BACKGROUND OF THE INVENTION The present invention relates to circuitry for processing electrical signals, and more particularly to circuitry for the frequency compression of such signals to enhance their subsequent further processing and analysis.

Over the years, numerous instruments have been developed for measuring physical phenomena by detecting energy generated due to the phenomena and converting the energy to electrical signals. The electrical signals are processed to obtain information on a wide variety of physical characteristics, e.g. the location, size and velocity of an object, or its index of refraction, transmissivity or surface characteristics. In many applications, signals occur intermittently and randomly and vary in their amplitude and duration. Further, the signals may be expected to occur over a wide frequency bandwidth in which frequencies near the upper end of the bandwidth are at least several orders of magnitude greater than frequencies at the lower end.

While such applications can involve radar and sonar, the present invention is more directly concerned with laser Doppler velocimetry (LDV) , also known as laser Doppler anemometry. LDV systems are known for their utility in measuring instantaneous velocities of discrete elements in two-phase flows, e.g. liquid sprays in air, or fine particles in fluid streams. In a conventional LDV

apparatus, two spatially separated laser beams intersect one another and interfere with one another to form a measuring volume. The measuring volume is typically quite small, and the concentration of particles sufficiently low, so that at any given time no more than one particle is within the measuring volume. Several photodetectors receive the coherent light scattered by particles passing through the measuring volume. The Doppler frequency, obtained by measuring the electrical signal generated as a function of light received by the photodetector, measures particle velocity and thus the velocity of the particle-carrying medium. Doppler frequency is proportional to particle velocity.

Laser phase Doppler systems are closely related to LDV systems and employ a phase difference between two separate Doppler frequency signals to determine the size of a moving particle. More particularly, several photodetectors receive light scattered by a particle passing through a measuring volume. The photodetectors are spaced apart from one another to receive scattered light. The difference in phase between signals generated by any two of the photodetectors provides the particle size information.

Particles, especially spherical particles, tend to scatter light in all directions and thus lend themselves well to analysis by laser Doppler techniques. However, physical characteristics of systems under analysis can lead to problems in analyzing the resultant electrical signals.

Given the typically low particle density and small size of the measuring volume, particles traverse the measuring volume individually and in intermittent, random fashion. The resultant electrical signal is a composite of background noise and occasional coherent frequency components, known as "bursts" superimposed on the background noise. Thus, signal processing circuitry should be capable of distinguishing the coherent frequency bursts from noise to avoid wasteful attempts to analyze the noise. The coherent frequency bursts tend to be non-uniform in amplitude, frequency and duration. Signal amplitudes vary with particle size, but also with varying tendencies of particles to absorb rather than scatter the laser energy. Signal frequencies can vary over orders of magnitude, particularly in turbulent flow systems. The length or duration of the bursts can vary considerably, even under conditions of uniform particle size and velocity, depending upon whether a particle is substantially centered as it traverses the measuring volume. Known analog processing techniques, e.g. involving LDV counters, are capable of processing coherent frequency bursts in real time. However, the signal is prone to distortion and noise that depends on signal frequency, amplitude and processing bandwidth. Thus, the nature and degree of distortion is difficult to predict.

Alternatively, the electrical signals can be sampled and converted to corresponding digital signals and then

processed digitally. While digital sampling and processing do not eliminate distortion, the distortion is more predictable. Typical digital processing techniques include signal correlation and fast Fourier transform. A disadvantage of this approach is that sampling and processing time can considerably exceed burst durations, seriously decreasing the rate at which the burst signals can be processed.

Given the wide range of possible burst signal frequencies, circuitry for processing the signals must be capable of functioning over a broad frequency band. Thiε requirement calls for circuitry which is more complex, more costly and less reliable than circuitry tailored to handle a narrower frequency bandwidth. To address this difficulty, analog and digital processors have been used to heterodyne, or translate, the signal frequency. Typically, analog mixing techniques are employed to limit the frequency range of signals subject to further processing, thus to reduce the complexity of processing circuitry. Analog mixing, however, is subject to reduced signal-to-noise ratio due to mixer insertion loss and post-mixer amplification and filtering. Intermodulation distortion and harmonic distortion also increase noise. Therefore, it is an object of the present invention to provide circuitry for real time processing of analog signals, while minimizing noise and signal distortion.

