Okay.. Vgs(ON) is the voltage where a mosfet starts to conduct at all. Below that, the drain-source resistance is infinite (well, a few hundred megohms). Above that, the drain-source resistance is significantly below infinite.

Vgs(ON).MAX is a cover-your-ass value to protect the manufacturer. Datasheets are legal documents, and a part that doesn't live up to the published spec is a product liability lawsuit waiting to happen.

If you don't know a lot about semiconductor fabrication, the tolerances are fairly wide. Actually, the fabs do a great job of controlling the process, but the values they're controlling are incredibly small. The average doping level is about 10^15 atoms per cubic centimeter, which sounds like a lot until you realize that there are about 10^23 atoms of silicon per cubic centimeter. The actual dopant concentration is around 10 parts per billion. If the level happens to reach 15-20 parts per billion, it changes the way the device behaves, but it's hard to measure such a small difference in concentrations. It's called 'process variation', and it's something you learn to work around.

In this case, they design the IRLB8721 for a Vgs(ON) of 1.8v, but process variations mean they'll get chips whose Vgs(ON) is 'around' 1.8v. The Vgs(ON).MAX value is a guarantee that every chip they sell will go ON to some degree if you feed 2.35v into the gate. Industrially, it means the standard deviation is probably about a tenth of a volt, and the odds are less than one in a million that you'll get one that's .6v off the nominal value.

peopleselectronics wrote:If the absolute Max VGS is 2.35V why is 4.5V given here?

Mosfets work by pulling electrons close to the gate. The more voltage you apply to the gate, the more electrons you get, and the better they conduct.

For the IRLB8721, putting 2.35v into the gate will get you an Rds of about 2.5 ohms. That's pretty good, but it can get better. If you take the gate voltage up to 4.5v, Rds drops to about 20 milliohms. If you take Vgs up to 10v, the device is guaranteed to have a channel resistance no higher than 8.7 milliohms.

.. if you take Vgs higher than 20v, you'll blow a hole through the gate and all the electrons will fall out.

The graph seems to stop at 4v because the vertical scale is in milliohms. Below 4v the curve flies off the top of the chart.

peopleselectronics wrote:I feel like the answer must lie in the graph for VGS and Gate Charge, which shows a flat response from around 2 - 6V. But can someone explain what that means?

You're right, but it's a physics lesson.

In macroscopic terms, imagine a box with pipes feeding into the opposing sides. The pipe that feeds water in is the drain and the one that lets water out is the source.. I know the names are backwards, but this model uses pressure in place of electrical attraction, so all the values get reversed.

Now tilt the box at a 45 degree angle with the drain (which feeds water in) at the bottom. To make the units meaningful, let's put a distance of 10' between the drain and the source.

To get water all the way from the drain to the source with the box sitting at 45 degrees, you'd need enough pressure to generate about 7' of head. Let's say the pressure in our system isn't anywhere near that strong.. it can manage maybe a foot of head. What we end up with is a foot-deep pool of water in the bottom of the box near the drain.

Applying voltage to the gate is like lowering the angle of the box. The farther you lower the angle, the farther the water extends across the gap between the drain and the source. When you get the angle down to about 5-1/2 degrees, the edge of the puddle reaches all the way across the box to the source, and water can start flowing.

That 5-1/2 degree angle corresponds to Vgs(ON). The Vgs(ON).MAX spec is equivalent to saying 5-1/2 degrees is our design goal, but it will definitely work at 4-1/2 degrees.

The amount of water you can send from drain to source depends on the amount of water above the source. The pool is still only 1' deep, but as the angle drops below 5-1/2 degrees the source goes underwater. The deeper it goes, the more water can flow.

Mosfets work roughly the same way using electrical attraction rather than water pressure. The drain has a positive voltage, so it attracts electrons from the surrounding silicon. The halo of electrons around the drain doesn't extend across the channel though, so there's no path for current to reach the source. Applying positive voltage to the gate pulls electrons from the drain toward the bottom surface of the gate, which is also the upper surface of the channel.

The flat zone in the Vgs/charge graph shows the process of 'tilting' electrons from the drain to the gate. Once you get above about 6v (for this device), the effect of the gate's electrical field outweighs the effect of the drain's electrical field. At that point you start pulling electrons up from deeper parts of the substrate, and to do that you need more electrons (charge) in the gate.

peopleselectronics wrote:I think I am basically asking - how do I know for sure when a MOSFET is gonna turn fully ON from the datasheet

It depends on what you mean by 'fully ON'.. for lower current applications where a couple of ohms won't hurt you, 2.5v is plenty. If you're trying to pump 30 amps through the thing, a couple of ohms would be completely unacceptable.. the voltage drop would be 60v (and the power dissipation would be 1800 Watts). In that case, you'd want to go to at least 4.5v, where Rds(ON) is about 20 milliohms. Then your voltage drop would only be about 600mV (power dissipation 18 Watts). If you want to be even more efficient, take Vgs all the way up to 10v and get that 8.7 mohm guarantee. At 30 amps, the voltage drop would only be about 260mV (power dissipation 7.8 Watts).

Broadly speaking, for this device, 2.5v works, 4.5v is good, 6v is great, 10v is the point where it isn't worth trying any harder, and 20v is the point where you have to go buy a new one.

When you void a product warranty, you give up your right to sue the manufacturer if something goes wrong and accept full responsibility for whatever happens next. And then you truly own the product.