Extended bandwidth signal-to-noise ratio enhancement methods and means

Dec. 16, 1969 c. r. trimble 3,484,591

EXTENDED BANDWITH SIGNAL-TO-NOISE RATIO ENHANCEMENT METHODS AND MEANS

Filed July 18, 1966 3 Sheets-Sheet 1

<img file="US3484591A_D0001.tif" />

Dec. 16, 1969 c. r. trimble 3,484,591

EXTENDED BANDWITH SIGNAL-TO-NOISE RATIO ENHANCEMENT METHODS AND MEANS

Filed July 18, 1966 3 Sheets-Sheet 2

<img file="US3484591A_D0002.tif" />

FEEDBACK SIGNAL AVERAGING MEANS

Dec. 16, 1969

C. R. TRIMBLE 3,484,591

EXTENDED BANDWITH SIGNAL-TO-NOISE RATIO ,, ENHANCEMENT METHODS AND MEANS

Filed July 18, 1966 ₃ Sheets-Sheet 3

SAMPLE NUMBER

<img file="US3484591A_D0003.tif" />

ATTORNEY

United States Patent Office _P , _n³’^484,591

Patented Dec. 16, 1969

3,484,591

EXTENDED BANDWIDTH SIGNAL-TO-NOISE RATIO ENHANCEMENT METHODS AND MEANS

Charles R. Trimble, Palo Alto, Calif., assignor to Hewlett- 5

Packard Company, Palo Alto, Calif., a corporation of California

Filed July 18,1966, Ser. No. 566,192

Int. Cl. G06f; GOlr 23/16

U.S. Cl. 235—152 15 Claims

ABSTRACT OF THE DISCLOSURE . A selected interval of a recurring input signal is repeti- 15 tively sampled in amplitude at different time positions during m successive recurrences of the input signal to obtain X samples completely representing the selected input interval. This process is repeated during each* of a plurality of signal averaging cycles. During the first 20 cycle the amplitude information obtained from each sample is stored as an average of one recurrence of the selected input interval in a memory location associated with the time position of that sample. During each succeeding cycle the difference in amplitude between each 25 sample and the average stored in the associated memory location during the preceding cycle is divided by a selected factor and the resultant quotient signal employed as a correction factor to update that average.

This invention relates to methods and means for enhancing the signal-to-noise ratio of an electrical input having a high-frequency signal component so as to clearly differentiate that signal component from the noise ^3δcomponent.

It is the principal object of this invention to provide methods and means for effectively extending the useful bandwidth of signal-to-noise ratio enhancement systems such as, for example, those disclosed in Clynes’ U.S. ⁴⁰Patent 3,087,487 and in my own copending patent application Ser. No. 557,167, entitled Signal-to-Noise Ratio Enhancement Methods and Means and filed on June 13, 1966.

The upper frequency limit of these systems is set by ⁴⁵the minimum time interval T required for obtaining a data signal related to the amplitude of the input at a selected time position and for processing that data signal so as to provide a corresponding output display point. Accordingly, it is another object of this invention to ⁵⁰provide sampling methods and means for making the effective data-pulse obtaining and processing time interval τ represented by each output display point smaller than T so as to significantly increase this upper frequency limit. -The upper frequency limit may be increased by as much as m, where m=T/r. Since the sampling means determines how small τ can be made (the sample time must be short compared to τ in order to approximate ideal sampling), the upper frequency limit may be extended to substantially the upper frequency limit of the sampling means.

These objects are accomplished according to the illustrated embodiments of this invention by amplitude sampling m recurrences of a selected interval of a recurring input at the maximum sampling and processing rate 1/T of 65 the signal-to-noise ratio enhancement system so as to obtain X amplitude samples as required to completely represent the selected input interval. These X amplitude samples comprise what is hereinafter termed a sweep, and the amplitude samples obtained during each of the m recur- 70 rences of the selected input interval comprise what is hereinafter termed a pass. The average of a number of sweeps more accurately represents the signal component of the input than any single sweep because the signal component contributes consistently to the average while the unrelated noise component adds to or subtracts from the average randomly. Thus, a number of sweeps necessary to provide the desired signal-to-noise ratio enhancement are taken. After the first pass of each sweep each subsequent pass is delayed relative to the preceding pass by one effective sampling and processing time interval, T=T/m, until the m passes have been taken and the sweep completed. Each amplitude sample obtained during a sweep is processed so as to provide a related average signal which is stored in a memory location associated with the time position of that amplitude sample. An online, X point output display of the selected input interval is provided by driving a cathode ray tube in synchronism with this signal average process. The time interval represented by the output display of a sweep is merely Xt, whereas the time interval required to take the sweep is XT (m times greater).

