High-precision digital decimation filter

Through the four-stage filter structure and variable decimation magnification design, the shortcomings of digital decimation filters in the prior art in terms of high precision and low power consumption are solved, and efficient and flexible signal processing is realized, suitable for high decimation ratio and high Application scenarios for accuracy requirements.

CN118944635APending Publication Date: 2024-11-12NO 24 RES INST OF CETC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411134781.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Existing digital decimation filters are difficult to ensure the accuracy of signal processing under high accuracy requirements, and there are problems of high power consumption and processing delays. Especially in the case of high decimation ratio, existing filter designs lack flexibility and are difficult to adapt to different types. Requirements for application scenarios.

Method used

Using a four-stage filter structure, including a multi-class fixed decimation magnification integrated comb filter, a variable decimation magnification integrated comb filter, a compensation filter and a semi-band filter, it is designed through a cascade architecture and variable decimation magnification. , optimize the functions and performance of each level, and reduce computing resource requirements and power consumption.

Benefits of technology

It realizes low-power and high-precision signal processing, adapts to the needs of different application scenarios, significantly improves signal integrity and accuracy, reduces processing delays, and improves system response speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118944635A_ABST
    Figure CN118944635A_ABST
Patent Text Reader

Abstract

The invention discloses a high-precision digital decimation filter which comprises four stages of filter structures which are sequentially connected in series, the first-stage filter structure is a multi-stage cascaded fixed decimation multiplying power integrator comb filter, the second-stage filter structure is a multi-stage cascaded variable decimation multiplying power integrator comb filter, and the fourth-stage filter structure is a multi-stage cascaded variable decimation multiplying power integrator comb filter. The third-stage filter structure is a compensation filter, and the fourth-stage filter structure is a half-band filter. According to the invention, by introducing the compensation filter, signal distortion is effectively corrected, high precision of signal processing is ensured, a structure of combining the CI C filter and the half-band filter is adopted, and a variable extraction ratio is adopted, so that power consumption of the system is remarkably reduced, the filter can adapt to different application requirements, design flexibility is improved, and cost is reduced. Problems existing in the prior art are solved, and remarkable improvement is provided in the aspects of high precision, low power consumption and design flexibility.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of integrated circuits, and in particular relates to a high-precision digital extraction filter. Background Art

[0002] With the rapid development of electronic technology and computer science, digital signal processing technology has been widely used in many fields such as communication, audio processing, image processing, radar, biomedicine, etc. Digital decimation filter is a key component in the field of digital signal processing. Its main function is to extract low-rate signals from high-rate input signals while maintaining the integrity and accuracy of the signals.

[0003] In the prior art, digital decimation filters usually use multi-stage filtering and decimation structures to achieve efficient signal processing, such as a decimation structure based on a FIR (finite impulse response) filter, which achieves signal downsampling by cascading multiple filters and decimators. However, this structure has certain limitations in achieving high-precision decimation, mainly manifested in the need for a large amount of computing resources, especially in the case of high decimation ratios, which will lead to increased system power consumption and processing delays. In addition, due to the quantization error of the filter coefficients and the rounding error in the operation process, the multi-stage structure is difficult to ensure the accuracy of signal processing under high-precision requirements. At the same time, the existing filter design methods often lack sufficient flexibility and are difficult to adapt to the specific needs of different application scenarios. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a high-precision digital decimation filter.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A high-precision digital extraction filter comprises a four-stage filter structure connected in series in sequence, wherein the first-stage filter structure is a multi-stage cascaded fixed extraction rate integrating comb filter, the second-stage filter structure is a multi-stage cascaded variable extraction rate integrating comb filter, the third-stage filter structure is a compensation filter, and the fourth-stage filter structure is a half-band filter.

[0007] Furthermore, the order of the fixed decimation rate integral comb filter is at least one order greater than the order of the modulator of the previous stage of the digital decimation filter; the order of the variable decimation rate integral comb filter is less than or equal to the order of the fixed decimation rate integral comb filter.

[0008] Furthermore, the decimation rate of the fixed decimation rate integrating comb filter is 64 times.

[0009] Furthermore, the variable decimation rate integrating comb filter can select decimation rates including 6 times, 12 times and 36 times.

[0010] Furthermore, the order of the fixed decimation rate integrating comb filter is k1, and the fixed decimation rate integrating comb filter includes a first integrating module, a first digital decimation module and a first differential module connected in series in sequence, the first integrating module includes k1 first integrating delay units connected in series, and the first differential module includes k1 first differential delay units connected in series.

