Connecting a resistor to Mosfet's gate

[MrAl]

Hi Mike,


You must have missed app note 941A

Part of the reason for using a gate resistor (Rg in series with the gate) is due to the small inductance in the source, which as im sure you know increases with the circuit layout. This is of concern when paralleling MOSFETs but could be a concern for single devices too. The source inductance acts as a counter voltage to the gate (turn on) voltage, which could cause the MOSFET to turn on, turn off, turn on, turn off, etc., which of course is not a good idea for a number of reasons. Maybe we can think of this as the 'Miller' inductance Now in any circuit that contains inductance we can damp out oscillations by simply adding a series resistor, but that usually causes a decrease in efficiency so we try to minimize that while still getting a more well behaved circuit out of the deal.

In a perfect world with perfect components and no stray capacitance or inductance we would want to use a pure voltage source to drive the gate so we can turn on as fast as possible and thus reduce switching power losses. In this world however that doesnt happen because we get side effects that we have to *also* deal with, and that leads us down paths that may cause a slight deterioration in our perfect world design. It's like anything else, when we try to get good performance we many times have to do something that has negative side effects to make the whole thing work in the real world.

To investigate this further, you could try to find the notes that talk about the actual switching waveforms that occur with various types of loads and those values. Those notes include an Rg value too. Otherwise you might try some simpler circuit simulation with a perfect voltage controlled switch and some 'source' lead inductance and see what happens...you'll also have to add gate and drain capacitances too i guess.

In short, the MOSFET is a very high gain device that requires special techniques to control. Think about trying to control a switching regulator circuit with a *pure* inductance (no series resistance) and pure capacitance, keeping in mind that even a small amount of energy into the circuit will cause the LC to oscillate forever. Then add even a small series resistance and see how much more easily it is controlled.

 

[MikeMl]

Ok, careful reading of all of slup169 still does not come up with a blanket recommendation: In every case, add a resistor in series with the gate. Read the checklist at the end; it does not say to unconditionally add a gate resistor!

Note in the discussion on page 10, the calculation to determine what Rg (if any) to add has to do with the source inductance resonating with the gate to source capacitance, after accounting for the intrinsic gate resistance, and the output impedance of the driver. In most of the circuits I have played with, the source impedance of the driver was sufficiently high (in fact I always wished it was lower) such that no additional gate resistor is necessary.

The AN-941 app note does not apply to this discussion because it specifically covers what to do when paralleling Power FETs, which is not what we are discussing here.

 

[MrAl]

Hello again,

This, just like many other electronic circuits, will not have a blanket policy but knowing the theory behind it tells you what *can* happen and knowing what to do about it you'll be able to solve it perhaps. You also noted that many real life circuits already show a gate resistor so that should tell you something. Yes, of course the driver impedance works into the equation too. It can also depend on the type of load being driven.

Paralleling MOSFETs is one use and i thought you might be interested in that but if you're not that's ok too, but someone else may be so i felt it was worth mentioning.

I think by now you know what a gate resistor is there for and if you talk to some other people that have built power circuits you'll see that they too like to use gate resistors because they help damp oscillations. The typical value ends up being around 10 to 22 ohms usually.

As i said before, there are other reasons too though, one being when you wish to use an asymmetrical drive, and doesnt have as much to do with oscillations as it does pure dynamic transistor control.

BTW, app note 941 also talks about the dynamics for a single transistor.