Good work, Ray. Isolating the antenna from direct connection with the noise source (in this case, the utility ground via the circuit board) is an essential requirement for any noise reducing antenna. A lot of folks forget that the coax shield IS PART OF the antenna, especially when there is no other ground or counterpoise for the vertical portion to "work against."Half the battle with a phasing-type noise canceller is getting to where both antennas are receiving the same type of QRM at nearly the same level in each. That's evidently easier when neither antenna is coupled strongly to the noise source.
he other half of the battle is much harder when the wavelength is significantly longer than the space available to separate the antennas! Unless there is enough time difference between the arrival of the desired signal between the two antennas, it's impossible to cancel the noise without ALSO canceling the signal.
A spacing of 100 feet is ample difference in time-of-arrival at 120 meters and above for reasonable cancellation to occur (assuming that both the signal and noise are not arriving from exactly the same direction, anyway). But as you tune down lower in frequency, the maximum difference between noise and signal cancellation decreases. For a spacing of 100 feet, or just under 5 ° of phase difference at 136 kHz, you'll cancel nearly as much signal as noise. Even if one achieved theoretical 100% noise cancellation at 2200 meters with 100' antenna separation, it would be accompanied by 98+ % signal cancellation. In practice, the S/N ratio will seldom improve by more than 0.3 - 0.5 dB in that band and such short spacing, and no change in component values on the board will entirely overcome that situation.
The solution would seem to be either (1) greatly increase the spacing, or else (2) create some difference between signal and noise in both the time and amplitude domains through use of one well-balanced broadband loop for LF, along with one vertical antenna. (Still keeping grounds isolated, of course.) It would add to the complexity of achieving a good null, and require trimming both gain and phase every time the loop is reoriented. But if one cannot achieve enough time and/or amplitude difference between the desired and undesired signals at the two inputs of the device, then nulling one must of necessity null the other.
John