,
which could be a recargeable battery pack, regular batteries, such as
lithium or alkaline batteries, or other self-contained power sources. The
power source could also be a power supply operated from the power line,
such as a plug-in transformer type. The power source generally will
provide a single voltage, which varies somewhat with the state of the
battery charge. The output of power source 134 goes to a power supply 136,
which takes the variable voltage basic power source and converts it to
regulated fixed voltages of various magnitude and polarities 138, which
are required by the other parts of the analyzer.
Power supply 136 may be an inverter, rectifier, filter capacitor, and
linear regulator type, or various pulse width modulated switched mode
power supply techniques may be used. These devices are available as
commercial modules from various manufacturers or may be custom designed
for the application, using techniques well known to those skilled in the
power supply art.
The combination of a self-contained power supply, oversampling A/D
converter, real-time digital signal processor integrated circuit, FFT
spectrum analysis programs and digital-to-analog converters provide a
complete signal processing system in a hand-held miniature package.
DESCRIPTION OF FIG. 2
FIG. 2 includes all of the components of the already described
self-contained real-time spectrum analyzer, but also includes the feature
of a continuously variable analysis bandwidth. This function is extremely
valuable, since it allows the analyzer to zoom in on a particular
frequency range of interest. For example, in the study of speech, it may
be desirable to observe only the low frequency components below, say, 500
Hz in order to accurately ascertain pitch. One could then increase the
analyzer bandwidth and analyze with 3 kHz bandwidth to see the format
structure of the speech. By using the full audio bandwidth of 20 kHz, high
frequency components, such as silibants and fricatives, may be observed.
The prior art analyzers had the limitation of requiring a different
anti-aliasing filter for each frequency range. In addition, many of the
A/D converter and sample-and-hold techniques had a limited range of
frequencies over which they would provide high performance. The use of an
oversampling A/D 104 eliminates all of these problems.
Variable frequency oscillator (VFO) 200 controlled by a frequency control
signal 204. The frequency control signal is a DC voltage or current whose
magnitude proportionally controls the frequency of output 202 of the VFO
200.
The preferred embodiment utilizes an oversampling A/D converter 104. This
type of converter includes a clock input, which determines the sample
frequency of the A/D converter, the sample instant of the internal
sample-and-hold, and the bandwidth of the digital anti-aliasing filter,
all of which are included in the A/D converter integrated circuit. By
varying the master clock frequency to the A/D converter, the bandwidth of
the internal digital filter is adjusted automatically to be one-half that
of the sample frequency, thus, automatically and continuously satisfying
the Nyquist criteria as described in the reference Schwartz, op. cit.
This means that a variable frequency spectrum analysis may be obtained for
any desired bandwidth within a wide range, merely by selected the master
clock frequency. The necessity for a bank of anti-aliasing filters is
thereby eliminated.
Variable frequency oscillator techniques are common in the art of analog
design. Variable frequency output 202 is applied to a sample command
intput 105 of A/D converter 104. This allows the sample rate of the input
signal and, hence, the bandwidth and resolution of the subsequent signal
analysis to be continuously varied.
In a model of the invention, a continuous frequency range from 100 Hz to 50
kHz has been analyzed in real-time at resolutions of 512 and 1024 points,
using the Motorola DSP56001 digital signal processor for block 108 and the
Motorola ADSP16 oversampling A/D converter for block 104.
A DC frequency control 204 input to oscillator 200 may be obtained from a
manual control, such as a potentiometer, for block 212 or from an external
input 208. A tachometer could be used to provide this external input 208
so that the frequency bandwidth of the analyzer tracks some external
event, such as the speed of rotation of a piece of machinery whose
vibration is to be analyzed. A switch 206 is used to select either of
these inputs.
Alternatively, A/D sample command 105 can come directly from an external
frequency source, which could also be from a rotating shaft via the use of
a magnetic switch or contact. The sampling frequency could be generated
using frequency multiplication techniques, such as phase lock loops, which
are well known to those skilled in the art.
Sampling frequency 105 may also by automatically controlled by digital
signal processor 108 so as to adapt to a changing signal situation. This
may be highly useful in certain types of analysis, such as speech,
acoustics, underwater sound, or analysis of intelligence information,
which is rapidly changing its spectral bandwidth. In also allows
synchronization to an external varying clock, such as the output of a
tachometer on the shaft of a machine being analyzed.
