It may not be what you need or want Neil (?) but I'll give a few thoughts with some info here regarding PIs. (It's a slow day at the Covid office.) Their job is not done in the way most people think. This may even help others in this sort of area or at least make them think.
The power amp is usually comprised of two stages. The second is the well known output stages of EL84s and Tx. The first is the Phase Inverter. In simple diagrams these are represented by the generic amplifier symbols, the large triangles. The PI has the first task of taking a single input signal and generating two amplified output signals, one the same and one inverted, which are each used to drive an output valve of the pair during different +ve and -ve sections of the signal. The Tx then takes the currents from both sides of the output stage and knits them back together again into a single output signal, the same as the input but much larger and with higher current capability, (that's the same as saying at a lower impedance). The power amp has global feedback applied around it! This means a bit of theory if you are interested in getting to the core of how your power amp really works. Let's make a very simple example. (That gives me the chance to draw pretty pictures and I always love that.) You can skip anything that you can't be bothered with but you won't get the full picture and completely follow what I am saying.
Have a look at the attached pic. It shows a power amp made up of a PI and an Output Stage hooked up exactly like our own power amps are in guitar work. This is pretty much universal for us, (more of that later). For simplicity in our example the PI has a gain of 10x and the Output stage also has a gain of 10x. Two feedback resistors R1 and R2 are connected so the PI sees a voltage picked from the output and dropped down to 1/20 of that level. This is attached at the -ve input of the PI, (the grid of the other triode). Pretty simple so far.
The first interesting fact is that the input to the PI is NOT the 1V which the preamp feeds it. The true input signal which is what the PI actually sees and acts on is (1V - 1/20*Vout) and Vout is a bit bigger so that second part is very significant. We need to do a little bit of maths to generate some figures. Don't worry it is only arithmetic!
The feedback equation for this sort of configuration is: GAIN = A / (1 + A*β) where A is the gain of the entire power amplifier without the feedback in place, 10x (PI) and 10x (OS) = 100x, and β is the "feedback factor" set by the two resistors of 1/20 (= 0.05).
This means that the gain of the amplifier with feedback in place is not (10 x 10 = 100x), it is actually 100/(1 + 100/20) = 100/6 = 16.67x. That's some drop! So with 1V from the preamp we end up with only 16.67V at the output even though the two stages both have gains of 10x. That's what feedback does! And in doing that it also decreases distortion, (not that from the preamp), decreases noise and increases bandwidth all for free.
Now we can see that the voltage at the PI -ve input fed back by the resistors is actually 1/20 of that 16.67V. That makes it 0.834V. Here is the crunch...
YOUR PI IS NOT SEEING THE VOLTAGE FROM THE PREAMP AS ITS INPUT, IT IS SEEING (THE PREAMP VOLTAGE - THE FEEDBACK VOLTAGE). In our case this is (1V - 0.834V) = 0.166V. This must be true because if we now multiply this by the real gains of the stages we get 0.166V x 10 x 10 = 16.6V. That matches up with our original equation's predictions.
The point is that the PI is only amplifying a signal of 0.166V not 1V from the preamp as most people assume. If you drop the PI gain by substituting a different valve this is another bigger system you are changing and it may change in ways you don't expect soundwise. In reality, what you are doing with this substitution is to drop the gain of the PI stage while leaving the output stage gain the same. The output stage requires the same input voltage as before to generate the same output level but the PI now has to supply that from the same input signals but with less gain. It can't do that. In reality the PI output voltage will drop, the output stage will see less voltage so its output level will drop. This will decrease the feedback signal to the PI -ve input which will in turn increase the PI input signal, (remember this is actually (1V - feedback signal) ). So the voltages will self adjust until the system does reset to a new equilibrium but with a different perspective from the PI's view of what its own distortion contribution will be.
One area this affects is crossover distortion. Guitar amps are very bad at eliminating it and it is not good sounding distortion. Its contribution is pretty much a constant level across the 0V line no matter what the signal level is so it also has the property of INCREASING ITS PROPORTIONATE LEVEL AS THE SIGNAL LEVEL DROPS. That's another reason why low level playing can sound rough. So we set bias in the output valves so as to minimise it - to a degree, but we can't get it down to anything like hifi standards, that's valves for you. It isn't really necessary when you actually want to generate >5%, (sometimes as much as square wave level depending on your genre) distortion in the rest of the circuit. However, because it is generated inside the feedback loop the negative feedback does the job of actually reducing that level of crossover distortion just as it reduces the gain of the power amp overall. Reducing the gain of your PI will reduce the overall level of feedback you are applying and will increase the level of crossover distortion.
Ok, so this is what people seem to want to explore. Fine, but I have always been curious to get accurate figures of how many people try out new stuff like valve swaps, RAVE about how fantastic they are when they first do them, then a little further down the line get dissatisfied and go back to their original setup BUT SAY NOTHING ABOUT THAT REVERSAL! I know that this does go on as I talk to many people about their amps and some are honest. In my own case, I use a JCM800 2210. You know, the ones that sound shit because they have clipping diodes built in? I have read posts from and talked to a lot of people who have removed those diodes completely or replaced them with their own little diode networks to get away from the simple "rectifier bridge around a 1N4007" that Marshall use. They are all on record as saying how great the amp then sounds with their changes in place. I have found a significant number of people who are honest enough to admit they simply reverted back after a time as the change got wearing. And I strongly suspect it is the same with valve swaps, both model changes and changes of manufacturer. If the differences are so great and the improvements people find are so huge why do they keep on looking for the next "ultimate" valve choice. Looking for improvements in that direction is a genuine dead end as it is only giving in to your FOMO, fear of missing out, and of course your overriding desire to be seen other musos as a fully paid up and licenced member of
The Priesthood ! It isn't The Illuminati we should be fearing, it is the mojo fuelled Cork Sniffers.