An analog-to-digital converter system [50D] processing an input signal, g, which can be either a discrete-time or a continuous-time signal. A first quantizer  generates a first digital signal, d0(k), representing the sum of the input signal, g, and a dithering signal, y0. A digital-to-analog converter  generates an analog feedback signal, alpha, representing accurately the first digital signal, d0(k). The DAC  may be linearized by the use of mismatch-shaping techniques. A filter  generates the dithering signal, y0, by selectively amplifying in the signal band the residue signal, r0, defined as the difference of the input signal, g, and the analog feedback signal, alpha. Optional signal paths  are used to minimize the closed-loop signal transfer function from g to y0, which ideally will be zero. An analog compensation signal, m0, which is described by a well-controlled relationship to the residue signal, r0, is extracted from the filter . Ideally, the closed-loop signal transfer function from g to m0 will be zero, or at least small in the signal band. A second quantizer  converts the analog compensation signal, m0, into a second digital signal, dm0(k). The two digital signals, d0(k) and dm0(k), are filtered individually and then added to form the overall output signal, dg(k). The second digital filter  has a low signal-band gain, which implies that the sensitivity to signal-band errors caused by the second quantizer  will be low. The output signal, dg(k), is a highly-accurate high-resolution representation of the input signal, g. Circuit imperfections, such as mismatch, gain errors, and nonlinearities, will cause only noise-like errors having a very low spectral power density in the signal band.The invention facilitates the implementation of uncalibrated highly-linear high-resolution wide-bandwidth A/D converters [50D], e.g., for use in digital communication systems, such as xDSL modems and other demanding consumer-market products for which low cost is of the essence.