[0011] Further, the first integral delay unit includes a first adder and a first delay device, the first adder has a first positive input terminal, a second positive input terminal and an output terminal, the first positive input terminal of the first adder serves as the input terminal of the first delay unit, the second positive input terminal of the first adder is connected to the output terminal of the first delay device, the output terminal of the first adder is connected to the input terminal of the first delay device, and the output terminal of the first delay device serves as the output terminal of the first integral delay unit;

[0012] The first differential delay unit includes a second adder and a second delay device, the second adder has a positive input terminal, a negative input terminal and an output terminal, the positive input terminal of the second adder serves as the input terminal of the first differential delay unit, the output terminal of the second adder serves as the output terminal of the first differential delay unit, the positive input terminal of the second adder is connected to the input terminal of the second delay device, and the output terminal of the second delay device is connected to the negative input terminal of the second adder.

[0013] Furthermore, the order of the variable decimation ratio integrating comb filter is k2, and the variable decimation ratio integrating comb filter includes a second integrating module, a second digital decimation module and a second differential module connected in series in sequence, the second integrating module includes k2 second integrating delay units connected in series, and the second differential module includes k2 second differential delay units connected in series.

[0014] Furthermore, the system function of the fixed decimation rate integral comb filter and the variable decimation rate integral comb filter is expressed as

[0015] H(z)=(1+z -1 ) N (1+z -2 ) N (1+z -4 ) N …(1+z 1-2k ) N

[0016] Among them, H(z) represents the system function of the integral comb filter; z is a complex variable used to represent the variable of the Z transform in discrete-time signal processing; N represents the decimation factor of the integral comb filter, and k represents the order of the integral comb filter.

[0017] Furthermore, the gain of the transfer function of the compensation filter after cascading with the fixed decimation rate integral comb filter and the variable decimation rate integral comb filter is 1.

[0018] Furthermore, the transfer function of the half-band filter is as follows

[0019] H HBF (z) = h(0) + h(1) * z -1 +h(2)*z -2 +……+h(M-1)*z -M

[0020] Among them, H HBF (z) represents the transfer function of the half-band filter; z is a complex variable used to represent the variable of the Z transform in discrete-time signal processing; h(0), h(1), h(2)...h(M-1) represent the coefficients of each order of the half-band filter; M represents the order of the half-band filter.

[0021] Compared with the prior art, the high-precision digital decimation filter of the present invention has the following significant advantages and positive effects:

[0022] (1) CIC filter and half-band filter are used. These two filter structures have the characteristics of low power consumption due to their simplicity and high efficiency. In addition, the design of variable sampling rate further optimizes power consumption management, so that the present invention can achieve low power consumption operation while maintaining high performance, which is particularly important for portable devices and energy-constrained applications.

[0023] (2) By introducing a compensation filter, the signal distortion and error introduced by the previous CIC filter can be effectively corrected, thereby achieving high-precision signal processing; compared with the existing technology, it provides a significant improvement in signal integrity and accuracy, and is particularly suitable for application scenarios with extremely high requirements for signal quality.

[0024] (3) The four-stage filter cascade architecture achieves a significant improvement in computing efficiency by optimizing the functions and performance of each stage; especially in the case of high decimation ratios, it can effectively reduce the demand for computing resources, reduce processing delays, and improve the overall response speed of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0026] Figure 1 It is a structural block diagram of an embodiment of a high-precision digital decimation filter of the present invention.

[0027] Figure 2 It is the structural block diagram of the 5th order fixed decimation rate integrating comb filter.

[0028] Figure 3 This is the structural block diagram of a 4th-order variable decimation rate integrator comb filter.

[0029] Figure 4 Schematic diagram for comparing the amplitude-frequency characteristics of a 5-order cascaded integral comb filter and a single-order integral comb filter.

[0030] Figure 5 The amplitude-frequency characteristic curve of the 4th-order variable decimation rate integrating comb filter with different decimation rates.

[0031] Figure 6 This is the amplitude-frequency characteristic curve of the cascaded first-stage filter structure and the second-stage filter structure.

[0032] Figure 7 This is the amplitude-frequency characteristic curve of the third-stage filter structure (compensation filter).

[0033] Figure 8 This is the amplitude-frequency characteristic curve of the fourth-order filter structure (half-band filter).

