I've been playing with filter design lately in the process of making PCB's for a few prototype transmitters and the old debate is raging. This is what I found.
Among hams there is a notion that toroids are self-shielding, that is the magnetic flux of the winding is contained to within the core and thus very little outside influence can be expected. With air-wound inductors, this is not true, and every pole in a filter is usually designed and build with adjacent coils at right angles to each other.
This lack of interaction in toroids is somewhat true I found, but measurable interaction does occur.
I built a 7 element MW lowpass filter of shunt type Chebyshev design with 0.1dB ripple and a cutoff frequency of 1.9 MHz. In this design, all inductors were built broadside to each other to save space, cores were T200-2 spaced about 0.6"
Anticipated rejection (Fc = 1720 KHz)
Fc -0.5dB
2Fc -50dB
3Fc -76dB
Measured rejection
Fc -1.9dB
2Fc -51dB
3Fc -77dB
I then changed the winding direction of the middle inductor and measured again.
Measured rejection
Fc -0.26
2Fc -51
3Fc -76
I found it interesting that the rejection of the harmonics was little affected, but reversing the winding direction had a large effect on the passband loss. With the inductors wound the same way, passband power loss was around 35%, meaning 100W in, 65W out! With the inductor wound the other way, losses were reduced to just 6%.
All measurements were done with a HP signal generator and spectrum analyzer. First, reference measurements were made with the generator set at 10dBm on each frequency, then the filter measured, and the loss calculated.
+-RH