Another object of the invention is to provide a process for rapidly compressing the frequency bandwidth of electrical signals before further processing of the signals. A further object is to provide a means for receiving signals over a broad range of frequencies and conditioning the signals for further processing within a considerably narrowed frequency bandwidth.

Yet another object is to provide simpler, less costly and more reliable circuitry for frequency compression and frequency translation of coherent frequency bursts.

SUMMARY OF THE INVENTION To achieve these objects, there is provided a procesε for measuring a coherent frequency analog signal, including the following steps: a. receiving an analog coherent frequency signal, and generating a frequency range value indicating an approximate frequency of the received analog coherent frequency signal; b. predetermining a plurality of sampling rates, each sampling rate corresponding to a different one of a plurality of frequency bandwidth segments that together comprise a frequency bandwidth for receiving the analog coherent frequency signal; c. selecting one of the plurality of sampling rates, based on the frequency range value;

d. converting the analog coherent frequency signal to a first digital signal by sampling the analog coherent frequency signal at the selected sampling rate; e. storing the firεt digital signal to a memory as a sequence of digital valueε, each digital value corresponding to one sample of the analog coherent frequency signal; f. reading the first digital signal out of the memory in the sequence and at a predetermined read frequency to generate a second digital signal; and g. converting the second digital signal to an analog signal, thus to generate a reconstructed analog signal corresponding to the analog coherent frequency signal.

Preferably, each bandwidth segment is predetermined as to range, such that the ratio of its highest frequency to its lowest frequency is at moεt two. Accordingly each bandwidth segment has a range of one octave or less. The full bandwidth typically is a decade, i.e. with the highest frequency being about ten times the lowest frequency of the range. A controller can be provided to select one of several sampling rates, each rate associated with one of the bandwidth segments. Consequently, despite a broad range of initial signal frequencies, appropriate matching of sampling rates and estimated frequencies controls and confines the relationship of the signal and sampling frequencies. For example, the number of samples per cycle

of the analog coherent frequency signal can be confined to within a predetermined range such as 5 - 10 samples/cycle. Conversion to the first digital signal proceeds at the selected sampling rate, and preferably comprises generating multiple eight-bit binary words, each binary word corresponding to one of the samples and representing a digital value. The first digital signal iε stored as a sequence of the binary words.

The second digital signal is generated by reading the digital values out of the memory, in the sequence, and at the predetermined read frequency. While the read frequency most directly controls the number of digital values or samples read per second, it also controls the cycles per second, based on the aforementioned relationship of samples and cycles of the original signal. More particularly, if appropriate selection of sampling rates confines the number of samples per cycle to a range of 5 - 10, the resulting digital signals read out of the memory fall within a one octave frequency range. The reconstructed analog signalε, likewise, lie within this limited frequency bandwidth. The predetermined read frequency further positions the narrowed frequency bandwidth, i.e. at least approximately determines the maximum and minimum frequencies within the bandwidth.

To further enhance signal processing accuracy, the reconstructed analog signal can be low paεε filtered (preferably approximately at the Nyquist value) before it is processed. The available range at which coherent

frequency bursts are proceεεed can be enhanced conεiderably by band-pass filtering of the incoming analog coherent frequency signal. In a specific application, sixteen band¬ pass filters are used in combination to sense frequencies varying from 300 Hz through 100 MHz.

The manner in which the reconstructed analog signal is processed depends upon the requirements of the physical system. In a velocity measurement syεtem, the reconstructed analog signal is processed to accurately determine its frequency. This measuring frequency is combined with the known selected sampling rate to accurately determine the frequency of the original analog coherent frequency signal, i.e. the burst frequency. Then, particle velocity iε determined aε a function of the burεt frequency.

Alternatively, particle εizeε can be determined based on differences in phase between several burst frequencies based on the same particle. This requires two or more photodetectors angularly spaced apart from one another, each generating an analog coherent frequency burst responsive to receiving energy scattered by the particle. The burst signals are separately converted and reconstructed as described above, for procesεing within a narrowed frequency bandwidth. A phaεe-detecting meanε is employed for determining a phase difference between the reconstructed signals. The particle size is determined as a function of the phase difference.

