These are typically very large, very deep battery banks. 50AH is=20 wimpy. In my other life, I have a location which has an array running=20 at 24VDC, 480AH. There is a *lot* of current there, admittedly. Fortunately there are also a lot of connectors, small-guage wires,=20 fuses, and the like typically between my gear and the array, all which=20 serve to limit the instantaneous current. In this application FQB22P10's are pretty much jellybean parts. -88A=20 pulsed drain current. Eas of 710Mj, Ear of 12.5Mj. Admittedly it's all a race of 'what will melt first', or maybe better=20 'will we get it all shut down before something exceeds a rating'. =20 Fortunately, this is also a very rare occurance, and so statistcally=20 we're just trying to make the product more bulletproof in those cases -=20 I.E. adding a lot of cost isn't in the cards in this case. I'm actually eyeing the 22P10 I'm using on the front end of this for=20 reverse polarity protection and thinking that with a few truly jellybean=20 parts it could become not only reverse polarity protection, but also a=20 rough current limiter to a few amps. If I can get the impulse energy=20 available to the downstream shunts and FETS down a ways, then things=20 become much less critical, and the 22P10 only needs to dissipate at=20 worse case that initial (admittedly huge) pulse while it adjusts, and=20 then a bit of additional power at the lower amperage for however long it=20 takes the circuitry to decide there's a true overcurrent/short going on=20 here and shut things down. But as is mentioned below the challenge=20 might be really in the realm of not worth the effort. -forrest On 12/17/2011 10:05 AM, Sean Breheny wrote: > How accurately do you need to sense current? > > I am concerned about the ability of your MOSFET switch to stop the > short-circuit current from a large battery bank. Remember, during the > time that the FET is switching off, there is a huge power dissipation > because the voltage is rising at the same time that the current is > falling. I don't know what kind of battery bank you are talking about, > but let's say it's a 24V 50 Amp-Hour lead-acid battery. The short > circuit current could be 5000 Amps. At the worst point, there will be > about half voltage (12V) across the FET and about half current (2500 > Amps) through it. That's 30kW. Let's also say that the FET will switch > off in 500 nanoseconds and that the overall average power dissipation > is half of the worst case point. So, 15kW for 500 nanoseconds - that's > 7.5 milliJoules. Might not seem like much but given that the actual > die is quite tiny, and the bond wires are even smaller, it might well > be enough to destroy them. > > This is ignoring the additional heat which will come from just I^2 R > heating during the time when the current is ramping up before your > system trips and decides to turn the FET off. It is also ignoring the > inductive energy in the leads which will show up as a huge voltage > pulse across the FET, increasing the power dissipation and possibly > exceeding Vds max. Finally, it is also ignoring the difficulty in > keeping a FET turned on fully during the very high current conduction > - the voltage drop across the FET (drain-source) will likely exceed > Vgs max which will prevent you from applying enough gate drive to keep > the FET in the ohmic region. > > What I am saying is that a FET swtich that is designed to catch and > stop the short circuit current from a large battery is NOT a small or > inexpensive thing :) > > Sean > > > On Sat, Dec 17, 2011 at 6:40 AM, Forrest Christian w= rote: >> On 12/17/2011 3:09 AM, Mike Harrison wrote: >>> Schottkys get very leaky at higher temperatures, so probably not a >>> good choice. Is the current really so high that you can't use a more >>> chunky sensing resistor? If you really want to bypass the shunt, a >>> MOSFET is probably a better bet - arranged such that it turns on >>> quickly when the measured current exceeds a threshold, witha >>> monostable to hold it on for a while, limiting the current duty cycle >>> through the resistor to the monostable time / the turn-on time. >> Just to be clear what we're typically talking about here... >> >> The normal load is around 250mA. Inrush currents can be extremely high, >> in the range of amps, but not close to high enough to exceed the >> resistor rating. >> >> The issue comes where our customer is using a large battery bank with >> virtually unlimited current available, and decides to dead short the >> output. I think we've actually got the firmware in the PIC doing the >> overcurrent protection tweaked so failures just aren't likely - It's a >> fine line between being slow enough that the inrush doesn't trip the >> protection, and a short or long-term overcurrent will. With some >> creative integration (in the mathematical sense), things seem to be >> working well - but I still hate not protecting something I know is close >> to the edge. >> >> At this point, I think I might look at using a mosfet on the input of >> the whole device to limit current to something reasonable - dozens of >> amps for instance - probably a more reasonable solution. >> >> And yes, we're already using about as chunky of a sense resistor as is >> practicable. >> >> -forrest >> -- >> http://www.piclist.com PIC/SX FAQ& list archive >> View/change your membership options at >> http://mailman.mit.edu/mailman/listinfo/piclist --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .