Delay generation filter

Abstract

An N.sup.th order infinite impulse response digital filter for long-delay filtering of signals representing a digital input sequence. The N.sup.th order infinite impulse response digital filter is synthesized by cascading in series N first order infinite impulse response filter stages. A filter stage operable on a digital computer with floating point capability is provided and comprises a multiplier which multiplies an input signal with a weighting factor. A summing element provides an output for the filter stage by summing the weighted input signal with a weighted and delayed output signal. A filter operable on a digital computer with fixed point capability is provided and comprises a summing element which sums an input signal with the output of a first multiplier, and the delayed output of a mask element. The output of the summing element is coupled to the input of a second multiplier and to the input of the mask element. The second multiplier provides an output for the filter stage which is also used, in a delayed form as the input of the first multiplier.

Description

AUTHORIZATION

A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent documents or the patent disclosure, as it appears in the Patent or Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.

FIELD OF THE INVENTION

The present invention relates to the field of signal processing and in particular to filtering of received signals to generate a delayed value of the received signal.

BACKGROUND OF THE INVENTION

In signal processing it is sometimes necessary to utilize a value of a received input signal as it existed at a prior time in order to perform a desired control or information processing function. Digital systems generally sample incoming signals at a very high rate which is on the order of many times a second. In known systems which perform control or information processing functions based on received signals, a large amount of storage is required in digital systems and a large number of delay elements are required in analog systems if the value of the incoming signal as it existed in the past is needed. For example, an electronic automatic temperature controller (EATC) used on an automotive vehicle utilizes a temperature sensor to generate a signal indicative of ambient temperature for use by the EATC. Unless the sensor is placed in a position far removed from the engine compartment, heat generated by the engine, exhaust or other components is likely to affect the temperature detected by the sensor. When the sensor is located in the engine compartment at a position to be exposed to air entering the engine compartment, it has been found to provide an accurate reading when the vehicle is in motion. However, when the vehicle is standing the temperature sensor gives a much higher reading than normal because of the increased air temperature in the engine compartment. In such a circumstance, it is helpful to have the value of temperature as it existed several minutes in the past, such as before the vehicle stopped, so that the EATC can accurately determine the existing ambient temperature. With certain finite impulse response filters, if a five minute delay is desired for a signal sampled every second, a system utilizing a finite impulse response filter of known type would need 300 storage or delay elements. One solution is for the EATC to utilize a vehicle velocity signal which is indicative of vehicle velocity in order to determine when the signal from the temperature sensor should be ignored. However, if the vehicle velocity signal is not already being used by the EATC, then use of such a signal adds to the cost and physical complexity of the EATC. In addition, even if the vehicle velocity signal is available to the EATC, an accurate value of ambient air temperature is still required. If the vehicle is stopped momentarily, then a temperature value indicative of the temperature before the vehicle stopped, will still be required.

Accordingly, there is a need for a filter which generates a delayed value of an input signal without requiring the large number of storage or delay elements required by known systems.

SUMMARY OF THE INVENTION

It is a primary object of the present invention to generate a delayed value of a value of a received input signal without utilizing a large number of storage or delay elements.

In accordance with the primary object, in one aspect of the invention an input signal is delayed a predetermined amount of time with a first order infinite impulse response (IIR) filter which comprises a summing element which generates an intermediate value as a function of an input value, a first value and a second value. A first multiplier generates an output value as a function of the intermediate value and a first weighting value. The IIR filter also comprises a residual generation element responsive to the summing element for generating a numerical residual value indicative of a difference in resolution between the output value and the intermediate value. A first delay element, which is responsive to the residual generation element delays the numerical residual value by a predetermined amount of time to generate the first value. A second delay element, which is responsive to the first multiplier, delays the output value by the predetermined amount of time to generate a delayed output value, and a second multiplier, which is responsive to the second delay element, generates the second value as a function of the delayed output value and a second weighting value.

In another aspect of the invention, a plurality of filters utilizing the principles of the aforementioned IIR filter are cascaded to generate a substantial time delay.

