pcb - Is parallelling inductors a viable solution?

I am doing a space constrained board layout, in terms of surface area. So I found an inductor which is \$15\mbox{ }\mu H\$, \$2.3\mbox{ }A_{RMS}\$. I need about \$3.5\mbox{ }A_{RMS}\$, so I was thinking of paralleling two \$8.2\mbox{ }\mu H\$ (\$2.7\mbox{ }A_{RMS}\$) to get \$16.4\mbox{ }\mu H\$ at \$5.4\mbox{ }A_{RMS}\mbox{ (max)}\$, with each inductor on opposite sides of the board. Is this a viable solution?

Answer

If build a "super-component" out of 4 identical inductors (or 4 identical resistors) in 2 chains of 2 components each,

   +--X1--X2--+--
   |          |
---+--X3--X4--+

(assuming negligible mutual inductance, which is true for many common "shielded" inductors), then if each of the four components has identical impedance X, all 4 of them considered as a whole as a single super-component also have the same net impedance X and can handle twice as much current and dissipate 4 times as much power. (This is related to the idea of measuring sheet resistance in "Ohms per square".)

There may have been times ;-) ;-) where I've already bought a bulk bag of exactly the impedance I need, and then discover it can't handle the power. The square arrangement allows me to continue prototyping with the components I have on hand, where each component has exactly the desired net impedance X, while I'm waiting for the "right" component(s) to ship.

Sometimes the current rating is limited by thermal considerations -- higher currents will make the wires overheat and something will melt and cause permanent damage or melt the solder. In those cases, the power dissipated is proportional to the surface area. Using N components rather than one big component makes it easier to cool and can save net space. (Sometimes N components in parallel, each with N times the desired impedance X, uses the least space. N components in series, each with 1/N of the desired impedance X, has the least parasitic capacitance).

Sometimes the current rating is limited by core saturation -- higher currents will saturate the ferrite core, causing the inductance to drop out of spec. In those cases, the maximum energy (temporarily) stored in the core is proportional to the volume of the core. Using one big component that holds all the necessary volume usually uses less board area than using the same volume of core divided up into a bunch of smaller components.