DISCLAIMER: I am not responsible for any damage caused by using the information below. Use it at your own risk. Please remember that there are lethal voltages inside the amplifier while it is turned on and after that. Please do not try to make these modifications if you do not have enough skills to deal with high voltages or if you do not understand something in the schematics. Though all of these mods are (or were) implemented in my amp and work successfully I could make a mistake plotting the schematics.

NOTE1: This modification requires 8 Ohm speakers.
NOTE2: Some parts are missing names and values in the schematics on the pages below. That means that the original parts are used at that place.
NOTE3: All resistors without explicit power rating are supposed to be 1W.
NOTE4: Mounting tips drafts suppose you are looking at the amplifier internals from the bottom with the front turned to you unless otherwise specified.

Parts (PSU and bias for class A):

• A piece of prototyping PCB
• 2x 100uF 450V electrolytic caps
• 2x 0.47uF 630V polypropylene film caps
• 200 Ohm 25W pipe-shaped wirewound resistor with mounting rod (I've used an adjustable 220 Ohm).
• 15k 2W, 100 Ohm 1W, 3.9k 1W resistors

• 2x 0.47uF 630V polypropylene film caps
• 2x 0.1uF 63V or more polystyrene or polypropylene film caps
• 2x 470uF 10V electrolytic cap
• 2x 1k 1W, 2x 330 Ohm 1W, 2x 22k 1W, 2x 15k 2W, 2x16.5k 2W, 2x5.6k 3W resistors

Schematics (bias power supply):

The only change is a different R5 value.

Schematics (output stage biasing, one channel):

You should connect 8 Ohm speakers to 4 Ohm taps and set bias to 55mA for class A operation.

Mounting:
C1, C2, R1, R2 of the power supply are mounted on a piece of prototyping PCB attached to the leftmost PSU PCB stand with one (lower) screw and glue. Another screw is removed and R6 mounting rod is screwed into it's hole:

Please note that R6 gets really hot. So use teflon insulated wires to connect it. An ordinary wire insulation may smoke and smell, I don't think you'd feel comfortable seeing your amp smoking. :)

An important note:
Biasing is a bit tricky with this schematic because the plate voltage significantly depends on the bias current. The recommended procedure is the following: set all the bias pots to the minimum current and power up the amp. After a couple of minutes, when the tubes warm up a bit, start increasing the bias graduately for all tubes at once. E.g. set all the tubes to 10mA, then to 30mA and 50mA. When you increase the bias of one tube other biases will decrease and visa versa, so you have to repeat bias adjustments several times before all the tubes reach the required bias. Leave the amp powered for half an hour so the tubes really warm up and set the bias to 55mA.

Removing the NFB and moving to class A uncovers the stock driver weakness. EL34 in triode mode requires rather low driver output impedance, much lower than stock 12AT7 stage has. The preamp/driver below can fix the problem.  Also it reduces the amp gain to a reasonable limit.
The driver shown here uses (6N6P) Russian tube. It is similar to 5687 tube (but is bigger, rated for more power, has higher transconductance, requires less heater current and has a different (6DJ8-like) pinout) so I think a properly connected 5687 should work here as well. Also tube may give even better results, it has the same pinout as . The preamp is a single-ended 12AU7 stage. It sounds very natural and 3-dimensional to me compared to stock SRPP stage.

Schematics (one channel):

C6, C7 are common for both channels. But surely separate caps for left and right channels placed close to the tubes will not make sound worse.
A Black Gate N or K series cap can be used instead of C4 and C5.

Mounting:
All the parts are mounted on the pre/driver PCB instead of original parts. It's good to rewire at least the preamp part point to point.