An advantage of at least certain embodiments is that a delay on the order of several minutes may be generated without utilizing a large number of storage or delay elements.

These and other features and advantages of the present invention may be better understood by considering the following detailed description of preferred embodiments of the invention. In the course of this description, reference will be made to the attached drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram of a first embodiment of a filter stage which embodies the invention.

FIG. 2 is a schematic diagram of a plurality of cascaded filter stages.

FIG. 3 is a schematic diagram of an alternative embodiment of a filter stage which embodies the invention.

FIG. 4(a) is a graph showing the response of a preferred embodiment to an input signal.

FIG. 5 is a schematic diagram of an electronic automatic temperature controller utilizing the principles of the present invention.

DETAILED DESCRIPTION

FIG. 1 shows a block diagram of a single filter stage which is operable on a digital computer which utilizes floating point arithmetic. The filter stage is a first order infinite impulse response (IIR) filter with an input port 2 and an output port 4. The first order IIR has a sample time represented by a variable h, which in a preferred embodiment has a value of one second. Multiplier 6 multiplies an input signal by a weighting factor h/.tau., where .tau. represents a time constant in seconds. The output from multiplier 6 is coupled to an input of a summing element 8 which provides an output value of the filter stage through output port 4. Storage element 10 has its input coupled to output port 4 and operates to provide a delayed version of the value existing at output 4, to multiplier 12. Multiplier 12 multiplies the output of storage element 10 by a weighting factor (1-h/.tau.), and provides the output to summing element 8.

Multipliers 6 and 12, summer 8, and delay element 10 are preferably implemented by a programmable digital computer. Alternatively, the filter stage may be implemented by specialized digital or analog circuits.

A preferred embodiment advantageously creates an N.sup.th order infinite impulse response digital filter by cascading in series N stages identical to the stage of FIG. 1. The result is filter 220 of FIG. 2, where the input port to the first stage 221 of filter 220 is coupled to the output port of source 201 and the output port of the last stage 223 of filter 220 is coupled to utilization circuit 240. The input port of any other stage, besides the first and last in FIG. 2, such as stage 222, is coupled to the output port of a preceding stage. The resulting z-transform equation H(z) of filter 220, ignoring any numerical round-off errors, is given by the following relationship: ##EQU1## where z.sup.-1 is a unit time delay, .tau. represents the time constant of each of the filter stages and N represents the number of filter stages.

Each of the filter stages 221, 222 and 223 of FIG. 2 advantageously provides a time delay, with the cascaded delay of all of the filter stages of FIG. 2 generating a substantial time delay without utilizing a large number of delay elements. In FIG. 2 source 201 provides an output in the form of a sequence of digitized signal values. In an alternative embodiment, source 201 may provide an analog continuous-time signal. In such an embodiment if filter 220 is a digital filter, a sample-and-hold circuit and an analog-to-digital converter are required to condition the source signal into a sequence of sampled signal values.

FIG. 3 shows an alternate embodiment, which is preferred for fixed point, or integer, arithmetic. This stage receives an input signal through input port 32 and provides an output through output port 34. Summing element 40 receives input signal 33 from input port 32, and provides an output 41 which is the sum of the value of input signal 33, and signals 37 and 48. Multiplier 44 multiplies signal 41 with a weighting factor 1/.tau. to generate output signal 45. As used in the fixed point filter shown in FIG. 3, .tau. is a time constant in terms of the sample period. Storage element 46 receives output signal 45 and generates signal 47 which is a delayed version of output signal 45. Multiplier 36 multiplies signal 47 with a weighting factor .tau.-1 to generate signal 37. The operations performed by elements 42 and 39 advantageously perform a resolution enhancement function which improves the accuracy of the filter stage. Masking element 42 masks signal 41 with a mask value .tau.-1 to generate signal 43 which is delayed by storage element 39 to generate signal 41. As will be appreciated by those skilled in the art in view of the present disclosure, the masking element generates a numerical residual value which is indicative of the difference in resolution between the value at 34 and the value at 41. Values 41, 43 and 48 are preferably generated using double-precision storage locations to provide enhanced resolution. The residual generation function is preferably performed by a masking element such as seen in FIG. 3. In an alternative embodiment, the residual generation function is performed by a modulo operation using a modulo value of .tau., where .tau.=2.sup.m, is an integer which represents a value selected to achieve a desired delay with a desired number of filter stages.