[0034] The accompanying drawings in the specification are numeraled as follows:

[0035] Fixed decimation rate integrating comb filter-100; Convert modules-101, 102, 103, 201, 202, 203; First integrating module-110; First adder-111; First delay-112; First digital decimation module-120; First differential module-130; Second adder-131; Second delay-132; Variable decimation rate integrating comb filter-200; Second integrating module-210; Second digital decimation module-220; Register-221; Decoder-222; Decimation unit-223; Second differential module-230; Compensation filter-300; Half-band filter-400. DETAILED DESCRIPTION

[0036] The following describes the implementation methods of the present invention through specific examples. The illustrations provided in the following embodiments are only used to schematically illustrate the basic concept of the present invention. The following embodiments and features in the embodiments may be combined with each other without conflict.

[0037] See also Figure 1 , Figure 1 The block diagram is a structural diagram of an embodiment of a high-precision digital decimation filter of the present invention. The high-precision digital decimation filter of this embodiment includes a four-stage filter structure connected in series, wherein the first-stage filter structure adopts a multi-stage cascaded fixed decimation rate integrating comb filter 100, the second-stage filter structure adopts a multi-stage cascaded variable decimation rate integrating comb filter 200, the third-stage filter structure adopts a CFIR compensation filter 300, and the fourth-stage filter structure adopts the simplest half-band filter 400.

[0038] The decimation rate of the fixed decimation rate integrating comb filter 100 is a fixed value. In this embodiment, the decimation rate of the fixed decimation rate integrating comb filter 100 is 64 times. The order of the fixed decimation rate integrating comb filter 100 is k1, and the order of the fixed decimation rate integrating comb filter 100 is at least one order higher than the order of the modulator of the previous stage of the digital decimation filter, that is: k1≥g+1; wherein g represents the order of the modulator of the previous stage of the digital decimation filter. In this embodiment, since the order of the modulator of the previous stage of the digital decimation filter is 2+2, a 5-order fixed decimation rate integrating comb filter 100 is used as the first-stage filter structure, that is, k1=5. In this embodiment, the CIC filter (i.e., integrating comb filter) used has the advantage of low power consumption due to its simple structure and no need for multipliers and coefficient storage; the basic efficiency and stability of signal processing are ensured by fixing the decimation rate.

[0039] The decimation rate of the variable decimation rate integral comb filter 200 is variable. In this embodiment, the variable decimation rate integral comb filter 200 supports three different decimation rates of 6 times, 12 times, and 36 times, and can be selected from the three different decimation rates of 6 times, 12 times, and 36 times as needed. This design provides flexibility, allowing the filter to adapt to different application scenarios and signal processing requirements. The design of the variable decimation rate helps to further reduce the computational complexity and power consumption. The order of the variable decimation rate integral comb filter 200 is k2, and the order of the variable decimation rate integral comb filter 200 is less than or equal to the order of the fixed decimation rate integral comb filter 100, that is: k1≥k2. In this embodiment, since the first-stage filter structure is a 5-order fixed decimation rate integral comb filter 100, a 4-order variable decimation rate integral comb filter 200 is used as the second-stage filter structure, that is, k2=4.

[0040] The compensation filter 300 is used to correct the signal distortion and error introduced by the first two stages of CIC filters (i.e., the fixed decimation rate integrating comb filter 100 of the first stage and the variable decimation rate integrating comb filter 200 of the second stage). The design of the compensation filter 300 ensures high-precision processing of the signal, and effectively improves the integrity and accuracy of the signal through precise adjustment of the filter coefficients.

[0041] As an efficient filter structure, the half-band filter 400 is particularly suitable for further reducing the signal sampling rate. The half-band filter 400 has coefficient symmetry, which can reduce multiplication operations, thereby reducing power consumption and computational complexity. In addition, the half-band filter 400 further optimizes the overall performance of the system while maintaining signal quality.

[0042] With the above structure, the analog signal outputs a 1-bit high-frequency signal through the modulator before the digital decimation filter, and is transmitted as the input signal of the first-stage 5th-order fixed decimation rate CIC filter (i.e., the fixed decimation rate integrating comb filter 100). After 64 times high-multiple decimation by the first-stage fixed decimation rate CIC filter, it is output to the second-stage 4th-order variable decimation rate CIC filter (i.e., the variable decimation rate integrating comb filter 200). The decimation rate of the second-stage 4th-order variable decimation rate CIC filter can be selected to be 6 times, 12 times or 36 times. The decimated signal is output to the third-stage CFIR compensation filter 300, and after being down-converted by two times, it is output to the fourth-stage half-band filter 400. Finally, the half-band filter 400 is down-converted by two times to output the final output signal.

