We illustrate the principles of digital quadrature detection and call
attention to its various benefits (ghost-free spectra and high immunit
y to low-frequency interference) and its intrinsic capability of gener
ating data sets with different aliasing behaviors. A function describi
ng the filtering efficiency is introduced, and the digital filters of
our detector are compared with their analog counterparts of convention
al nuclear magnetic resonance spectrometers. With an appropriate analo
g-to-digital converter (ADC), our digital detector has a dynamic range
which is essentially limited by the analog noise, and increases when
the spectral bandwidth is reduced. These nearly ideal performances are
achieved through dithering, which randomizes the quantization error a
nd oversampling, which reduces the quantization noise in the band of i
nterest. We introduce a ''figure of merit'' for AD converters which es
timates the noise performances of ADCs, and allows to compare products
which achieve different compromises between speed and accuracy. The d
istortions due to the nonlinearities of the ADCs are analyzed through
simulations. We find that the majority of the spurious signals (i.e.,
the errors other than noise) occur outside the band of interest, and a
re disposed through digital filtering. An unexpected result of the sim
ulation is that, in some circumstances (e.g., large-scale narrowband d
ithering), an increase in the number of bits of the ADC may actually r
educe the distortion-free dynamic range. In Sec. VIII we analyze pract
ical problems like the role of the aperture jitter and the selection o
f the sampling frequency. (C) 1996 American Institute of Physics.