Push-Pull using only N-Channel MOSFETs

[Kashmire]

Here’s a design inspired by Arthur R. Bailey’s 1968 Wireless World article “The Bailey 30-30”. The intro to the article had a little picture of a bipolar transistor power stage driven by an interstage transformer. I’ve also seen bits and pieces of this type of design on John Broskie’s excellent TubeCad website.

However, JRB isn’t a fan of interstage transformers, and I’m expecting a lot of people on this forum aren’t either. Hold back the knee-jerk reaction when you see this topology!

I ordered a bunch of iron for my new tube amp. My Lundahl LL1660S interstage transformers showed up early, and I still have five weeks to wait for my Plitron output transformers. What to do, what to do?

Use MOSFETs in the interim, of course! Using my 6C45pi tubes to drive my interstage transformer, I can prototype the design in the schematic below. Don’t concentrate on the tube part; look at the transistor output stage.

Why would I do this instead of heading over to the Pass forums? For a big reason: a push-pull topology can have much higher efficiencies than the Pass designs. This amp can work in class A or AB, and it doesn’t have to heat my whole house to do it.

Should I waste my time with an interstage transformer? Maybe, for a number of reasons:
(1) Use the tube to do what a tube function – voltage gain. Concentrate using the transistors for a transistor’s meaning of life - current gain.
(2) Excess gain in the tube stage can be reduced through the transformer’s winding ratio, therefore, reducing the effect of the transistor’s gate capacitances.
(3) Two N-channel FETs can be used, getting rid of P-channels.

This amp is not complete. The bias may be difficult to tune as the amp heats and cools. Any bias mismatch between the top and bottom transistor will cause a DC offset at the output. In the Zen version 5 (push-pull Zen), Nelson Pass uses a potentiometer to adjust bias current of the two devices. John Broskie has an example using an op-amp in a servo loop. Both of these examples are conventional push-pull with N-channel on top and P-channel on bottom.

This design may work with a potentiometer to fine-tune the bias, but an op-amp servo loop would probably be the best worry-free solution. Some safety features must be looked at, also. If one of the FETs fail, the output slams to a rail, potentially destroying the loudspeaker. This is where a good feature of the Zen amps comes into play. If the current source FET fails, the amp just doesn’t work. If the amplification FET fails, then the current source just shunts to ground. Either way is safe.

Also note there is no feedback in the output stage. Some local feedback may be useful, but I’d steer away from global feedback in this amp.

So, I’m asking you to resist the knee-jerk reactions, and objectively look at the topology. This is not a thread to debate the evils/advantages of interstage transformers and/or tube driver stages, there’s probably a better place for that debate…

[subwo1]

Go for it, and let us know how it turns out.

[x-pro]

Hi Kashmire,

The circuit should work, thought you need to be careful with the biasing divider design othewise the power supply and temperature variations will change the idle current dangerously . I've seen this topology published before, about 15 years ago, with a transistor driver stage and MOSFET's on the output, with good results achieved.

I always was particularly interested in a PP design using same transistors on the output - my very first proper amplifier was a JLH class A circuit. Have a look here for my own design with N-channel MOSFETS push-pull class AB output stage:

N-channel only output devices in a power follower.

(not using a transformer - probably not useful for you right now )

Cheers

x-pro

[Circlotron]

Run the bottoms of the driver transformer secondaries to the drains instead of to the sources. This will give the output stage a gain of ~1 as it will be a pair of source followers, with all the usual advantages.

Shift the lower end of the upper fet bias resistor from earth to the upper fet source. The supply to the top end of the top bias resistor should have it's negative end to the top fet source as well.

[Circlotron]

Almost forgot - put a few hundred ohms of resistor *right* at the gate lead. Trim the gate lead to ~4mm length. The resistor damps the Q of the gate lead inductance among other things, making the fets less likely to oscillate.

The sky is the limit for N-channel fets you know.

[millwood]

this transformer is essentially the phase splitter in the jlh class a design. yes, yours should work, and there is a mosfet version of the jlh that uses power mosfet as output, very much the same as yours in terms of the output / transformer "driver" stage, but with global feedback.

[traderbam]

Hi,
What is the amp meant to do? It appears to be a transconductance amp - so you'd need a fixed resistive load or global feedback to make it into an accurate voltage amp. Why do you warn against global feedback?
BAM

[Kashmire]

Thanks for responding to the thread. I'm going to cook up a schematic with Circlotron's suggestions. Circlo's right: I made a mistake by not drawing the amp as source follower.

For now, I'll address the feedback. I've had success using interstage transformers in tube amps. Usually I'll include local feedback at the driver stage to incorporate the interstage into a feedback loop. There's some tricks with tubes that naturally take the transformer into account.

I perfer local feedback with my tube amps - the driver stage has its own feedback that takes the interstage into account, and the output stage has feedback that takes the output transformer into account. I do this because of the natural ability of tubes and transformers. Since this amp will have a driver stage with local feedback, the obvious direction (for me) was for the output to have it's own local feedback.

I suppose global feedback can be used in this case.

Regarding the bias: an op-amp DC servo loop may be the way to go. In my experience, an amp like this will drift like crazy....

[Kashmire]

On this issue of feedback:

I am a firm believer in building the device to desired performance without feedback, and then later add some feedback to correct for small things.

Feedback should be for fine-tuning or fine-correction, not to "fix" problems that could have otherwise been addressed with good design. This is why the tube amps and solid-state of the 1970s sounded so bad. Amp designers were saving costs with cheap designs, and relying on gross feedback to correct for all the problems. Great amps sound good without feedback, because the designs generally don't have gross problems.

Summary: this amp should work without global feedback to desired (or close to) desired performance. Only after that can global feedback be justified.

[jacky]

Think it would be better to do without the coupling cap when you use transformer coupling.

See e.g.
http://www.homestead.com/whaan/files/page3.html