[0043] Please continue reading Figure 1 ,from Figure 1 It can be seen that the input signal is a 1-bit signal after the analog signal passes through the modulator, and the sampling frequency is 921.6KHz. It first passes through the first 5th-order CIC filter with a fixed decimation ratio of 64, and outputs a first-order signal of 14.4KHz, which is used as the input signal of the second-order 4th-order CIC filter with a variable decimation ratio. After the optional decimation ratio: 6 times, 12 times or 36 times, it is used as the input signal of the compensation filter 300. The compensation filter 300 is mainly used for passband compensation and double decimation after the second-stage CIC filter. Finally, it is used as the input signal of the last-stage half-band filter 400 to complete double frequency reduction, and the final output signal is 100Hz~600Hz.

[0044] See also Figure 2In this embodiment, according to the Noble identity, by changing the position of the decimator, a 5th-order CIC filter structure with the least resources can be obtained. This filter structure is called a Hogenauer decimation filter, also known as a recursive structure. In this structure, the integrator is placed in front of the decimator to work in the high-frequency region, and the comb filter is placed behind the decimator to work in the low-frequency region, reducing the overall power consumption of the filter.

[0045] Specifically, the fixed decimation rate integrating comb filter 100 in this embodiment includes a first integrating module 110, a first digital decimation module 120 and a first differential module 130 connected in series in sequence, the first integrating module 110 includes 5 (i.e., k1) first integrating delay units connected in series, and the first differential module 130 includes 5 (i.e., k1) first differential delay units connected in series.

[0046] In order to facilitate the change of the data type of the signal, in this embodiment, the fixed extraction rate integration comb filter 100 also includes three Convert modules, namely Convert module 101, Convert module 102 and Convert module 103, wherein the Convert module 101 is connected in series between the input end of the fixed extraction rate integration comb filter 100 and the first integration module 110, the Convert module 102 is connected in series between the first digital extraction module 120 and the first differential module 130, and the Convert module 103 is connected in series between the first differential module 130 and the input and output ends of the fixed extraction rate integration comb filter 100.

[0047] The first integral delay unit includes a first adder 111 and a first delay 112. The first adder 111 has a first positive input terminal, a second positive input terminal and an output terminal. The first positive input terminal of the first adder 111 serves as the input terminal of the first delay unit. The second positive input terminal of the first adder 111 is connected to the output terminal of the first delay 112. The output terminal of the first adder 111 is connected to the input terminal of the first delay 112. The output terminal of the first delay 112 serves as the output terminal of the first integral delay unit.

[0048] The first differential delay unit includes a second adder 131 and a second delay 132. The second adder 131 has a positive input terminal, a negative input terminal and an output terminal. The positive input terminal of the second adder 131 serves as the input terminal of the first differential delay unit. The output terminal of the second adder 131 serves as the output terminal of the first differential delay unit. The positive input terminal of the second adder 131 is connected to the input terminal of the second delay 132, and the output terminal of the second delay 132 is connected to the negative input terminal of the second adder 131.

[0049] The Convert module is used to change the data type of the signal in the simulation model. Specifically, it can convert one data type of the input signal to another specified data type. For example, double, int, boolean, etc., from floating point to fixed point, or from integer to Boolean. In this filter model, because there are multiple modules that require specific data type input, the use of the Convert module can ensure that the output signal is consistent with the input requirements of these modules, avoiding the problem of incompatible data types, to ensure that the signal can be transmitted correctly, so as to better manage and control the data type of the signal, and improve the robustness and simulation efficiency of the model.

[0050] See also Figure 3 In this embodiment, the second-stage 4th-order CIC filter also uses a recursive structure, including 4 integration units, 1 digital extraction module, and 4 differential units. Specifically, the variable extraction ratio integration comb filter 200 in this embodiment includes a second integration module 210, a second digital extraction module 220, and a second differential module 230 connected in series in sequence. The second integration module 210 includes k2 second integration delay units connected in series, and the structure of the second integration delay unit can be the same as that of the first integration delay unit. The second differential module 230 includes k2 second differential delay units connected in series, and the structure of the second differential delay unit can be the same as that of the first differential delay unit.

