| ??? 04/12/00 15:29 Read: times |
#2195 - RE: I2C EEProm Discussion |
I decided to include a validation mark at the endof every record that I write to the serial EEProm.
The reason isn't data integrity but rather the vulnerability to an incomplete write. Without going into the details here, a reset, watchdog timer, or power glitch can result in an incomplete record written, surprisingly even when using a single page-write command. The latter is somewhat surprising as a quick consideration would lead on to think that due to its write-all-at-once nature, page writes would be the safest. But the write buffer may be partially loaded when a non-power outage occurs. Many software initializations recommend a STOP condition be sent before a START, to guarentee you have the EEProm's attention. In the example above, the initializing STOP would signal the chip to write its incomplete page buffer and would result in a write cycle time delay before acknowledging any of the rest of the initialization block. So the flag in need is merely the last byte written to indenify that it belongs with the record instead of being an old byte under a partially written record. A checksum or CRC should do this. -Jay C. Box |
| Topic | Author | Date |
| I2C EEProm Discussion | 01/01/70 00:00 | |
| RE: I2C EEProm Discussion | 01/01/70 00:00 | |
| RE: I2C EEProm Discussion | 01/01/70 00:00 | |
RE: I2C EEProm Discussion | 01/01/70 00:00 |