The filter stage seen in FIG. 3 is preferably implemented with a digital computer executing a stored program. As will be appreciated by those skilled in the art in view of the present disclosure, the filter stage of FIG. 3 may be implemented with digital circuitry which comprise either discrete components or are integrated onto a single integrated circuit. As will be appreciated by those skilled in the art in view of the present disclosure, multipliers 36 and 44 involve simple bit shifts and subtraction and can be implemented with relatively few program steps or logic elements and shift registers. Summing element 40 and delay elements 39 and 46 may also be implemented with logic elements and shift registers.

An N.sup.th order filter for fixed point, or integer, arithmetic is synthesized by cascading in series N stages identical to the stage of FIG. 3 to generate a series of delay stages as seen in FIG. 2 and described above.

FIG. 4 of the drawings shows an input 401 to filter 220 with a corresponding output at 402 of filter 220. Each of the stages of filter 220 may be of the type shown in FIG. 1 or FIG. 3. In FIG. 4, ten filter stages are used with the value .tau. for each filter having a value of 32. For the floating point filter in FIG. 4, h preferably has a value of one second. As can be seen in FIG. 4, the cascaded stages of filter 220 delay and smooth the output of source 201. It is understood that line 402 represents an output of a cascaded filter utilizing filters of the type shown either in FIG. 1 or FIG. 3. The delay time can be significant and is approximately given by the product N.tau.h, where h is the sampling rate. The approximation improves as N increases. In order to provide a delay of N.tau.h time shifts, filter 220 advantageously requires only N delay or storage elements, whereas a finite impulse response filter would require [N.tau.] delay or storage elements.

FIG. 5 of the drawings shows an Electronic Automatic Temperature Controller (EATC) installed in a vehicle which utilizes the principles of the present invention. In FIG. 5, EATC 500 receives a temperature signal from temperature sensor 535. Temperature sensor 535 is of known type and is positioned in the engine compartment to be directly exposed to air as it enters the engine compartment while the vehicle is in motion. EATC 500 responds to signals from temperature sensor 535 and from user inputs entered at 560 to generate control signals for changing the temperature and volume of air entering the interior compartment of the vehicle. In a preferred embodiment, the interior compartment is a passenger compartment. In alternative embodiments, the interior compartment may be a storage compartment which may be a refrigerated compartment or other type of compartment kept within a predefined temperature range. As can be seen in FIG. 5, control signals from EATC 500 control the operation of a plurality of temperature alteration systems which take the form of ventilation 510, heating 520 and air conditioning 530 units of the vehicle.

EATC 500 preferably comprises a microcomputer including a central processor unit (CPU) 541, input and output (I/O) port 540, read only memory (ROM) 542 for storing control programs, random access memory (RAM) 543, for temporary data storage which may also be used for counters or timers, and a conventional data bus. EATC 500 operates under stored program control to receive input signals, and to generate output signals to perform the above described control functions.

As previously indicated, in a preferred embodiment, the present invention is advantageously implemented by means of a microprocessor and memory containing suitable programs and data structures for providing the desired control functions. The following C language program listing illustrates to those skilled in the art, an implementation of a preferred embodiment.

