Here is my little EL95 amp in my bedroom system.
The circuit worked out well. I'm using a 12AT7 as a differential amplifier and driver circuit. It has a constant current source for a cathode circuit, tied to B- of -300V, so it is very well balanced.
The output section is ultralinear. Feedback is taken off the 8 ohm tap and applied to the grid of one side of the differential amplifier, audio input on the other. This way there is no IMD caused by the feedback node mixing with the audio. The EL95 is a easy to use because you can use a 1M grid leak resistor. They are cathode biased, using a common 220 ohm cathode resistor tied to both cathodes, in this way increasing the differential operation of the whole circuit. I optimized the 12AT7 circuit before adding feedback; it was plenty linear and easy to listen to before adding the feedback. The latter was added as much to clean it up a bit as it was to simply reduce gain; I also wanted the amp to behave a bit more like a voltage source.
Bandwidth appears limited by the output transformer. I was surprised to see it rolling off at about 100KHz (-1/2dB @90KHz)- better than I expected. The amp is very quiet and is nicely suited driving headphones- no buzz or hiss. I tried several SETs (including one using the type 45 power tube) and this amp was faster, more detailed, plays better bass and is overall smoother and more musical. It could easily be expectation bias but this seemed to confirm my theory that the main reason people think SETs sound better is because they can't get their hands on a push-pull amplifier of similar power, built to the same level of build quality. Apples and oranges... When the amp is the same power level, its first watt is spectacularly good.
It did need some time to break-in; I ran it for a couple of weeks and it seemed to sound OK, but lately I've noticed its sounding a bit better, more palpable.
