February 2016
16
C
hips
& C
omponents
Depending on the flash generation, flashes
require a certain number of bits to be correct-
able by the controller. U9 corrects up to 96 bit
errors per 1KB data unit. During Read, the
ECC information is used to identify decreas-
ing data quality, wrong bits are corrected and
flawless information is send to the host. As
soon as a certain threshold of bits that need
to be corrected is reached, the entire block is
refreshed immediately after error correction
and without impacting the overall device per-
formance. This so-called Near-Miss ECC fea-
ture is standard in the U9.
Bit errors occur for several different physical
reasons: Data retention and program disturb
are most common but more recent flashes
also need to be protected against read disturbs.
Therefore, Hyperstone developed an algorithm
(threshold and triggers can be defined by ven-
dors) that frequently checks the data of the
entire flash system. Depending on error pat-
terns, individual blocks, areas, or the entire sys-
tem is refreshed. The U9 firmware analyses the
error pattern in order to optimally refresh data
without unnecessary performance or wear-
out impact. This ensures longest lifetime, data
integrity and reliability of the device. For the
unusual case that the ECC is not able to cor-
rect errors in the data read from a flash, Read
Retry is applied. With this function it is pos-
sible to recover data by repeating reads on the
same cells but with different flash threshold set-
ups. Reading with dynamic read thresholds can
dramatically increase flash endurance.
5-stage error handling describes a combina-
tion of single features of the U9 to make the
flash drive as safe as possible. It therefore com-
bines the basic error correction with additional
features like Near-Miss ECC, Dynamic Data
Refresh, Read Retry and additional CRC (cyclic
redundancy check) protection. The process of
the error handling is optimized in a way that:
1) Standard ECC corrects up to 96 bits depend-
ing on flash requirements; 2) Near Miss ECC
identifies a certain level of errors; 3) Data on
the flash is refreshed by the controller and cor-
rect data is ensured by double-checking with a
CRC check; 4) If the number of errors is greater
than the maximum capacity of the ECC, Read
Retry is started; 5) To cope with extreme read
intensive situations, which may have negative
influence on the flash, Dynamic Data Refresh
is used to refresh data based on the actual error
situation on the whole drive.
The U9 feature Device Health Data - SMART
allows users to retrieve data about the health
status of the drive. Important parameters
related to wear-out or erase cycles, bit error
occurrences, read counts etc explain the cur-
rent health situation on the drive. It gives clear
indications whether immediate actions are
required or not. This allows the user to run
Figure 1. Hyperstone U9 - USB 3.1 Block Diagram
Table 1. Comparing USB 3.1 to USB 2.0
(Source: Universal Serial Bus 3.1 Specification, Rev. 1.0, July 26. 2013, chapter 3.1.4)
Figure 2. Hyperstone hyReliability firmware features API and Security.