______________________________________/*********************************************************************Five-minute ago filter: critically damped IIR filter --** Copyright (c) 1993 Ford Motor Company**********************************************************************#include <stdio.h>#include <stdlib.h>#include <math.h>#define E 2.718281828/* Function prototypes: *//* float camel (double, double, double, int, int, int) ; *//* float icamel (int, int, int, int, int, int); *//* globals: */float camel (sys.sub.-- in, tint, tau, order, init.sub.-- flag, details)double sys.sub.-- in, tint, tau;int order, init.sub.-- flag, details;{ /* tint: sample time interval, tau: time constant, (n-th) order filter*/ /* -- This is a series-cascaded (order times) single pole filter. */ int i,j;static int first.sub.-- time = 1;static int noa.sub.-- wts, nob.sub.-- wts;static float * casc.sub.-- in, * a.sub.-- wts, * casc.sub.-- out, *b.sub.-- wts;static float s.sub.-- out;static float alpha, beta; /* Initialization: */ if (first.sub.-- time) { if (details > 1) fprintf (stderr,"-- starting up camel--.backslash.n"); alpha = tint / tau; beta = 1.0 - alpha; /* Forward weights: */ noa.sub.-- wts = order + 1 ; /* no. of cascade sections */ if (details > 3) fprintf (stderr,"-- allocating casc.sub.-- in &a.sub.-- wts --.backslash.n"); casc.sub.-- in = (float *) calloc (noa.sub.-- wts, sizeof(float)); a.sub.-- wts = (float *) calloc (noa.sub.-- wts, sizeof(float)); if (details > 3) fprintf (stderr,"-- order = %d --.backslash.n",order); if (details > 3) fprintf (stderr," a.sub.-- wts @%p .backslash.n",a.sub.-- wts); for (i=0; i < noa.sub.-- wts ; i++) { casc.sub.-- in[i] = 0.0; /* coupling between sections: */for unity pulse response */ gamman; a.sub.-- wts[i] = alpha; for unity step response */ if (details > 3) fprintf (stderr," a.sub.-- wts[%d] = %7.3f.backslash.n", i, a.sub.--wts[i]); } /* Recurrent weights: */ if (details > 3) fprintf (stderr,"-- allocating casc.sub.-- out &b.sub.-- wts --.backslash.n"); nob.sub.-- wts = order + 1; /* addl. for offset (not used) */ casc.sub.-- out = (float *) calloc (nob.sub.-- wts, sizeof(float)); b.sub.-- wts = (float *) calloc (nob.sub.-- wts, sizeof(float)); b.sub.-- wts[0] = 0.0; /* - not used - */ casc.sub.-- out[0] = 0.0; for (j=1; j < nob.sub.-- wts; j++) { casc.sub.-- out[j] = 0.0; b.sub.-- wts[j] = beta; } first.sub.-- time = 0; if (details > 1) { fprintf(stderr,"--- %dth order weights initialized: ---.backslash.n", order); fprintf(stderr," a.sub.-- wts @%p.backslash.n", a.sub.-- wts); fprintf(stderr," b.sub.-- wts @%p.backslash.n", b.sub.-- wts); for(i=0;i < noa.sub.-- wts;i++) { fprintf(stderr," a.sub.-- wts[%d] = %7.4f.backslash.n", i, a.sub.--wts[i]); } for (j=0; j < nob.sub.-- wts; j++) { fprintf(stderr," b.sub.-- wts[%d] = %7.4f.backslash.n", j, b.sub.--wts[j]); } } } if (init.sub.-- flag) { if (details > 0) fprintf (stderr,"-- (re-)initializing ccamel--.backslash.n"); for (i=0; i < noa.sub.-- wts; i++) { casc.sub.-- in[i] = 0.0; } for (j=0; j < nob.sub.-- wts; j++) { casc.sub.-- out[j] = 0.0; } } /* Starts here: (above is mostly overhead) ************** */ crasc.sub.-- in[0] = sys.sub.-- in; for (i=1; i < noa.sub.-- wts; i++) { casc.sub.-- in[i] = b.sub.-- wts[i]*casc.sub.-- out[i] + a.sub.--wts[i-1]*casc.sub.-- in[i-1]; casc.sub.-- out[i] = casc.sub.-- in[i]; } s.sub.-- out = casc.sub.-- out[order]; if(details > 4) { fprintf(stdefr,"-- Input Cascade:"); for(i=0; i < noa.sub.-- wts;i++) { fprintf(stderr,"%8.4f", casc.sub.