TABLE OF CONTENTS As we learned earlier in this chapter, the achievable accuracy of a Nyquist-rate data converter is limited by the matching accuracy of the analog components (e.g., capacitors or current sources) in the circuit, which normally has a minimum relative error of about 0.02%. As a result, most Nyquist-rate converters are not capable of achieving ENOBs over 12-bit calibrations. However, applications such as digital audio and instrumentations usually demand data converters with ENOBs as large as 20 or even 24 bits. Even with effective calibrations, the integrating data converter, which is the most accurate (and the slowest) Nyquist-rate data converter, can hardly achieve an accuracy as high as 24 bits.
In contrast to the Nyquist-rate data converters, oversampling data converters are capable of achieving ENOBs over 20 bits without any special calibration circuitries; however, this occurs at the cost of applying sampling rates much higher than the Nyquist rate (typically by a factor between 16 and 512). In other words, oversampling data converters trade speed for accuracy.
In addition, for a Nyquist-rate ADC, undesired out-of-band signals tend to reside near the desired signal cutoff frequency. As the result, a highly selective antialiasing filter (AAF) is often required to precede the ADC. In comparison, an oversampling ADC uses a sampling frequency that is much higher than the signal cutoff frequency, thus the possibility of aliasing is trivial, and the use of a high-order AAF is not mandatory. Furthermore, the decimation filter, which can be...
