CH552 flash writes maximum number far greater than 200

Tests like this are not very expressive, 200 write cycles are the specified amount of write cycles, where the manufacturer specified data retention time should be valid if programmed according the specification.

Additionally a single fail during data verification doesn‘t mean the memory cell is defective, it is also possible that something went wrong during programming.

Your test ignores that flash writing is a 16 bit operation and therefore allways 16 bit are written on even addresses. So its more likely the flash may fail earlier especially when writing 0xFFFF and 0x0000.

Note that flashing a app with wchisptool allways eats 2 cycles because for security reasons some cells have to be erased to enable the flash programming. This erase is part fo the bootloader and more serious since v2.31

https://www.wch.c n/bbs/thread-83327-1.html

I'm using flash memory as a large non volatile storage space for variables.

The main CH552 application software will only get programmed a few times but when using the application some variables need to be stored when the user changes them. There are about 32 bytes that need storing.

Since the internal data flash is limited to 10,000 writes I'm using the normal flash storage area instead. Currently limiting the writes to 256 per location then moving the start location of the non-volatile flash storage.

Bert V. wrote:

Basically I found that you can write to the same location more than 2000 times before an error occurred.

Did you do a power-off-power-on cycle after each write?

Bert V. wrote:

Currently limiting the writes to 256 per location then moving the start location of the non-volatile flash storage.

There are about 32 bytes that need storing.

Sounds like you need some simple wear-leveling on the flash cells.

Thomas Z. wrote:

So its more likely the flash may fail earlier especially when writing 0xFFFF and 0x0000.

A "flash erase" is "change all 0 bits to 1". An empty flash has all bits set to 1. So the flash will fail much faster, when writing 0 to it, because then the silicon insulation barrier will be stressed in both ways always: first by deleting to 1, then by changing the bits to 0. There is no stress to the falsh when writing 0xffff to an empty cell.

Lothar M. wrote:

A "flash erase" is "change all 0 bits to 1" ...

I know that. CH552 flash is not the usual flash. The cells dont need any erase cycle. You can simply write 0xFFFF (or any other value) even when all bits are clear. Other controller from WCH have 64 bytes pages then empty is all zero programmed all óne.