Pipelined analog-to-digital converters (ADCs) tend to be sensitive to compo
nent mismatches in their internal digital-to-analog converters (DACs). The
component mismatches give rise to error, referred to as DAC noise, which is
not attenuated or cancelled along the pipeline as are other types of noise
. This paper describes an all-digital technique that significantly mitigate
s this problem. The technique continuously measures and cancels the portion
of the ADC error arising from DAC noise during normal operation of the ADC
, so no special calibration signal or auto-calibration phase is required. T
he details of the technique are described in the context of a nominal 14-bi
t pipelined ADC example at both the signal processing and register transfer
levels. Through this example, the paper demonstrates that in the presence
of realistic component matching limitations the technique can improve the o
verall ADC accuracy by several bits with only moderate digital hardware com
plexity.