Differential evolution design of polyphase IIR decimation filters

• US 8,111,791 B2
• Filed: 12/20/2006
• Issued: 02/07/2012
• Est. Priority Date: 12/20/2005
• Status: Expired due to Fees

• Current Assignee: CSR Technology Incorporated (Qualcomm, Inc.)
• Original Assignee: SIRF Technology Incorporated (Qualcomm, Inc.)
• Inventors: Sturza, Mark Alan
• Primary Examiner(s): Payne, David C.
• Assistant Examiner(s): FILE, ERIN M
• Application Number:
  US11/614,001
• Publication Number:
  US 20070153946A1
• Time in Patent Office: 1,875 Days
• Field of Search: 375/346, 375/350, 455/307, 708/300
• US Class Current: 375/350
• CPC Class Codes: H03H 17/0277 comprising recursive filters
• CPC Class Codes: H03H 17/045 where the output-delivery frequency is lower than the input sampling frequency, i.e. decimation

Abstract

A method for designing a digital filter, includes: (a) selecting a predetermined number of current sets of coefficients for the digital filter; (b) selecting a metric for evaluating coefficients of the digital filter; (c) computing a metric for each current set of coefficients; (d) deriving a next set of coefficients based on a subset of the current set of coefficients; (e) computing the metric for the next set of coefficients; (f) replacing a selected one of the current set of coefficients based on comparing the metric for the next set of coefficients with the metric for the selected current set of coefficients; and (g) iterating steps (a) to (e) until a termination criterion is met. In one embodiment, the selected metric represents a desired stop band response. The next set of coefficients may be derived by adjusting a first current set of coefficients by a weighted difference between a second current set of coefficients and a third current set of coefficients. The weighted difference may be obtained by multiplying a predetermined factor to the difference between the second current set of coefficients and the third current set of coefficients. In one embodiment, a scaling is performed such that the next set of coefficients does not include a pole outside of the unit circle.

Claims

1. A method, implemented in an electronic computing device, for designing a digital filter, comprising:

(a) selecting a predetermined number of current sets of coefficients for the digital filter;

(b) selecting a metric for evaluating coefficients of the digital filter;

(c) scaling the selected current sets of coefficients by a function of the current sets of coefficients;

(d) computing a metric for each of the scaled current set of coefficients;

(e) deriving a next set of coefficients based on a subset of the scaled current set of coefficients;

(f) computing the metric for the next set of coefficients;

(g) replacing a selected one of the current set of coefficients based on comparing the metric for the next set of coefficients with the metric for the selected current set of coefficients; and

(h) iterating steps (a) to (f) until a termination criterion is met.

2. A method as in claim 1, wherein the selected metric represents a desired stop band response.

3. A method as in claim 1, wherein the next set of coefficients is derived by adjusting a first one of the current set of coefficients by a weighted difference between a second one of the current set of coefficients and a third one of the current set of coefficients.

4. A method as in claim 3, wherein the weighted difference is obtained by multiplying a predetermined factor to the difference between the second one of the current set of coefficients and the third one of the current set of coefficients.

5. A method as in claim 4, wherein the predetermined factor is less than 1.0.

6. A method as in claim 3, further comprising scaling the next set of coefficients each by a function of a maximum magnitude of the current set of coefficients.

7. A method as in claim 3, wherein the function is selected such that none of the next set of coefficients exceeds 1.0 after the scaling.

8. A method as in claim 1, wherein the precision of the current sets of coefficients and of the next set of coefficients is limited by a predetermined number of bits.

9. A method as in claim 1, wherein the digital filter comprises a N-path, polyphase IIR decimation filter.

10. A method as in claim 1, wherein the replacing comprises:

comparing the metric for the next set of coefficients with the metric for a selected one of the current set of coefficients; and

replacing the selected one of the current set of coefficients with the next set of coefficients if the metric for the next set of coefficients is less than the metric for the selected one of the current sets.

