Idler current is the general term for current flowing in the "multiplier" that is not at either the input or desired output frequency. Generally, one wants to reactively terminate the major idler frequencies, so that output power is not squandered there. Don't confuse the various types of multipliers. I would group them in four groups: 1) passive rectifying multipliers, like shottky diode ones, that are either push-push for even order harmonics, or antiparallel (clip tops and bottoms of sine waves) for odd harmonics. These multiply simply based on the fact that the rectified or clipped output waveform has a fourier content at the frequency of interest. Typically very lossy, at least 10 dB. Terminating the unwanted frequency idler currents in these multipliers is probably not worth it, because the lossy diodes do not allow you to recover that idler energy into the desired frequency output. 2) active device like bipolar or transistor, who because of the added DC power can operate with less conversion loss. In these multipliers, you can gain some efficiency by reactively terminating the unwanted frequency currents (idler currents). A time-domain way of thinking about it is if you are trying to get a X3 output, you can "short out" any content at X2 and X4, leaving the remaining horsepower of the active device to only produce the desired X3. 3) reactive multipliers, like varactor diodes, which are very efficient but demand that you reactively terminate the unwanted frequencies. I have made passive doublers that had 2 dB conversion loss. The idea, valid because the varactor is low loss, is one of conservation of energy. If you put in energy at one frequency, but only allow it to resistively dissipate at the desired output frequency, the conversion will be very efficient. Penfield wrote a good book on the effect a coon's age ago. 4) Impulse type devices, like step recovery comb generators or digital gates, that generate a wide number of odd, or even, or sometimes both harmonics at very high conversion loss by forming a spike out of a sinewave input. Here idler currents are of no factor. Since the spike width is very narrow in time, it contains very little energy; and since there are many frequency "teeth" sharing that energy, the total conversion loss is very high (25 dB or so).
