Question about Mosfet Threshold Voltage (Vt)

[Bill Mountain]

So I've read this again for the 100th time and it is starting to make sense:

http://www.geofex.com/Article_Folders/mosboost/mosboost.htm

I have some questions about the threshold voltage:

Does it matter how high the voltage is once the threshold is met?  I plan on using the mosfets for boost, phase splitting, and mixer stages which will all have varying voltages at the source. 

Some possible biasing ideas I've had:

1. Setting up a Vref that is high enough to turn on the mosfet regardless of how it's used and what the source voltage is and using Rgate resistors to set bias.
2. Using a voltage divider on the drain to get half of the drain voltage at the gate and sorting mosfets to find the ones that will turn on.  Should be higher than the 2V typical rating but you never know.
3. Direct coupling after boost stages which will put a minimum of V/2 on the gate and keep the source voltage of the following stage 1V or less.
4. Direct coupling after a phase splitter stage which will put either 1/4 or 3/4 of the voltage at the gate and sorting mosfets to find ones that will turn on at 1/4 V (hopefully 2.25V).  I would have to keep the source voltage low on stages following the non-inerted output and sort until I find one that has a smaller Vt.
5. A couple of D cells in series.
6. Etc.

Would I have any issues with these ideas?

I plan on starting with 9V but if don't get the headroom I need then I plan to move up to 18 or 24 volts.  I imagine biasing at higher voltages will be easier because the gate bias requirements become a smaller variable (if it doesn't matter what voltage is used once I exceed the Vt).  My bench supply can go up to 60V so I have a lot of options to experiment with.  I would think that at a certain voltage many baising issues no longer exist.

My only other question is whether the voltage on the gate affects the size of the signal at the drain or source.  For example:  If I had 7V on the gate but wanted 4.5 at the drain will I have any issues or does the mosfet simply not care as long as Vt is exceeded?

[Bill Mountain]

RG? Gus? PRR? AMZ?

[kingswayguitar]

3. Direct coupling after boost stages which will put a minimum of V/2 on the gate and keep the source voltage of the following stage 1V or less.

all i can add with certainty is you should experiment with direct coupling because I've used it a few times with great results

[R.G.]

I have some questions about the threshold voltage:
Does it matter how high the voltage is once the threshold is met?  I plan on using the mosfets for boost, phase splitting, and mixer stages which will all have varying voltages at the source. 

It does matter, a lot. Here's how:
Pretend you have an N-channel MOSFET with its source tied to ground. That is, Vsource = 0. Vt is actually the difference between the gate voltage and the source voltage, Vgs=Vgate-Vsource. Until Vgs is at least as big as Vt, no current flows.

When Vgs>Vt, increasing current flows. How much flows is the transconductance. Transconductance is defined as the amount of channel current that flows per additional volt above Vt.

So when Vgs<Vt, channel current is zero. When Vgs>Vt, then the channel current Id = k * (Vgs-Vt), where k is the transconductance, in amperes per volt (i.e. Siemens).

When Vgs>20V or so, the gate insulation breaks, and the device dies.

The trick in biasing MOSFETs is much like biasing bipolars - pick a channel current you want to flow in the source and drain. Pick resistors for the source and drain to make the voltages appear there that you want to make happen. From that, find the source voltage, and then bias the gate at Vt above the source. Negative feedback will hold the bias there.

The problem with this is although the uncertainty on a bipolar base's Vbe is about 0.1V, the uncertainty of Vt is maybe 2V. Vt for some common MOSFETs may be 1-4V. This means that is it much more difficult to bias a MOSFET as a linear amplifier all by itself. Usually a trimmer pot or test-and-set methods are needed. As another viewpoint, the amount of variation in Vt is much bigger than the amount of Vgs change needed to make most circuits operate. That's the problem. You have to figure out a way to make the Vt differences from device to device not matter.

Some possible biasing ideas I've had:
1. Setting up a Vref that is high enough to turn on the mosfet regardless of how it's used and what the source voltage is and using Rgate resistors to set bias.

If I understand you correctly, this won't work. There is no current into the base, so any series Rgate has no voltage drop, so no adjustment ever gets done.

2. Using a voltage divider on the drain to get half of the drain voltage at the gate and sorting mosfets to find the ones that will turn on.  Should be higher than the 2V typical rating but you never know.

Sorting of some kind will work. But you probably should sort by drain current in a simple circuit similar to your application.

3. Direct coupling after boost stages which will put a minimum of V/2 on the gate and keep the source voltage of the following stage 1V or less.

