| ??? 05/20/10 04:52 Read: times |
#176047 - skew/asymmetry Responding to: ???'s previous message |
Neil Kurzman said:
Assuming the UART syncs to the center of the start bit you get up to half a bit over the total number of bits. ( More bits require more accuracy)
You get less since most UARTs read 3 points then do majority rules. Then you split the error budget over the sender and receiver. [minor addition]usually 3 samples out of 16, i.e. sampling is 16x overclocked and UART votes majority between samples 7, 8 and 9. Some MCU UARTs can be set to 8x only overclocking, though (a.k.a. x2 mode UART) which decreases the budget a bit, too.[/minor addition] I'd also like to stress the "split the error budget between transmitter and receiver", which effectively halves the 4.something% from Kai's calculation. In cases where the baudrate error approaches the margin, e.g. in MCU driven off its internal RC/ring oscillator, the proponent often takes a quiet assumption, that the "other party" is perfectly precise, though. Another oft forgotten factor is asymmetry in the transmission chain, i.e. "1" bits may be longer/shorter than "0" bits ("SPACE" and "MARK" in the long forgotten RS232 terminology). This is due to potentially different rise/fall times and asymmetry in receiver's decision level (and maybe other factors). The percentual error due to this depends on baudrate of course, as it is a fixed number in a given setup. Usually not a big issue until you start to run at high baudrates, 115200 and up, when it starts to be in the range of several tenths of %. JW |
| Topic | Author | Date |
| Baud RateTolerance? | 01/01/70 00:00 | |
| acceptable baud-rate error | 01/01/70 00:00 | |
| 100% ?? | 01/01/70 00:00 | |
| 2.5% | 01/01/70 00:00 | |
| whoa!! | 01/01/70 00:00 | |
| I prefer < 0.5% | 01/01/70 00:00 | |
| the method | 01/01/70 00:00 | |
| skew/asymmetry | 01/01/70 00:00 | |
| Asymmetry | 01/01/70 00:00 | |
| ZERO | 01/01/70 00:00 | |
| So do I... | 01/01/70 00:00 | |
| Not always possible to select crystal just for UART | 01/01/70 00:00 | |
| In reference to the above: | 01/01/70 00:00 | |
| Thank you all. | 01/01/70 00:00 | |
| Try a test with a PC and Sqr Sig Gen | 01/01/70 00:00 | |
| likely a wromg approach | 01/01/70 00:00 | |
| Maxim AN-2141 | 01/01/70 00:00 | |
link and comment | 01/01/70 00:00 |