Method and apparatus to optimize adaptive radio-frequency systems
View Patent ↗A performance optimizing circuit is provided for a signal processing system which is parameterized by a set of coefficients that vary the operational characteristics of the signal processing system. The performance optimizing circuit receives as input a reference signal and an output signal of the signal processing system. The performance optimizing circuit may include (a) a cost computation circuit that receives the reference signal and the output signal and provides as output a cost signal representing a cost function computed using a set of current values for the set of coefficients, the output signal and the reference signal; and (b) a cost optimizer circuit that, at each of a plurality of successive time intervals, evaluates one or more values of the cost signal in the cost computation circuit and provides to the signal processing system a new set of values for the set of coefficients. The cost optimizer circuit implements two or more of the random search, parabolic interpolation and hill climbing techniques.
1. A performance optimizing circuit for a signal processing system parameterized by a set of coefficients that vary the operational characteristics of the signal processing system, the performance optimizing circuit receiving as input a reference signal and an output signal of the signal processing system, the performance optimizing circuit comprising:
a cost computation circuit receiving the reference signal and the output signal and which provides as output a cost signal representing a cost function computed using a set of values for the set of coefficients, the output signal and the reference signal; and
a cost optimizer circuit that, at each of a plurality of successive time intervals, evaluates one or more values of the cost signal from the cost computation circuit and provides to the signal processing system a new set of values for the set of coefficients, the cost optimizer circuit implementing two or more of the random search, parabolic interpolation and hill climbing techniques.
2. A performance optimizing circuit as in claim 1 , wherein the random search technique selects a set of incremental values for the set of coefficients using a random process.
3. A performance optimizing circuit as in claim 2 , wherein each incremental value is less than a predetermined integer multiple of a resolution of the corresponding coefficient.
4. A performance optimizing circuit as in claim 2 , wherein the parabolic interpolation technique provides the new set of values for the set of coefficients based on varying the set of current values using the set of incremental values selected by the random search technique.
5. A performance optimizing circuit as in claim 4 , wherein the set of current values are varied by adding each coefficient to an integer multiple of the corresponding incremental value.
6. A performance optimizing circuit as in claim 5 , wherein the parabolic interpolation technique computes a set of corrective values to adjust the set of current values for the set of coefficients, such that adjusted values provide a minimum in the cost function.
7. A performance optimizing circuit as in claim 6 , wherein the new set of values correspond to adjusting the set of current values each adjusted by the products of an adjustment factor and the corresponding corrective value, wherein the adjustment factor is less than 1.0.
8. A performance optimizing circuit as in claim 7 , wherein the adjustment factor decreases over time.
9. A performance optimizing circuit as in claim 6 , wherein the new set of values result from adjusting a product of a leakage term and each current value.
10. A performance optimizing circuit as in claim 4 , wherein the hill-climbing technique rejects the new set of values for the set of coefficients when a figure of merit obtained from the new set of values and a set of current values for the set coefficients exceeds a predetermined threshold value.
11. A performance optimizing circuit as in claim 10 , wherein the hill-climbing technique rejects the new set of values at a predetermined probability, when the figure of merit exceeds a second predetermined threshold value.
12. A method for optimizing a signal processing system parameterized by a set of coefficients that vary the operational characteristics of the signal processing system using a reference signal and an output signal of the signal processing system, the method comprising:
at each of a plurality of successive time intervals, using two or more of the random search, parabolic interpolation and hill climbing techniques to:
(a) evaluate one or more values of a cost function based on a set of current values for the set of coefficients, a value of the reference signal and a value of the output signal; and
(b) provide the signal processing system a new set of values for the set of coefficients.
13. A method as in claim 12 , wherein the random search technique selects a set of incremental values for the set of coefficients using a random process.
14. A method as in claim 13 , wherein each incremental value is less than a predetermined integer multiple of a resolution of the corresponding coefficient.
15. A method as in claim 13 , wherein the parabolic interpolation technique provides the new set of values for the set of coefficients based on varying the set of current values using the set of incremental values selected by the random search technique.
16. A method as in claim 15 , wherein the set of current values are varied by adding each coefficient to an integer multiple of the corresponding incremental value.
17. A method as in claim 16 , wherein the parabolic interpolation technique computes a set of corrective values to adjust the set of current values for the set of coefficients, such that adjusted values provide a minimum in the cost function.
18. A method as in claim 17 , wherein the new set of values correspond to adjusting the set of current values each adjusted by the products of an adjustment factor and the corresponding corrective value, wherein the adjustment factor is less than 1.0.
19. A method as in claim 18 , wherein the adjustment factor decreases over time.
20. A method as in claim 17 , wherein the new set of values result from adjusting a product of a leakage term and each current value.
21. A method as in claim 15 , wherein the hill-climbing technique rejects the new set of values for the set of coefficients when a figure of merit obtained from the new set of values and a set of current values for the set coefficients exceeds a predetermined threshold value.
22. A method in claim 21 , wherein the hill-climbing technique rejects the new set of values at a predetermined probability, when the figure of merit exceeds a second predetermined threshold value.