Used as a source follower, the source will be Vt+Id/k lower than the gate (k is the transconductance again). This is generally fine, but again, the variation of Vt is such that two devices in the same circuit will be as much as 2-3V apart because the Vt is different.

4. Direct coupling after a phase splitter stage which will put either 1/4 or 3/4 of the voltage at the gate and sorting mosfets to find ones that will turn on at 1/4 V (hopefully 2.25V).  I would have to keep the source voltage low on stages following the non-inerted output and sort until I find one that has a smaller Vt.

I'd have to see the circuit. Again, sorting is going to be tough.

Elegant designs involve finding ways to use parts of a type number without (much!) sorting, and also without trimmers or other adjustments. But it's hard to be elegant in the face of big variations.

5. A couple of D cells in series.
6. Etc.

Not good.

My only other question is whether the voltage on the gate affects the size of the signal at the drain or source.  For example:  If I had 7V on the gate but wanted 4.5 at the drain will I have any issues or does the mosfet simply not care as long as Vt is exceeded?

I think I'd say it differently. The gain of a vacuum tube changes from low plate currents to high plate currents. This is the basis for many tube amp tremolos. But for a MOSFET, the changes in transconductance are much smaller with changes in drain current.

The case you are stating kind of doesn't apply. If you had a MOSFET that had a Vt and biasing such that it was stably biased at 4.5V on the drain, the MOSFET would work find. The gate is isolated from the entire channel by glass 20 volts thick. This includes the drain part of the channel. It's not like a bipolar where the base is separated from the collector by a reverse biased diode junction which turns on when the collector drops below the base.

I hope I didn't misunderstand the questions. Ask more where I've muddied things instead of clearing them.

[Bill Mountain]

My posts are usually quite confusing so I applaud your effort in trying to answer my questions.  Which you have done excellently as always.

I guess most of this is up to experimentation but my main question is...

If I were to use a bias voltage that I know would exceed Vt (and provide a large enough Vgs for all types of amplifier stages in my circuit) would mosfet always turn on or is it possible to have too high of a bias voltage?  You may have explained this and I may have missed it.  I guess 20 volts would be a practical maximum.

Some subsequent questions are:

When using higher than 9V the Vt will stay the same and Vgs and the Source voltage would get much bigger than Vt so I should be able to bias easier???  I sort have asked this but maybe not so clearly.

As for the phase splitter stage.  I wanted to do a high impedance phase splitter like the BJT stage on the Brassmaster.  I thought mosfets would be a good bet because of the high headroom and high input impedance.  I was looking at ways to simplify my circuit and I considered generating the bias voltage of each stage based off of the output voltage of the direct coupled previous stage.  I'll have to draw up a schematic.  I just haven't done so because I want to make sure I understand mosfet biasing well enough.

[Bill Mountain]

You have to figure out a way to make the Vt differences from device to device not matter.

This is something I've mentioned a view times in this thread but I think this is what will help me the most.  It seems to me that if I do commit to higher than 9 volts and keeping Vgs less than 20V I should be able to turn the mosfet on with a Vbias of any voltage between Vgs >Vt and 20V.

I guess this rule of thumb would work for 9 V too.  So this takes me back to my very first question...how close to V supply can V bias be and still turn on???

[PRR]

> use a bias voltage that I know would exceed Vt (and provide a large enough Vgs for all types of amplifier stages in my circuit) would mosfet always turn on

The dumb answer is: yes, it will turn on, often turn _ON_, slam the Drain down to zero, and won't amplify worth a rat's knee.

[Bill Mountain]

so...to sum up...

I will bias each stage individually to have Vgs>Vt...but not by much.

I'll try and put together a rough schematic this weekend.

[R.G.]

Yes - you will have to bias each stage separately.

Each MOSFET has its own unique Vt and g_m. "g_m" is the datasheet term for transconductance. It's in units of amperes per volt.

The "volts" in the transconductance is the amount that the ACTUAL Vgs exceeds Vt. If your circuit has to have a drain current of 1ma, and you use a 100ma/V  MOSFET, then Vgs has to exceed Vt by 1ma/(100ma/V) = 0.01V. And THAT is sitting on top of the Vt, which can vary by a few volts.

You either have to know exactly what the Vt and g_m are for each device and tinker in each device's bias, or figure out some way for local negative feedback to correct the errors for you.

Bipolars have the same issue, but the forward voltage of the base-emitter is so small that it's easy to correct the change in Vbe with emitter negative feedback. MOSFETs have such a large Vt, and more importantly such a large VARIATION in Vt that is it much harder to get it to self-correct.

