Look-ahead delta sigma modulator with quantization using natural and pattern loop filter responses

• US 20050156768A1
• Filed: 11/22/2004
• Published: 07/21/2005
• Est. Priority Date: 01/16/2004
• Status: Active Grant

• Current Assignee: Cirrus Logic Incorporated
• Original Assignee: Cirrus Logic Incorporated
• Inventors: Melanson, John L.
• Granted Patent: US 7,170,434 B2
• Time in Patent Office: Days
• US Class Current: 341/143
• CPC Class Codes: H03M 7/3011 of non-linear distortion, e.g. by temporarily adapting the operation upon detection of instability conditions avoiding instability by structural design H03M7/3035

Abstract

A look-ahead delta sigma modulator reduces and simplifies the computations used to generate quantizer output values. Superposition can be applied to a loop filter response of the look-ahead delta sigma modulator. By superposition, the complete loop filter response for each output candidate vector equals the difference between a forced pattern response and a natural input signal response. The forced pattern response of the loop filter can be determined from the response to each output candidate vector with an input signal set to zero and loop filter state variables initially set to zero when determining the loop filter response for each output candidate vector. The natural input signal response of the loop filter can be determined from the response to each input signal vector with feedback data set to zero. The forced pattern response is independent of the input signal data and can be determined once for all input signal vectors X_t. The natural input signal response is independent of the feedback data and, therefore, need only be determined once for each output data value.

Claims

1. A method of processing input signal data using a look-ahead delta sigma modulator of depth N using natural and pattern loop filter responses, wherein N is greater than or equal to two, the method comprising:

determining a set of forced pattern responses of a loop filter of the look-ahead delta sigma modulator to a set of vectors;

determining a natural response of the loop filter of the look-ahead delta sigma modulator to a set of input data signal samples; and

quantizing each input signal data sample by applying predetermined decision criteria to determine a best match between each forced pattern response and the natural response and selecting output data from an output candidate vector associated with the forced pattern response used to determine the best match.

2. The method as in claim 1 wherein:

applying the predetermined decision criteria to determine the best match comprises determining a lowest power of the difference between one of the forced pattern responses and the natural response; and

selecting output data from the output candidate vector associated with the pattern response used to determine the best match comprises selecting the output candidate vector associated with the lowest power difference.

3. The method as in claim 2 wherein determining a lowest power match between one of the forced pattern responses and the natural response comprises determining a dot product for each forced pattern response and the natural response.

4. The method as in claim 1 wherein:

applying the predetermined decision criteria to determine the best match comprises determining a lowest power match between one of the forced pattern responses and the natural response wherein at least one element of the pattern response is weighted with a weight different than any other element of the forced pattern response; and

selecting output data from the output candidate vector associated with the forced pattern response used to determine the best match comprises selecting the output candidate vector associated with the lowest power match.

5. The method as in claim 1 wherein the set of pattern responses excludes pattern responses that provide duplicate best match determinations.

6. The method as in claim 1 wherein determining the set of forced pattern responses of the loop filter of the look-ahead delta sigma modulator further comprises precomputing the set of forced pattern responses before quantizing each input signal data sample.

7. The method of claim 6 further comprising:

storing the precomputed set of forced pattern responses in a memory for use in subsequently quantizing each input signal data sample.

8. The method as in claim 1 further comprising:

computing the natural response only once for quantization of each filtered input signal data sample.

9. The method as in claim 1 wherein the set of input data signal samples used to determine the natural response of the loop filter equals N.

10. The method as in claim 1 wherein the set of candidate vectors is reduced to less than 2^N without decreasing quantization accuracy.

11. The method as in claim 1 wherein the input signal data sample comprises audio input signal data.

12. The method as in claim 1 further comprising:

recording quantized input signal data on storage media.

13. The method as in claim 1 further comprising:

for each time t, determining a dot product of each forced pattern response with the natural response using substantially only addition operations.

14. The method as in claim 1 wherein the set of vectors used to determine the set of forced pattern responses includes a set of output candidate vectors from which the output candidate vector associated with the forced pattern response is selected.

15. The method as in claim 1 wherein each forced pattern response is determined using initial loop filter state variables set to zero.

16. The method as in claim 1 wherein determining a set of pattern responses of the loop filter includes determining a set of pattern response of the loop filter by setting signal input data to the loop filter of the delta sigma modulator to at least approximately zero and filtering a set of output candidate vectors.

17. The method as in claim 1 wherein determining a natural response of the loop filter of the look-ahead delta sigma modulator to a set of input data signal samples includes setting feedback data to the loop filter to at least approximately zero and filtering the set of input data signal samples.

