In this work, media noise is considered as the magnetization derivative (MD) noise, because it is the MD, not thr magnetization, that produces external magnetic field that is sensed by any readhack head. This is in contrast to the previous treatments of media noise as additive, where the concepts of cross-track statistical averaging of magnetization and Poisson statistics for magnetic grains were employed [ I , 2, and references therein]. These works contained an implicit assumption that transitions in magnetization are uniform in shape and are written at their intended locations. However, in new advanced disk drives, the size of a magnetic bit has s h k dramatically, and contains only about 1-2 dozens of sfill stable magnetic grains. Furthennore, hit dimensions along the track have shrunk to -3-4 grains. In such an environment the traditional mapping technique "write c u r r e n f a e d i a magnetization" results in significant mapping errors. As pointed out in 131, when the sizes of magnetic sources are comparable with the track-width and hit wavelength, the concept uf statistical cross-track averaging becomes meaningless and Poisson statistics become irrelevant. Indeed, magnetic transitions (and pulses of MD) cannot be written at their intended locationsthey need to follow boundaries between magnetic grains, and the positions of MD-pulses deviate from their intended positions. Also, transitions (and pulses of MD) do not remain uniform; they fluctuate in shape and width, depending on grain distributions. A readback head, while reacting to an averaged MD across the track, integrates ragged deviations in transition contours. Thus, the media noiqe has a multiplicative characterit is a random multi-dimensionul mrdulotion of MD-pulses fhaf occurs in recording, hut ir sensed during readback. As was demonstrated in [4-61, an understanding of multiplicative media noise mechanism requires a separate analysis of MD behavior during recording and readback cycles.