SPRUJ28E November 2021 – September 2024 AM68 , AM68A , TDA4AL-Q1 , TDA4VE-Q1 , TDA4VL-Q1
Host 32-bit word read and write access requests to a chip-select associated with a 16-bit-wide NAND device are split into successive read and write 16-bit word accesses to the NAND memory device. 16-bit word access is ordered according to little-endian organization.
Host byte read and write access requests to a 16-bit-wide NAND device are completed as 16-bit accesses on the device itself, because there is no byte-addressing capability on 16-bit-wide NAND devices. This means that the NAND device address pointer is incremented on a 16-bit word basis and not on a byte basis. For a read access, only the requested byte is given back to the host, but the remaining byte is not stored or saved by the GPMC, and the next byte or 16-bit word read access gets the next 16-bit word NAND location. For a write access, the invalid byte part of the 16-bit word is driven to FF, and the next byte or 16-bit word write access programs the next 16-bit word NAND location.
Generally, byte access to a 16-bit-wide NAND device must be avoided, especially when ECC calculation is enabled. 8- or 16-bit ECC-based computations are corrupted by a byte read to a 16-bit-wide NAND device, because the nonrequested byte is considered invalid on a read access (not captured on the external data bus; FF is fed to the ECC engine) and is set to FF on a write access.
Host requests (read/write) issued in the chip-select memory region are translated in successive single or split accesses (read/write) to the attached device. Therefore, incrementing 32-bit burst requests are translated in multiple 32-bit sequential accesses following the access adaptation of the 32-bit to 8- or 16-bit device.