Look-ahead delta sigma modulators with quantizer input approximations

• US 7,084,798 B2
• Filed: 01/18/2005
• Issued: 08/01/2006
• Est. Priority Date: 01/16/2004
• Status: Expired due to Fees

• Current Assignee: Cirrus Logic Incorporated
• Original Assignee: Cirrus Logic Incorporated
• Inventors: Melanson, John L.
• Primary Examiner(s): Barnie, Rexford
• Assistant Examiner(s): LAUTURE, JOSEPH J
• Application Number:
  US11/037,316
• Publication Number:
  US 20050156772A1
• Time in Patent Office: 560 Days
• Field of Search: None
• 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 signal processing system includes a look-ahead delta sigma modulator having an approximation generator to approximate quantizer input signals. A look-ahead delta-sigma modulator can be implemented as a noise shaping filter and a quantizer. The quantizer can be implemented as a function generator. State variables of the noise shaping filter provide the input data from which the function generator determines a quantizer output signal. Latter state variables are more dominant in determining the quantizer output signal. Accordingly, earlier state variables can be approximated to a greater degree than earlier state variables. The approximations can result in slightly lower output signal accuracy but can significantly decrease implementation cost. Additionally, latter state variables can completely dominate (i.e., be deterministic) the quantizer output signal. This situation can result in a further, slight increase in the non-linearity of one or more quantization region boundaries.

Claims

1. A signal processing system comprising:

a look-ahead delta sigma modulator comprising;

a noise shaping filter to process a signal and generate N state variables, wherein N is an integer greater than or equal to two;

an approximation generator coupled to the noise shaping filter to generate approximated state variables determined from at least a subset of the N state variables; and

a quantizer coupled to the approximation generator to quantize quantizer input data determined from the approximated state variables and determine a quantization output value with an objective of minimizing quantization error at a current and at least one future time step.

2. The signal processing system as in claim 1 wherein the quantizer comprises a look-up table having records of possible approximated state variables and corresponding records of quantization output values.

3. The signal processing system as in claim 2 wherein each record of possible approximated state variables comprises a concatenation of approximated state variables.

4. The signal processing system as in claim 2 wherein each ith approximated state variable record represents a range of non-approximated state variable values generally evenly distributed on either side of the ith record and for each ith record the corresponding quantization output value is determined using the non-approximated state variable located in the center of the range of non-approximated state variable values centered around the ith record.

5. The signal processing system as in claim 2 wherein:

each ith approximated state variable record is determined using an actual test signal X(n) to drive the look-ahead delta sigma modulator using non-approximated state variables;

the actual test signal comprises a set of test signal samples associated with the ith approximated state variable record;

the number of quantization output values=+1 decisions and the number of quantization output values=−

1 decisions are recorded;

for the set of test signal samples associated with the ith approximated state variable record that had a majority of +1'"'"'s recorded, the quantization output value corresponding to the ith approximated state variable record is +1; and

for the set of test signal samples associated with the ith approximated state variable record that had a majority of −

1'"'"'s recorded, the quantization output value corresponding to the ith approximated state variable record is −

1.

6. The signal processing system as in claim 2 wherein any time that a dominant approximated state variable is (i) greater than A, the quantization output value is a +1 and (ii) less than −

B, the quantization output value is a −

1, wherein A and B represent predetermined threshold values.

7. The signal processing system as in claim 6 wherein the noise shaping filter comprises multiple, cascaded integrators and the dominant approximated state variable is derived from a state variable of a last integrator of the multiple, cascaded integrators.

8. The signal processing system as in claim 2 wherein the records of quantization output values comprises records of indeterminate quantization output values and the quantizer determines the quantization output value using non-approximated state variables of the noise shaping filter each time one of the generated approximated state variable corresponds to an indeterminate quantization output value.

9. The signal processing system as in claim 1 wherein the approximation generator is configured to generate the approximated state variables by rounding the subset of the N state variables to smaller bit sizes, using more rounding for less influential state variables.

10. The signal processing system as in claim 1 wherein the look-ahead delta sigma modulator is a one-bit look-ahead delta sigma modulator.

11. The signal processing system as in claim 1 wherein the quantizer comprises a look-up table having records of all possible combinations of approximated state variables and corresponding records of quantization output values.

12. The signal processing system as in claim 1 further comprising:

a preprocessor coupled between the noise shaping filter and the approximation generator to preprocess the at least a subset of the N state variables.