Thus in accordance with the present invention, analog coherent frequency signals can be sensed over a broad frequency bandwidth, converted to digital information, then reconverted to provide reconstructed analog signals confined to a narrow frequency bandwidth for processing. When so confined aε to their frequencieε, the reconstructed signals lend themselves to processing by analog techniques employing simpler, less costly circuit components, high processing speed, and enhanced reliability due to reduced distortion and noise. This considerably improves the accuracy of measurements obtained in laser phase Doppler and laser Doppler velocimetry systemε, and similarly can improve radar and sonar applications involving wide frequency ranges. IN THE DRAWINGS

For a further appreciation of the above and other features and advantages, reference is made to the following detailed description and to the drawings, in which:

Figure 1 is a schematic view of a laser Doppler measurement system including signal processing circuitry;

Figure 2 is a more detailed view of the optical elements of the laser Doppler system;

Figure 3 is an enlarged partial view of Figure 2;

Figure 4 is a schematic view of a bandpass filtering device shown in Figure 1;

Figure 5 is a more detailed schematic view of a frequency translator shown in Figure 1;

Figures 6 and 7 are timing diagrams, showing an electrical signal at various stages of its processing by the frequency translator;

Figure 8 is a schematic view of an alternative embodiment frequency translator employing selectively phase-shifted A/D converters and memories;

Figure 9 schematically shows part of a size-measuring device constructed in accordance with the present invention; and Figure 10 schematically illustrates a part of velocity measurement device constructed in accordance with the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Turning now to the drawings, there is shown in Figure 1 a laser Doppler system 16 for generating electrical signals based on particles or other light-scattering elements moving through a particle measurement volume. The system processes the signals to determine sizes, velocities, or other information about the moving particles. The system includes optical apparatus 18 for generating an electrical signal as a function of sensed particles. Typically, particles are conveyed by a medium, usually a gas or liquid, through a meaεuring volume defined by interεecting coherent energy beamε. The particle concentration is sufficiently low, and the measuring region sufficiently small, to virtually eliminate the posεibility of more than one particle being within the measuring region

at any given time. Accordingly, the signal generated by apparatus 18 is a composite signal that includes analog coherent frequency components known as "bursts", and background noise. Particles cause bursts of widely varying amplitudes, frequencies and durations, depending on their size, speed, index of refraction, absorption characteristics and the extent to which they are "centered" as they traverse the measuring volume. The background noise occurs when no particle is present within the measuring volume.

The composite signal, including the bursts and background noise, is provided to an amplifier 20 and the amplified signal then provided to a bandpass filtering device 22, where the signal is bandpass filtered to limit its frequency range. The filtered signal is provided to a burst detector 24 and a frequency translator 26.

Burst detector 24 can be of the type described in U.S. Patent No. 4,973,969 (Jensen), issued November 27, 1990 and incorporated by reference herein. Burst detector 24 provides a signal burεt gate 27 to frequency tranεlator 26 indicating that a burst has been detected, and a frequency range signal 29 indicating an approximate frequency of the burst.

Based on the eεtimated frequency, frequency tranεlator 26 selects a sampling rate for sampling the analog output of filtering device 22 and converting that output to a digital signal. Within the frequency translator, the

digital signal is selectively modified as to its frequency and converted to a reconstructed analog signal corresponding to the filtering device output.

The reconstructed analog signal, confined within a frequency bandwidth of about one octave, is provided to signal processing circuitry 28 for determining the signal frequency, phase or other characteristic correεponding to the physical phenomenon being measured. Processing circuitry 28 provides an output to a microprocessor 30, which can be a personal computer. The corresponding physical measurement values can be shown on a display terminal 32 connected to the microprocesεor.

Figure 2 shows the optical apparatuε in greater detail. A diode laser 34 generates the coherent energy in a beam 36 that diverges as it approaches a collimating lens 38. Beyond lens 38, a beam splitter 40 receives the collimated beam and generates a pair of collimated laser beams 36a and 36b, directed longitudinally parallel to a beam axis 41, toward a focusing lens 42. Beams 36a and 36b are transversely spaced apart and define a beam plane.

