Method and device for improving convergence time in correlation-based algorithms

Abstract

A method and a corresponding device reduce the convergence time of a correlation algorithm that uses random signals injected into an analog-to-digital converter (ADC) as input to the algorithm. The method and device involve, at a processor of a pipelined ADC, injecting a random signal into each of a plurality of stages in the pipeline and obtaining digital values generated in response to the random signals. Noise components of residue signals in the plurality of stages are calculated as a function of the digital values and values of the random signals. The noise components correspond to the random signals.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application claims the benefit, under 35 U.S.C. §119(e), of U.S. Provisional Patent Application No. 61 / 791,279 filed Mar. 15, 2013, the content of which is hereby incorporated by reference in its entirety.BACKGROUND INFORMATION[0002]Random signal sequences are sometimes injected into circuits. The randomness of the sequences may refer to the amplitude of the signals in the sequences, which may, for example, randomly vary between a set of discrete, predefined values. These sequences may be uncorrelated with the input signals to the circuits, i.e., the useful signals. The sequences may also be uncorrelated with each other. The sequences can be used for a variety of purposes, which are known in the art. For example, random sequences may be used to calibrate one or more stages of a pipelined analog-to-digital converter (ADC), e.g., to reduce a degree of mismatch between the devices in the stages or correct for non-linearity errors. Rando...

AI Technical Summary

Problems solved by technology

This may be difficult because dither injected into one stage may propagate through subsequent stages, thereby influencing the signals output from one stage to the next, and may ultimately influence the overall output as well.

Therefore, removal of the dither signals is not as straightforward as simply subtracting the dither injected into a stage from the output of that stage.

However, the convergence time of the correlation algorithm can be very long, even in the absence of any input signal, e.g., when the input is disconnected and only the random sequences are injected.

This leads to long start-up times and long production test times, which can be prohibitive.

Images