Other and incidental objects of this invention will become apparent from a reading of this specification and an inspection of the accompanying drawing in whibh:

FIGURE 1 is a block diagram illustrating an extended bandwidth signal-to-noise ratio enhancement system according to one embodiment of this invention; '

FIGURE 2 is a block diagram of a signal-to-noise ratio enhancement system according to another embodiment of this invention;

FIGURE 3 is a waveform diagram illustrating the operation of the systems shown in FIGURES 1 and 2.

Referring to FIGURE 1, there is shown feedback signal averageing means 10 for enhancing the signal-to-noise ratio of an input having a signal-component frequency sufficiently low that a sweep of amplitude samples can be obtained during a single pass of an input interval Of interest. Time expansion means 12 is provided for extending the bandwidth of the signal averaging means 10 by enabling it to make m passes of the selected input interval as required to obtain the sweep of amplitude samples. A recurring input waveform 14 (see FIGURE 3α) having a high frequency signal component of interest embedded in a heavy background noise component is applied to a signal input 16 of the signal averaging means 10 at time T_o. Although any interval of this input waveform 14 from a portion of a recurrence to one or more recurrences may be selected for processing and output display, a single recurrence is selected here for purposes of illustrating my invention. An externally generated synch pulse 18 (see FIGURE 3b) which is time locked to the input waveform 14 is applied to a synch input 20 of the time expansion means 12 at or before the beginning each of recurrence of the input waveform 14. As hereinafter described in detail, these synch pulses 18 are applied from the time expansion means 12 to the signal averaging means 10 for initiating each of the m passes of each sweep of the signal averaging process.

The feedback signal averaging means 10 may comprise, with wiring modifications to accommodate the time expansion means 12, stable averaging apparatus such as that shown and described in detail in my aforementioned copending patent application. A sample and hold circuit 22 is connected to the signal input 16 and is responsive to a train of timing pulses 24 (see FIGURE 3c) for sampling the input waveform 14 to produce an analog sample signal related to the amplitude of the input waveform at each sampling time position. A timer 26 is connected to the sample and hold circuit 22 and is responsive to each synch pulse 18 that occurs at the beginning of the selected input interval for supplying a train of sample timing pulses 24 to the sample and hold circuit at the

3,484,591 maximum sampling and processing rate 1/T of the signal averaging means 10. Thus, (Τ'-f- 1) amplitude samples are obtained during each pass of the selected input interval, where Y represents the largest integer in the absolute value of the quantity |(X—l)/m|.

A multiple channel memory 28 comprising a series of memory locations, each of which is associated with a different sampling time position, is provided for storing the average amplitude information obtained by processing a sweep of amplitude samples. An address register 30 is connected to the memory 28 and is responsive to a timing pulse from the timer 26 for selecting the memory location associated with the time position of each amplitude sample as that sample is being taken and processed.

The signal averaging means 10 operates to enhance the signal-to-noise ratio of the selected input interval by processing each amplitude sample of a plurality of sweeps as generally indicated below for the Jth sample of the Nth sweepj where the sample number is indicated by a superscript and the sweep number is indicated by a subscript. The memory 28 is connected to the timer 26 and is responsive to a timing pulse therefrom at a time T^J, the time position of the Jth amplitude sample, for supplying a digital average signal, Mn_i^j, to a digital-toanalog converter 32 and to an add or subtract circuit 34. This digital average signal represents the average amplitude information stored during the (N— l)th sweep in the Jth memory location selected by the address register 30 during processing of the Jth amplitude sample. It is zero when N is the first sweep. The digital-to-analog converter 32 converts this digital average signal to an equivalent analog average signal, ΜΝ_/, and supplies it to one input of a differential amplifier 36. The sample and hold circuit 22 supplies the analog Jth amplitude sample signal, In^j, of the Nth sweep to the other input of the differential amplifier 36. An analog difference signal, ZN^J— indicating the difference between the amplitude sample signal, IN^J, and the avetage signal, M_n_i^j, is thereby provided at the output of the differential amplifier 36.