DESCRIPTION OF FIG. 3
The analog or digital output of the embodiment of FIG. 1 and 2 requires an
external device to provide a display or to utilize the information. In
many applications, the sole use will be to provide a graphic display so
that the user can immediately ascertain the spectrum. In this case, a
self-contained display capability can be added to the digital signal
analyzer within the scope of the current invention.
This system is shown in FIG. 3, which includes the previous system.
However, the output of digital signal processor 108 on lines 122 now
enters a display register and driver 300. This is an integrated circuit or
other type of circuit that provides the storage capability for an entire
screen of information in the form of a matrix or "bit mapped" display. The
organization of an internal random access memory (RAM) is an image of an
X/Y matrix display. The output of the digital signal processor 108 now
includes a frame synchronization signal, which indicates the beginning of
a new set of spectra. Generally this signal will begin at the zero
frequency or DC line of the spectrum.
Display register and driver 300 will begin to sequentially fill in RAM
memory locations with the spectral data, starting with the zero (DC)
spectrum line. Display register and driver 300 includes circuitry that
provides a row and column scanning of the RAM and level shifting, which
can operate a visual display 304 via lines 302. Commercially available
products can perform these functions, such as the Densitron LM656 display
controller. Display 304 can be an electroluminescent or light emitting
diode (LED) dot matrix or in the preferred embodiment, it can be a liquid
crystal device (LCD) matrix display.
The LCD uses very low power and can present graphic information at high
resolution. For example, a commercially available display panel from
Toshiba provides 640.times.400 point resolution. The hand-held embodiment
of the invention could use a somewhat smaller display of, say,
128.times.128 or 128.times.256 pixel. The liquid crystal display uses a
flat panel of specially prepared chemicals sandwiched between glass or
plastic with a matrix of transparent electrodes deposited on it. By
putting appropriate voltages on these electrodes, a point on the X/Y
display may be made opaque or transparent. By rapidly scanning the display
and energizing various X/Y locations, a graphic visual display may be
obtained. The flat panel liquid crystal device is driven by a display
register and driver 300, specially designed for the particular LCD 304.
FIG. 3 shows typical types of programs that can be used for the digital
signal processor. Stored program 112 will access a fast-Fourier transform
306, which provides the basic spectrum analysis function. The spectrum
from an FFT will include certain undersirable artifacts, such as side
lobes. These are due to sharp edges in the block of data analyzed and may
be greatly reduced by the application of a "windowing" function using
techniques well known to those of skill in the digital signal processing
field. The windowing multiplies the input signal by a function, such as a
raised cosine, which is stored as a set of points in a look-up table. This
windowing program is shown as 310 in FIG. 3 and is also accessed by stored
program 112.
The output of a fast Fourier transform is real and imaginary linear
amplitude components. Input signals to system 100 may occupy a very wide
dynamic range. In this case, a display of the log of the magnitude of the
real and imaginary component provides a better indication, especially for
wide dynamic range amplitudes.
The inclusion of a log power spectral density program 314 solves this
problem. This program takes the raw FFT data from the FFT program 306,
which is complex points (real and imaginary pairs) and first performs a
square root of the sum of the squares operation to obtain an magnitude. By
taking the logarithm of this magnitude, an output presentation directly in
the form of decibels may be obtained. The scale of the presentation on
visual display 304 can be controlled by manual controls 118. Window
function 310 may also be manually controlled in order to allow the user to
vary trade off between side lobe suppression and analysis resolution.
SUMMARY, RAMIFICATIONS AND SCOPE
The reader will thus see that we have provided a self-contained real-time
spectrum analyzer capable of providing high resolution frequency spectra
of audio bandwidth analog signals. It is self-contained, battery-operated,
and hand-held. Its analysis bandwidth and resolution are continuously
variable over a wide range of frequencies through the manipulation of a
single control. Also, it has a variable frequency capability, which may be
controlled by an external input, such as that derived from a tachometer.
It includes a flat screen visual display in a hand-held package or a
digital output suitable for interfacing to a computer. It includes a
window algorithm, a log power spectral density program, and a fast-Fourier
transform program. Its power source and power supply are self-contained
within a hand-held package. Its analog-to-digital conversion process
utilizes an oversampling A/D converter so as to achieve smaller size and
weight by eliminating a separate input anti-aliasing low-pass filter and
sample-and-hold circuit. It provides analysis of a plurality of channels
simuiltaneously.
Although we have described our inv | | |