Focusing lens 42 causes beams 36a and 36b to intersect one another at a measurement volume 44. Lenε 42 alεo bringε the beams to a focus at the measurement volume. As the beams cross, they interfere with one another to form interference fringes, i.e. alternating regions of higher and lower intensity. Downstream of measurement region 44,

bea s 36a and 36b encounter a second focuεing lenε 46 which directε the beamε to a beam εtop 48.

The optical εystem further includes optical receivers for sensing laser energy scattered by particles or other elements travelling through the measurement volume. One of the optical receivers includes a convex lens 50 for collecting and columinating scattered light, a photodetector 52, and a convex lens 54 between lens 50 and photodetector 52 for focusing the collected light onto the photodetector. Photodetector 52, which can be an avalanche photodiode, generates an analog voltage as a function of received light. The other optical receiver is similar, and includes a collecting lens 56, a focusing lenε 58 and a photodetector 60. Both of the optical receivers are positioned off of the beam axis to insure in each case that scattered light reaches the collecting lens via a non-longitudinal path. As shown, collecting lenses 50 and 56 are positioned to receive forward scattered light. The lenses can be positioned to receive back scattered light, if desired. Each collecting lens is poεitioned so that its focal point occupies the measurement volume. Further, lenεes 50 and 56 are positioned such that the respective paths for scattered light to these lenseε do not form the εame angle relative to the beam axiε.

While two optical receivers are shown in Figure 2, it is to be appreciated that a single optical receiver would

suffice for measuring velocities based on Doppler frequencies. Conversely, three or more optical receivers are employed in situations of directional ambiguity, or where measurements are taken in two or three dimensions. Examples of these situationε include aerodynamic and turbulent flow studieε and studies of atomized εprayε.

The optical system detectε light-scattering elements conveyed through the measurement region by a liquid or a gas. Figure 2 shows a nozzle 62 for admitting particle- containing air, into an enclosed chamber (not shown) . The air is drawn out of the chamber through an exit conduit 66 for a continuous flow downward as viewed in the figure. A pump (not shown) reduces preεεure in conduit 66 to generate the flow, and is controlled to maintain a steady velocity of the air within the chamber. As seen in Figure 3, the air flow causes a stream of particles 68 to flow through measurement volume 44.

While Figure 2 illustrateε a εtream of εolid particleε carried by air, it is to be appreciated that the optical system can be used to senεe different types of discreet elements carried by different media, for example liquid droplets within a gas, solid particles within a liquid medium, or even light-εcattering εolid elements carried on a moving, light-absorptive solid medium. In any event, the concentration of discreet elements in the medium and the size of the measurement volume insure the presence of at most a single discreet element within the measurement

volu e. For example, a fluid may include particles at a concentration on the order of io ⁴ particles per cubic centimeter, with an appropriate measuring volume on the order of 100 X 1,000 microns. Frequently, εufficiently low concentrationε occur naturally in fluid flows under test. If not, such flows are diluted before testing, e.g. with ultrapure water or gas. Of course, the optical system elements can be controlled to a certain extent to enlarge or reduce the size of measurement volume 44 in accordance with expected concentrations.

The photodetector outputs are composite signalε including background noise segments correεponding to timeε when the meaεurement volume iε unoccupied, and active segments corresponding to a light-scattering element's travel through the measurement volume. The active segments comprise the only useful or meaningful information of the photodetector output signals. These active segments, called coherent frequency burstε, occur intermittently and at random. As noted above, the coherent frequency bursts can vary considerably in their amplitude, frequency and duration. Accordingly, signal processing circuitry must distinguiεh coherent frequency burεtε from background noise, respond rapidly enough to detect short bursts, and handle wide-ranging frequencies and amplitudes. Filtering device 22 provides a broad frequency band for burst detection. As seen in Figure 4, the filtering device includes sixteen bandpass filters 70a - 70p,

encompassing different and overlapping frequency bandwidths. One of these filters is selected for bandpass filtering of the incoming signal. An operator can select a filter by providing a filter selection input as indicated at 72. Alternatively, filter εelection logic 74 automatically selects one of filters 70 by comparing data rates and selecting the filter generating the highest data rate. In either event, the selected bandpasε filter eliminates high frequency noise and the pedestal of the burst. Table I lists the overlapping frequency ranges for the bandpass filters in a preferred version of filtering device 22:

