Gene Heskett wrote: > On Tuesday 21 May 2013 21:53:39 Jon Elson did opine: > > >> Gene Heskett wrote: >> >>> Humm, is there an IC that will function as the switch driver, and >>> whose output could be controlled with a PWM mode? Something that >>> would start at 15% duty and taper to 100% when the applied voltage >>> had dropped to say 20 volts. Then the Q is how fast can these be >>> switched, I get the impression I'm still looking at hexfets I can >>> steal out of dead computer PSU's. These things speed (lack thereof) >>> will cause their own SOA self destruction I expect. >>> >> I use the IR2113 and similar chips to drive FETs. You'd need some op >> amps and/or comparators to generate the PWM. Lack of speed? The >> IRFB31N20D was so fast it caused problems with the dV/dT (and dI/dT) >> affecting the rest of the >> circuitry. Easily 100 ns turn-on/turn-off times, even with resistors in >> the gate circuit. >> > > Those are problems you should like Jon, they are telling you that your > ground & power planes need to be wider, and need more bypassing. Both of > those translate to reduced switching losses & cooler devices. Often you > will need both a .05 ceramic on very short leads or better yet, surface > mount, AND a 10uf in order to get enough bypassing. > I went to a slightly larger transistor (moved up from the fast IRFB31N20D to the slightly slower IRFB260N) and the problems disappeared. The servo amp runs cool at rated current, so losses are not a problem. I've got about as wide traces as the board can stand. The boards is nearly all surface mount, the IR drivers are right next to the transistors and the loop area of the power section is very small. But, the output conductors run right under the logic on the bottom of the board. I've sold 275 of these, now, so they are a pretty mature design. > Of course I'm far enough away its not worth getting in the car to come and > beat some manners into me. :) I think. > > FWIW, resistors in the gate to slow the rise & fall times will raise the > switching losses, in some cases enough to destroy the device long before > your finger says its even getting warm. > The big problem with this is the IR driver chips have a parasitic diode in them that can't tolerate excessive negative voltage at the junction between the high and low side transistors. That was what I had to tame, that junction going below ground. The body diode in the low-side transistor takes several MICROseconds to turn on with 12 V forward bias! I put ultra-fast power diodes across the low-side transistor to stop it from going negative, but still wanted to slow down the fall-time to get a bit more safety margin.
Jon ------------------------------------------------------------------------------ Try New Relic Now & We'll Send You this Cool Shirt New Relic is the only SaaS-based application performance monitoring service that delivers powerful full stack analytics. Optimize and monitor your browser, app, & servers with just a few lines of code. Try New Relic and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
