I Need to Squish my Inductor

I went on

I'm getting a hint that I might have to redesign with higher frequencies for smaller magnetics.

D from BC

What sort of voltage and power levels? The numbers I just ran on the info you posted don't align very well with my sense of reality...

80kHz these days is a pretty low SMPS frequency. People that are making very small POL regulators are running in the MHz range. Have a look at Enpirion, for example, or at the Linear Technology LTM4600. You can get "planar" cores that are intended to snap around a multi- layer circuit board where the transformer or inductor windings are PC traces. These can work well for "thin".

Cheers, Tom

Oh no..I think I goofed on the specs.. I'm still new at smps design.. :P

And bummer,....I'm designing with ancient speeds that are out of style. I'm so glad I don't do this for a living...The stress :(

The inductor is being used in a continuous mode 180Watt supply. The power supply converts 170VDC to 90VDC @2A

I took a look at the LTM4600...that's a shocking example. 850khz operation using internal IC inductors! It can operate up to 1.7Mhz for

5V output. I can't use this IC in my app but it's still a good example of how small the magnetics can get with those frequencies.

And here I am playing with toroids relatively the size of boat anchors.

I like the PCB trace and snap on core idea...I'll look into that..

D from BC

I am assuming you have looked at the various surface mount inductors and found them lacking in inductance at those curents (and size constraints). Seems to me that using a toroid is the best way to go; if you can get flat wire that will help to reduce the height by a fair amouny. Failing flat wire, maybe 4 or more small wires side-by-side would fill the bill; the problem being is to keep a flat profile as it is wound . Perhaps lay them out first on sticky tape and either add epoxy or more tape on top? Slit tape for width?

I wonder why you need that much inductance. The designs I'm familiar with run with a triangle wave of current that comes close to going to zero at the bottoms of the triangle. If you do that, you have a delta of close to 4 amps. For simplicity, let's just say 3 amps. V=L*di/ dt. Even at 80kHz, you have about 6 microseconds at 90 volts, so the required L would be about 90*6/3 uH = 180uH. Obviously if you go to higher frequency, the required inductance drops. Maybe you need to use larger inductance to keep it continuous at lower output currents. Are you using an active half-bridge driver, or just a catch diode?

The Enpirion parts are even smaller--and thinner--than that LTC part, and run at higher frequencies. They aren't in the power class you want to run, though, as you noted with the LTM4600.

Robert's idea of using multiple strands of wire is a good one also from the standpoint of more surface area in the wire. If the number of turns is not too great, it shouldn't be difficult to just wind one winding, spacing the turns out evenly, and add more windings in parallel. But--where to get a thin toroid?? Ferrite isn't easy to machine.

Cheers, Tom

Not recently, but I faintly recall hacking away at this problem some months ago and yes flat and mH inductance at the same time I've yet to see. I think I tried a bunch of button inductors for another design.. That got toasty..Too much heat from Rdc due to thin wire used.

This time around I think I'll figure out what the flat inductors can do then redesign the controller for the best frequency..

D from BC

I've done a low ripple design where the inductor triangle current is wayy less than 2A..I think about 400mApp... In other words, the inductor is nearly creating clean DC current at

2amps....with very little ripple. I'll have to look at the math again when I'm more awake..

Thanks I guess no belt sanding toroids this weekend ..Bummer :( Seemed like fun.. D from BC

OK, well, it should be charging the inductor for 9/17 of the time at

80 volts and discharging for 8/17 of the time at 90 volts... 80V*6.6usec/2.5mH = 211mA p-p. You may find it's a lot more efficient use of parts if you just break the filtering into two sections, especially given that you need to make things thin. That's assuming the reason for the low ripple current is low output ripple voltage. Compare, for example, a single 2.5mH into 2uF and 45 ohms, with 0.2mH series, 1uF shunt, 0.2mH series, 1uF shunt, 45 ohm load. You should find the second gives about half the output ripple that the first does. The first cap must be able to handle the ripple current, of course.

Cheers, Tom

IIRC ferroxcube makes a flat bar ferrite, perhaps you could use a tile drill to cut out a washer-shaped toriod.

also going to a higher frequency reduces size requirements

Bye. Jasen

Ferrite E-cores are available with a thickness of