Table of Contents
Advertisement
At the start of a write operation, the ECC syndrome is zeroed.
DICEY
computes the syndrome as each byte is written.
After the last byte of
data, DICEY appends the syndrome to the end of the data on the sector.
Later, when the sector is read, DICEY calculates another syndrome from
the data and the ECC bytes originally written on the sector.
If the
syndrome is zero, no errors have occurred in writing and reading.
A
non-zero syndrome indicates an error, and the user's Error Recovery
Parameters (previously selected with the MODE SELECT command) deter-
mine the action to be taken.
See the Q200 Series Programmers Manual
for full details.
The basic strategy coded in firmware is to continue retrying a read
operation until a stable syndrome is found (i.e., the same syndrome
is obtained when the sector is read twice in a row).
This strategy
avoids invoking correction on soft errors and dramatically reduces the
probability of miscorrection.
When a stable syndrome is found, the
drive attempts to use ECC to correct the data, if ECC correction is
enabled.
If the number of retries reaches the maximum specified by
the user, the drive again attempts to correct the error, even if a
stable syndrome has not been found (again, only if ECC correction is
enabled).
4.2.2
Allocating Replacement Sectors
Traditionally, replacement sectors are put in a reserved area of the
drive far from the inside or outside diameter of the disk.
Thus, when
a defective sector is found, a long seek to the replacement sector is
required.
Q200 Series drives use an alternate approach: two spare sectors are
reserved at the end of every cylinder for the replacement of defective
sectors.
Long seeks to the replacement sectors are eliminated.
When a drive is formatted, either in the factory or later in the
field, an "in-line sparing" method is used: defective sectors are
replaced with adjacent sectors, and all subsequent sectors are shifted
up (see Figure 4-3).
In the rare instance where more than two defects
are found within the same cylinder, the "In-Line Sparing Overflow"
method is used: additional defects are mapped into the nearest adja-
cent cylinder with spare sectors available (see Figure 4-4).
With this approach, cylinder boundaries are well defined, with a sim-
pIe relationship between each cylinder's physical and logical block
addresses--thus providing the host a method to ensure that file
records are stored within cylinder boundaries.
When Automatic Read Reallocation is enabled (with the MODE SELECT
command), field-found or "grown" defects are automatically mapped
directly into spare sectors without shifting subsequent logical
blocks, as shown in Figure 4-5.
4-6
Advertisement
Table of Contents
