Effect of erasing an already-erased Flash page - Wear?

[Andrew Neil]

Does erasing an already-erased Flash page contribute to wear?

Reason for asking: the EEPROM emulation code erases pages which should already be erased without checking that they are already erased - is that just laziness, or is it because there is actually no/negligible wear from erasing an already-erased page?

 

EDIT:

Looks like it probably does:

https://community.st.com/t5/stm32-mcus-products/erasing-flash-what-does-really-affect-the-wear/td-p/623282

 

[Tesla DeLorean]

Tends to depend on the exact implementation, some of the more self-timed one aren't discharging/charging if the cell isn't in need of that occurring. Micron QSPI NOR for example complete a lot faster if the blocks/part are already blank.

ST doesn't allow for multiple knock-down writes, where you zero out the cells before erasing, so the charge levels are equalized, a technique used to reduce "memory" / ghosting, where prior content persists at some much reduced level. 

In ST's designs a lot of the wear/stress seems to be with the charge-pump and the life there with the stresses from higher voltages.

@STOne-32 any white papers on the various flash arrays, geometries/processes?

 

Probably laziness or simplicity. Now, with much of my update code I tend to explicitly blank check before doing an erase, but this is more to improve the perceived speed of operation, a read/check is very fast, and will terminate early if content seen, whereas erasing usually takes an age, especially with larger blocks/sectors.

The ECC methods may imped the ability to check erased/virgin state on flash lines.

[Andrew Neil]

---

@Tesla DeLorean wrote:

a read/check is very fast, and will terminate early if content seen, whereas erasing usually takes an age

---

Indeed. Especially in the case of the EEPROM emulation, where the whole thing relies on content being written from low-to-high addresses - so any content should be seen early in the search.

 

And now another question occurs: I guess a blank-check would consume (a lot?) less power than an erase?