[0051] The second digital extraction module 220 includes a register 221, a decoder 222 and an extraction unit 223. The main function of the register 221 is to store and control the setting of the extraction ratio. By changing the value of the register 221, the system can quickly switch between different extraction ratios to meet the needs of real-time signal processing. When the system is turned on or reset, the register 221 will be initialized to a default value to ensure that the filter is in an effective working state immediately after startup. When the value of the register 221 is updated, the extraction ratio selection logic in the filter will respond immediately and select the corresponding extraction ratio. The binary value of the register 221 is converted into a corresponding selection signal by a decoder 222 provided with a simple decoding circuit to drive the operation of the extraction unit 223. The register 221 can be designed to be 3 bits wide to support up to 8 kinds of encoding of extraction ratios. However, in this embodiment, the actual magnifications used are 6x, 12x, and 36x, so the value of register 221 can be mapped as follows: value 000: decimation magnification = 6x, value 001: decimation magnification = 12x, value 010: decimation magnification = 36x. Other values ​​can be reserved or used for future expansion. The setting and updating mechanism of register 221 is intuitive and clear, which simplifies the operation process. Only a new magnification value needs to be provided, and register 221 can automatically handle the corresponding changes. Rapid adjustment of the decimation magnification according to real-time needs greatly enhances the adaptability of the filter.

[0052] In order to facilitate the change of the data type of the signal, in this embodiment, the variable extraction rate integration comb filter 200 also includes three Convert modules, namely Convert module 201, Convert module 202 and Convert module 203, wherein the Convert module 201 is connected in series between the input end of the variable extraction rate integration comb filter 200 and the second integration module 210, the Convert module 202 is connected in series between the second digital extraction module 220 and the second differential module 230, and the Convert module 203 is connected in series between the second differential module 230 and the input and output ends of the variable extraction rate integration comb filter 200.

[0053] If the length of the cascaded filter is an integer power of 2, the system function of the fixed decimation ratio integral comb filter 100 of the first stage and the variable decimation ratio integral comb filter 200 of the second stage can be expressed as:

[0054] H(z)=(1+z -1 ) N (1+z -2 ) N (1+z -4 ) N …(1+z 1-2k ) N

[0055] Among them, H(z) represents the system function of the integral comb filter (i.e., the fixed decimation rate integral comb filter 100 and the variable decimation rate integral comb filter 200); z is a complex variable used to represent the variable of the Z transform in discrete-time signal processing; N represents the decimation rate of the integral comb filter; k represents the order of the integral comb filter.

[0056] Figure 4 Schematic diagram of the comparison of the amplitude-frequency characteristics of a 5-stage cascaded integral comb filter (i.e., a 5-stage CIC filter) and a single-stage integral comb filter (i.e., a single-stage CIC filter). Figure 4 It can be seen that compared with the single-order CIC filter, the first sidelobe suppression ratio of the 5-stage CIC filter is about 70dB lower than the main lobe, and under the condition of the same CIC filter extraction multiple, if the order of the CIC filter is larger, the first sidelobe suppression ratio will be larger, and the edge of the passband will be steeper, that is, the 5-stage CIC filter exhibits stronger attenuation ability in the stopband and can more effectively suppress unwanted frequency components. At the same time, the frequency roll-off is usually steeper than that of the single-stage CIC filter, which helps to quickly reduce unwanted frequency components, and the transition band of the 5-stage CIC filter is usually narrower than that of the single-stage CIC filter, that is, the passband and stopband can be divided more accurately.

[0057] Figure 5 is the amplitude-frequency characteristic curve of the 4th-order variable decimation rate integrating comb filter 200 with different decimation rates, Figure 5 The three curves in the figure correspond to decimation ratios of 6, 12, and 36, respectively. Figure 6 This is the amplitude-frequency characteristic diagram of the cascaded 5th-order fixed decimation rate CIC filter and the 4th-order variable decimation rate CIC filter. Figure 6 The three curves in are the amplitude-frequency characteristic curves when the decimation ratio of the 5th-order fixed decimation ratio CIC filter is 64 and the decimation ratio of the 4th-order variable decimation ratio CIC filter is 6, 12, and 36, respectively. Figure 5 It can be seen that when the decimation factor of the variable decimation factor CIC filter is larger, the digital frequency at the amplitude of -3dB is smaller, which means that the -3dB bandwidth of the filter is narrower and the signal amplitude decreases gradually faster. When the decimation factor (i.e., the product of the decimation factor of the first-stage fixed decimation factor CIC filter and the decimation factor of the second-stage variable decimation factor CIC filter, expressed as k1*k2) is 64*6, the passband ripple is 0.297dB, and when the decimation factor reaches 64*36, the passband ripple reaches 0.368dB. The overall passband attenuation of the designed digital decimation filter is required to be 0.01dB. Therefore, it is necessary to use compensation filter 300 to compensate for the passband attenuation characteristics of the CIC decimation filter.

