EEprom and data lost

Fabri · Joined: 22 Aug 2005 Posts: 275

Hi,
I need a suggestion regarding the use of the internal EEPROM of a PIC18F4525. I use it to save the calibration data for probes J and for system data. The cases are really remote, but I would like to increase reliability ad understand when happen.
I already use write routines in the EEPROM with access via an enable bit. At the end of the writing I do a reading on a fixed location so that the last operation is the reading of a location not important. The system performs an update every 10 seconds a few bytes of EEPROM state. After each A/D conversion I read the calibration data in memory to display the temperature. I wonder if a reset occurs during a write system what makes the PIC? can be a dangerous situation for my data?
Note that I have performed various reset conditions by injecting noise burst without getting a data corruption.

Have you anything to suggest?

Regards,
Fabrizio

Geps · Joined: 05 Jul 2010 Posts: 129

In my current application I write the data twice and then use a 16 bit CRC. I can then recover the data if a single piece of data is corrupted.

They're spread across the EEPROM for further protection.

Fabri · Joined: 22 Aug 2005 Posts: 275

Really interesting but every time you change one datum you need to calculate CRC of all data.
I cannot think to write all data two times but just calibration data is possible.
I have a database by default in ROM. In case of error in the calculated CRC I can restore the default data.
Any other suggestion ?

Geps · Joined: 05 Jul 2010 Posts: 129

No, I don't have to store alot of data so I use a CRC for each piece of data.

PCM programmer · Joined: 06 Sep 2003 Posts: 21708

Fabri · Joined: 22 Aug 2005 Posts: 275

Yes PCM programmer.

But I write only if data changed and not all data every 10 seconds. So the write operation of single eeprom location, in the life of product, is under eeprom life. This data are only status of electronics and not really important because I restore it at start-up.

In any case I take care of your consideration and think about. With your experience, for improve my knowledge, can you suggest me other better ways ?

Regards,
Fabri

Ttelmah · Joined: 11 Mar 2010 Posts: 19504

OK.
The classic way of dealing with this possibility is to hold two sets of data.
Each set has as well as the CRC, a counter at the end.
You write a new block of data, and only once you have written it, increment the counter.
Each time you write, you work on the opposite block to the last one used.
On 'wake up', you read the counters, and take the one with the higher counter, as the 'last block'.
This way if a write fails, and the counter has not updated, you get the last value. If the counter has updated, then you know the data has already been verified as written.

Best Wishes