hesela.devRacetrack position error
In a racetrack read-channel model, a position error is an incorrect shift that causes a fixed read head to skip or reread stored domains.
In the read-channel model of Chee et al., overshifting skips a domain and produces a deletion in the observed bit sequence. A failed shift rereads the same domain, producing a sticky insertion: a repetition, not an arbitrary extra bit. The stored domain is not erased by the modeled deletion. These are synchronization errors, not simply flipped bit values. [1]
Multiple heads do not automatically provide independent observations. Their errors are correlated through the shared shift and head spacing; adjacent heads can return the same ambiguous sequence within a long run. The paper's bounded-error correction guarantees depend on coding constraints and head spacing, not merely the presence of a second head. [1]
Physical positioning error is broader: continuous displacement from the intended position need not produce a discrete read error. Ishibashi et al. studied a Pt/GdFeCo wire about 5 micrometers wide using Kerr microscopy. Figure 4C reports 1,000 trials for Gd concentration x = 30.5%: bit-center position variance grew approximately linearly with average shift-pulse count at 14 V and 20 V, but nonlinearly at 8 V. The number of pulses per shift operation was adjusted between amplitudes to keep the average bit size near 5 micrometers. The shift pulse width was 2 ns, with 10 ms between pulses in the protocol. These are not a bit-error rate or sustained throughput measurement. [2]
Relations
- related
- Racetrack memory
Limitations
- The cited work is a coding model, not a measured device failure rate. [1]
Citations
- Y. M. Chee, H. M. Kiah, A. Vardy, V. K. Vu and E. Yaakobi: Coding for Racetrack Memories, arXiv:1701.06874v1 (2017), sections I-III
- M. Ishibashi et al.: Decoding the magnetic bit positioning error in a ferrimagnetic racetrack (2024), Fig. 4C; DOI 10.1126/sciadv.adq0898
Sources & caveats
- Full-text review of a theoretical read-channel model; no hardware measurement, raw-data reanalysis or independent code-theorem proof is claimed.
- c2: full-text Results, Fig. 4C and Discussion; laboratory spatial variance, not raw-data reanalysis or measured read-error probability. Dense nanoscale behavior is not established by the micrometer-scale experiment.
- The distinction between continuous displacement, discrete read-channel events and measured event probabilities is a source-based interpretation, not a fitted conversion model.