Coupling into the loop can either be by using a small loop within the main loop, or as shown, by tapping around the loop(via an isolating transformer, to allow the loop to float relative to ground). Using the tapping method gives a more obvious idea of loop impedance, and so has been used here. Calculating the capacitance value is easy enough: C(pF) = 0.0885A(square cm)/d(spacing in cm), giving 4425pF at 7MHz and 16 594pF at 3.7MHz. Ignoring effects due to the capacitors finite length, this corresponds to reactances of 5 ohms at 7MHz, and 2.72 ohms at 3.7MHz.
Assuming a constant current around the loop, the voltage across the capacitance is 5/1.45 x that across the feedpoint, which is 50 ohms. Thus for 100W drive, V = 70.7 x 5/1.45 = 244v. This gives the following loop current:
7MHz = 244/5 = 48.8A 3.7MHz = 244/1.7 = 100A
High currents are inevitable since the loop is not only much shorter than a quarter wave monopole, but being in phase, the two verticle section (say) currents are also in opposition (the radiatedsignal being a result of the small though finite difference between path lengths).
Return loss for loop tuned to 7MHz [-25dB at resonance]
Return loss for loop tuned to 3.7MHz [-32dB at resonance]
From these plots, you can see the 3dB bandwidth points (ie 6dB return loss) are 120KHz for the 7MHz antenna, and 30KHz for the 3.7MHz one. For solid state transceivers that have no output match tuning, this corresponds to an un-retuned operating range of about 60KHz at 7MHz (wide enough for the entire UK ssb segment) and 15KHz or so on 3.7MHz.
Using a remote tuner, these narrower figures could be trebled without incuring too much loss, but it would be sensible to use a coupling loop, not a ferrite transformer, if this is contemplated. A coupling loop circumference of 2.2m gave a good 50 ohm match when tried.