If the goal is to address only the impedance rise around 2 kHz (as MoFi did with the SP-10 ME), a more refined approach is required.
There is a Audiogoner who PM me to ask about this. Instead of replying back privately, I will share here for the benefit of the group. In order to notch down the impedance spike at a specific frequency, such as 75ohm at 1.6khz in the case of SP-8 (see the above imp.-phase chart), you need a so-called LCR serial components in parallel with the driver, which consist of an inductor, a capacitor and a resistor. Here is how you design the LCR serial components. First, since the target impedance we try to notch down is the nominal impedance of 8 ohm, use the voltage divider formula to back track the required resistor value, i.e., R = (75x8) / (75-8) = 9 ohm. Then select a proper capacitance value, preferably from a higher quality part commonly available in the market, say, C = 10 micro Farad (uF) and use peak freq. = 1/[2 pi (LC)^.5] to back track the required inductance, L, which can be calculated as 1/C/(2 pi peak freq.)^2 = 1.0 milliHenry (mH). So now you have acquired the 1st set of trial values for LCR, i.e.,
L = 1.0 mH; C = 10 uF and R = 9 ohm;
Next check the resulted Q value which define the shape of spike = peak freq. / bandwidth. The Q = 1/R x (L/C)^.5 = 1/9 x (0.001/0.00001)^.5 = 1.1, meaning the resulted bandwidth of the notch filter is approximately 1.45 khz which seems to cover the shape of the spike reasonably well. So this seems a good starting point. It will take some iterations to ultimately iron out the peak as shown in the SP10 ME but this process should give you a pretty good idea how to design a proper notch filter.
