| ??? 03/26/07 17:46 Read: times |
#135921 - signal integrity Responding to: ???'s previous message |
The first rule is:
Clock frequency is not relevant. Rise time is the important parameter. It really doesn't matter if your signal toggles once every 10 ns or once every week. If the rise time of the signal is "too fast" and if the trace is "too long," then the signal will ring. If the signal settles by the time you need to look at it, fine. If not, you need to take care of things, perhaps using terminations, perhaps by minimizing trace lengths, perhaps by changing the driver family to something slower, perhaps by all of the above. I recommend Johnson and Martin's High Speed Digital Design: A Handbook of Black Magic. Quite the eye-opener. (That's a joke for those of you doing communications things.) -a |
| Topic | Author | Date |
| Track lengths and widths (8051) | 01/01/70 00:00 | |
| Controlled impedance | 01/01/70 00:00 | |
| It wont work out | 01/01/70 00:00 | |
| See the edited version of my original reply. | 01/01/70 00:00 | |
| an approximation is OK. | 01/01/70 00:00 | |
| Hard to really guide but... | 01/01/70 00:00 | |
| thanks for the width | 01/01/70 00:00 | |
| This is why Erik goes ballistic | 01/01/70 00:00 | |
| Clock signals | 01/01/70 00:00 | |
| thank you | 01/01/70 00:00 | |
| fast signals don't care about width | 01/01/70 00:00 | |
| "Clock signals"? | 01/01/70 00:00 | |
| clock | 01/01/70 00:00 | |
| be careful, Mike | 01/01/70 00:00 | |
| 1983 Fourth Edition | 01/01/70 00:00 | |
| Length is more important than width | 01/01/70 00:00 | |
| Try this, Mike | 01/01/70 00:00 | |
| I've tried to explain this. | 01/01/70 00:00 | |
| What Mike wants | 01/01/70 00:00 | |
| I meant "Dave and Kai" | 01/01/70 00:00 | |
| I understand... | 01/01/70 00:00 | |
| Join the club! | 01/01/70 00:00 | |
| Interesting note, Kai... | 01/01/70 00:00 | |
| signal integrity | 01/01/70 00:00 | |
Very cute | 01/01/70 00:00 |