-- in[i]); } fprintf(stderr,".backslash.n"); } if(details > 4) { fprintf(stderr,"-- Output Cascade:"); for(j=0; j < nob.sub.-- wts;j++) { fprintf(stderr,"%8.4f", casc.sub.-- out[j]); } fprintf(stderr,".backslash.n"); } return(s.sub.-- out);}int icamel (sys.sub.-- in, tau, order, accumulate, init.sub.-- flag;details)int sys.sub.-- in, tau, order, accumulate, init.sub.-- flag, details;{ /* tint: not used, tau: time constant in steps, (n-th) order filter */ /* -- This is a series-cascaded (order times) single pole filter. */ int i,j;static int first.sub.-- time = 1;static int noa.sub.-- wts, nob.sub.-- wts;static int * casc.sub.-- in, *a.sub.-- wts, * casc.sub.-- out, * b.sub.--wts, * accum ;static int s.sub.-- out;static int alpha, beta; /* Initialization: */ if(first.sub.-- time) { if (details > 1) fprintf(stderr,"-- starting up icamel--.backslash.n"); alpha = 1; beta = tau - 1; /* Forward weights: */ noa.sub.-- wts = order + 1; /* no. of cascade sections */ if (details > 3) fprintf(stderr,"-- allocating casc.sub.-- in &a.sub.-- wts --.backslash.n"); casc.sub.-- in = (int *) calloc(noa.sub.-- wts, sizeof(int)); a.sub.-- wts = (int *) calloc(noa.sub.-- wts, sizeof(int)); if(details > 3) { fprintf(stderr,"-- order = %d --.backslash.n", order); fprintf(stderr,"a.sub.-- wts @%p:.backslash.n", a.sub.-- wts); } for (i=0; i < noa.sub.-- wts; i++) { casc.sub.-- in[i] = 0; /* coupling between sections: */ a.sub.-- wts[i] = alpha; /* for unity step response */ } /*Recurrent weights: */ if (details > 3) fprintf(stderr,"-- allocating casc.sub.-- out &b.sub.-- wts --.backslash.n"); nob.sub.-- wts = order + 1; /* addl. for offset (not used) */ casc.sub.-- out = (int *) calloc(nob.sub.-- wts, sizeof(int)); accum = (int *) calloc(nob.sub.-- wts, sizeof(int)); b.sub.-- wts = (int *) calloc (nob.sub.-- wts, sizeof(int)); b.sub.-- wts[0] = 0; /* - not used - */ casc.sub.-- out[0] = 0; for (j=1;j < nob.sub.-- wts;j++) { casc.sub.-- out[j] = 0; accum[j] = 0; b.sub.-- wts[j] = beta; } first.sub.-- time = 0; if (details > 1) { fprintf(stderr,"--- %dth order weights initialized: ---.backslash.n", order); fprintf(stderr," a.sub.-- wts @%p :.backslash.n", a.sub.-- wts); for (i=0; i < noa.sub.-- wts; i++) { fprintf(stderr," a.sub.-- wts[%d] = %4d.backslash.n", i, a.sub.--wts[i]); } fprintf(stderr," b.sub.-- wts @%p :.backslash.n", b.sub.-- wts); for (j=0; j < nob.sub.-- wts; j++) { fprintf(stderr," b.sub.-- wts[%d] = %4d.backslash.n", j, b.sub.--wts[j]); } }if (init.sub.-- flag) { if (details > 0) fprintf(stderr,"-- (re-)initializing icamel--.backslash.n"); for (i=0; i < noa.sub.-- wts; i++) { casc.sub.-- in[i] = 0; } for (j=0;j < nob.sub.-- wts;j++) { casc.sub.-- out[j] = 0; }}/* Starts here: (above is mostly overhead) ************** */casc.sub.-- n[0] = sys.sub.-- in;for (i=1; i < noa.sub.-- wts; i++) { accum[i] += b.sub.-- wts[i]*casc.sub.-- out[i] + a.sub.-- wts[i-1]*casc.sub.-- in[i-1]; casc.sub.-- out[i] = accum[i]/tau; if (accumulate) accum[i] = accum[i] % tau; else accum[i] = 0; casc.sub.-- in[i] = casc.sub.-- out[i];}s.sub.-- out = casc.sub.-- out[order];if(details > 4) { fprintf(stderr,"-- Input Cascade:"); for(i=0; i < noa.sub.-- wts; i++) { fprintf(stderr,"%4d", casc.sub.-- in[i]); } fprintf(stderr,".backslash.n"); fprintf(stderr,"-- Accumulator:"); for (j=0; j < nob.sub.-- wts; j++) { fprintf(stderr,"%4d", accum[j]); } fprintf(stderr,".backslash.n"); fprintf(stderr,"-- Output Cascade:"); for (j=0;j < nob.sub.-- wts;j++) { fprintf(stderr,"%4d", casc.sub.-- out[j]); } fprintf(stderr,".backslash.n"); } return (s.sub.-- out);}______________________________________

