The analog-to-digital (A/D) interface is generally considered to be the most critical part of any signal acquisition and processing system. Because of the difficulty in achieving high-resolution and high-speed A/D converters, this interface has been a barrier to the realization of high-speed, high-throughput systems. Recently, there has been renewed interest in new and innovative approaches to A/D conversion, with a significant emphasis on photonic techniques. Interleaving is a common approach applied to high-speed photonic A/D conversion; it reduces the wide-bandwidth input signal to one that can be converted using conventional high-speed A/D converters. The high-speed sampled input is interleaved to individual channels with each channel operating at of the sampling rate. These channelization techniques are known to suffer from performance degradations due to channel-to-channel mismatch. Within the electronic A/D converter community, temporal oversampling and spectral noise shaping have become common practice in high-fidelity audio applications. Here, a low-resolution quantizer is embedded in a feedback architecture in an effort to reduce the quantization noise through spectral noise shaping. A large error associated with a single sample is diffused over many subsequent samples, and then linear filtering techniques are applied to remove the spectrally shaped noise, thereby improving the overall SNR of the converter. The approach to photonic A/D conversion described here leverages the 2-D nature of an optical architecture to extend the concept of spectral noise shaping to include 2-D spatial noise shaping. The proposed approach uses a mode-locked laser to generate the optical sampling pulses, an interferometer to modulate the electronic analog signal onto the optical pulses, and a 2-D smart-pixel hardware implementation of a distributed error-diffusion neural network. © 2002 Society of Photo-Optical Instrumentation Engineers.