Well, the fact is that power tubes do NOT sound different. They do not have any intrinsic tone.

"But I can hear the difference when I change to a different type of power tube. How can that be?"

A power tube has a very flat frequency response and they all clip roughly the same. If you put a resistive dummy load on a tube power amp (assuming it doesn't have any intentional frequency shaping) it will measure very flat. However a speaker is not a resistive load. A speaker is a highly reactive load. As I've mentioned in the other threads in this forum section a speaker has an impedance that is sort of scooped at the midrange frequencies.

It is the impedance of the speaker that affects the tone of the amp and different types of power tubes react differently with that impedance. As I've mentioned before a power tube is nearly a current source. The operative word here is "nearly". No power tube has an infinite plate impedance and that's why power tubes sound different. A current source has infinite output impedance, an actual power tube has a finite output impedance.

The output impedance of a power tube (or any active device for that matter) is defined as delta V / delta I which is the change in voltage vs. the change in current.

Let's take a 6L6 for example. Let's assume that the tube has a quiescent operating point of 300V and let's assume we swing +/- 100V around that point. If we look at the plate graphs for a 6L6 at a bias of -10V we see that the plate current at 200V is 95 mA and at 400V it's 105 mA (roughly). Using our formula for impedance we get 200/0.01 = 20 Kohms.

Now let's take an EL34. At 200V the current is 130 mA and at 400V the current is 150 mA. The plate impedance is therefore 10 Kohms which is half that of the 6L6.

This lower output impedance "de-Q's", or flattens, the speaker impedance. Essentially the EL34 has a higher damping factor than a 6L6. This higher damping factor reduces the mid-scoop due to the speaker impedance. This makes the tone have more midrange.

There's a little more to it as the output transformer plays a role as well and 6L6 power amps typically have a slightly higher impedance ratio. There's also different operating voltages and bias points but I'm trying to keep this simple.

You can simulate changing power tubes in the Axe-Fx by simply increasing or decreasing the LF and HF resonance values.