Table I

Filter Frequency Range Filter Frequency

Range

70a 0.3 to 3 kHz 70i 3.0 to 20 MHz

70b 1.0 to 10 kHz 70j 5.0 to 35 MHz

70c 3.0 to 30 kHz 70k 10.0 to 50 MHz

70d 10.0 to 100 kHz 701 20.0 to 60 MHz

70e 30.0 to 300 kHz 70m 30.0 to 70 MHz

70f 0.1 to 1 MHz 70n 40.0 to 80 MHz

70g 0.3 to 3 MHz 70o 50.0 to 100

MHZ

70h 1.0 to 10 MHz 70p 10.0 to 100 MHz

Thus, coherent frequency bursts within a broad range of about 0.3 kHz - 100 MHz are selectively bandpasε filtered to enhance the signal-to-noise ratio.

The output of filtering device 22 is provided to a resistive power splitter 76. Amplifiers 78 and 80 provide the amplified output to burst detector 24 and to frequency translator 26, respectively. In the manner explained in the aforementioned U.S. Patent No. 4,973,969, burεt detector 24 distinguisheε a coherent frequency burst from the background noise of the signal. Upon detecting a coherent frequency signal, burst detector 24 generates two output signals: burst gate signal 27, and frequency range signal 29. The burst gate signal is a one-bit εignal active ("high" logic level) when burst detector 24 is receiving a coherent frequency signal from amplifier 78. Frequency range signal 29 is a four-bit digital word indicating an approximate frequency of the burst. The estimated frequency is obtained with multiple, incremented delay lines, in the manner described in the '969 patent.

With reference to the bandwidth over which burst detector 24 receives signals, the four bit frequency range signal is capable of dividing that bandwidth into sixteen different bandwidth segments. However, satisfactory results have been achieved by dividing the burst detector bandwidth into four such bandwidth segments. For example,

an incoming decade bandwidth is divided into four one- octave bandwidth segments. Table II illustrates the segmenting of the bandwidth of filter 7Oh, i.e. 1 - 10 MHz. Table II

Selected filter: 1-10 MHz

Base sample frequency: F = 50 MHz

Range (binary) : 0000-0011 0100-0111 looo-ioil 1100-1111

Bandwidth .625-1.25 1.25-2.5 2.5-5 5-10

Segment (MHz) :

Divide F: F/8 F/4 F/2

F

Sampling 6.25 MHz 12.5 MHz 25 MHz 50 MHz

Frequency:

In the same manner, each of the other bandpasε filterε 70 haε an aεεociated baεe εampling frequency F, which is either applied directly or divided in accordance with the appropriate one of four bandwidth segments. Table III illustrateε baεe frequencies and bandwidth segments for bandpass filters 70b through 70i, with the final four columns indicating the frequency range signal (numerical rather than binary range) and the factor by which the base frequency is divided.