13. The signal processing system as in claim 1 wherein the look-ahead delta sigma modulator comprises a jointly non-linear function generator.

14. The signal processing system as in claim 1 wherein the noise shaping filter comprises N integrators and the state variables represent output values of at least a subset of the integrators during successive, discrete times.

15. The signal processing system as in claim 14 wherein the subset of the integrators consists of the Nth integrator through the N−

x^th integrators, wherein 2≦

x≦

N−

1.

16. The signal processing system as in claim 1 wherein the noise shaping filter comprises an infinite impulse response filter.

17. The signal processing system as in claim 1 further comprising:

a processor; and

a memory coupled to the processor and storing processor executable code to implement the noise shaping filter and the quantizer.

18. The signal processing system as in claim 1 wherein the quantizer input data is derived from audio input signal data.

19. The signal processing system as in claim 1 further comprising:

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

20. A method of processing a signal using a look-ahead delta-sigma modulator that includes a noise shaping filter having N state variables, wherein N is an integer greater than or equal to two, the method comprising:

filtering an input signal to the look-ahead delta-sigma modulator using the noise shaping filter to generate the N state variables;

generating approximated state variables determined from at least a subset of the N state variables; and

quantizing the quantizer input data determined from the approximated state variables to determine a quantization output value with an objective of minimizing quantization error at a current and at least one future time step.

21. The method as in claim 20 wherein quantizing the quantizer input data determined from the approximated state variables comprises accessing a look-up table to retrieve a quantization output value record corresponding to a record of the generated approximated state variables.

22. The method as in claim 21 wherein the record of the generated approximated state variables comprises a concatenation of approximated state variables.

23. The method as in claim 21 wherein each ith approximated state variable record represents a range of non-approximated state variable values generally evenly distributed on either side of the ith record and for each ith record the corresponding quantization output value is determined using the non-approximated state variable located in the center of the range of non-approximated state variable values centered around the ith record.

24. The method as in claim 21 further comprising:

driving the look-ahead delta sigma modulator using an actual test signal X(n), wherein the test signal X(n) comprises a set of test signal samples associated with the ith approximated state variable record;

quantizing the test signal X(n) using non-approximated state variables to determine quantization output values for each test signal sample;

recording the output values for each test signal sample;

for the set of test signal samples associated with the ith approximated state variable record that had a majority of +1'"'"'s recorded, recording the quantization output value corresponding to the ith approximated state variable record as +1 in the look-up table; and

for the set of test signal samples associated with the ith approximated state variable record that had a majority of −

1'"'"'s recorded, recording the quantization output value corresponding to the ith approximated state variable record as −

1 in the look-up table.

25. The method as in claim 21 wherein any time that a dominant approximated state variable is (i) greater than A, the method further comprises recording the quantization output value as a +1 and (ii) less than −

B, the method further comprises recording the quantization output value as a −

1, wherein A and B represent predetermined threshold values.

26. The method as in claim 25 wherein the noise shaping filter comprises multiple, cascaded integrators, the method further comprising deriving the dominant approximated state variable from a state variable of a last integrator of the multiple, cascaded integrators.

27. The method as in claim 21 wherein the records of quantization output values comprises records of indeterminate quantization output values, and the method further comprises determining the quantization output value using non-approximated state variables of the noise shaping filter if the generated approximated state variable corresponds to an indeterminate quantization output value.

28. The method as in claim 21 wherein generating approximated state variables comprises rounding the subset of the N state variables to smaller bit sizes, using more rounding for less influential state variables.

29. The method as in claim 21 wherein quantizing the quantizer input data determined from the approximated state variables to determine a quantization output value comprises quantizing the quantizer input data determined from the approximated state variables to determine a one-bit quantization output value.

30. The method as in claim 21 further comprising:

preprocessing the subset of N state variables prior to generating the approximated state variables.

31. The method as in claim 20 wherein the input signal data sample comprises audio input signal data.

32. The method as in claim 20 further comprising:

recording quantized quantizer input signal data on storage media.

33. An apparatus comprising:

means for filtering an input signal to a look-ahead delta-sigma modulator to generate N state variables, wherein N is an integer greater than or equal to two;

means for generating approximated state variables determined from at least a subset of the N state variables; and

means for quantizing the quantizer input data determined from the approximated state variables to determine a quantization output value with an objective of minimizing quantization error at a current and at least one future time step.