Does anyone here know where I can get some toroids or other small cores in
3F4 material. I need about a 3/8 inch OD core.
I have a DC-DC that is working well but most of my losses are in core losses. I need to fit into a very tight space. The height of the finished inductor/transformer has to be under 0.3 inches.
Type 77 core material gets too lossy above about 100KHz. It isn't as good as the 3F3 material I can easily get. I can easily give up a bit on the Al number for the core, since my copper losses are quite low.
Wow. I am not sure 3F4 is right for that. Best would be to talk to an engineer at one of the ferrite manufacturers about which material may be best. I usually went to #43 when going into the MHz range but there is some stuff in between as well, most likely more expensive though.
There aren't too many power inductor/transformer applications that will unequivocally suit a toroidal ferrite core - a small gap is often beneficial.
Although cost may be no object, availability generally dictates the use of parts commonly used elswhere.
`3/8 OD toroids are metric T8 orT10 (TDK), R9.5 or R10 (EPCOS), TN9 or TN9.5 (Ferroxcube).
Low profile bobbin and core assemblies (eg ER14.5) are available from EPCOS, TDK and others in low loss material, gapped or not.
How have you established that core losses are dominant (zero load rises)? Can you reverse this imbalance, to prove it?
Tolerance of core loss in production of larger parts generally prevents you from allowing it any sign of dominance in a prototype - it can easily double. In a smaller part, the responsibility for heat removal is likely shared by the total assembly - it's harder to pinpoint. This can even be useful.
Yes, in general, the core used in a DC-DC converter is normally gapped in some way. Toroids can have a "distributed gap" where the magnetic material is mixed with some non-magnetic material to provide the gapping. In my case, the availability of the part off the shelf is driving the selection of the core. If the core was optimized, it could be quite a bit smaller.
[...]
The design is basically a SEPIC converter and an isolated CUK converter sharing a core to make two outputs. As a result, I can't use the no-load losses method to do an in application measurement.
I this situation the measurement is made by measuring the temperature rise of the copper and the core in the circuit and then reproducing the rise out of the circuit. Because the inductor is in a confined space, it doesn't take much power to cause the rise I'm seeing. My best estimate is that just about 0.3W is being lost in the inductor when I'm making 3W of output power. The semiconductors losses are about another 0.3W.
And it's dirt cheap. The variety is larger than that of the beer aisle at Safeway.
I have used #43 material for pulse transformers in the several 10nsec range. Also for EMI. The FT-37-43 core is 0.375" OD and 0.125" high. If that saturates in your application you could stack two for experiments.
Our tax Dollars at work? Lucky you. My toroid inductors often have to come in under 50c, custom wound and all. But that is in huge quantities.
Converters like the SEPIC generally do not need gapped cores. Also, as Harry pointed out, in a SEPIC you can "core share" (ain't that a nice word?) and wind both inductors on the same core.
0.3W core loss in a 3W converter is huge. Typically core loss would contribute a percent or so, if that. The thing to watch out for is saturation at the end of the charge cycle. If it isn't a current mode concept you are using install a very small resistor in the source leg of the FET. 100milliOhms or so, something a current mode concept already has. Look at it with a scope. You should ideally see a nice linear ramp. If you see a marked rise at the end, like a shark fin, then saturation is setting in. This is also a reason why continuous mode may not be such a good thing when the core is on the skimpy side.
I went to the store that doesn't have beer and chips mixed in with the sockets and chips[1], and, *bummer*, They are out of them. I will have to order some. I meant to do the order today but I got busy putting the prototype in a box and then trapped in a meeting.
[1] ie: it isn't Fry's
Right now, this is privately funded. The product will sell for about the same as a new car so a few bucks on a core won't matter.
I have a $300 OCXO, a $100 CPLD and a couple of $50 micros so most other things get lost into the noise.
The numbers say that I'm only at about 1/4 saturation. I'm at home and the real numbers are at work but that much I rememeber. This is why I decided it was just the core material that was at fault.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.