Table III

Filter Band Base F 0- 4- . 8-11: 12- width 3:F/8 7.F/4 F/2 15:F

70b 1-10 50 kHz .625- 1.25- 2.5-5 5-10 kHz 1.25 2.5 kHz kHz kHz kHz

70c 3-30 .15 1.87- 3.75- 7.5-15 15-30 kHz MHz 3.75 7.5 kHz kHz kHz kHz

70d 10-100 .5 MHz 6.25- 12.5- 25-50 50-100 kHz 12.5 25 kHz kHz kHz kHz

70e 30-300 1.5 19- 37.5- 75-150 150- kHz MHz 37.5 75 kHz kHz 300 kHz kHz

70f .1-1 5 MHz 62-125 125- 250- .5-1 MHZ kHz 250 500 MHz kHz kHz

70g .3-3 15 MHz 187- 370- .75- 1.5-3 MHZ 375 750 1.5 MHz kHz kHz MHz

7Oh 1-10 50 MHz .62- 1.25- 2.5-5 5-10 MHZ 1.25 2.5 MHz MHz MHz MHz

70i 3-20 100 N/A 2.5-5 5-10 10-20 MHZ MHz MHz MHz MHz

Thuε, for the εelected bandpasε filter 70, frequency range signal 29 indicates one of four bandwidth segments and causes selection of the appropriate sampling frequency rate. Upon comparing each base sampling frequency F (or the result of the indicated division of F) with its associated bandwidth segment, it will be appreciated that so long as the actual frequency lies within the indicated bandwidth segment, the incoming signal will be sampled at a rate of 5-10 samples per cycle. Figure 5 shows frequency translator 26 in greater detail. Translator 26 includes an analog-to-digital converter 82 that receives an analog voltage signal V from amplifier 80, and converts V _; to a binary word or digital value representing the voltage. A/D converter 82 is an eight-bit device that generates an eight-bit binary word upon each sampling of V.. The A/D converter output is a sequence of the binary words, each representing the voltage at a different sampling time. The binary words can represent voltage valueε over 256 increments, with 255 (binary representing the highest expected voltage and binary zero representing the most negative voltage, with zero voltage corresponding to 128.

The output of A/D converter 82 is stored to a sequential first-in-first-out memory 84. The memory has an eight-bit width (eight channels) corresponding to the length of the binary words and has a length (in number of bit positions) sufficient to encompasε the maximum expected

burst length. In the preferred embodiment, the length of memory 84 is 1024.

The contents of memory 84 are read out to a digital-to-analog converter 86 having eight input lines for converting the sequence of digital values into an analog signal. The signal from D/A converter 86 is provided to a low pass filter _88 with a cutoff frequency set at the Nyquist value, i.e. at a maximum of one-half of the frequency at which D/A converter 86 converts the digital value sequence. The output V,, of low pass filter 88 is a reconstructed, frequency-translated analog signal representing the burst.

Frequency translator 26 further includes a controller 90. Inputs to the controller include burst gate signal 27 and frequency range signal 29 from the burst detector; a bandwidth filter selection input from microprocessor 30; and two clocking inputs of 30 MHz and 100 MHz, reεpectively indicated at 92 and 94. Based on these inputs, controller 90 provides a sampling clock input to A/D converter 82 and to memory 84 via a line 96. The controller provides a read frequency via a line 98 as a clocking input to the memory and to D/A converter 86. Further controller outputs include a reset signal 100 to memory 84 and a bandpass filter select signal 72 to filtering device 22. Controller 90 preferably is a programmable sequencer or a programmable logic device.

A salient feature of the present invention is that despite an initial burst capture range of several orders of magnitude in frequency, and a bandpass filter output V that can range over a decade in frequency, the translated analog signal V ₀ is confined to a frequency range of one octave. This result is achieved through selection and control of the clock rate at which V, is sampled by A/D converter 82 and stored to memory 84, in conjunction with the clock rate at which the series of digital values is read out of the memory and converted by D/A converter 86.

Semiconductor logic in controller 90, actuated when the burst gate signal goes high, selects one of four available sampling clock rates as a function of: (1) the four-bit binary frequency range signal; and (2) the selected bandwidth filter, whether such selection was automatic or by the operator. The available sampling clock rates are generated by clocks 92 and 94, and depend upon the bandwidth filter selected. For example, the bandwidth of 30 - 300 kHz has sampling rates that correspond to the following adjacent frequency segments or ranges: 19 - 37.5 kHz; 37.5 - 75 kHz; 75 - 150 kHz and 150 - 300 kHz. Likewise, each of the remaining burεt detector bandwidths corresponding to remaining bandwidth filters 70, is divided into four bandwidth segments. Each preferably encompasses one octave.

Although coherent frequency bursts can differ considerably with one another as to their frequencieε, each

burst considered alone is subεtantially uniform in its frequency. Consequently, a sufficiently accurate estimate of burst frequency is obtained within a few cycles of initial detection. The controller logic selects the sampling frequency based on a predetermined desired number of samples per cycle of the incoming analog signal. For example, if filter 70p was the selected filter (10 - 100 MHz) , and the frequency range signal indicated a burst frequency in the range of 10 - 20 MHz, then selecting a sampling frequency of 100 MHz would predetermine a number of sampleε per cycle ranging from 5 to 10, depending upon the actual burst frequency. The number will lie within the range of 5 - 10, so long as the actual burst frequency is within the range of 10 - 20 MHz. The range of 5-10 is preferred, although as few as 2.5 samples per cycle can be εatiεfactory in certain velocity meaεurement applications.

