US 12,444,439 B2
ML assisted dynamic decoding gear selection
Eran Sharon, Rishon Lezion (IL); Ran Zamir, Ramat Gan (IL); Omer Fainzilber, Herzeliya (IL); and Idan Alrod, Herzliya (IL)
Assigned to Sandisk Technologies, Inc., Milpitas, CA (US)
Filed by Sandisk Technologies, Inc., Milpitas, CA (US)
Filed on Jul. 26, 2023, as Appl. No. 18/359,147.
Claims priority of provisional application 63/481,056, filed on Jan. 23, 2023.
Prior Publication US 2024/0249749 A1, Jul. 25, 2024
Int. Cl. G11B 20/10 (2006.01); G11B 20/18 (2006.01)
CPC G11B 20/1037 (2013.01) [G11B 20/1833 (2013.01); G11B 2020/185 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A controller, comprising:
a first decoder, wherein the first decoder operates at a first power efficiency level;
a second decoder, wherein the second decoder operates at a second power efficiency level above the first power efficiency level; and
a decoder manager coupled to the first decoder and the second decoder, wherein the decoder manager comprises a machine learning (ML) classifier, the decoder manager is configured to:
direct encoded data to the first decoder for decoding;
decode, using the first decoder, the encoded data directed to the first decoder;
during the decoding, calculate a point in the decoding where the encoded data is sufficiently decoded to be able to be decoded by the second decoder, wherein the point is not calculated prior to the decoding;
determine whether the point has been reached by the first decoder;
direct the sufficiently decoded data to the second decoder for decoding;
determine whether the second decoder successfully decoded the sufficiently decoded data;
trigger a recovery flow based on the determination that the second decoder did not successfully decode the sufficiently decoded data; and
if the recovery flow is triggered, bias a conservative working point, wherein the conservative working point is more conservative than the point.