Gary,

 You say these are in electric vehicles.  Is it possible that
 they are in a place and position such that high currents that
 operate these vehicles are causing very strong electric
 fields, and thereby, erasing the EEPROM?  If you read the
 EEPROM, is the data all zeros by chance?  Or is it a mixture
 of ones and zeros?  If it's all zeros, you may want to try
 either moving the PIC or placing it inside some sort of Faraday
 shield to potect it from the electric fields that are no doubt
 present.   Just a thought.

                                           Regards,

                                              Jim



>> -----Original Message-----
>> From: Jeff Hunsinger [SMTP:jeff@SILICONENGINES-LTD.COM]
>> Sent: Wednesday, October 13, 1999 4:10 PM
>> Subject:      Re: PIC12CE518 EEPROM
>>
>>
>        [Gary Schroeder] Jeff, Thank-you for your response
>> A couple things that may be part of your EEPROM corruption problem:
>>
>> - Do you provide protection against load dumps (24-50V+) and other
>>   transients?
>>
>        [Gary Schroeder]  I use a very simple voltage regulator that
>includes a transient
>        voltage suppressor
>
>> - Does your circuit run properly over the normal operating range of
>>   the battery (generally 9-16V).
>>
>        [Gary Schroeder] Yes.
>
>>  Do you have any built-in detection
>>   of under/over battery voltage? Does the code enter a safe mode
>>   when this condition arises?
>>
>        [Gary Schroeder]  I am using brown out protection circuitry.
>When the alarm
>        1st turns on, it reads a value from its EEPROM and uses this
>value to control the
>        pitch of the alarm. I use no code that reacts to under/over
>battery voltage. When the EEPROM
>        is erased the alarm has an extremely low pitch. I can not
>reproduce this failure with my power
>        supply in the lab. It only happens on certain electric vehicles.
>If I take a failed alarm to the lab
>        and power it up, it continues to fail. I am assuming the code is
>behaving O.K., it's just that some
>        type of noise erased the EEPROM value.
>
>        Also, I have added code so that if the EEPROM read fails, a
>nominal pitch controlling value is used
>        in place of the EEPROM value so that the alarm will sound at a
>decent pitch.
>
>
>> - Are the lines going to your serial EEPROM short and guard banded?
>>
>>
>        [Gary Schroeder]   the PIC12CE518 has an on board EEPROM.
>        What is guard banded?
>
>
>> - Do you keep checksums for both the EEPROM data and the data that
>>   is to be written?
>>
>        [Gary Schroeder]  I did not know that you could have checksums
>for EEPROM data.
>        How do you obtain these checksums and how do you use them?
>
>> I've had problems with EEPROM corruption before. It can generally
>> be traced back to either noise on the serial lines or the MCU
>> getting "lost", especially during normal EEPROM access.
>>
>        [Gary Schroeder] Again, I think the code is acting O.K., it's
>that some kind of noise erased the
>        EEPROM. I suspect the solution will be better protection for the
>PIC against spikes. I guess the problem is getting to the EEPROM through
>the Vdd pin. Or could other pins be allowing some signal to get to the
>EEPROM?
>> Jeff
>