The sampling frequency enables A/D converter 82, controls the rate at which the A/D converter samples the incoming analog signal V , and also sets the same rate for storing the digital valueε to memory 84. To inεure that memory 84 is cleared for accepting the digital values, controller 82 provides reset pulse 100 to the memory, responsive to the burst gate signal going high. So long as the gate signal remains high, εignal εampling and storage proceed at the selected εampling frequency.

The signal sampling mode is illustrated in Figures 6 and 7, respectively for a high frequency analog burst signal 102a and a low frequency analog burst signal 102b. Respective burst gate signals 104a and 104b go high within the first several cycles of the detected burst. Respective sampling frequencies are shown at 106a and 106b. In each case, the sampling frequency is selected to yield a desired number or limited range of samples per cycle of the burst. Since the frequency of analog burst 102a is approximately four times the frequency of analog burst 102b, clocking signal 104a is likewise about four times the clocking frequency 104b.

Returning to Figure 5, burst gate signals 104 go inactive or "low" in response to detecting the end of the coherent frequency burst. In response, controller 90 shifts from the burst signal sampling mode to a read mode in which the stored binary words are read out of memory 84 and provided to D/A converter 86, in the same εequence in which they were stored. Reading occurs at a read frequency predetermined by the controller. Accordingly, respective digital read signals 108a and 108b have the same frequency, preferably about 40 MHz. Respective frequency-translated analog signals V are shown at 110a and 110b in Figures 6 and 7. Analog signals 110a and 110b do not necesεarily have the same frequency. However, the higher frequency signal will have a frequency at most double the lower frequency,

regardleεε of the ratio of frequencieε for incoming analog signals 102a and 102b.

The selected bandpass filter 70 and the selected sampling frequency 106 are stored in microprocessor 30. Accordingly, despite the similarity in the frequencies of reconstructed signalε 110a and 110b, icroproceεsor 30 calculateε and causes terminal 32 to display the true frequencies of the original bursts.

Figure 8 discloses an alternative embodiment frequency translator in which two A/D converters, phase-adjusted relative to one another, are employed to increase the frequency at which incoming εignal V. is sampled. More particularly, signal Vi is provided to an A/D converter 101 and also to an A/D converter 103. A delay line 105 is configured to phase-shift a sample clock, so that the sampling of the signal provided to A/D converter 103 lags the sampling of the signal provided to A/D converter 101 by 180 degrees. A/D converter 101 provides its digital output to a sequential memory 107, while A/D converter 103 similarly provides its output to a sequential memory 109. A controller 111 receiveε two clocking inputε of 15 MHz and 50 MHz from clock oscillators indicated at 113 and 115, respectively. The output of delay line 105 is provided to sequential memory 109 as well as A/D converter 103. Accordingly, memory 109 is cauεed to sample the signal at a 180 degree phase shift with respect to such sampling by sequential memory 107.

Considering just oscillator 115 (50 MHz) , the maximum frequency of the resultant controller clocking signal to the A/D converters and the sequential memories is 50 MHz. However, because of the half-cycle phase shift of the clocking signal provided to converter 103 and memory 109 relative to the clocking signal provided to converter 101 and memory 107, the effective rate at which V _; is sampled, is 100 MHz. In the same manner, the frequency of oscillator 113 is effectively doubled to a sampling rate of 30 MHz.

Controller 111 alternately leads εequential memories 107 and 109 to a D/A converter 117. The D/A converter provides its analog output to a low-pass filter 119, the output of which is the frequency-translated burεt εignal V . At any given time, only one of the εequential memory outputs is enabled to provide input to D/A 117. In other words, the outputs of sequential memories 107 and 109 are interlead to the D/A converter. Each sequential memory is read at a frequency of 20 MHz, resulting in a D/A converter input frequency of 40 MHz.

At present, this alternative translator is preferred, in spite of the requirement of an additional A/D converter and sequential memory. Oscillators 113 and 115 are less expensive than their counterpart oscillators 92 and 94. Further, the phase-shifted, lower frequency signals are more robust for enhanced accuracy and reliability.

