I made this graph from the 9.csv file from yesterday. I notice some differences between your graphs, can you explain how you modified that data so it gives similar output as yours?
Basically I normalize the fuchsia data to the max peak, assuming that this peak is as close to the zero impedance as this gets. Anything below this peak I treat as an attenuation of the primary LC circuit. Next, I divide the blue measurement by the primary attenuation just computed. The above applies to linear magnitudes. To do the same with logarithmic magnitudes (dBmV), I normalize the fuchsia peak to 0dB and subtract it from the blue instead of dividing by it, because subtracting logarithmic magnitudes is equivalent to dividing linear magnitudes. See here. Also, note that the magnitudes in the red plot, which I have posted, were linear. There is a way to do all this without the csv math by using some fake calibration of the SA. To do that, just normalize/calibrate it using the fuchsia setup and then do the blue measurement with this fake calibration*. This way, the frequency response of the primary LC circuit will be automatically cancelled out by the SA's correction/calibration algorithm and the result will be the red measurement. I preferred to calculate the csv data by hand to see what was happening and to have the current probe's frequency response calibrated out, too. A much better way would be to put a 10MHz Hall magnetic sensor between the pancake and toroidal coil, because the data from such sensor (when properly oriented) would not include the effects of current flowing between the turns of the pancake coil (through the interwinding capacitance) but such HF magnetic sensors are very hard to obtain. The usual ones top out at 1MHz. * _{This assumes that the current probe has a constant amplitude vs. frequency response.}