As can be seen by those skilled in the art, the above program implements a method for generating an output value which is delayed for a predetermined length of time from a received time varying input signal by first generating an input value corresponding to the received input signal. Second, the input value is multiplied with a first weighting value to generate a first intermediate value. Third, the first intermediate value is added to a third intermediate value to generate an intermediate output value. Fourth, the input value is replaced with the intermediate output value to generate an input value for a subsequent delay stage. Fifth, the intermediate output value is stored to a storage location, for use in the subsequent delay stage. Sixth, the third intermediate value is generated by retrieving the stored intermediate output value from the storage location and multiplying the intermediate output value with a second weighting value. In a preferred embodiment, the storage location containing the intermediate output value is initialized with a predetermined value. Seventh, a different storage location for storage of a subsequent intermediate output value is selected. This different storage location will be used as the storage location by the subsequent delay stage to store the intermediate output value. Eighth, the output value is generated by periodically repeating steps two through seven a predetermined number of times, to execute a predetermined number of delay stages. As can be seen from the above description, such an embodiment utilizes a limited number of storage locations while generating an output value which is delayed a substantial amount of time from the input value.

It is to be understood that the specific mechanisms and techniques which have been described are merely illustrative of one application of the principles of the invention. Numerous modifications may be made to the methods and apparatus described without departing from the true spirit and scope of the invention.

Claims

1. The method of delaying a time varying input signal represented by a sequence of binary integer values to supply an output signal to a utilization circuit, said output signal being a delayed version of said time varying input signal,

said method comprising

a sequence of like binary integer arithmetic processing operations each of which converts an input value expressed as a binary integer into a binary result integer passed to the next one of said operations as the input to the next one of said operations as its input integer until the last of said processing operations produces a sequence of final integers representing said delayed version of said time varying input signal, wherein each given one of said operations comprises the steps of:

sequentially accepting each said input value expressed as a binary integer;

adding each accepted input integer and a previously stored value to produce an intermediate sum integer;

shifting the bits in said intermediate sum integer to thereby divide said intermediate sum integer by the quantity T to form said binary result integer passed to the next of said operations as its input integers;

masking said intermediate sum integer with the binary number (T-1) to form a masked integer;

performing bit shifts and subtraction on said result value to thereby form a second product integer equal to (T-1) times said binary result integer;

temporarily storing said masked integer and said second product integer; and

thereafter utilizing said masked integer and said second product integer as said previously stored values added to the next input integer processed by said given operation.

2. The method of delaying a time varying signal as set forth in claim 1 wherein the steps of said method are performed by a programmed digital processor capable of performing integer arithmetic and logical operations on binary numbers stored in memory locations and wherein each of said processing operations is assigned a different memory location for storing the combination of said masked integer and said second product integer.

3. The method of delaying a time varying signal as set forth in claim 2 wherein T is a binary number which defines a time constant value, N is the number of like processing operations performed, and h is the rate at which input integers are formed by sampling said time varying input signal to be delayed, said method providing an effective delay between signal events in said time varying input signal and the corresponding events in said delayed version which is equal to the product of N times T times h, and wherein N storage locations in said memory are employed to store intermediate values for said N processing operations.