Linder's circuit

Doesn't a higher voltage mean that the direct coupling must be eliminated? First of all, the safety enhancement garnered from using a coupling capacitor may easily outweigh any potential feedback instability or even any sonic liability the capacitor might bring to the circuit. As the amplifier is now configured, a blown fuse will probably blow your speakers as well. Removing the tube while the unit is use might do the same. Coupling capacitors break the DC domino effect's path and thus are highly recommended. But a coupling capacitor is not necessarily necessary and DC coupling can be retained even with the higher voltages.

Resorting to our friend, the transistor, we can implement Linder's circuit, which combines the differential amplifier's two outputs into one. While the output of this circuit is single-ended, it is linear and the slewing ability we lost buy forgoing the push-pull arrangement is more than made up the sevenfold increase in idle current through this stage. Erno Borbely referenced this circuit in his 60 watt MOSFET amplifier article in the second issue of the 1982 set of The Audio Amateur. The circuit works better than most transistor circuits because it tends to cancel out much of the transistor's squirrellines in AC terms, but does require some tweaking in DC terms, i.e. matching and tight tolerance resistors.

(I played with Linder's circuit back in the 80s after reading Borbely; the task I wanted to accomplish was to build a DC hybrid line stage that sandwiched transistors in between triodes. The input was a 6DJ8 based differential amplifier and the output tube was a triode connected EL84 used as a cathode follower. The project was abandoned not because of Linder circuit but because I had played with some zero feedback alternatives that provoked a major rethink and this hybrid had lots of feedback. Although, if redesigned, it might make a good article.)

Since the input to this circuit is at the top of the amplifier, the output must be taken at the bottom of the amplifier, which means reversing the roles between top and bottom MOSFETs. Something that might be for the good, as I believe that the IRF540 is slightly more linear than the IRF9540. (Actually, if you can source some Hitachi 2SK134s and 2SJ49 all the better. One caution is that these devices are only rated for 100 watts dissipation.)

Still, I am not convinced that this is best way to go. Something simpler and less linear might be better. The higher voltage and current are absolutely desirable and should be retained at all costs. So how do we implement something like the plate resistor loading transformation while still retaining DC coupling?