How to Bias a MOSFET Gain Stage

[FredB]

Does anybody know were a simple walk though of how to bias a typical MOSFET gain stage can be found?

[tubeswell]

Well I know that if you're using a high-voltage mosfet which is a drop-in replacement for a triode, like an IRF820, then the circuit values are the same as for a triode gain stage. (Although you may need a 100-200R resistor in series with the gate - to limit current, and a 12-15V zener between the source and the gate - to protect against voltage spikes). But that's my limit of knowledge about mosfets

[FredB]

It's more than that. I need something that explains the quadratic transfer function in general, and how to keep the MOSFET in the active region.

Also in all the data sheets for MOSFETs I've looked at I seem to have missed whree that function is listed. Perhaps is a general function, in any case I need a reference on that.

[Steve Conner]

The threshold voltage of a MOSFET, Vgs(th), is hugely temperature sensitive and varies massively between parts and between batches. This makes them pretty difficult to bias as linear amplifiers. It's also why no manufacturer will print a transfer function- he can't guarantee that his devices will even land in the same county. The general transfer function is printed in many places, including The Art Of Electronics, but it would be a miracle if it matched the actual MOSFET you were holding in your hand. Sure, you could measure your MOSFET and find the coefficients to make the transfer function fit, but if the temperature changed a few degrees it would be useless again.

The best way to bias one is with heavy negative feedback to manhandle it to the operating point you want, but this cancels out the distortion and duty cycle shifts that you're probably trying to achieve, if you're using one in a musical instrument application.

The most "musical" ways of applying JFETs and MOSFETs I know are demonstrated in the solid-state "amp models" published on runoffgroove.com. These have trimmers that you tweak to suit whatever MOSFET you put in there. They tend to do it by making the drain resistor variable.

JFETs are a bit less temperamental than MOSFETs. You can use them in pairs: one to do the amplifying and another as a current source to bias the first one, in place of the drain resistor. They will track each other reasonably well over temperature. But I think this might also cancel the even harmonics.

Or you could buy a Blackstone Overdrive and take it apart to see how it works. It's all black magic

[R.G.]

Does anybody know were a simple walk though of how to bias a typical MOSFET gain stage can be found?

Does this help?
Designing MOSFET boosts

[FredB]

Steve,

That explains a little.

I've learned how to use them in JFET/MOSFET cascodes. Although now I wanted to learn how to use them as stand alone gain stages.

I want to be able to use them as clean gain stages. I know where thermionics are necessary, and where clean FET gain stages can be convenient.

Lots of NFB aye, good thing they have lots of gain.

***

Hey RG,

That looks like just what I was wanting.

Lemme give that a read...

[defaced]

No quadratic transfer function here, but this is my personal experience with playing with high voltage MOSFETs.

As a gain stage the input capacitance can be an issue, the IRF820 is a little high, the LND150 and ZVN0545 are lower. The LND150 (depletion mode device) can drop into a tube gain stage with the addition of the aforementioned gate stopper resistor. Other devices, IRF820, ZVN0545, enhancement mode, are biased differently and will require modifications to the circuit (see MOSFET follies). I've used the LND150 as a gain stage and phase inverter in a typical cathodyne setup and it has worked well - no audible loss in high end or distortion - I've not verified this with a scope because I don't have one yet. I've used the ZVN0545 in a typical LTPI and it's also worked well. I have not used the IRF820 as a gain stage. RG has made several threads here that add to what's on the MOSFET follies page. It's well worth the searching and reading.

[FredB]

Hey Mike,

I've already read some of those threads. I especially liked the one about the MOSFET phase inverters.

By the time I had found that one, I had already designed a solid state phase inverter based on JFET/MOSFET cascode. It worked extremely well for a clean phase inverter, although it was sensitive to changes in supply voltage.


The article RG mentioned was just the thing. After reading that I whipped this up:

MOSFET Gain Stage.jpg (Supply = 250V)

I based the DC source voltage to be just over the worst case threshold voltage to see just how much gain I could still get out it. With a source resistor of 1k and an un-bypassed drain resistor of 24k5 I was expecting a gain of 24.5 although it simmed out to be just over 20. I have to figure out why. With the source resistor fully bypassed the gain went up to a satisfying 251.

I wonder just how sensitive these particular values are to device and temperature variations. I can see how raising the value of the source resistor will increase the negative feedback, increasing the tolerance to variations in those factors.

I guess the moral of the story is setup the circuit for just as much gain as necessary, and use as much NFB as possible.

Cheers RG

[FredB]

I got another question, or more questions:

How does Rsource help in handling variations in Vgs?

And, does that still work if Rsource is bypassed?

I'm guessing it does because were talking about the DC operating point here.

[R.G.]

How does Rsource help in handling variations in Vgs?

It introduces a voltage in series with Vgs compared to the input, which does not vary. Variations in Vgs are then "watered down" as seen from a biasing standpoint. If you add a big enough constant to a variable, the sum becomes nearly constant. This is Rsource.

You can also think of it as DC feedback, which it is.

And, does that still work if Rsource is bypassed? I'm guessing it does because were talking about the DC operating point here.

You're correct, it does.

[FredB]

I measured the Vgs for the model I am using. Funny, it's already out of spec from the data sheet.

The data sheet says, 3.75 typ, with 3 to 4.5 as the range.

Without a source resistor that's a 1:1.5 ratio or +/-20%.

Now if I set things up so that source is at 5V, then typ would be 8.75, and the range would be 8 to 9.5, and that would round off to +/- 9%.

I'll have to tweak the model to get some best and worse cases and see how that effect the sim.

I've worked out a nifty MOSFET Concertina for a balanced line driver. It draws only 8ma, and lacks a need for a heater hook up. I gotta say that's much easier than trying to generate +/-15 rails for opamp.