RailMax - Power Amplifier with Rail to Rail output

I wanted to create a "Rail to Rail" power amplifier.
This is what I came up with.
I use dual parallel output transistors.
Supply is +/-33V.
Output at 8 Ohm is +/-32.5V = 66W.
Output at 4 Ohm is +/-32.1V = 128W

RailMax MJL4281A_22.jpg
 
One serious issue. You should replace the three 1N4148 with one BD139 mounted on the heat sink to better temperature compensation.

Not so serious: You can replace U1 and U2 with 1 p channel mosfet to get lower distortion. It needs a back diode G - D

If you put an 18k resistor from VCC to emitter of a new KSA992 with the base to ground and collector to R16 you will not burn a loudspeaker if positive rail fuse trips.
With no current limit i think it is a good choice.
 
One serious issue. You should replace the three 1N4148 with one BD139 mounted on the heat sink to better temperature compensation.
Agree.

Not so serious: You can replace U1 and U2 with 1 p channel mosfet to get lower distortion. It needs a back diode G - D
Dalington VAS also needs a back diode at that position if it lacks other current limit means.

If you put an 18k resistor from VCC to emitter of a new KSA992 with the base to ground and collector to R16 you will not burn a loudspeaker if positive rail fuse trips.
With no current limit i think it is a good choice.
Good point, this is often overlooked.
 
If you put an 18k resistor from VCC to emitter of a new KSA992 with the base to ground and collector to R16 you will not burn a loudspeaker if positive rail fuse trips.
With no current limit i think it is a good choice.

Always a good idea to make the current sources completely independent. No need to treat it like it’s an IC op amp. And if you do fuse the rails, do not fuse U4 and U12 or anything in front of it. If R18 burns open or TO-92’s explode it will not set fire to the house. If that happens the output stage is already lost and it’s fuse(s) should have blown. If what happens is a rail fuse blowing due to momentary overload you don’t want to lose DC feedback. Those type transistors, at these low voltages will usually survive 10 ms short circuits with a typical power supply. Sustained short, no. High voltages, hell no. But you could sense current in the emitter resistors and trip a relay in time. TIP35C’s might not fare so well.
 
PMA has good point, collector values seem low for the drivers, shouldn't need 3 diodes.
Something I'm guilty of in sim with this topology, you end up roasting the driver and getting strange bias behavior.

Last time I got a better hold on a design, the phase margin changes with temp. Big jumps
It is a odd topology and have had my fascination with it.
Driving close to clipping at 10k or 20k is where it shows its ugly.
Wonder what THD is for this sim at high frequency. the numbers seem little good to be true at 1k.

I do slightly like the concept of low gain for the CFP, just enough to ride the rails better.
Maybe add internal resistance to the voltage sources, sim assumes infinite power.
Work well for vertical mosfet, to overcome the 4 to 6 volt losses to the rail, low gain is all that is needed.
2 pairs of outputs is the way to go, even at low voltages/ power ratings.
I think with internal resistance added in sim and decoupling of the power supply, its closer to 90 watts at 4 ohms in real life.