The principle is simple: The coils in series with the pitch antenna and the static capacitance (vs environment) of the pitch antenna itself have to form a series resonant circuit which is tuned a little bit lower than the oscillator's parallel resonant tank circuit itself. The static capacitance of the pitch antenna increases with its thickness and length.
The thickness has also an effect on the tone spacing: A thicker antenna will create a bigger capacitance change for the same hand movement as a thinner one.
Linearization by series coils has (if well done) most effect on the field when playing close to the antenna - the intention is to stretch the field there. When I look at the prototype at the bottom of the page, I understand that there much static capacitance from the big metal front and bottom. In a less metallic environment, I'd rather go towards 60mH (6x10mH in order to reduce the windings' parallel capacitance) in order to match the antenna circuit at ~172kHz (if the osc will really run that low, check with a frequency counter), at least with a "standard" pitch antenna which has a length of 16" (40cm) and a diameter of 3/8" (1cm).
On the other end of the field (lower tones, far away from the antenna) other factors have more importance and effect: The thickness of the antenna and the coupling between both pitch oscillators.
Coupling occurs through the common connection to the mixer, so increasing the value of both 330k resistors towards the mixer input (you might try 390k, 470k, 560k) will stretch the tone spacing in the lower register and allow you to reach still lower tones than before.
Coupling occurs also through the common power supply of both oscillators (not for this circuit, there is a decoupling through the 1k resistors and the 0.033u capacitors) and through the air. The latter can be reduced or eliminated by creating a distance between the oscillators, i.e. by putting the mixer circuit in between.