[0058] The output of the variable decimation rate CIC filter is used as the signal input of the compensation filter 300. The passband amplitude-frequency characteristic of the compensation filter 300 must be complementary to the passband amplitude-frequency characteristic of the CIC filter, so that the gain of the transfer function after the compensation filter 300 is cascaded with the 5th-order fixed decimation rate integrating comb filter 100 and the 4th-order variable decimation rate integrating comb filter 200 is 1.

[0059] Figure 7 The amplitude-frequency characteristic curve of the compensation filter 300 is shown in FIG. 1 . The output sampling frequency of the variable decimation rate CIC filter is the input sampling frequency of the compensation filter 300. The decimation rate of the variable decimation rate CIC filter is configurable. As the decimation rate of the variable decimation rate CIC filter changes, its corresponding output sampling frequency will also change, that is, the input sampling frequency of the compensation filter 300 will also change. In order to save chip area and hardware consumption, the coefficients of the compensation filter 300 can be designed as fixed coefficients instead of programmable coefficients, which only compensate for the attenuation of the variable decimation rate CIC filter at the maximum bandwidth. Figure 6 It can be seen that the bandwidth is maximum when the decimation rate of the variable decimation rate CIC filter is configured to be 6 times. Therefore, the frequency 921600 / (64*6) is used as the frequency compensation point, and the design indicators of the compensation filter 300 are shown in Table 1:

[0060] Table 1. Design specifications of compensation filter 300

[0061]

[0062]

[0063] The coefficients are shown in Table 2:

[0064] Table 2 Compensation filter 300 design coefficients

[0065] h'(1)=h'(81) -0.000016792572 h'(22) = h'(60) 0.0054545966280597 h'(2) = h'(80) -0.000024962967 h'(23)=h'(59) 0.0088083900425636 h'(3)=h'(79) -0.000002985910 h'(24)=h'(58) 0.0074692110966615 h'(4)=h'(78) 0.0000800122594 h'(25)=h'(57) 0.000803282820405 h'(5)=h'(77) 0.0002205652953 h'(26) = h'(56) -0.008487667057213 h'(6)=h'(76) 0.00035169556832 h'(27)=h'(55) -0.015123492931419 h'(7)=h'(75) 0.00035077694898 h'(28) = h'(54) -0.013937313984866 h'(8)=h'(74) 0.00010793562158 h'(29)=h'(53) -0.003286932783157 h'(9)=h'(73) -0.0003664828042 h'(30) = h'(52) 0.0128277364444077 h'(10) = h'(72) -0.00086720478688 h'(31)=h'(51) 0.0256513705026164 h'(11)=h'(71) -0.00103830155239 h'(32) = h'(50) 0.0258098992001777 h'(12) = h'(70) -0.00056857143512 h'(33)=h'(49) 0.0089919729576737 h'(13)=h'(69) 0.000540463581141 h'(34)=h'(48) -0.019853210993468 h'(14)=h'(68) 0.001812857129142 h'(35)=h'(47) -0.046596703328572 h'(15)=h'(67) 0.002423154738356 h'(36)=h'(46) -0.053159507581729 h'(16) = h'(66) 0.001644950386710 h'(37)=h'(45) -0.026008964029071 h'(17) = h'(65) -0.000575067834766 h'(38)=h'(44) 0.0360644310189661 h'(18) = h'(64) -0.003295147081404 h'(39)=h'(43) 0.1190477283423599 h'(19) = h'(63) -0.004848728262050 h'(40)=h'(42) 0.1977673318734161 h'(20) = h'(62) -0.003739481299916 h'(41) 0.2456060317402789 h'(21)=h'(61) 0.0002509481048945