A sweep counter 38 is connected to the timer 26 and is responsive to a timing pulse therefrom at time T_N, the beginning of each sweep, for sequentially changing states to provide an output signal indicating the number of the sweep. The sweep counter 38 is connected for supplying the Nth sweep number output signal to a divide circuit 40. In response to a timing pulse from the timer 26 at time T^J+1 this divide circuit 40 is set by the Nth sweep number output signal to divide a subsequently applied digital difference signal by N. An analog-to-digital converter 42 is connected to the differential amplifier 36 and is responsive to a timing pulse from the timer 26 at time T^J+² for converting the analog difference signal, ZN^J—to an equivalent digital difference signal. The analog-to-digital converter 42 is connected for supplying this digital difference signal to the divide circuit 40, which then produces a digital quotient signal, (ZN^J—

The divide circuit 40 supplies this digital quotient signal to the add or subtract circuit 34. This add or subtract circuit 34 is responsive to a timing pulse from the timer 26 at time T^J+³ for algebraically adding the digital quotient signal, (ZN^J—Μν_₁·¹)/Ν, to the digital average signal, Mn_!^J, which was stored in the add or subtract circuit at time T^J. The output of the add or subtract circuit 34 is connected to the memory 28 for storing the resultant average signal, MN^J=MN_1^J+(ZN^J—M_n_i^j)/ZV, back in the Jth memory location when a timing pulse is supplied to the memory from the timer 26 at time T^J+⁴.

Processing each amplitude sample of each sweep in the above-described manner efficiently enhances the signal-to-noise ratio of the selected input interval of the waveform 14 by correcting during each sweep the average amplitude information stored in each memory location during the preceding sweep. Moreover, a full-scale, on-line output display of the selected input interval may be obtained by driving an oscilloscope 44 in synchronism . ⁴ with this signal averaging process as hereinafter described in detail.

The time expansion means 12 includes inhibit means comprising an AND gate 48 and a flip-flop 50 which are _g connected between the synch input 20 and a variable time delay 52 to prevent extraneous synch pulses, such as may occur when the selected input interval comprises several recurrences of the input waveform 14, from interrupting a pass of the selected input interval once it 1q is initiated. Thus, when the first synch pulse 18 (see FIGURE 3b) is passed through the AND gate 48, the flip-flop 50 is set to an inhibit state for preventing subsequent synch pulses from passing through the AND gate until the pass initiated by the first synch pulse is com15 pleted.

The variable time delay circuit 52 is connected between the signal output of AND gate 48 and the timer 26 for delaying application of selected passed synch pulses 18 to the timer 26. A MOD m counter 54 is cori20 nected to the delay circuit 52 for varying its time delay at the end of each pass so that each passed synch pulse 18 after the first one is delayed by the effective sampling and processing time interval τ relative to the preceding passed synch pulse 18 associated with the same sweep. The 25 MOD m counter 54 counts the m passes of a sweep by changing states as follows, 0, 1 . . . m—1, 0, and accordingly varies the time delay of the delay circuit 52 so that the passed synch pulses 18 and, hence, the m passes of each sweep are successively delayed from the 30 first to the mth pass by 0, τ . . . (m—l)r. Thus, the first passed synch pulse 18 passes without delay through the variable delay circuit 52 to the timer 26 for initiating the first pass of the first sweep. During this first pass the timer 26 supplies the first train of sample timing pulses 35 24 to the sample and hold circuit 22 for producing (Y+l) amplitude samples of the selected input interval at time positions spaced the minimum sampling and processing time interval T apart from the beginning of the selected input interval. The timer 26 is connected to 40 an add m circuit 56 for supplying each train of sample timing pulses 24 thereto at the same time it is being supplied to the sample and hold circuit 22. This add m circuit 56 is connected to the address register 30 for incrementing the selected address of the address register 45 by m in response to each of the applied sample timing pulses, except the first one of each sample timing pulse train 24. Thus, since the first memory location is initially selected to correspond to the “0” state of the MOD m counter 54, the resultant average signals obtained 50 by processing the (Υψί) amplitude samples of the first pass are stored in the memory 28 in locations spaced m apart from the first memory location.

An overflow output of the add circuit 56 is connected to the timer 26, to the counting input of the MOD m 55 counter 54, and to the flip-flop 50 for supplying an overflow signal thereto after the last of the (Y-f-1) amplitude samples of each pass has been processed. This overflow signal stops the train of sample timing pulses 24 being supplied to the sample and hold circuit 22 and reinitializes 60 the timer 26 in preparation for the next passed synch pulse 18. It causes the MOD in counter 54 to change states so as to provide the variable delay circuit 52 with the appropriate time delay, for example, τ when the MOD m counter changes from the “0” to the “1” state at the 65 end of the first pass. Additionally, it sets the flip-flop 50 to a pass state so that the next synch pulse 18 applied to the AND gate 48 is passed to the variable delay circuit 52. The address register 30 is also connected to the overflow output of the add m circuit 56 and to the output 70 of the MOD m counter 54 for selecting a memory location which corresponds to the new state of the MOD m counter in response to the overflow signal from the add m circuit, where the initially selected first memory location corresponds to the “0” MOD m counter state, the 75 second memory location to the “1” MOD m counter

3,484 5 state . , . and the mth memory location to the MOD m counter state.

