PowerElecHub

Gate Drive Calculator: Gate Power, Resistor Split and Miller Time

Gate-drive power Q_g·ΔV·f_s for unipolar or bipolar rails, how it splits between the driver, the external turn-on and turn-off resistors and the MOSFET's internal gate resistance, peak and Miller-plateau gate currents, and the energy each resistor takes per edge.

Gate drive and charge

Read Q_g from the gate-charge curve between your two rails, not the headline value at a different V_GS.

Resistances

With a single gate resistor, enter the same value for R_on and R_off. R_HI and R_LO are the driver's output resistances from its datasheet.

Total gate-drive power

209 mW

Q_g·ΔV·f_s, ΔV = 19.0 V

Dissipated in the driver

29.3 mW

14 % of the total

Peak gate current on / off

2.00 / 3.33 A

ΔV over the path resistance, no loop inductance

Miller time on / off

35.6 ns / 15.5 ns

Q_gd / plateau current

Where the power goes

Driver, turn-on (R_HI)
16.5 mW
Driver, turn-off (R_LO)
12.8 mW
External R_on
55 mW
External R_off
36.7 mW
Inside the MOSFET (R_g,int)
88 mW

Currents and energy

Plateau current, turn-on
842 mA
Plateau current, turn-off
1.93 A
Energy in R_on per edge
550 nJ
Energy in R_off per edge
367 nJ
Average current from each rail
11 mA

Turn-on resistor sweep

Raising R_on slows the voltage transition at turn-on and moves dissipation out of the driver into the resistor. The total gate power does not change.

Resistive (MOS-output) driver model. Loop inductance, driver current limits and the shape of the gate-charge curve are not modeled; the on/off split of the total is taken as half and half.

How it works

Every switching cycle the gate is charged from V_off to V_on and discharged back again. Whatever resistors sit in the path, the gate supply delivers Q_g·ΔV of energy per cycle, where Q_g is the gate charge over that swing and ΔV = V_on − V_off. All of it ends up as heat in the resistances the gate current flows through, so the gate-drive power is Q_g·ΔV·f_s. Choosing a larger gate resistor does not reduce this number. It only moves where the heat goes.

Within one path the same current flows through the driver's output stage, the external resistor and the MOSFET's internal gate resistance, so the loss in that path divides in proportion to their resistances. That part is exact. Dividing the total between the turn-on and turn-off paths is not: TI's application report SLUA618A assigns half to each, which is exact for a linear gate capacitance. With a real gate-charge curve the energy stored in the gate when it is on, which the turn-off path later dissipates, differs from ½·Q_g·ΔV, so treat the on/off split as an estimate and the total as exact.

The peak current at the start of each edge is ΔV divided by the path resistance, because the gate still sits at the opposite rail. Loop inductance between driver and gate slows the rise, so the measured peak is usually lower. The current that sets the switching speed is the plateau current: while the gate sits at V_pl the drain voltage moves, and that interval lasts Q_gd divided by (V_on − V_pl)/R at turn-on and (V_pl − V_off)/R at turn-off. A negative turn-off rail increases the second current without changing the first.

Each rail delivers an average current of Q_g·f_s. For an isolated gate supply the power to budget is Q_g·ΔV·f_s plus the driver's own quiescent consumption.

Design procedure

  1. 1Read Q_g, Q_gd and the plateau voltage from the MOSFET's gate-charge curve for your actual rails and a drain voltage close to your bus.
  2. 2Enter the driver's pull-up and pull-down resistances from its datasheet, and the MOSFET's internal gate resistance.
  3. 3Choose R_on for the turn-on speed you can tolerate in EMI and ringing, and R_off, often smaller, for a fast and robust turn-off.
  4. 4Check that the driver's share of the gate power stays well inside its package dissipation at the highest ambient temperature.
  5. 5Size the external resistors for both average power and the energy per edge, and the isolated gate supply for Q_g·ΔV·f_s plus quiescent current.

