| ??? 02/14/07 05:30 Modified: 02/14/07 05:31 Read: times |
#132870 - The risetime isn\'t all that fast. Responding to: ???'s previous message |
Kai Klaas said:
Richard said:
The Vcc is distributed through a solid layer of nothing but Vcc and GND is on a solid layer as well. The bypass is a 1 nF on EVERY Vcc and EVERY GND connection on every IC. Me thinks that your supply voltage routing is a bit too fast and by this prone to resonances? SMPS can generate unsane narrow voltage spikes, which when not being sufficiently filtered out can then cause lots of trouble (ringing!). I would increase the 1nF caps, add some ferrites or low ohmic resistors in the supply lines and use a post filter at the SMPS. By the way, I never use a switcher to directly power the digital section, but always a linear 5V regulator (LM7805 type). Switchers I only use for pre-regulating, if needed. Yes, that's what I did, too, but probably not for the same reasons. I have a hefty 12-volt supply feeding the really small circuit that's fed from a 5-volt linear regulator (LM340-5). I don't see any spikes or anything annoying. I do see about 40-50 mV of power-gnd noise when the thing is running normally, i.e. from the linear 12-volt regulated supply feeding the linear 5-volt regulator. I don't see much more than that with the switcher either, but I suspect there's some ripple feeding through while the LM340 is starting to work. What troubles me is that the wall-wart, still a pretty hefty one, causes the same sort of trouble as the switcher. I think, it's better to have a clean, spikes-free but smooth power-up, than a fast but ringing one. In any case, I know, that the DS1232 has lots of problems (even total lock-up!) with fast spikes on Vcc. The MAX1232 can handle them much better. But the best is to avoid these fast spikes and too steep edges at power-up.
Kai I think what makes the difference is the lack of ripple on the 5-volt locally regulated supply when I'm using the linear 12-volt regulated supply to feed the LM340. I don't have any switchers that don't produce what I'd consider an intolerable amount of (noise+ripple) under load. The noise is worse with no load. I've had a few switchers catch fire when I have a lot of capacitance on their output, so I don't do that any longer, but that would probably filter out some of the ripple. I wish they'd build some really fast switchers some day. The ones you see most of the time operate between 10 and 100 KHz. I'd like to see somewhere between 5 and 50 MHz. Unfortunately, that would cause other problems ... <sigh> ... RE |