11. A method, implemented in a electronic computing device, for designing a digital filter, comprising:

(a selecting a predetermined number of current sets of coefficients for the digital filter;

(b) selecting a metric for evaluating coefficients of the digital filter;

(c) scaling the selected current sets of coefficients by a function of the current sets of coefficients;

(d) computing a metric for each the scaled current set of coefficients;

(e) deriving a next set of coefficients based on a subset of the scaled current set of coefficients;

(f) computing the metric for the next set of coefficients;

(g) for each of the current set of coefficients, comparing the computed metric for the respective current set of coefficients with the computed metric for the next set of coefficients and replacing the respective current set of coefficients with the next set of coefficients if the metric for the next set of coefficients is less than the metric for the respective current set of coefficients; and

(h) iterating steps (a) to (f) until a termination criterion is met.

12. A method as in claim 11, wherein the deriving comprises:

adjusting a first one of the current set of coefficients by a weighted difference between a second one of the current set of coefficients and a third one of the current set of coefficients.

13. A method as in claim 12, wherein the weighted difference is obtained by multiplying a predetermined factor to the difference between the second one of the current set of coefficients and the third one of the current set of coefficients.

14. A method as in claim 11, further comprising scaling the next set of coefficients each by a function of a maximum magnitude of the current set of coefficients.

15. A method as in claim 14, wherein the function is selected such that none of the next set of coefficients exceeds 1.0 after the scaling.

16. A method, implemented in a electronic computing device, for designing a digital filter, comprising:

(a) selecting a predetermined number of current sets of coefficients for the digital filter;

(b) selecting a metric for evaluating coefficients of the digital filter;

(c) scaling the selected current sets of coefficients by a function of the current sets of coefficients;

(d) computing a metric for each scaled current set of coefficients;

(e) deriving a next set of coefficients based on a subset of the scaled current set of coefficients;

(f) computing the metric for the next set of coefficients;

(g) comparing the computed metric for the respective current set of coefficients with the computed metric for the next set of coefficients;

(h) replacing the respective current set of coefficients with the next set of coefficients if the metric for the next set of coefficients is less than the metric for the respective current set of coefficients; and

(i) iterating steps (a) to (f) until a termination criterion is met.

17. A method as in claim 16, wherein the deriving comprises:

adjusting a first one of the current set of coefficients by a weighted difference between a second one of the current set of coefficients and a third one of the current set of coefficients.

18. A method as in claim 17, wherein the weighted difference is obtained by multiplying a predetermined factor to the difference between the second one of the current set of coefficients and the third one of the current set of coefficients.

19. A method as in claim 16, further comprising scaling the next set of coefficients, each by a function of a maximum magnitude of the current set of coefficients, if the maximum magnitude of the next set of coefficients exceeds a predetermined threshold.

20. A method as in claim 16, wherein digital filter comprises a N-path, polyphase IIR decimation filter.

21. A system for designing a digital filter, comprising:

(a) means for selecting a set of current coefficients;

(b) means for selecting a metric for evaluating the set of current coefficients;

(c) means for scaling the set of current coefficients by a function of the set of current coefficients;

(d) means for computing a metric for each of the scaled coefficients in the set of current coefficients;

(e) means for deriving a set of next coefficients based on a subset of the scaled set of current coefficients;

(f) means for computing a metric for the set of next coefficients;

(g) means for comparing the computed metric for the respective set of current coefficients with the computed metric for the respective set of next coefficients;

(h) means for replacing the respective set of current coefficients with the respective set of next coefficients if the computed metric for the respective set of next coefficients is less than the computed metric for the respective set of current coefficients; and

(i) means for iterating steps (a) to (f) until a termination criterion is met.

22. The system as in claim 21, wherein the digital filter comprises a N-path, polyphase IIR decimation filter.

23. The system as in claim 21, wherein the function in the means for scaling the set of current coefficients is a maximum magnitude of the set of current coefficients.

24. The system as in claim 21, wherein the function in the means for scaling the set of current coefficients is selected such that none of the set of next coefficients exceeds 1.0 after the scaling.

25. The system as in claim 21, wherein the metric in the means for selecting a metric represents a desired stop band response.