Design of comb-based decimators for sigma-delta analog-digital converters

Gordana Jovanovic Dolecek

Publicado: 2023-11-30

Formato de livro: Digital PDF (Portable Document Format)

Correspondência: gordana@ieee.org

ABSTRACT

Oversampled sigma-delta (sd) converters have become widely used as a valid alternative to conventional a/d (analog/digital) converters, operating at the nyquist frequency, due to many advantages like high resolution of reconstructed data and easy implementation in a single high–speed very large scale integration (vlsi) chip, among others.

In contrast to conventional a/d converters, an oversampling sd-a/d converter samples the analog input signal with a frequency much larger than the nyquist frequency (the minimum sampling frequency necessary to preserve the information in the sampled analog signal).

The sd architecture robustness and properties are the result of this oversampling. Later, the rate of the oversampled signal must be decreased to a nyquist frequency. This process is performed in a digital form and is called downsampling.

Therefore, the sd-a/d converter consists of a sigma-delta modulator and a decimation stage, which decreases the oversampled frequency of the modulator to the nyquist frequency. However, the downsampling introduces the aliasing, which deteriorates the signal and must be eliminated. This is why we need a digital filter before the downsampling. The filter is called the decimation filter

The key part of the decimation stage is a decimation filter responsible for the aliasing rejection. However, this filter cannot be a bottleneck of the total design and it is very important to design a simple decimation filter with high performances: high alias rejection and a high preservation of the signal.

The simplest decimation filter is the comb filter, which has all coefficients equal to unity and consequently does not need the multipliers for its implementation. However, the comb filter has a limited aliasing rejection and passband droop, which may deteriorate the decimated signal. We present the methods for improving the passband, aliasing rejection, and simultaneously improving both passband and aliasing rejection.