Challenges

Flash memory statistics is inherently different and much more difficult to handle when compared to the communication channel case. Largely speaking, there are four key dividing issues:
  1. Noise model. The Gaussian model, often used in communications, isn't good enough when targeting extreme reliability like Flash memory require. DensBits employs a unique statistical model for that noise, and its technology is optimized for handling it.
  2. Varying statistics. Flash memory statistics changes over time, physical space, and logical space. As direct consequence, one must employ an inherently adaptive system solution should performance be kept optimal. DensBits applies such an inherently adaptive approach.
  3. ECC and DSP efficiency. For Flash memory typically only 5% out of the array might be used as overhead for ECC and DSP purposes, whereas for communication applications one might use 25% to 50% out the bandwidth. Thus, ECC and DSP solutions for Flash memory must dramatically be more efficient.
  4. Required reliability. Flash memory require extremely high reliability of 1e-15 Bit Error Rate (BER) post coding, as opposed to typically only 1e-5 for communication applications. This is due to the fact that communication systems have option for retransmission, which isn’t available for Flash memory.

 

Flash memory statistics is inherently different and much more difficult to handle when compared to the communication channel
design: J&J+A | dev: entry.
Copyright © 2011 DensBits. All rights reserved. | Memory Modem is a trademark of DensBits Technologies Ltd. | Terms Of Use | Privacy Policy