|
Multi-track Recording System
|
|
|
Personal research presented at poster session for TMRC '98 (Boulder CO, August 17-19, 1998; Designation CP8)
|
|
One of Chuck's personal interests is research into detectors for multi-track recording systems. Chuck has proposed a novel write narrow - read wide (G)MR recording head. This is, of course, exactly opposite to the current design philosophy of write wide - read narrow. The new head is NOT an array head.
The idea is to write one track of data, then on the very next revolution write an adjacent track with no guardband in between the tracks. This creates a pair of tracks. Guardbands exist between pairs of tracks only. Both tracks of a pair are written one immediately after the other to minimize spindle speed variation between the writing of the tracks.
On readback, pairs of tracks are read simultaneously by the wide (G)MR reader. For example, in a typical disk drive the head reads sector 1 then sector 2 then 3 and then 4. In this multi-track system, the head reads sectors 1 and 2 simultaneously -- one in each of the tracks of the pair. Then sectors 3 and 4 are read simultaneously. By ensuring a particular phase difference between the tracks exists during writing, a properly constructed detector can correctly detect both tracks even when read simultaneously.
|
|
|
|
Based on simple calculations and simulations in additive white gaussian noise, this method should
|
|
- Improve capacity by about 35 to 45%
- Double readback data rate without correspondingly increasing the write frequency
- Greatly improve servo burst SNR to allow for following smaller tracks
- Suppress off-track interference with a simplified version of this idea
- NOT require an array head
|
|
|
|
The challenges of this method include
|
|
- Maintaining the phase relation between the tracks during writing
- Added detector complexity
- Difficult to achieve and maintain synchronization between the tracks on readback (patterned media people will also be faced with this problem)
- Radically different from traditional systems
|
|
|
|
Chuck is in the process of writing a paper on this topic for submission as a journal article. Here is the current abstract of the paper.
Abstract— A novel (G)MR head architecture and associated multi-track detection system is introduced. The head’s write element is nominally about half the size of the read element, completely contrary to the traditional write-wide read-narrow head design. It is not an array head. Capacity increases from 35% to over 45% are predicted. Further, the user read data rate can effectively double without correspondingly faster write drivers or write field switching requirements.
The data in pairs of tracks must be written in two consecutive passes to minimize frequency differences between the tracks. Each pair of tracks is read simultaneously by the wider read element and accurately decoded using digital communication techniques for co-channel detection. This is accomplished without resorting to a complicated array head structure or multiple pre-amps and read channels.
Experimental performance of such a system is approximated by writing two overlapping tracks and capturing the readback waveform that results when an AMR read element is positioned over both tracks. The digitized waveform is then processed through a software co-channel detector. Promising performance is demonstrated at typical signal-to-noise ratio conditions.
Three different detectors are presented: a joint maximum likelihood sequence detector (JMLSD) implemented as a Viterbi algorithm, a two-stage joint maximum a posteriori symbol detector (JMAPSD) implemented as a fixed delay tree search with decision feedback (FDTS/DF) and a modified JMAPSD that dramatically improves performance when both tracks have about equal signal power.
There are many potential benefits to this type of recording system. The wider active read element postpones the onset of manufacturing problems associated with extremely narrow (G)MR heads. The effective user data rate can theoretically be double the data rate of each track. Internet servers especially can benefit from this increase. This also provides a way to get higher data rates in slower spinning, small form factor drives. Further, this increase in effective data rate can be traded for a slower write frequency on each track, thereby easing write process limitations on fast head field switching and on non-linear transition shift and partial erasure. The elimination of one inter-track guardband between every other track automatically improves track density. In addition, all of the written transition width can be read, not just a portion as in traditional write-wide read-narrow designs. Simpler versions of this technique can be used to estimate and cancel off-track interference. Further, the detected differences between track pairs can be used for servo demodulation. These techniques are also applicable to tape and optical recording systems.
If you would like to find out more, please contact us.
|
|
|
|
|
|
|
|
|
|
|
|
50th Anniversary of the Hard Disk Drive: Keynote Speech
|
|
|
|
|
|
|
Place screen shot of PRMLpro here, but don't have it say "for Infiniium"
|
|
|
|
