Section I: IntroductionAs the conventional film media is working towards its ``superparamagnetic limit'', patterned media emerges as a new research thrust to achieve the ultra high magnetic recording density [1]- [5]. The basic idea is to transform the film media from its continuous form into the discrete form. By making the grains within one single island coupled with each other, the promising recording densities can be as high as 1 Tbits/in 2 . However, this new technology also brings a lot of new challenges and problems as pointed out in [1]. In this work, we focus on the read channel for patterned media. By building up appropriate read channel models, we illustrate how the various coding techniques affect the read channel performance for patterned media storage. In this digest, we first introduce the channel model in the next section. In Section III, we show part of the simulation results which employ and compare different coding schemes for perpendicular recording channels. Section II: Channel Model Based on recording physics magnetostatics, we can model the readback pulse as given in [2]-[5]. Generally, different head and recording structures produce different readback pulses. For longitudinal recording, the readback pulse is a dipulse, while it is a unipulse for perpendicular recording with soft underlayer and triplet pulse for vertical recording without no soft underlayer. The overall system model includes an encoder, a channel interleaver at the transmitter end and a sampler, an equalizer, a channel detector, a channel deinterleaver and a decoder at the receiver end. The equalizer is designed to be a 21-tap FIR filter to equalize the received signal to a desired PR target with the coefficients updated using LMS algorithm. Specifically, we consider the perpendicular recording. Section III: Simulation Results In all the simulations, the pole length is chosen as 4nm and the shield to shield distance is fixed at 90nm. The fly height and the bit thickness are both 20nm. The PR target is chosen as [-0.2, 1, -0.2] [2], which is a close match to the triplet pulse under consideration. We consider three different coding techniques here. Convolutional codes with generator (33,31) octal are implemented for convolutional coded system. Turbo codes with the same encoder generator are used for turbo coded systems. Regular LDPC codes with variable degree 3 are employed for LDPC coded systems. The number of decoder iteration is 15 for turbo decoder and 50 for LDPC (belief propagation) decoder. We compare both the uncoded and coded system performance in Fig.1. The length of information block is 10000. The code rate is fixed at R c =8/9. It is clearly demonstrated that, similar observations can be obtained for both with and without underlayer. At bit error rate of 10 -5 , simple convolutional codes can provide about 3dB coding gain compared to uncoded system. The two capacityapproaching codes (i.e. turbo codes and LDPC codes) have similar performances and can beat the uncoded system even more, about 6dB. As another example,...