MOSFET Attenuator?

[lowell]

Does this exist on the market? Popped in my head yesterday that this could be a great way to achieve variable attenuation. Not too different from RGs MOSFET Follies regarding the B+ reducer, only on the secondary of the output transformer, in series with and before the "top" of the speaker. Could adjust attenuation with a couple components and a pot - changing the Vg with said pot. Yes the impedance would change, but IME attenuation can still sound fine when voltage division is achieve without impedance compensation.

Thoughts?

[nickb]

Maybe. But it always seemed to me that resistors are linear, cheap and reliable.

That said, I did mess around with this on paper for a while, just to test the idea. It's not a complete design. Now, since it needs rectifiers to make it work, there is a dead band of about 1Vpp. I guess a different approach is required. I never did any more work on it as it just didn't seem to be worth the effort.

[lowell]

Thanks Nickb. Could you please explain what is happening there? Where's the signal from the OT secondary entering? Why the rectifiers? I'm quite confused haha.

This is what I was thinking. It's crude I'll admit.

[nickb]

What I showed was an electronic load rather than an attenuator. It leapt into mind as had FETS and speaker loads. Sorry - that wasn't what you asked for at all Never mind...

V1 is the input. It it full wave rectified and applied to the FET. The opamp monitors the input voltage and adjusts the current through the FET to be proportional to it. The effective resistance is 4 x R3 i.e 4 ohms in this case.

Switching back to attenuators, you need to present a roughly constant load to the transformer What you drew will not do that. There are other problems with it. A FET is unipolar and you have an AC signal and a FET is highly non linear to name a couple that spring to mind.

PS:
If R3 were replaced by a speaker, what I showed would work the way that I think you had mind. The 1 meg resistor R5 could be replaced with 4,8 16 or whatever ohm. That would give you an approximately constant transformer load so long that the effective resistance of the FET section was fairly high. You then vary the R1.R2 ratio to change the attenuation. The big problem is the the dead band would introduce a nasty crossover distortion so it's really not a solution at all

The only linear method I can think of for now involves trans-impedance amplifiers but that is a different discussion, I think.

[raiken]

Thanks Nickb. Could you please explain what is happening there? Where's the signal from the OT secondary entering? Why the rectifiers? I'm quite confused haha.

This is what I was thinking. It's crude I'll admit.

You're neglecting to take into account the intrinsic reverse body-drain diode of a MOSFET, which will conduct from the source-to-drain in one direction. It takes two back-to-back MOSFETs to control AC signals.

[lowell]

You're neglecting to take into account the intrinsic reverse body-drain diode of a MOSFET, which will conduct from the source-to-drain in one direction. It takes two back-to-back MOSFETs to control AC signals.

Good point - the "flyback diode" as I believe it would be called but correct me if I'm wrong. The negative going signal will be a short.

[R.G.]

I'm just guessing here, but I think you're going to run into linearity problems.

First: the advice earlier is right about flyback (body or substrate diodes) on the MOSFETs and this meaning that you need some way to present the MOSFET as a variable impedance to both halves of the AC signal. Source-to-source connections and an overall bridge are ways to do this, but as pointed out, both of these will suffer from crossover issues.

I think that a way to deal with the crossover issues could be concocted, but that would leave you still confronting the inherent non-linearity of the MOSFET when used as a variable resistor. MOSFETs have a very limited variable resistance region, and then start acting like current limiters.

What MOSFETs do very well is act as switches. If you could stand a switched attenuator, using fixed resistors for linearity and MOSFETs as AC switches would do a good job, I think.

[Chuck H]

I can't help thinking that this is more a design exercise than an improvement over existing attenuators? A typical resistive attenuator can work fine. I used Mr. Aiken's tuned circuit as a load for one of my designs and it was an improvement. The Weber units use an actual speaker motor and the Hoh/Ultimate attenuator uses a resistive load and re amps SS. Is there any reason a MOSFET load would be better than any of these?

[Mike Sulzer]

But an old fashioned mechanical switch works. Or if you want low level remote control control, there are always relays.

I'm just guessing here, but I think you're going to run into linearity problems.

First: the advice earlier is right about flyback (body or substrate diodes) on the MOSFETs and this meaning that you need some way to present the MOSFET as a variable impedance to both halves of the AC signal. Source-to-source connections and an overall bridge are ways to do this, but as pointed out, both of these will suffer from crossover issues.

I think that a way to deal with the crossover issues could be concocted, but that would leave you still confronting the inherent non-linearity of the MOSFET when used as a variable resistor. MOSFETs have a very limited variable resistance region, and then start acting like current limiters.

What MOSFETs do very well is act as switches. If you could stand a switched attenuator, using fixed resistors for linearity and MOSFETs as AC switches would do a good job, I think.