It is easier if you can allow a big change in source voltage from device to device. If you think about it, correcting Vgs variations for Vt and other quirks requires the source voltage to vary at least as much as the variation in Vt. With bipolars, this can be about 0.1-0.2V. With MOSFETs it's usually 1-4V. JFETs have the same issue because Vgsoff varies even more than the Vt of MOSFETs - these things can vary 3-10 volts in bad cases. The popularity of the J201 is at least partially because its Vgsoff is anomalously small - 0.1 to 1.5V. They can be biased almost like bipolars.

In a small power supply voltage like 9V, having to cover up as much as 3V of source voltage variation is a big deal. It's easier with bigger power supplies. But then you run into current drain and power issues. It's easy to make a device self bias at high currents - but it may become a Smoke Emitting Diode too, at the necessary currents.

The response of the electronics industry to device-to-device variation since the 1930s at least was to first try to make devices more consistent, then to use negative feedback to cover up the variations. Making consistent devices was made to work OK-ish with tubes. Semiconductors had big consistency problems that were solved at first with selcetion/binning, and later with vastly improved manufacturing processes. But in the early 1900s, negative feedback was invented, and was found to be able to make variations in devices largely irrelevant. And today, almost no one in the industry would think of doing audio circuits without negative feedback to clean things up.

With pedals, we're dealing with almost archaic circuits, and putting new devices like MOSFETs into them.

[Bill Mountain]

Here is the overly complicated pedal I wanted to build before I started this thread:



Here's the more sane version I bread-boarded after starting this thread:



If I need more volume on the output I'll add a boost.  If I need a clean channel to blend with I'll just add some simple blender circuity.

I guess I was trying to be witty with all of the direct coupling but I understand now how it won't work as I had intended.

[R.G.]

Here is the overly complicated pedal I wanted to build before I started this thread:
...
Here's the more sane version I bread-boarded after starting this thread:
...
If I need more volume on the output I'll add a boost.  If I need a clean channel to blend with I'll just add some simple blender circuity.

I guess I was trying to be witty with all of the direct coupling but I understand now how it won't work as I had intended.

OK, that helps me understand.

First, you can do the phase splitter thing with MOSFETs. You would first determine the voltage you want on source and drain. In most cases, you will want these to be 1/4 and 3/4 of the power supply, to give you the biggest undistorted signal swing.

With that determined, you want Vgate = 1/4 of the power supply +Vt. So you pick a MOSFET, find the range of Vts from the datasheet, and arrange your biasing network from V+ to ground to be adjustable from 1/4 of V+ plus the lowest Vt to 1/4 of the power supply plus the highest Vt. There are several ways to do this, but a simple one is to make the top resistor in the biasing string be a fixed resistor plus a trimmer pot connected as a variable resistor. Ohm's law gives you the values, as the gate will pull exactly zero current. From a noise standpoint, it makes sense to make this be a "Vbias" generator, with a cap to ground from the bias voltage point, then a series resistor of some high value to the gate. The signal comes into the gate on an input capacitor as you show. You can then either put in a MOSFET, measure Vsource and adjust the trimmer to get 1/4 of V+ on the source, or swap in many MOSFETs noting the voltage they produce, and pick several MOSFETs that make almost the same source voltage at the same trimmer setting. These are then matched for Vth and can all be used from the SAME Vbias by using separate gate resistors from Vbias.

[PRR]

> Here's the more sane version http://i40.tinypic.com/1678tg3.jpg

I once found dozens of 47 ohm resistors scattered along the road.

Looks like you found the 100 ohm resistors?

[kingswayguitar]

> Here's the more sane version http://i40.tinypic.com/1678tg3.jpg

I once found dozens of 47 ohm resistors scattered along the road.

Looks like you found the 100 ohm resistors?

me, when i first started about 3 years ago i bought an 8" square box of 3.9 decade resistors
to this day 3k9 is sooooo common in my builds

[Gus]

I think MOSFETs are overrated in effects
I would have used a BJT(s)

Are the resistors that are shown as 100, 100ohm or 100K or some other value(s)?

[Bill Mountain]

They are place holders.  It's what the software defaults to.  It's still early in the design phase so the values are not set in stone.

[PRR]

IMHO, circuit design *starts* with resistors. They are often more important than the device.

So an "amplifier plan with "default" resistor values, which clearly "won't work" (has no voltage gain or current gain), and has no guesstimated currents or voltages, is a non-starter.

There's many workable MOSFET amplifier plans. Start by stealing one. This will get your Ks and Megs in the right ballpark.