Digital correlators have been analyzed in considerable detail for cases in which the components behave ideally. Here we consider the effects of the major deviations from ideal behavior which are encountered in practice. Attention is restricted to three-level quantization, but the fairly general case of the complex cross correlator is studied. (The real autocorrelator, for example, is a special case of this.) The errors analyzed include quantization threshold errors, quantizer indecision regions, sampler timing skews, quadrature network errors, and numerical errors in algorithms for converting digital measurements to equivalent analog correlations.