Figure 9 shows signal procesεing circuitry of a phase Doppler device constructed according to the present invention. Device 112 is useful in determining the εizes of particulates, aerosols and droplets in sprays. With reference to Figure 2, device 112 translates frequencies of analog frequency burstε generated by photodetectorε 52 and 60, then measures the phase difference between the respective frequency-translated signals as an indication of particle or droplet size. For providing the necessary inputs to its frequency translating circuitry, device 112 employs a bandpass filtering device and burst detector with the photodetectors, each similar to its counterpart shown in Figure 1. The amplified and bandwidth filtered analog signals are provided, respectively, to A/D converters 114 and 116. A/D converter 114 samples the input and stores the digital values to a memory 118. In the read mode, the contents of the memory are provided to a D/A converter 120. The D/A converter output is low-pass filtered at 122 and provided to a comparator amplifier 124.

Likewise, A/D converter 116 stores its output to a memory 126, whose contents are later read into a D/A converter 128, whereupon the D/A converter output is low¬ pass filtered at 130 and provided to a comparator amplifier 132. A controller 134 provideε a εampling frequency in common to both A/D converterε and both memories, and likewise provides a common read frequency to the memories

and the D/A converters. The controller responds to the same input signals (not shown in Figure 9) as are provided to controller 90 in Figure 5.

Each of amplifiers 124 and 132 haε a threshold voltage input V _t . The amplifier outputs are provided through respective capacitors 136 and 138 to a phase detector 140 for phase detection. The analog phase detector output is provided to an A/D converter 142, which also includes an enabling input 144 from controller 134. The A/D converter output is provided to a microprocessor 146 via an interface 150. A display terminal 148 connected to the microprocessor displays information on the size of particles, droplets or aerosols, based on the phase difference input. Microprocessor 146 is coupled to controller 134 through the interface.

Figure 10 shows signal processing circuitry of a velocity measurement syεtem 152 which includes optical componentε εuch as those εhown in Figure 2, but requires only one of photodetectors 52 and 60. The system alεo utilizes a burst detector and filtering device as previously discussed. The bandwidth filtered and amplified input V _x is sampled by an A/D converter 154, and the resulting digital values are stored in sequence to a memory 156. A controller 158 governs the sampling and storing rate based on the selected bandwidth filter and the estimated burst frequency range. The controller further governs the rate at which the sequence of binary words is

read out of the memory to a D/A converter 160. The D/A converter output is low-pass filtered at 162, and provided to a comparator amplifier 164. A threεhold input V. is also provided to the amplifier. The amplifier output is provided to a shift register 166, governed by a clock rate from the controller via a line 168. The shift register output is provided to an autocorrelation circuit 170, where a digital correlation of the burst signal is constructed. The output of the autocorrelator is provided to a microprocesεor 172 via an interface 174.

While the disclosed embodiments are preferred, it iε to be recognized that alternative components can be selected to provide several of the discloεed functionε. For example, thoεe skilled in the art are aware of means other than the burst detector, for estimating the burεt frequencies and frequency ranges. One-bit digitizing could be employed in lieu of the multibit digitizing of the A/D converters. However, this subεtitution would reεult in reduced phase resolution and would require a greater number of samples per burst cycle, e.g. at least ten samples per cycle. An addresεable memory regiεter could be employed in lieu of the sequential FIFO memory. Also, digital values read out of the memory could be subject to direct procesεing, e.g. a faεt Fourier tranεfor . However, translation back to an analog signal is preferred at preεent, εince less time is required for signal procesεing.

Thuε in accordance with the preεent invention, analog coherent frequency εignalε received over a broad range of frequencieε are converted to digital information, then reconstructed as analog signals confined to a much narrower frequency bandwidth. The reconstructed signals can be processed with simpler, lesε costly circuit components, yet provide more accurate and reliable measurements of physical characteristics, because signal processing is confined to the narrowed frequency bandwidth. Thus, analog and digital processing techniques are combined in a unique manner that affords rapid frequency bandwidth compresεion and better preserves the signal-to-noise ratio.

What is claimed is: