OK the short answer is yes....
Essentially the problem with bass in the power amp is blocking distortion. Either the PI or the power valves will block when they are driven into grid conduction. When this happens the coupling capacitors from the previous stage charge up (and in a cathode bias amp the cathode resistor bypass capacitor also charges up). When the signal excursion returns back to a "non-blocking level" the coupling capacitors need to discharge substantially before signal can pass again, which as you can imagine sounds terrible!
By changing the values in the PI you can effect the time constant looking into the PI, through which the coupling cap into the PI must discharge, and also the output impedance through which the coupling caps to the power grids must discharge. The impedance looking into the PI is very high due to the grid resistor being boot strapped to the tail resistor. This is even higher in an amp with negative feedback, so blocking into the PI is very much to be avoided as you will invariably have a very large time constant here. Varying the grid resistors and coupling cap into the PI will affect the time constant and hence recovery from blocking. Reducing the plate load resistors will not only reduce the gain into the power valve grids, which again will reduce the tendency to block, but also will reduce the discharge time constant so recovery from blocking should be quicker. Fender used this approach as well as reducing grid leak resistor size in their silverface amps — Fender amps, much as I love them, are notorious for blocking.
Blocking can be avoided altogether by DC coupling the offending stages thus eliminating the coupling caps. This is now what we do, and it works very well.
The PI can be DC coupled directly of the plate of the preceding stage (obviously this is now a fixed bias PI). This approach is used in several hi-fi amps from the 50s, however you have to be fairly careful about choosing the DC operating point. An other approach is to use a fixed bias cathode follower to drive the PI, which what we've done with good results. This circuit is used by Hiwatt in there 400 W slave amps. If you have a spare 1/2 valve this is quite easy to implement.
In a fixed bias amp you can use a cathode follower tied to a negative rail DC coupled to the grids of the power valves a la Ampeg SVT. Again this works very well but might be hard to implement in a existing design as it requires a significantly more negative bias voltage than is found in most amps. Also if you do this I would recommend you use a 12AU7 driver as not only will it have a very low output impedance it also has the best cathode/heater voltage compliance. Unfortunately it will also draw quite a lot of current which might drag down the B+ to the pre-amp. Now you could do this with MOSFETs......
In a cathode biased amp, DC coupling can be done but is hard. Another alternative is to place zener diodes + regular diodes across the power valve grids reduce the amount the cathode bypass cap charges up. I've not tried this so can't comment on its efficaiousness.
I suspect you are seeing earlier blocking in a non -ve feedback amp as the gain into the power amp is not being tamed by the feedback, although in theory the negative feedback will make blocking worse when it happens.
Hope you've managed to plough through this OK.
