FET "Turn on" speed

11 posts / 0 new
Author
Message

I'm looking for N-channel FETs for a PWM based regulator. To reduce heat dissipated during switching events I want to use an FET that can be turned on and off as fast as possible. When comparing FET specs, what affects the rate at which the FET can be turned on or off?
I thought it was the Total Gate Charge (specified in nC @ a given Vgs), but then there's input capacitance (usually in pF at a given Vds).
Given 2 FETs whose Vds is sufficient and the same FET drive, what spec should I look at regarding how fast the FET can switched? Which one is "faster", one with the lower Gate Charge or lower Input Capacitance?

Thanks.

Jim M., Rank amateur AVR guy.

A useful link for mosfet characteristics
http://www.irf.com/technical-inf...

geoelec wrote:
A useful link for mosfet characteristics
http://www.irf.com/technical-inf...

Thanks. That's great reading. Exactly what I was looking for. And very in depth.

Jim M., Rank amateur AVR guy.

To give you a quicker answer: if you can find identical test conditions, your gate charge is what you care about. The gate charge is what it takes to charge up your input capacitance, but the capacitances involved are not constant with respect to voltage, so it's better to use the charge rather than the capacitance, IMHO. You care most about your miller charge - as that is a measurement of how much charge it takes to take your FET from off to on.

nleahcim wrote:
To give you a quicker answer: if you can find identical test conditions, your gate charge is what you care about. The gate charge is what it takes to charge up your input capacitance, but the capacitances involved are not constant with respect to voltage, so it's better to use the charge rather than the capacitance, IMHO. You care most about your miller charge - as that is a measurement of how much charge it takes to take your FET from off to on.

Right. The graph in the doc that shows a ramp-plateau-ramp... I've seen that on my scope and kind of wondered why that plateau existed. Now I know.

Jim M., Rank amateur AVR guy.

For a given FET, the faster the switching RATE (number of turn-ons per second), the higher the power. This is because a certain amount of charge gets put into the gate on every turn-on. The driver gets rid of the charge (usually dumps it to ground) on the turn-off. Charge per unit time is current. Current times supply voltage is power.

If your driver is really good, then it delivers the charge efficiently and it does not add to things. So, "driver power" simply becomes (mostly) the gate charge tines switching frequency times the supply voltage.

Gate capacitance ADDs to this. There is MORE required charge due to the charging of the gate capacitance. This is Q = C*deltaV where deltaV is the Vmax-Vmin seen by the gate capacitance. This can be small compared to gate charge, or it can be significant.

So, the earlier question asking (effectively) whether Cgate is more important or less important than Qgate, the answer is: "it depends". You can determine it for your self for your application. The Q needed to charge the gate capacitance, alone, is Q = C*deltaV. Compute this Q and compare it to Qgate. THAT will tell you.

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

ka7ehk wrote:
For a given FET, the faster the switching RATE (number of turn-ons per second), the higher the power. This is because a certain amount of charge gets put into the gate on every turn-on. The driver gets rid of the charge (usually dumps it to ground) on the turn-off. Charge per unit time is current. Current times supply voltage is power.

If your driver is really good, then it delivers the charge efficiently and it does not add to things. So, "driver power" simply becomes (mostly) the gate charge tines switching frequency times the supply voltage.

Gate capacitance ADDs to this. There is MORE required charge due to the charging of the gate capacitance. This is Q = C*deltaV where deltaV is the Vmax-Vmin seen by the gate capacitance. This can be small compared to gate charge, or it can be significant.

So, the earlier question asking (effectively) whether Cgate is more important or less important than Qgate, the answer is: "it depends". You can determine it for your self for your application. The Q needed to charge the gate capacitance, alone, is Q = C*deltaV. Compute this Q and compare it to Qgate. THAT will tell you.

Jim

Jim - I believe the gate charge is a measurement of the charge it takes to charge the gate capacitance (Ciss). It sounds like you are saying that you have to put in the gate charge as well as charge up the input capacitance, and I don't think that is right. The two quantities are given separately because Ciss is not a constant term, and the two components of Ciss (Cgs + Cgd), have different deltaVs typically, and are very voltage dependent.

But if I'm mistaken please correct me.

nleahcim wrote:

Jim - I believe the gate charge is a measurement of the charge it takes to charge the gate capacitance (Ciss). It sounds like you are saying that you have to put in the gate charge as well as charge up the input capacitance, and I don't think that is right. The two quantities are given separately because Ciss is not a constant term, and the two components of Ciss (Cgs + Cgd), have different deltaVs typically, and are very voltage dependent.

But if I'm mistaken please correct me.

Nope. I think you're right. The charge is given at a specific Vgs, but the capacitance Icc changes based on the conditions at the drain/source Vds. But I don't need to know exactly. Because 2 FETs I compare would see the same conditions in use. I just want to know which of 2 possible FETs would allow the faster turn on or off time. Thus, the Gate Charge would be the better indicator.

Jim M., Rank amateur AVR guy.

Yes, there are TWO charges involved. Qg involves little or no change in the gate voltage. You have to supply this charge just to get the device switched.

Cg involves charge from gate input Voff to the switching threshold and from the switching threshold to gate input Von.

Usually, Qg dominates. But, if you have larger than average input swing, then just getting the gate capacitance charged can become significant.

YMMV!

Jim

Until Black Lives Matter, we do not have "All Lives Matter"!

ka7ehk wrote:

YMMV!

Jim

Thanks, Jim.
YMMV. Does it ever!.

Jim M., Rank amateur AVR guy.

ka7ehk wrote:
Yes, there are TWO charges involved. Qg involves little or no change in the gate voltage. You have to supply this charge just to get the device switched.

Cg involves charge from gate input Voff to the switching threshold and from the switching threshold to gate input Von.

Usually, Qg dominates. But, if you have larger than average input swing, then just getting the gate capacitance charged can become significant.

YMMV!

Jim

Jim, what do you mean by Cg? Do you mean Ciss?

Again, I'm pretty sure that Qg is a measure of the charging of Ciss. IRF seems to agree with me: http://www.irf.com/technical-inf... (see page 10)