Digital halftoning is required whenever a grayscale image must be displayed or rendered on a device with fewer levels of gray, such as a laser printer or a low-cost video display. Halftoning converts shades of gray to patterns of dots that give the illusion of grayscale. In this talk, we present some common methods of halftoning, and then explain error diffusion in detail. Error diffusion produces excellent halftones, but the effect of the error filter at its heart is unpredictable. We show that error diffusion is equivalent to a noise shaping feedback coder (NSFC), a form of delta-sigma modulator, and use this correspondence to explain certain features seen in error diffused imagery. Using a linearized NSFC model of the halftoned image, we derive image quality metrics for halftones.