Hi Gordy,
Formula is C (in uF) = 159000/Xc.f
Xc is impedance desired in ohms
f is frequency in Hertz.
Example:
Let's say you want your coupling cap to be 6dB down (that's half voltage) at 80 Hertz. The input impedance of an AKSA is around 43K ohms, so the equation give us:
C = 159000/43000 x 80
This is 0.04622uF.
You'd likely go up to the nearest preferred value to further reduce LF phase shift, and this 0.047uF, or 47nF.
Recalculating for the preferred value gives us 78.6Hz, close enough.
This simple first order filter is flawed, however. The Zin of the AKSA is not a constant 43K; it varies very slightly with input amplitude because of the base admittance on the semiconductor, T1. But this is likely only about 10%, so it's not all bad. Furthermore, there will be considerable phase shift at the corner frequency, extending upwards to about 300Hz, which is significant. The only way around this is to increase C1 to large levels, and the stock value is 0.47uF, ten times larger. But for the frequencies at which you are using the amp, it's unlikely all this is too important, because it beats using an active filter, with all the attendant complexity, to secure just a first order function.
There is a swathe of math around filter design, something I don't fully grasp. Others might know the fine detail; but I can assure you that this will be around 6dB down at 80Hz, and 12dB down at 40Hz, and 18dB down at 20Hz. This is useful attenuation, as it reduces out of band operation of the midbass driver.
Cheers,
Hugh