[0066] The half-band filter 400 is a low-pass filter that reduces the maximum bandwidth of the sampled data by two times. When implementing the circuit, half of the coefficients are 0, and no hardware implementation is required. Therefore, the efficiency of the half-band filter 400 is derived from this. Its amplitude-frequency characteristic curve is as follows: Figure 8 As shown, for an M-order HBF filter (i.e., half-band filter 400), its transfer function is shown in the formula:

[0067] H HBF (z) = h(0) + h(1) * z -1 +h(2)*z -2 +……+h(M-1)*z -M

[0068] Among them, H HBF (z) represents the transfer function of the half-band filter 400; z is a complex variable used to represent the variable of the Z transform in discrete-time signal processing; h(0), h(1), h(2)...h(M-1) represent the coefficients of each order of the half-band filter 400; M represents the order of the half-band filter 400.

[0069] Please continue reading Figure 1 Taking the sampling frequency Fs in the prior art as an example, which is reduced from 921.6KHz to the final Fs=100Hz~600Hz, the passband ripple of the filter group is less than 0.01dB, and the stopband attenuation is greater than 10dB, wherein the normalized passband is 0 to 0.4.

[0070] The design indicators of the half-band filter 400 are shown in Table 3:

[0071] Table 3. Design specifications of half-band filter 400

[0072]

[0073] According to the parameters of the relevant half-band filter 400 in the above table, the half-band filter 400 is modeled using the relevant function in MATLAB, and the order of the half-band filter 400 is 48, and the non-zero coefficients of the half-band filter 400 are shown in Table 4 (that is, the coefficients not listed in the table such as h(0) and h(2) are all "0"):

[0074] Table 4. Design coefficients for half-band filter 400

[0075] h(1)=h(47) -0.000506942954431066 h(15)=h(33) 0.023776095371118 h(3)=h(45) 0.001092558380791026 h(17)=h(31) -0.035868570532065 h(5)=h(43) -0.002191163497502963 h(19)=h(29) 0.056465434971747 h(7)=h(41) 0.003932501145620288 h(21)=h(27) -0.101644507775142 h(9)=h(39) -0.006551263329511143 h(23)=h(25) 0.316799499860653 h(11)=h(37) 0.010362054903603973 h(24) 0.5 h(13)=h(35) -0.015833491458646

[0076] The coefficients of the CIC filter are all 1, which is convenient for encoding without considering the quantization problem. However, the coefficients of the compensation filter 300 and the half-band filter 400 obtained by MATLAB modeling are all decimals, and direct encoding is very complicated. In order to facilitate encoding and save hardware resources, the coefficients need to be quantized. The quantization bit width b of the coefficients and the filter order L and passband ripple are shown in the formula:

[0077]

[0078] Where b represents the quantization bit width, L is the order of the filter, |E(e jw )| represents the passband ripple of the filter, and the minimum quantization bit width b of the filter coefficient can be obtained as shown in the formula:

[0079]

[0080] The minimum quantization bit width calculated by the formula only serves as an estimate, and may not be the most appropriate in actual design. The passband ripple of the half-band filter 400 is 0.01dB, and the order is 47. The minimum quantization bit width obtained is ≥12. Considering the requirement that the filter stopband attenuation is greater than or equal to 100dB in this design, combined with the analysis results of MATLAB fixed-point quantization, the quantization bit width b of the filter coefficient in this design is determined to be 17 bits. However, when performing RTL coding, it was found that there were many non-zero bits in the quantization coefficients, especially the negative filter coefficients. In order to save hardware resources, a more reasonable coding method is selected to reduce the non-zero bits in the coefficients. Therefore, CSD (Canonical Signed Digit) coding is used to optimize the filter coefficients.

[0081] For ordinary binary coding, each αi (n-1≥i≥0) is in the set {0, 1}. In CSD coding, each αi is in the set {-1, 0, 1}, that is, the coefficients are encoded using three numbers -1, 0, and 1. CSD coding is optimized based on binary coding using the formula:

[0082] 2 n +2 n-1 +…+2=2 n -2

[0083] The number of non-zero bits in the multiplication coefficient is reduced to a minimum, and any two non-zero numbers are not adjacent, thereby reducing the number of partial products caused by non-zero bit operations, thereby reducing the number of adders required in the multiplication operation, and simplifying the entire multiplication operation. The CSD encoding of the half-band filter 400 is shown in Table 5:

[0084] Table 5 CSD coding of half-band filter 400

[0085]

[0086]

[0087] In this embodiment, by introducing the compensation filter 300, the signal distortion is effectively corrected, the high precision of signal processing is ensured, the structure combining the CIC filter and the half-band filter 400 and the design of the variable decimation ratio are adopted, the power consumption of the system is significantly reduced, and the design of the variable decimation ratio also enables the filter to adapt to different application requirements and improves the flexibility of the design. It not only solves the problems existing in the prior art, but also provides significant improvements in high precision, low power consumption and design flexibility, and provides a more efficient, accurate and flexible solution for the field of digital signal processing.