The above-described cycle of operation is repeated in response to each of the subsequently passed synch pulses 18, which initiate the remaining m-1 passes of the first sweep. Since the time delay between each subsequently passed synch pulse 18 and the onset of the corresponding sample timing pulse train 24 is increased by τ relative to the preceding time delay, by the end of the mth pass amplitude samples are taken of the selected input interval ₁₀at X sampling time positions spaced spaced τ apart. These amplitude samples are processed by the signal averaging means 10 to provide corresponding resultant average signals which are interlaced into the memory 28 from the first to the Xth location in the order of in- 15 creasing sample time position relative to the beginning of the selected input interval (see FIGURE 3d). The addresses of these X memory locations increase in the same order as the memory locations. Thus, an on-line output display of the selected input interval (see FIGURE 20 3e) may be provided during the second sweep by driving an oscilloscope 44 in synchronism with the signal averaging process. This is done by connecting the output of the digital-to-analog converter 32 for supplying each analog average signal to the vertical input of the oscil- 25 loscope 44 at the same time it is being supplied to the differential amplifier 36. A digital-to-analog converter 59 is connected to the output of the address register 30 for converting each digital address, when that address is selected, to an analog address signal which is supplied to 30 the horizontal input of the oscilloscope 44 as the time base for the corresponding analog average signal supplied to the vertical input by the digital-to-analog converter 32. During each pass the contents of (F+l) different memory channels are read out in the order of increasing 35 sample time position relative to the beginning of the selected input interval so as to form (F+l) output display points 60 on the face of the cathode ray tube 61 of the oscilloscope 44. Since the persistence of the cathode ray tube 61 is much longer than the time interval ⁴⁰XT required to take and process a sweep of amplitude samples, all of the X output display points 60 are visible at the end of the sweep. Thus, at the end of the second sweep a full-iscale output display of the selected input interval is provided which does not change during the 45 signal averaging process except to the extent of the attenuation of the noise component. Although an oscilloscope 44 is shown as the output device, virtually any analog or digital readout device may be used.

The MOD m counter 54 is connected to the timer 26 50 for supplying an overflow signal thereto When the MOD m counter changes from the “m-1” to the “0” state at the end of each sweep so as to reinitialize the timer in preparation for the next sweep which is taken in the same manner as the first sweep described above. A multiple 55 position switch 58 is connected to the MOD m counter 54 and to the add m circuit 56 for selecting the value of m necessary, to obtain a sweep of amplitude samples for the given high frequency input waveform 14. The m selection switch 58 may be constructed to permit selec- 60 tion of the value of m from any one of a binary sequence of numbers such as 1, 2, 4, 8, 16 . . . . Alternatively, it may be constructed to permit selection of the value of m from any one of a decade sequence of numbers such as 1, 2, 4, 10, 20 . . . or 1, 2, 5, 10, 20 . . . for the 65 convenience of the operator. The selected value of m determines the value of r since the amplitude samples are taken and processed at the maximum rate 1/T of the signal averaging means 10.

Referring now to FIGURE 2, there is shown feedback 70 signal averaging means 62 which is constructed in the same manner as the signal averaging means 10 shown in FIGURE 1, except for minor modifications as indicated to accommodate the new time expansion means 64. In accordance with these modifications the timer 26 is con- 75 ,591 nected directly to the address register 30 instead of first tiiroUgh the time expansion means, and the horizontal input of the oscilloscope 44 is connected to the time expansion means instead of to the output of the address register 30. The signal averaging means 62 enhances the signal-to-noise ratio of the selected interval of the input waveform 14 (see FIGURE 3a), which is applied to the signal input 16, in the same manner as described in connection with FIGURE 1.

