2020/09/29 14:01:15
f.musso
Hello,

I've used a 24LC16B memory on an industrial digital regulator. The purpose of the memory is to store the configuration of the system, so is basically written on few hundred times upon user commissioning and read out at every system startup. data integrity is checked with CRC.
After 6 years of continuous service it happened that 4 regulator at same time (few days between each other) the data in memory become corrupted and after a rewrite on memory become corrupted again after few hours if the power supply is present or few minutes if power supply is not present.

The chip is supplied by 3.3V nominal (3.5V directly measured on terminals) by a switching regulator (LM2576S-ADJ) but no other component is damaged.

What can be happened? where I can find information about memory failure not related to write cycles?

Any suggestion is appreciated



Thanks in advance





Federico
2020/09/29 15:48:01
ric
How fast does the power supply to the chip rise, and how soon after it has risen do you read from the chip?
I had problems years ago, where reading an eeprom chip too soon after power on (i.e. will Vdd was still rising) was interpreted as a write rather than a read. Just adding a few milliseconds delay before the read was all it took to fix it.
 
 
2020/09/29 22:39:29
jtemples
Your CPU is browning out and doing random writes to the EEPROM.
2020/10/02 11:49:38
f.musso
thanks for the reply
 
The cpu is hold reset by an external by a TPS3307-33 wich monitor all power supply in the board, the 3.3V is  monitored up to 2.93V. If the voltage drop below this threshold the cpu is reset.
 
The first access (read) of the memory is hundreds of milliseconds after the release of reset.
 
It must be something else
2020/10/02 14:02:22
NKurzman
Are you sure you are not writing to the Chip more than you think and wearing out the cells?
your code has an issue and is not doing what it is supposed to do.
You Power supply is not Stable.
Your decoupling caps are not placed correctly.
H/W or S/W interface to the Chip?  if S/W check the Timing.
 
2020/10/02 15:46:43
f.musso
NKurzman
Are you sure you are not writing to the Chip more than you think and wearing out the cells?

Yes, only few times on user interaction.
NKurzman
your code has an issue and is not doing what it is supposed to do.

Checked, no strange behavior
NKurzman
You Power supply is not Stable.

Power supply is stable
 
NKurzman
Your decoupling caps are not placed correctly.

There is one single 100nF cap near the chip, maybe its not enough? 
 
 
NKurzman
H/W or S/W interface to the Chip?  if S/W check the Timing.
 

HW interface,  correctly set
 
2020/10/02 20:35:48
GeeWhizz
You could start by doing an extensive analysis of the power rails.
 
However, I'd start by replacing all electrolytic capacitors.  They typically only have a rated life of a few thousand hours.  Particularly, look at all low ESR electros in any switch mode power supplies.  Replace them with the best quality ones you can find.
2020/10/05 09:33:57
f.musso
Thanks, I will start to investigate on that
© 2021 APG vNext Commercial Version 4.5

Use My Existing Forum Account