I did make a successful SE amp with a CF output stage. It DID use a 6EM7 for the driver.....then I decided to make it bigger and push pull. That's when things started to go wrong.
The driver is NOT the big problem....in fact I found a rather easy solution......a typical push pull pentode amp, say with a pair of EL34's running on the same plate supply that you are going to use for the CF output stage. Use a decent, but not super expensive OPT on the amp, with a resistor in the 20 something ohm range loading it's 8 ohm output. Then cap couple the CF stage directly from the plates of the EL34's. Tweak the resistor value on the OPT secondary for minimum distortion.
I get something like 1000 V P-P of drive voltage at like 0.5% THD with this setup.
A push pull CF output stage has the CT of the OPT tied to ground. The cathodes of the output tubes need to swing from B+ voltage to a NEGATIVE voltage equal to B+ when not conducting. This makes bias and screen grid supplies a bit more challenging......I had a pile of burnt parts and a lot of head scratching before I figured this out.
The mosfet buffer that feeds the grids of the output tube needs to swing more than 1000 volts. Back when I tried to build my first push pull CF output amp, these did not exist.
I see,
Nice solution for getting the voltage swing. Could be done using some choke loading I think.
It would be hard on the screen and bias supplies indeed.
So I think to keep the OPT in the anode.
Did some further simulation and using a 640v supply I'm getting almost 700w of clean power.
I used a gain stage direct coupled to the concertina phase splitter.
After this there is a gain stage followed bye a cathode follower for each side.
the gain stage is bootstrapped by the cathode followers.
Then after this is another cathode follower but the cathode resistor is connected to the bias supply.
There is also a trimmer for each tube in the cathode (look at the mi350 driver)
The grid is then coupled to a resistor divider to the bias supply
This gives a strong drive for driving 4 tubes and it even allows a little bit of grid current.
The anode of this last cathode follower is tied to the +180v screen supply.
I also added quite some amount of feedback.
Those lundahl's are quite good so some feedback shouldn't be a big problem
So far I have simulated this and results look good.
I tried some IMD simulating feeding 20khz and 19khz. The difference at 1kHz was around -90db down (I think this was more a simulation artifact as it was also in the mixed source)
The sidebands at 18kHz and 21Khz are around -63db down. (almost the same as the THD highest harmonics)
Output power was 600w at that point
THD was 0.056%
Changing the bias more to class B gives a rise in distortion
Tubes are dissipating 42w at this power.
I think they should be able to handle this. More so as we are listening to music and not sine waves....
Then last but not least it is still simulation so the real world would be less good for sure but it gives an idea
Nice solution for getting the voltage swing. Could be done using some choke loading I think.
It would be hard on the screen and bias supplies indeed.
So I think to keep the OPT in the anode.
Did some further simulation and using a 640v supply I'm getting almost 700w of clean power.
I used a gain stage direct coupled to the concertina phase splitter.
After this there is a gain stage followed bye a cathode follower for each side.
the gain stage is bootstrapped by the cathode followers.
Then after this is another cathode follower but the cathode resistor is connected to the bias supply.
There is also a trimmer for each tube in the cathode (look at the mi350 driver)
The grid is then coupled to a resistor divider to the bias supply
This gives a strong drive for driving 4 tubes and it even allows a little bit of grid current.
The anode of this last cathode follower is tied to the +180v screen supply.
I also added quite some amount of feedback.
Those lundahl's are quite good so some feedback shouldn't be a big problem
So far I have simulated this and results look good.
I tried some IMD simulating feeding 20khz and 19khz. The difference at 1kHz was around -90db down (I think this was more a simulation artifact as it was also in the mixed source)
The sidebands at 18kHz and 21Khz are around -63db down. (almost the same as the THD highest harmonics)
Output power was 600w at that point
THD was 0.056%
Changing the bias more to class B gives a rise in distortion
Tubes are dissipating 42w at this power.
I think they should be able to handle this. More so as we are listening to music and not sine waves....
Then last but not least it is still simulation so the real world would be less good for sure but it gives an idea