Online ordering:
Mouser: EPCOS/TDK cores and bobbins are fairly well stocked, considering. Also Fair-Rite, and the usual EMI parts (beads / snap-its / plates / etc.) from Steward, Laird, etc.
Amidon: relatively expensive; powdered iron toroids and various shapes of ferrites.
Online inventory, quoting:
Elna Magnetics
Adams Magnetic
I don't know of any supplier with searchable / selectable core parameters, so you'll have to check back and forth between manufacturers' catalogs and supplier inventories to select parts. PITA, but I guess the magnetics business goes so slowly that they have little need of modern computerized interfaces (one of those "be thankful they're even on the internet" things).
Manufacturers (you can usually guess from the types or codes who it is; Amidon doesn't list manufacturers on their up-front catalog):
Micrometals (powdered iron)
Arnold Magnetics, Magnetics Inc. (Sendust, MPP, etc. powder cores)
Magnetics Inc. (also stripwound cores, ferrites)
Fair-Rite (ferrites...if you can imagine)
Ferroxcube (ferrites, powder cores)
EPCOS/TDK (ferrites usually, I think other stuff too though)
And probably a bunch of others I can't think of at the moment.
My documentation: rather old, though.
http://seventransistorlabs.com/tmoranwms/Elec_Magnetics.htmlNever really finished the inductor section; I know better now. Works the same as with transformers, select core area based on turns and flux. Flux is amps * inductance (since inductance is Wb/A), and that's peak flux for whatever peak current you apply. Once you have enough flux, you can set the inductance anywhere you want by adjusting the air gap, which means you get flux first, then tune for the amps required. Don't worry too much about energy or energy density. You'll be in the right range when you find a core that ends up with a calculated gap that produces an effective permeability of 20-100.
Also, mind that the winding will generally benefit from using Litz wire, even if the wire is significantly smaller than the (free air) skin effect at the frequency you're using it at. Reason being, for each and every layer in the winding, proximity effect squeezes the current tighter and tighter, easily 2-10 times tighter than skin effect alone would suggest. You have some gain already from using just a few strands in parallel (multifilar), and from using >20 strands (usually as proper Litz), you can recover all the cross-sectional area of the copper you started with (the winding factor of Litz is lower, but more than made up by the efficiency gain). This assumes you're using a ferrite core because you need low losses -- if you're just smoothing DC, use powdered iron. (And, to select powdered iron, you don't usually get gapping as an option, so choose a grade based on losses, then select a size and calculate turns. Don't worry so much about using multiple strands, as you'll usually be seeing more core than copper loss, even with moderate levels of ripple.)
Tim