Do you trust the internal EEPROM ?

[VanHauser]

I want to store some configuration parameters (a few dozens of bytes) in the internal EEPROM of the 18F452. Is it worth making a checksum for the entire configuration block in order to detect storage errors or am I really paranoid? If I do so, every time I update a variable in the configuration I have to remake the checksum and rewrite it as well. Is this a waste of CPU time? Would it be enough if I store each variable without any error checking?

[ckielstra]

It depends on how critical your application and the parameters are. When lives are depending on them you for sure want to have some validation.

I think the EEPROM in the PIC is quiet a save way for storing data, but always there is the small chance of a power failure when writing data to the EEPROM.

A simple CRC-8 like the one used in the 1-wire protocol would suffice for most applications to detect an error.

What are you going to do in case a configuration error is detected? You could fall back onto a default configuration or think of some clever way of restoring the configuration (error correcting codes or storing the configuration multiple times).

Check the maximum number of writes you can do to the EEPROM of your PIC, especially the CRC-code will be the byte most frequently updates.

[VanHauser]

I could do writes with read-back for verification. If the original data matches what it is read back, the the write was successful. Otherwise, the system would say Contact Service and halt, because the EEPROM is faulty. But I cannot deal with power failures this way.

If I think about it, yes, lives and expensive equipment are depending on those parameters!

[ckielstra]

[newguy]

Is there any way you can incorporate an external eeprom? If so, an SPI ramtron chip would be ideal. SPI would give you fast access times and in cooperation with the pic's brownout detect, power failures shouldn't ever become an issue. Also with the ramtron chips, fatigue/wearout is not going to be an issue either.

[VanHauser]

The configuration data is not changed often. It is entered by hand or remotely and stays like that for a while, hours, days or even longer, but there is a lot of overhead when saving the configuration and EEPROM writing is really slow.

The datasheet for 18F452 says the endurance is 1,000,000 cycles.

It's kind of difficult to find FRAM here in Romania. The stocks for small serial FRAMs are empty and myy PIC's pins are almost all taken. It is an idea that's worth remembering in the future though, thanks for the tip.

[rnielsen]

One possible solution might be to write the data to multiple locations within the eeprom and then compare your main data with one or more backup sections. This might detect if your main section has become corrupt some how.

Ronald

[VanHauser]

Like they use voting for onboard computers on space ships? I should keep an odd number of the same record in order to detect and eliminate the error. That would be a good idea, if I could squeeze at least 3 configuration sets in the 256-byte EEPROM. I'll try that.

Realistically speaking, supposing no power failure occurs when writing the EEPROM and the write is successful, what are the odds for the data to get altered by itself? Pretty small I think.

[treitmey]

Just a note. We get the FRAM directly from Ramtron. Not from
Digi-key,Mouser,ect...

And they have samples
http://www.ramtron.com/doc/Products/productsOptInEmail.asp