The simplest reactance tube circuit to analyse, is a pentode with a capacitor Cr connected from plate to control grid, and a resistor Rr from grid to cathode. This RC network is a phase shift network such that if an AC voltage is applied between plate and cathode, you will get a signal at the grid that is phase shifted (leading). Now, the plate current is equal to the grid voltage times the tube's transconductance. Because the grid voltage leads the plate voltage, the resulting plate current will lead the plate voltage. Hence, if you connect the plate & cathode to two points in an external circuit, they will look like a capacitor. Now, if you change the grid bias, this will change the transconductance, and therefore the effective capacitance will also change. The relationship between plate to cathode current Ipk, and plate to cathode voltage Epk is: Ipk=Epk*(Gm*Rr)/(Rr-j(1/(2pfCr)) where Gm is the tube's transconductance. This means that the plate to cathode impedance Zpk is: Zpk=Ipk/Epk=(Gm*Rr)/(Rr-j(1/(2pfCr)) In the special case where the resistance Rr is much smaller than the reactance of Cr at the operating frequency, this simplifies to: Zpk= -j*[(1/2pfCr)/(Rr*Gm)] which is purely capacitive, and results in an effective plate to cathode capacitance essentially independent of frequency: Cpk=gm*Rr*Cr One point that was brought up in one of the earlier posts related to the difference between a triode and a pentode as a reactance tube. I mentioned that the pentode is easier to analyse. This is because, in a triode, the transconductance varies with both grid bias and plate voltage. The definition of transconductance is "The change in plate current caused by a change in grid voltage, when the plate voltage is held constant." That's tough to do with a triode. It doesn't mean that a triode won't work as a reactance tube, but it means that the published gm curves won't be much help in predicting behavior. With a pentode, it's easier; the plate current is relatively independent of plate voltage, as long as you keep the screen voltage constant, so the gm characteristic is fixed, and except for normal production variations, should follow the published curves. If you wanted to linearize the gm characteristic, you would have to unbypass the RF at the cathode, in order to sense the RF current. There is also another possibility. You can try different screen voltages to see how they affect linearity.