The time expansion means 64 includes inhibit means comprising an AND gate 48 and a flip-flop 50 which are connected and which operate in the same manner as already shown and described in connection with FIGURE 1. Thus, synch pulses applied to the synch input 20 after a synch pulse 18 (see FIGURE 3b) is passed through the AND gate 48, to initiate a pass are prevented from interrupting that pass. A ramp generator 66 is connected to the output of the AND gate 48 and is triggered bv each passed synch pulse 18 to provide a ramp signal, the duration of which varies with the value of m. An m selection switch 58 is connected to the ramp generator 66 and to a MOD m counter 54 for selecting the value of m necessary to obtain a sweep of X amplitude samples. This m selection switch ,58 controls the duration of the ramp signal in accordance with the selected value of m so that the ramp signal terminates after the (F+l) amplitude samples of each pass have been taken. The ramp generator 66 supplies this ramp signal to one input of a comparator 68. A digital-to-analog converter 70 is. connected for converting the digital state signal of the MOD m counter 54 to an equivalent analog signal which is supplied to the other input of the comparator 68. The comparator 68 supplies a pass-initiating signal to the timer 26 when the . ramp signal equals the analog signal representing the state of the MOD m counter 54, This comparator input signal equivalence is made to occur at the start of the ramp signal when the MOD m counter 54 is in the 0” state, τ later when it is in the 1” state . . . and (m—1)t later when it is in the (m—1) state. The MOD m counter 54 is initially set to the “0” state. Thus, in response to the first passed synch pulse 18 the timer 26 supnlies the first train of samnle timing pulses 24 (see FIGURE 3c) associated, with the first pass of the first sweep to the sample and hold cirpuit 22 at time T_n. the start of the ramp signal. The sample and hold circuit 22 thereby produces (F+l) amplitude samples of the selected;input interval at time positions spaced the minimum sampling and processing time T apart from the beginning of the selected input interval. Each sample timing pulse applied to the sample and hold circuit 22 is also applied to the address register 30 so that the resultant average signals obtained by processing the (F+l) amplitude samples of the first pass are consecutively stored in the first (F+l) locations of the memory 28 (see FIGURE 3f).

The ramp generator 66 is connected to the timer 26, to the counting input of the MOD m counter 54, and to the flip-flop 50 for supplying a control signal thereto upon termination of the ramp signal after the last of the (F+l) amplitude samples of a pass have been processed. This control signal stops the train of sample timing pulses 24 being supplied to the sample and hold circuit 22 and reinitializes the timer 26 in preparation for the next passed synch pulse 18. It causes the MOD m counter 54 to change states so that the comparator input signal equivalence required for the comparator 68 to initiate the next pass of the sweep occurs τ later than it did for the preceding pass. Additionally, it sets the flip-flop 50 to a pass state so that the next synch pulse applied to the AND gate 48 is passed to initiate the next pass of the sweep.

The above-described cycle of operation is repeated in response to. each of the subsequently passed synch pulses 18, which initiate the remaining m-1 passes of the first sweep. Since the time delay between each subsequently passed synch pulse 18 and the onset of the corresponding

3,484,591 sample timing pulse train 24 is increased by τ relative the preceding time delay, by the end of the mth pass amplitude samples are taken of the selected input interval at X sampling time positions spaced τ apart. These amplitude samples are processed by the signal averaging means 62 to produce resultant average signals which are sequentially stored in the memory 28 from the first to the Xth memory location in the order they are taken (see FIGURE 3f). The MOD m counter 54 is connected to the timer 26 for supplying an overflow signal thereto when the MOD m counter changes from the m-1” state to the 0” state at the end of each sweep so as to reinitialize the timer in preparation for the next sweep which is taken in the same manner as the first sweep described above.

The output of the ramp generator 66 is connected to the horizontal input of the oscilloscope 44 for supplying the ramp signal thereto as the time base for the analog average signals that are supplied to the vertical input of the oscilloscope during the associated pass. However, since all amplitude samples are processed and sequentially stored in the X memory locations, the memory 28 must be decoded for readout. This may be done by Z-axis modulation of the oscilloscope 44 in which the vertical input is blanked until an unblanking signal is applied to the Z-axis input. An on-line output display of the selected input interval (see FIGURE 3g) is obtained by connecting the output of the comparator 68 through a delay circuit 72 to the Z-axis input of the oscilloscope 44 so as to provide this unblanking signal to the Z-axis input when processing of the first amplitude sample of each pass is completed. During each pass the contents of (F+l) different memory locations are read out in the order of increasing sample time position relative to the beginning of the selected input interval until all of the X output display points 60 are obtained. Thus, by the end of the mth pass of the second sweep a full-scale output display (see FIGURE 3g) comprising X output display points 60 is obtained. The readout problem becomes more complex if the output device is to be a printer or an X—Y recorder instead of an oscilloscope. In this case an address decoding network must be used to map the real time at which each resultant average signal is stored in a memory location onto the expanded time scale. One method of doing this is to use the interlace technique for loading the memory locations as described above in connection with FIGURE 1.