A study on ECC for servo track identifiers

Abstract

N. Suhrahmuny".llck. M. Chue' (1) Western Digital Corporalim. SX63 Rue Ferrilri Sasi lohe. CA 92118. USA limoductionAs the recording industry has pushed towards higher areal densities, lhe MR sensor sensitivity has progressively increased with n concomilant decrease in the sensor size to suppoii higher SNR and TPI requirements. The.% changes have made the render sensitive I O noise phenomena and bwlinc popping (BLP) evenls 111. The observed effecl in all head instahilily cases is extra pulses. missing pulses and T A like events.The servo psllern in the drive usually benrs the hrunl o l the elkcis o f (he inslahiiily kcause il is the least r o b u l in terms d dealing with the instahilily effect.; and i d w due 10 the delayed rcliiriilion ar inslability cvcnis. So an clkcctive drivc dcsign and mmur:sluring slrnlcgy has Io cope with all forms of head iimlahility. In this paper we propose n method to characterize head isslabilitics and propox Ihc MC olGray Hanuning. Golay. and Rccd Solomon (RS) codcs IO dcal with Track ID errors in the servo sector during write and read operations. We tiral discuss the churucteriratinn method and then show sew" system perbrmance with different ECC coder. Chariiclcrizalion Mcl hodIn order to understand the nature of the instability events disk drives were characleiired. The characterinlion method accessed lhc d i k e in iiiltive Inode using lbe setup sliow~~ iii Figurrl. A random seek operation WI F performed to a drive data track and the Servo Gray Code (GCI data for that track were captured using an oxillowope. Scrvo GC values for a numher of sew" tracks were collected rimiliirly lhrough rilndoin wrilelreild aeeka. A wriwure prugram was lhen used Iu compaile the expected GC to the captured GC values and count the numher of errors and genet ate a histogmm. A typical histogram of lhc rcsults is shown in Figure 2. Details oflhe Error Correcting CodeAs Seen in Figure 2 most errors are single hit (missing and exlra bill errors. This suggests lhal 1hit and 2-hit error correcling code9 cm he used IO mitigate the effect of these errors on the l r x k ID decvding prwess. In lhis nule we consider Hitmming. Gul;iy and RS Cutler. These cudrs have k e n modified in order to mainlain Ihe gray properly or the Gray Code. This is done hy ensuring that adjiiccnt track id's arc mvppcd 10 codcd valucs in Ihc code domain that arc il minimum dktmce apan. When more Ihan n single codeword is used to rcpre3eilt adjacent track id's rhe firs1 aid second codewords arc inlapped in reverw order. The decoding ol t h e e codes iF done lhrougll