Note 20Power Rev 1 Watch the video · 0:19

FET choice

Is the FET with the lowest Rds(on) the best one?

Only with a fast gate driver. In moteus-c1’s 48 V, 10 A half bridge at 30 kHz, the 1.2 mΩ IRFS7530 loses 0.77 W and the 15 mΩ TPN19008QM 1.98 W with 1 A of gate current. At 60 mA they lose 9.08 W and 2.41 W.

The figure drives both FETs from the same gate driver and steps its current down from 1 A to 60 mA. The IRFS7530’s 236 nC gate takes longer to fill at every step, and its turn-on heat climbs from 13.0 µJ to 188 µJ.

The TPN19008QM holds 15 times less charge. Even at 60 mA it is half on for only 48 ns, and its leg’s loss barely moves.

One driver, two FETs

Gate current from 1 A down to 60 mA

Heat in each FET as it turns on, from the gate command, at 48 V and 10 A. The faint lines are the other gate currents. Below, the leg’s whole loss at 30 kHz: darker for its on-resistance, lighter for switching.

The numbers

Turn-on heat beyond plain conduction, and the leg’s loss at 30 kHz. 48 V, 10 A.
Gate current	IRFS7530, a turn-on	IRFS7530, leg loss	TPN19008QM, a turn-on	TPN19008QM, leg loss
1 A	13.0 µJ	0.77 W	0.98 µJ	1.98 W
570 mA	21.1 µJ	1.15 W	1.44 µJ	2.00 W
330 mA	35.3 µJ	1.82 W	2.14 µJ	2.02 W
260 mA	44.3 µJ	2.26 W	2.60 µJ	2.04 W
120 mA	94.5 µJ	4.66 W	5.14 µJ	2.17 W
60 mA	188 µJ	9.08 W	9.91 µJ	2.41 W

The half bridge and its gate driver

One phase of moteus-c1

A motor driver connects each motor wire to the supply or to ground through two FETs, one above the other. Switching between the two thousands of times a second sets the motor’s current.

On moteus-c1 the supply is 48 V, and the switching runs at 30 kHz, its firmware’s default. This page takes 10 A out of one leg.

A gate driver turns each FET on by pushing current into its gate, and off by pulling it out. The DRV8323 sets that current in 16 steps from 10 mA to 1 A. It powers up at 1 A, and moteus-c1’s firmware turns it down to 60 mA to turn on and 20 mA to turn off.

One leg and its driver

Each FET has its body diode drawn beside it. The shaded loop is the high-side FET’s gate current, out of GH and back on SH.

The numbers

Part	What it is
Q1, Q2	Both IRFS7530 or both TPN19008QM, n-channel
Driver	DRV8323, 60 mA to 1 A into a gate, twice that out of it
Supply	48 V
Motor	10 A out of the switch node

Two FETs suit this leg. VESC 4.12 uses the IRFS7530, a 60 V part with 13 times less resistance. moteus-c1 uses the 80 V TPN19008QM.

Low resistance takes a big die, and a big die has a big gate. The IRFS7530 holds 236 nC of gate charge against 16 nC, and its package takes 14 times the board.

The rest of the page puts each pair in the same leg in turn, with the same driver.

The two FETs, to scale

Drawn from each datasheet’s package drawing at one scale, so they compare at true size. The TPN19008QM is shown twice, the second turned over. Its drain pad, like the D2PAK’s tab, carries the drain into the board. Resistance and gate charge are typical, at 25 °C with 10 V on the gate.

The numbers

Sizes from each datasheet’s package drawing. Datasheet typical values at 25 °C and 10 V on the gate.
Part	Package	Size	Drain	Rds(on), typical	Gate charge	Rated	Board
IRFS7530	D2PAK-7	10.1 × 15.25 × 4.45 mm	Tab	1.15 mΩ	236 nC	60 V	VESC 4.12
TPN19008QM	TSON Advance	3.3 × 3.3 × 0.85 mm	Pad underneath	14.7 mΩ	16 nC	80 V	moteus-c1

Two ways to lose power

48 V, 10 A, 30 kHz

A FET loses power in two ways. While it is on, the motor’s current heats its resistance:

P_R=I²R

=(10 A)²×1.45 mΩ=0.15 W

That is the IRFS7530. The TPN19008QM’s 19.1 mΩ loses 1.91 W, 13 times more. R is each FET’s resistance at 100 °C, and one FET or the other always carries the current, so this is the whole leg’s.

Every cycle also costs a fixed energy, while a FET is half on and the current changes hands between the two:

P_sw=Ef

=21.0 µJ×30 kHz=0.63 W

E counts the whole cycle, with both edges, the dead time and the diode’s recovery. The 13.0 µJ turn-on in the opening figure is one part of the 21.0 µJ.

That is the IRFS7530 with 1 A of gate current. The TPN19008QM loses 2.43 µJ a cycle, so 0.07 W. At 60 mA the big FET’s switching costs 8.93 W.

Where the loss goes

Each FET in the same leg at 30 kHz, with 1 A and with 60 mA of gate current, twice that to turn off. Darker is the on-resistance’s share, lighter is switching.

