| ??? 08/16/07 10:31 Modified: 08/16/07 10:36 Read: times |
#143329 - Pitfalls Responding to: ???'s previous message |
* simple addition scheme without carry
* XORing all bytes together Both of these have one major downside (which can be easily removed, though): What happens if your data and the checksum gets set too all zeros ? Both of the algorithms will indicate a correct checksum. The XOR algorithm will even indicate a correct checksum if the data gets set to all 0xFFs and the number of bytes is odd. The addition is the better choice of the two, because one wrong bit in the data will most likely lead to more than one wrong bit in the checksum - unlike the XOR algorithm where one wrong data bit will lead to one wrong checksum bit. To avoid the all-zeros scenario, define the checksum as - (lsb of sum of data bytes), NOT (lsb of sum of data bytes) or something similar. |
| Topic | Author | Date |
| Simple EEPROM checksumming | 01/01/70 00:00 | |
| Pitfalls | 01/01/70 00:00 | |
| a checksum out of 2 bytes? | 01/01/70 00:00 | |
| It's an "either - or" ... | 01/01/70 00:00 | |
| code sketch | 01/01/70 00:00 | |
| a simple checksum | 01/01/70 00:00 | |
| CRC8 instead of CR32 | 01/01/70 00:00 | |
| I am talking about CRC-16 not CRC-8 | 01/01/70 00:00 | |
| oups | 01/01/70 00:00 | |
| how much data? | 01/01/70 00:00 | |
| approx 40-60 bytes | 01/01/70 00:00 | |
| I use 2 copies and a counter for that | 01/01/70 00:00 | |
| adding to Jan's post | 01/01/70 00:00 | |
| Here's one method | 01/01/70 00:00 | |
| another interesting scheme! | 01/01/70 00:00 | |
| Try the Dallas 1-wire CRC | 01/01/70 00:00 | |
| another good idea but... | 01/01/70 00:00 | |
| I just meant the CRC | 01/01/70 00:00 | |
oups | 01/01/70 00:00 |