General > General Technical Chat
EEPROM losing memory, heat?
woodchips:
My products use the 93C46 EEPROM to store calibration information. I have used many different manufactures but all were 5V versions after problems with the 3V version. These products were manufactured in the late 1990's into the 2000's.
I have just had the third enquiry since about Christmas time that the calibration information has corrupted, easily fixed in each case by getting the customer to reload the calibration information from the backup copy stored in the same EEPROM.
In the last 20 odd years this has scarcely ever happened before.
The question is why? My thoughts as the moment are that the roasting hot summer, north and south of the equator, have caused this. Is this reasonable? The data in the 93C46 is stored by a voltage on a floating gate and high temperatures could have drained it. What about all the rest of the calibration data stored in the chip, is that slowly draining away as well so whilst it still works, it could be in error?
The microcontrollers used are several from the Microchip PIC16C7x family, and there has been no failure there that has been reported. With the code going wrong then the device would simply stop working, not much error protection in those chips!
Data retention is guaranteed for 10 years from memory, but EPROMs are still good after 50+ years.
Any comments or ideas please.
m k:
Are backups always ok?
Are corrupted addresses known?
Jeroen3:
Ambient temperatures of 40 are not supposed to severely reduce the retention. But maybe your devices are getting very hot?
I suspect that the most likely culprit here is jittering of the spi lines during chip reset that can cause an accidental write when the stars align.
You will probably never have this happen on your desk, but once you've sold enough...
Some workaround could be to never use address 0x00, depending on the reset pin states.
Do you have sufficient pull-down/up resistors on the chip selects?
Or you know, find the jitteriest (is that a word) power switch and turn it off and on for a while to see if you can reproduce it?
RoGeorge:
- I would suspect first a software bug. Maybe there is a bug that keeps writing more often than it should, and it wears out the EEPROM page where that calibration values reside.
- Second, I would check if the brown out interrupt is enabled and properly handles a power outage during an EEPROM write.
- Could be as well a faulty batch of EEPROM chips
james_s:
You said data retention is guaranteed for 10 years and these products are over 20 years old, it's not too surprising this is happening. Elevated temperature certainly doesn't help, is it possible these devices are being left in direct sunlight, in a hot car, or a garage loft or something where the temperature could be MUCH higher than ambient?
Navigation
[0] Message Index
[#] Next page
Go to full version