SPRUIV7B May 2022 – September 2023 AM620-Q1 , AM623 , AM625 , AM625-Q1 , AM625SIP
All references to ECC in this subsection refer to the 1-bit error correction Hamming code.
The ECC is based on a 2D (row and column) bit parity accumulation known as the Hamming code. The parity accumulation is done for a programmed number of bytes or 16-bit word read from the memory device or written to the memory device in stream mode.
There is no automatic error detection or correction, and the software NAND driver must read the multiple ECC calculation results, compare them to the expected code value, and take the appropriate corrective actions according to the error handling strategy (ECC storage in spare byte, error correction on read, block invalidation).
The ECC engine includes a single accumulation context. It can be allocated to a single designated chip-select at a time, and parallel computations on different chip-selects are not possible. Because it is allocated to a single chip-select, the ECC computation is not affected by interleaved GPMC accesses to other chip-selects and devices. The ECC accumulation is sequentially processed in the order of data read from or written to the memory on the designated chip-select. The ECC engine does not differentiate read accesses from write accesses and does not differentiate data from command or status information. Software must ensure that only relevant data are passed to the NAND flash memory while the ECC computation engine is active.
The starting NAND page location must be programmed first, followed by an ECC accumulation context reset with an ECC enabling, if required. The NAND device accesses discussed in the following sections must be limited to data read or write until the specified number of ECC calculations is complete.