А вот что объясняет один из зарубежных гуру - R.G. :
Цитата:I guess I'd better reply then...
The MOSFET source follower replacement for a CF is a handy way to put in a very low impedance buffer/follower. The input side offers a very high impedance to whatever drives it, and the output is very low impedance, much lower than a CF.
On the input side, the thing suffers only from having a high gate-source capacitance, which is mitigated by the device running with 100% negative feedback as a follower. The follower connection prevents the gate-source capacitance from overloading the device it's buffering and makes the whole thing work. You're good to go up to the top end of the audio range with a 12AX7 driving a power MOSFET. The easiest way to do this is to collect the MOSFET gate directly to the triode plate.
If you don't have a plate handy to connect it to, then you need bias resistors, as in the thumbnail. The input impedance for low and mid audio range is essentially the parallel combination of the two biasing resistors, so 1/2M for the 1M resistors, 1.6M for the two 3.3Ms.
The output impedance is low, off the charts low. The IRF8xx transistors have a transconductance of about 1/2 A per volt on the gate and a threshold voltage of 2-4V. Once you get above the threshold gate-source voltage, the device conducts about 0.5A/V. So in a tube application with tens of volts available to drive the gate-source, the thing is always running just barely cracked open. A heavy load on the source can be driven easily, as long as it's heavy pulling it up.
The pull down is all done by the load resistor, which is the origin of the distortion.
If you really want to drive a heavy load with a CF in a tube amp, concoct a constant-current source load from another MOSFET to replace the source resistor. Now the load can be actively pulled down by a MOSFET as well as up, and you can probably drive a speaker directly from the MOSFETs. Heck, no probably about it. It will drive amps directly from the source follower. This is one way to make a solid state audio power amp.
However, for use in a tube amp, it's probably overkill.
Running the source resistor up makes it easier for the preceeding stage to drive the high gate-source capacitor at all frequencies and I think that's what happened for you.
The IRF800 series is not the only one to use. There are some dandy fine devices available from Zetex. The ZVN0545A comes in a TO-92 equivalent package and is rated for 450V, 90 ma, and 0.7W.
You have to be careful not to bias it at too high a current and exceed the power rating, but it's input capacitance gate-source is only 55pF, so it can be driven by a typical 12AX7 plate WITHOUT source feedback assist to over 17kHz. In the SF setup, it's good to far beyond audio. There are also complementary P-channel devices (e.g. ZVP0545A) that could be used to make a complementary MOSFET follower that would be a massive driver for any load you'd like in a tube amp.
The big power MOSFETs have big gate-source cacitances, up to 1000pF or so. All by itself, this would prevent the high-impedance plate of a tube from driving it at even modest audio frequencies. But with a big enough resistor in in the source, the voltage feedback effectively reduces the capacitance that the tube plate sees. In this case, the source-ground resistor does make a difference in what the driving stage sees. So it sounds like the 33k was too low. A higher resistor increases the local feedback that's hiding the gate-source capacitance from the tube plate.
Or you could go with one of those Zetex ZVN0545A's. Those have gate capacitance of only 55pF, much lower than the IRFs. But they can only dissipate 0.7W, so you have to keep the current through them down.
With 363V on the drain and 173V on the source, there is 363-173V= 190V across the MOSFET. The highest current you can use and not burn it up is 0.7W/190V =3.6ma, and that's a source resistor of 173V/3.6ma = 46.7k.
With the IRFs, they can dissipate (much!) more power, so you could run them with more current, but then you lose the audio response. So you still have to use a bigger resistor.
It looks like you're forced to 47K to 100K for the source resistor.
Подчеркнутое от меня.