[0088] The above embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A high-precision digital decimation filter, characterized in that: It includes four filter structures connected in series in sequence, the first filter structure is a multi-stage cascaded fixed extraction rate integrating comb filter, the second filter structure is a multi-stage cascaded variable extraction rate integrating comb filter, the third filter structure is a compensation filter, and the fourth filter structure is a half-band filter.

2. The high-precision digital decimation filter according to claim 1, characterized in that: The order of the fixed decimation rate integrating comb filter is at least one order higher than the order of the modulator of the previous stage of the digital decimation filter; the order of the variable decimation rate integrating comb filter is less than or equal to the order of the fixed decimation rate integrating comb filter.

3. The high-precision digital decimation filter according to claim 1, characterized in that: The fixed decimation rate integrating comb filter has a decimation rate of 64 times.

4. The high-precision digital decimation filter according to claim 1, characterized in that: The variable decimation ratio integrating comb filter can select decimation ratios including 6 times, 12 times and 36 times.

5. The high-precision digital decimation filter according to claim 1, wherein: The order of the fixed decimation rate integrating comb filter is k1, and the fixed decimation rate integrating comb filter includes a first integrating module, a first digital decimation module and a first differential module connected in series in sequence, the first integrating module includes k1 first integrating delay units connected in series, and the first differential module includes k1 first differential delay units connected in series.

6. The high-precision digital decimation filter according to claim 5, characterized in that: The first integral delay unit comprises a first adder and a first delay device, the first adder having a first positive input terminal, a second positive input terminal and an output terminal, the first positive input terminal of the first adder serves as an input terminal of the first delay unit, the second positive input terminal of the first adder is connected to the output terminal of the first delay device, the output terminal of the first adder is connected to the input terminal of the first delay device, and the output terminal of the first delay device serves as an output terminal of the first integral delay unit; The first differential delay unit includes a second adder and a second delay device, the second adder has a positive input terminal, a negative input terminal and an output terminal, the positive input terminal of the second adder serves as the input terminal of the first differential delay unit, the output terminal of the second adder serves as the output terminal of the first differential delay unit, the positive input terminal of the second adder is connected to the input terminal of the second delay device, and the output terminal of the second delay device is connected to the negative input terminal of the second adder.

7. The high-precision digital decimation filter according to claim 1, characterized in that: The order of the variable decimation ratio integrating comb filter is k2, and the variable decimation ratio integrating comb filter includes a second integrating module, a second digital decimation module and a second differential module connected in series in sequence, the second integrating module includes k2 second integrating delay units connected in series, and the second differential module includes k2 second differential delay units connected in series.

8. The high-precision digital decimation filter according to any one of claims 1 to 7, characterized in that: The system functions of the fixed decimation rate integrating comb filter and the variable decimation rate integrating comb filter are expressed as follows: H(z)=(1+z -1 ) N (1+z -2 ) N (1+z -4 ) N …(1+z 1-2k ) N Among them, H(z) represents the system function of the integral comb filter; z is a complex variable used to represent the variable of the Z transform in discrete-time signal processing; N represents the decimation factor of the integral comb filter, and k represents the order of the integral comb filter.

9. The high-precision digital decimation filter according to any one of claims 1 to 7, characterized in that: The gain of the transfer function of the compensation filter, the fixed decimation rate integral comb filter, and the variable decimation rate integral comb filter in cascade connection is 1.

10. The high-precision digital decimation filter according to any one of claims 1 to 7, characterized in that: The transfer function of the half-band filter is as follows H HBF (z)=h(0)+h(1)*z -1 +h(2)*z -2 +……+h(M-1)*z -M Among them, H HBF (z) represents the transfer function of the half-band filter; z is a complex variable used to represent the variable of the Z transform in discrete-time signal processing; h(0), h(1), h(2)...h(M-1) represent the coefficients of each order of the half-band filter; M represents the order of the half-band filter.