minimising heat in buck SMPS

[technogeek]

1.4 A off-line supply....

How do I select frequency, inductor size, and mosfet with minimising heat (or I guess maximising efficiency?) in mind?

If I run at 50khz, the current mosfet I have (I forget the part #) doesn't get warm at all. The diode and 1mH inductor hit 180F. The 1mH inductor is this part number: 2224-H-RC from JW Miller(bourns). The Rdc of the inductor is ~0.3 ohms, *1.4A^2 = 0.6 watts........ which doesn't make sense to me, because it seems like it is dissipating 2 or 3 watts.

Turn the frequency up to 200khz, mosfet gets hot (150F on the aluminum case). Diode stays about the same, the inductor got a little cooler, IIRC.

So, it seems like the ideal solution would be to minimise Rdc of the inductor and run at a low frequency. Problem is, 0.3ohm is the smallest Rdc of the 1mh inductor series. (without having to order 100+ of some less used line)

And when you run low frequency, you need a larger inductor. So it seems like either I'm going to have heat spewing out of the inductor, or the mosfet. Is that what Im getting here?

 

[dch222]

You don't say what your voltage conversion factor is which would help in any analysis.

In general a small component dissipating half a watt can easily get that hot - depends on physical size.

The MOSFET probably gets hot at higher frequencies because the rise and fall times during switching represent a greater percentage of the total switching period. During the the times the MOSFET is between on and off it will be dissipating power (V x I).

 

[technogeek]

The inductor is a toroid about the size of a quarter, maybe 3/4" thick.

Vout/Vin=38% (voltage conversion factor?)

The mosfet has an Rds of 0.5 ohms, rise time of 89ns, fall time 81ns "typical". (~200max)

So at 200khz, the period is 5000ns. Then the switching losses would be (170ns/5000ns)*((Vin-Vout)*Iout)? If so, then I've got 5 watts, which would explain a lot. Are ~80ns times typical?

 

[technogeek]

Well..... I think the solution is to increase the frequency to 200khz.

That means I can use a 600uH inductor with an Rdc of 0.15 ohms... basically 1/2 what it is now.

I found a mosfet with an 18ns rise, 15ns fall time, 0.48 ohm Rds. If the above formula is right, that's ~1W DC, ~1W switch losses, total ~2W. That's managable.

I still think there might be some kind of frequency related heating (or something) going on in the inductor. I can't believe 1/2 watt can heat a huge piece of metal up that much. Thoughts?

 

[Hero999]

Please post a schematic or if you're using a particular IC mention it otherwise it's pretty hard to help.

Have you simulated it useing SPICE software?

 

[Oznog]

Well you've struck upon the basic design issues of a buck converter.

200KHz is getting up there for frequency. There are 2 losses here. One is the rds-on steady-state I^2*R losses, the other is the switching losses. All other things remaining the same, switching losses more or less increase linear with freq. 4x the freq, 4x the losses and the MOSFET has to dissipate 4x the heat.

In the end switching losses usually do present a serious problem on how high you can run the freq, and thus present a limt on the smallest inductance that can do the job. Larger inductance in the same size package means a higher DC resistance so that also presents a limit on how physically small the inductor can be.

The first answer is to review your MOSFET driver design. Some can switch faster than other. "Ringing" on the gate voltage will increase switching losses substantially. There is often room for improvement which can dramatically reduce switching losses thus increase the optimum freq of the device. However, keep in mind that there are still serious limits to how much you can get the switching losses down, thus setting limits on the maxium practical freq and smallest inductor size, even with the best of designs.

 

[OutToLunch]

what's with all the secrecy? you need to give up some more info before you can get any meaningful help with your particular design. Output current is good, but vin/vout doesn't give much but the duty cycle. The very basic info would include a schematic, the input voltage, the target output voltage, the pwm controller ic you are using, the output capacitors and their esr, the gate driver you are using (if the pwm ic does not have gate drivers), the mosfet part number.

 

[technogeek]

Just to update this with some formulae I found after browsing 100s of technical pages about calculating mosfet power dissipation:

The DC Power dissipation of a mosfet = Iload^2*Rds*duty cycle
Where Iload = your output current, Rds is your drain-source resistance (see data sheet) and duty cycle is your Vout/Vin ratio.

The Switching Power dissipation of a mosfet = 0.5*(Iload*Vds*(Tf+Tr)*Fs)
Where Vds ~ your input voltage, Tf and Tr are the rise/fall times (see mosfet data sheet), and Fs is your switching frequency.

Add all that up, and my original transistor was putting out over 6W, which explains why it was heating up the metal case so well @200khz.

The new transistor has a quicker rise/fall time, so the calculated dissipation is 1.2W @200khz......... Much better. With the smaller inductor, that should minimise the heat dissipation by a lot.

Thanks for your help guys, especially dch222 for guiding the way to help me find those switching losses. I guess it turns out I just picked a bad mosfet. (Don't just look at the on resistance!!!!!)