18. A signal processing system comprising:

a look-ahead delta sigma modulator comprising;

a loop filter;

a pattern response generator coupled to the loop filter to determine a set of forced pattern responses of the loop filter to a set of vectors;

a natural response generator to determine a natural response of the loop filter to a set of input data signal samples; and

a quantizer to receive output data from the loop filter and the set of pattern responses, wherein the quantizer includes a function generator to quantize each input signal data sample by applying predetermined decision criteria to determine a best match between each forced pattern response and the natural response and selecting output data from an output candidate vector associated with the forced pattern response used to determine the best match.

19. The signal processing system as in claim 18 wherein the set of vectors used to determine the set of forced pattern responses includes a set of output candidate vectors from which the output candidate vector associated with the forced pattern response is selected.

20. The signal processing system as in claim 18 wherein the pattern generator determines the set of forced pattern responses of the loop filter to a set of vectors obtained by setting signal input data to the loop filter to at least approximately zero and filtering the set of vectors.

21. The signal processing system as in claim 18 wherein the natural response generator determines the natural response to a set of input data signal samples by setting feedback data to the loop filter to at least approximately zero and filtering the set of input data signal samples.

22. The signal processing system as in claim 18 further comprising:

a memory coupled to the loop filter to store the determined set of pattern responses of the loop filter for each output candidate vector.

23. The signal processing system as in claim 18 wherein:

applying the predetermined decision criteria to determine the best match comprises determining a lowest power match between one of the forced pattern responses and the natural response; and

selecting output data from the output candidate vector associated with the pattern response used to determine the best match comprises selecting the output candidate vector associated with the lowest power match.

24. The signal processing system as in claim 23 wherein determining a lowest power match between one of the forced pattern responses and the natural response comprises determining a dot product for each forced pattern response and the natural response.

25. The signal processing system as in claim 18 wherein:

applying the predetermined decision criteria to determine the best match comprises determining a lowest power match between one of the forced pattern responses and the natural response wherein at least one element of the pattern response is weighted with a weight different than any other element of the forced pattern response; and

selecting output data from the output candidate vector associated with the forced pattern response used to determine the best match comprises selecting the output candidate vector associated with the lowest power match.

26. The signal processing system as in claim 18 wherein the set of forced pattern responses excludes any pattern responses that provide duplicate best match determinations.

27. The signal processing system as in claim 18 wherein to determine the set of forced pattern responses of the loop filter of the look-ahead delta sigma modulator the natural response generator includes components to precompute the set of forced pattern responses before quantizing each input signal data sample.

28. The signal processing system as in claim 18 wherein the natural response generator generates the natural response only once for quantization of each filtered input signal data sample.

29. The signal processing system as in claim 18 wherein the set of input data signal samples used to determine the natural response of the loop filter equals N, wherein N is a look-ahead depth and is greater than or equal to two (2).

30. The signal processing system as in claim 18 wherein the set of output candidate vectors is reduced to less than 2^N without decreasing quantization accuracy, wherein N is a look-ahead depth and is greater than or equal to two (2).

31. The signal processing system as in claim 18 wherein the input signal data sample comprises audio input signal data.

32. The signal processing system as in claim 18 further comprising:

signal processing and recording equipment to process output data from the quantizer and record the processed output data on storage media.

33. The signal processing system as in claim 18 wherein the quantizer further comprises:

an add/compare/select system to combine each forced pattern response with the natural response using substantially only addition operations, compare the combinations, and select the output data.

34. An apparatus for processing input signal data using a look-ahead delta sigma modulator of depth N using natural and pattern loop filter responses, wherein N is greater than or equal to two (2), the apparatus comprising:

means for determining a set of forced pattern responses of a loop filter of the look-ahead delta sigma modulator to a set of vectors;

means for determining a natural response of the loop filter of the look-ahead delta sigma modulator to a set of input data signal samples; and

means for quantizing each input signal data sample by applying predetermined decision criteria to determine a best match between each forced pattern response and the natural response and selecting output data from an output candidate vector associated with the forced pattern response used to determine the best match.

35. The apparatus as in claim 34 wherein the set of vectors comprises a set of output candidate vectors.

36. A method of processing input signal data using a look-ahead delta sigma modulator of depth N using natural and pattern loop filter responses, wherein N is greater than or equal to two, the method comprising:

determining a response of a loop filter of the look-ahead delta-sigma modulator from a natural response of the loop filter to a set of N input data signal samples and a forced pattern response of the loop filter to a set of vectors; and

quantizing each input signal data sample by applying predetermined decision criteria to determine a best match between each forced pattern response and the natural response and selecting output data from an output candidate vector associated with the forced response used to determine the best match.