Real IAP Wear Out Data

[lpcware]

Content originally posted in LPCWare by Ex-Zero on Sat Mar 23 15:32:48 MST 2013
Using IAP to store data in flash has been an issue in several posts already. But how often can I erase a single sector :confused:

Datasheet shows:

---

Quote:

11.1 Flash memory
Number of program/erase cycles.
Endurance  Min:10000 Typ:100000 cycles

---

A simple IAP Erase - Blank Check - Write - Read cycle is showing real data :)
[INDENT] LPC1768-Sector 27: Blank Error after 2481520 cycles

LPC11C14-Sector 7: Blank Error after 1437980 cycles
[/INDENT]So obviously erasing is the critical part :)

After the first Blank Check Error it's possible to erase the sector with repeating the Erase command, but now Blank Check Errors are increasing violently :eek:

[lpcware]

Content originally posted in LPCWare by frame on Mon Mar 25 05:22:43 MST 2013

---

Quote:
And any results of any tests in your 70°C company to help us estimate a realistic value :confused::confused:

---

Not really, because this devices use 8 bit MCU's (and the worst kind available on the market, IMHO.)
But they have Flash/EEPROM and are relatively cheap ...

Second, this is a general test, that includes the whole device functionality.  and not a specific Flash wear-out test.
But just by coincidence, this kind of test revealed a silicon problem related to Flash/EEPROM of this 8-bit vendor.
After some discussions, the vendor fixed it, provided us some worst-case test results (we never saw before),
and updated the datasheet. Seems even such large companies not always do their homework ...

[lpcware]

Content originally posted in LPCWare by Ex-Zero on Mon Mar 25 05:12:21 MST 2013

---

Quote: frame
That means, for instance, tests at 70°C in my company.

---

And any results of any tests in your 70°C company to help us estimate a realistic value :confused::confused:

[lpcware]

Content originally posted in LPCWare by frame on Mon Mar 25 04:58:31 MST 2013

---

Quote: Zero
:confused:
This is a real IAP Wear out test. So temperature and voltage are stable. And, as usual, the critical part of the process is erasing (See: Floating gate) ;)

---

I'd argue that the voltage thresholds for erase_ok and erase_fail are much tighter than the readout thresholds (hig/low/invalid).
So failing cells will usually be noticed during erase.

---

Quote:
Usually I'm not working in worst case environment, so I doing my own tests

---

For a realistic test of a device, such a test needs to be made in the worst expected environment.
And in devices that heat up considerably, such test may avoid some rude surprises.
That means, for instance, tests at 70°C in my company.

---

Quote:
  For commercial projects, I would avoid EEPROMs at all and use faster FRAMS :eek:

---

Yes, good point. But only if the projected target costs allow for that.
We prefer MCU's with internal EEPROM, which is far cheaper than FRAM.

BTW:
I have lots of smileys left over - can I donate them to you ?[IMG]http://knowledgebase.nxp.com/images/icons/icon7.gif[/IMG]

[lpcware]

Content originally posted in LPCWare by Ex-Zero on Mon Mar 25 03:49:23 MST 2013

---

Quote: frame
Not actually.

---

:confused:

This is a real IAP Wear out test. So temperature and voltage are stable. And, as usual, the critical part of the process is erasing (See: Floating gate) ;)

---

Quote: frame
The actual number of cycles depend basically on temperature and voltage. A vendor usually  gives a worst-case number, which is guarantied over the whole temperature range. But generally, the hotter, and the lower the voltage, the faster it wears out.

---

Usually I'm not working in worst case environment, so I doing my own tests :)

---

Quote: frame
For commercial projects, I would avoid Flash for storing often-canging data, and use EEPROM instead. Beside of being byte-accessible, it usually specifies about 10 times the erase/write cycles.

---

For commercial projects, I would avoid EEPROMs at all and use faster FRAMS :eek:

[lpcware]

Content originally posted in LPCWare by frame on Mon Mar 25 03:30:21 MST 2013

---

Quote:
So obviously erasing is the critical part :)

---

Not actually.
But erasing involves a more stringent check of the voltage level, so it usually fails here.

The actual number of cycles depend basically on temperature and voltage.
A vendor usually  gives a worst-case number, which is guarantied over the whole temperature range.
But generally, the hotter, and the lower the voltage, the faster it wears out.

For commercial projects, I would avoid Flash for storing often-canging data, and use EEPROM instead.
Beside of being byte-accessible, it usually specifies about 10 times the erase/write cycles.

[lpcware]

Content originally posted in LPCWare by Ex-Zero on Sat Mar 23 18:02:47 MST 2013
Here it is: [ATTACH]990[/ATTACH]

[lpcware]

Content originally posted in LPCWare by micrio on Sat Mar 23 16:53:06 MST 2013
Could you share the project?
I have a bag of otherwise useless boards containing an LPC1114
that I would like to run your test on.

Pete.