Formulas

Gate-drive power
P_G = Q_g·ΔV·f_s, ΔV = V_on − V_off
Split within a path (SLUA618A form)
P_drv,on = ½·P_G·R_HI/(R_HI + R_on + R_g,int), P_Ron = ½·P_G·R_on/(R_HI + R_on + R_g,int), same for turn-off with R_LO and R_off
Peak current (no inductance)
I_pk,on = ΔV/(R_HI + R_on + R_g,int), I_pk,off = ΔV/(R_LO + R_off + R_g,int)
Plateau current and Miller time
I_pl,on = (V_on − V_pl)/R_on,path, I_pl,off = (V_pl − V_off)/R_off,path, t_Miller = Q_gd / I_pl
Energy per edge in a resistor
E_Ron = P_Ron / f_s
Gate supply
I_avg = Q_g·f_s per rail, P_supply = Q_g·ΔV·f_s + P_quiescent

Worked example

The defaults describe a SiC MOSFET driven from +15 V and −4 V at 100 kHz, with 110 nC of gate charge over that 19 V swing, 30 nC of Miller charge and a 7 V plateau. The resistances are illustrative: 3 Ω inside the MOSFET, 5 Ω external for turn-on and 2 Ω for turn-off through a diode, and a driver with 1.5 Ω pull-up and 0.7 Ω pull-down. The gate draws 209 mW. The driver dissipates only 29 mW of it; 55 mW goes into R_on, 37 mW into R_off and 88 mW into the MOSFET's own gate resistance. The turn-on edge starts at 2.0 A and the turn-off edge at 3.3 A. On the plateau the currents are 0.84 A and 1.93 A, so the drain-voltage transition takes about 36 ns at turn-on and 16 ns at turn-off. Removing R_on altogether shortens the turn-on Miller time to 17 ns but raises the driver's share to 48 mW and the initial current to 4.2 A; doubling it to 10 Ω gives 54 ns and leaves 24 mW in the driver. The total stays at 209 mW in every case. With a 0 V turn-off rail instead of −4 V, the turn-off plateau current would drop to 1.23 A and the Miller time would grow to 24 ns. These numbers were computed independently in Python and match the calculator.

Common mistakes

  • Using the headline Q_g at 10 V when the gate actually swings from −4 V to +15 V, which underestimates the power by tens of percent.
  • Expecting a bigger gate resistor to cut gate-drive power. It only moves the loss and slows the edge.
  • Forgetting the internal gate resistance, which can be several ohms and takes a large share at turn-off when the external resistor is small.
  • Sizing a small chip resistor by average power alone; check its pulse rating against the energy per edge.
  • Reading the driver's peak current rating as the current the gate will see on the Miller plateau.

Frequently asked questions

Why doesn't a larger gate resistor reduce the gate-drive power?

Because the energy per cycle is set by the charge moved and the voltage it is moved through, Q_g·ΔV. The resistor only decides how fast it moves and where the heat ends up. A larger resistor takes heat away from the driver package and slows the switching edge, which usually costs more in MOSFET switching loss than the driver saves.

Which Q_g should I enter?

The charge between your actual rails, read from the gate-charge curve. A datasheet headline Q_g is quoted at one V_GS, often 0 to 10 V for silicon or a bipolar swing such as −5 V to +18 V for SiC, and can differ from your swing by tens of percent. Many curves start at 0 V; for a negative rail the extra charge below 0 V can be estimated as C_iss·|V_off|, with C_iss taken at high drain voltage. That estimate is approximate.

Is the half-and-half split between turn-on and turn-off exact?

Only for a linear gate capacitance. The split within each path, by resistance, is exact because the same current flows through every element of the path. The split between the two paths depends on how much energy the gate holds when it is fully on, which a real gate-charge curve with its Miller plateau does not make exactly ½·Q_g·ΔV. TI's SLUA618A uses the half-and-half form, and so does this calculator.

Does the peak current tell me which driver I need?

Partly. The peak at the start of the edge is limited by the resistances and by loop inductance, and the driver's rated peak current is usually specified at a particular supply voltage. What actually sets the switching speed is the current on the Miller plateau, which is lower. SLUA618A makes the same point. Check that the driver's output resistances, not just its headline current, give the plateau currents you want.

How were the numbers checked?

The formulas were implemented independently in Python, including a numerical check that charging a linear capacitance through a resistance dissipates exactly ½·Q·ΔV, and the calculator's results match to the displayed digits. The driver-loss split follows the form in TI SLUA618A, section 2.7. A worked example from a manufacturer that matches number for number has not been found yet.

Results are engineering estimates. Verify your design with datasheets, simulation and measurement.

Related calculators