Plate-Grid feedback: How to get better than DHT performance from cheaper tubes

[SpreadSpectrum]

I'm not the first to do this sort of thing or anything, but I threw together a circuit here with some analysis to stimulate some discussion and hopefully help me learn some more.

The attached circuit uses a p-channel mosfet follower with a so-called 'Schade network' to introduce plate-grid feedback and make the pentode or beam tetrode output tube behave like a very efficient low-mu triode. I put in an n-channel follower to present a very low capacitance to the Schade network. This allows higher resistances to be used in the Schade network without rolling off the high end too much. I have been using FQPF2N90(900V, 5.5pF Crss). Another advantage is that the driver can supply grid current. Anyway, the schematic shows values for 20% plate-grid feedback. I calculated out a 170+kHz corner frequency for the feedback network which should keep things nice and flat through the audio range.

I have also attached curves for KT88 triode connected(blue), 300B(yellow), and KT88 with 20% plate-grid feedback (pink). Sorry, grid voltages are not labeled and are different increments for each of the three. If I recall correctly mu came out to a little less than 4 for the 300B, 5 for the plate-grid feedback KT88, and about 8 for triode connected KT88.

Anyway, I think that this circuit should best a 300B in linearity, gain, have a higher input impedance, lower plate resistance, and higher power output. It obviously has the disadvantage of needing a bunch of supply voltages, but I'm not gonna let that stop me.

Problems:
I am concerned about voltage ratings of p-channel devices. Maybe some protection circuit is in order? It shouldn't be a problem unless funky stuff happens on power up/down or oscillations or something.

Anybody see anything here that won't work or that I did wrong?

[Sch3mat1c]

With DC coupling, the downside is a dramatic shift in cutoff bias. Instead of driving in the 0 to -60V range, say, you need -50 to -150 or more. The reduced input impedance demands heavy drive, such as the follower. All these followers brings the question, is it more "FET follower" sound than "tube" sound, if any at all? (The latter is the correct answer, of course, since NFB removes "sound".)

MOS will be happy up to 70% of rated. More than that and you're risking transients (line and load variation, inductive kick for switching circuits, etc.). At these currents, avalanche shouldn't really be harmful, so you could supply a narrowly rated FET from higher voltage than the operating point, as long as it doesn't stay there forever. I make no guarantees on this, of course...

If you can't find HV PMOS, you can cascode them.

Tim

[defect9]

would any of what you're presenting be related to this?

SuperTriode

[smoking-amp]

From looking at the purple curves, it certainly looks like a very linear triode. A cascode could fix the P-chan voltage issue, but getting into a lot of parts then. I can see Tim's issue with the input follower causing some Fet characteristic, but not much due to it's high gm.

I like a little different approach to Schade feedback, using voltage or series feedback at the input. (see attached) Then the Fet operates at near or at constant current, so no Fet signature remaining (constant voltage across the Fet gate to source). Loses some voltage gain though, but tubes are good at voltage gain anyway. Some issue maybe with Drain to Gate capacitiance, but recent Fets are very good there if you keep 25 Volts or more across them. Main issue is tube bias stability, needs additional circuitry.

The super triode is a different animal, but a very nice scheme. Using the Fet for current gain and the tube for voltage gain control (via plate feedback) . With no Fet signature remaining either. Some Japanese sites have used that scheme for some time. You won't have much of any showy power tubes that way in most cases since the Fet is handling the power.

[Michael Koster]

Originally Posted by defect9

would any of what you're presenting be related to this?

SuperTriode

JRB's "Super Triode" uses a vacuum tube to control a MOSFET, which is a sort of signal current multiplier (or conversely an impedance multiplier as seen by the tube). The end characteristic should be like rescaling the current axis of the plate curve of the controlling triode.

The Schade feedback scheme presented above uses a P-MOSFET follower to drive a vacuum tube with a "Schade network" feedback divider. It looks like a nice circuit for varying the % of feedback and hence the output impedance as desired. The triode characteristic is derived from the pentode gm and external feedback. The end characteristic will be similar to a very high gm triode spud amp.

Michael

[tubelab.com]

would any of what you're presenting be related to this?
SuperTriode

Gee, that looks vaguely familliar. My web page is 5 years old:

Super Tubes

Japanese sites have used that scheme for some time. You won't have much of any showy power tubes that way in most cases since the Fet is handling the power.

My best sounding version used a 6W6 connected to a mosfet in a Simple SE board. Real tubes (6V6 sized) and a BIG heat sink on the fets. B+ was 150 volts and the power is limited by the OPT and the heat sink.

[duderduderini]

Hi Tubelab
How did your best version stack up soundwise against say a conventional se amp (either triode or triode strapped pentode/tetrode)
Is it worth it in otherwords or does it sound like sand?
Nick

[SpreadSpectrum]

A couple of questions:

What does a cascode follower look like?

I think that the p-channel follower would be going from abut .5 to 6mA on the extremes. I would hope that what Vgs variation there is would be pretty small especially compared to the swings we are talking about here (100Vrms or so). Would there be a significant amount of distortion because of this? I don't know. I think I will eventually build this and see.

Don,

I'm going to have to take your drawings to work tomorrow and think about them a bit. How will these do as far as input capacitance? One of my design goals with this was to keep input impedance high so as to be easily driven by tubes.

[tubelab.com]

How did your best version stack up soundwise against say a conventional se amp......Is it worth it in otherwords or does it sound like sand?

The experiments that I performed sounded quite tube like. The sound quality was limited by the budget pair of used Chinese "50 watt 600 ohm OPT's" that I was using. I found that the trioded 6W6 - mosfet pair and the paralelled sections of a 6336A sounded much the same except the 6W6 combo had more gain. The real power rating of the OPT's is about 10 watts and both combinations could easilly saturate the OPT's. More power from either combo could be achieved by a lower impedance OPT, which I don't have.

[smoking-amp]

The P channel cascoded follower is attached. The Rd/Rd voltage divider keeps half the Source to Drain V drop on each Mosfet.

Edit: OOPs! I just noticed I drew an N-chan cascoded follower with P-chan parts. For P-chan, the divider middle would go to the bottom gate, and the input drive to the top gate. And sources go on the top sides.

I just noticed some goofs on my earlier diagrams. The bottom one should have the "or CCS" on the top drain resistor, not the bottom source one.
I forgot to put real in Mosfet diagrams too (my shorthand instead). Top diagram is P-chan and the bottom one N-chan. Some gate stoppers needed too.

With a CCS instead of a resistor in the above diagrams, the gate to source voltage stays near constant, so removing that capacitance issue (390pF for FQP2N90). The reverse transfer capacitance from drain to gate with Miller gain factor is now the predominant issue. (5.5 pF * Miller for FQP2N90) For the N-chan case, one could just use some 5W video amp tube instead.

The local loop gain gets quite high with the CCS's also, so some freq. roll-off element may be required for stability.