The numbers

At 48 V, 10 A and 30 kHz, on-resistance at 100 °C.
FET, gate current	On-resistance	Switching	Leg loss
IRFS7530, 1 A	0.15 W	0.63 W	0.77 W
TPN19008QM, 1 A	1.91 W	0.07 W	1.98 W
IRFS7530, 60 mA	0.15 W	8.93 W	9.08 W
TPN19008QM, 60 mA	1.91 W	0.50 W	2.41 W

The catch is the gate

Both FETs turning on at 60 mA

Switching loss comes from the time a FET spends half on, and the gate sets that time. The driver fills the gate at a set current, so a gate that holds more charge takes longer:

t_g=Q_gI_g

=236 nC60 mA=3.9 µs

The TPN19008QM’s 16 nC fills in 0.27 µs.

Turning on at 60 mA

IRFS7530
TPN19008QM

The high-side FET from its gate command, at 48 V and 10 A. Half on runs from 0.5 A through the FET until the voltage across it is down to 4.8 V.

The numbers

Half on: from 0.5 A through it until the voltage is down to 4.8 V.
FET	Current starts	Half on	Peak heat	Turn-on heat
IRFS7530	886 ns	792 ns	538 W	188 µJ
TPN19008QM	71 ns	48 ns	561 W	9.91 µJ

392

Switching loss

Half on for 792 ns, or 48 ns

Field marks. At 60 mA the IRFS7530 waits 886 ns before any current flows. Then it carries the full 10 A for 792 ns while its voltage falls, heating at up to 538 W. The textbook sketch, E_on≈½VIt_half, gives ½ × 48 V × 10 A × 792 ns = 190 µJ, and the turn-on in Fig. 20d loses 188 µJ. The TPN19008QM is through in 48 ns.

Look-alike. The whole gate’s fill time. The FET heats only while its voltage falls, and that lasts as long as the gate-drain charge takes to fill: t_gd=Q_gd/I_g=73 nC/60 mA=1.2 µs. Fig. 20d has the IRFS7530 half on for 792 ns.

Range. At 1 A the IRFS7530 is half on for 46 ns and loses 13.0 µJ turning on, near the small FET’s 9.91 µJ at 60 mA.

Habitat. Drives that turn their gate current down to tame ringing. moteus-c1 turns its gates on at 60 mA, where an IRFS7530 would stay half on 16.5 times as long as its own TPN19008QM.

393

Which one wins

Every DRV8323 step, 10 A at 30 kHz

Add the two losses for the leg’s whole loss, and do it at every step the driver offers:

P=I²R+Ef

=0.15 W+298 µJ×30 kHz=9.08 W

That is the IRFS7530 at 60 mA, the left end of its line in Fig. 20e.

Leg loss against gate current

IRFS7530
TPN19008QM

Each dot is one of the DRV8323’s steps, with twice that current to turn off. 48 V, 10 A, 30 kHz.

The numbers

Leg loss at 30 kHz, 48 V, 10 A. Turn-off current is twice the turn-on current.
Gate current	IRFS7530	TPN19008QM
60 mA	9.08 W	2.41 W
80 mA	6.88 W	2.29 W
120 mA	4.66 W	2.17 W
140 mA	4.02 W	2.14 W
170 mA	3.36 W	2.10 W
190 mA	3.02 W	2.08 W
260 mA	2.26 W	2.04 W
330 mA	1.82 W	2.02 W
370 mA	1.64 W	2.02 W
440 mA	1.42 W	2.01 W
570 mA	1.15 W	2.00 W
680 mA	0.97 W	1.99 W
820 mA	0.86 W	1.98 W
1 A	0.77 W	1.98 W

394

FET choice

The low-resistance FET wins above 290 mA

Field marks. With 1 A of gate current the IRFS7530 loses 0.77 W and the TPN19008QM 1.98 W. At 60 mA, 9.08 W and 2.41 W. The lines cross at 290 mA.

Look-alike. Pick the lowest Rds(on). Part searches sort by it, and in this leg it is the right pick only above 290 mA of gate current.

Range. Over the driver’s steps the IRFS7530’s loss moves 12 times, from 0.77 W to 9.08 W. The TPN19008QM’s moves from 1.98 W to 2.41 W.

Habitat. VESC 4.12 drives its IRFS7530 from a DRV8302, a fixed driver rated 1.7 A, through 4.7 Ω. moteus-c1 turns its DRV8323 down to 60 mA on and 20 mA off, and uses the TPN19008QM.

395

Why a driver runs slow

The FET that is off, as the other turns on

If a fast gate lets the big FET win, why turn a driver down? A fast edge makes the loop’s inductance ring, and the FET that is off takes the spike. Note 4 follows it through moteus-c1’s copper.

At 1 A the TPN19008QM rings up to its breakdown, 81 V in the model, past its 80 V rating. At 60 mA it peaks at 47.8 V.

The IRFS7530’s 13.0 nF gate slows its own edge, so even at 1 A it peaks at 48.8 V. Its rating is 60 V, though, so a 48 V supply leaves it 12 V.

moteus-c1’s firmware runs 60 mA on and 20 mA off. With that setting the TPN19008QM’s leg loses 3.19 W. A pair of IRFS7530 would lose 23.9 W.

The spike on the other FET

IRFS7530, 1 A
TPN19008QM, 1 A
TPN19008QM, 60 mA

The voltage across the low-side FET as the high side turns on, lined up where the switch node passes 24 V. 48 V, 10 A. The dotted lines are each FET’s rating, the gray dashed line the 48 V supply. 81 V is where the model’s FET breaks down.

The numbers

Lined up where the switch node passes 24 V. 48 V, 10 A. 81 V is where the model’s TPN19008QM breaks down.
FET, gate current	Peak	Rated
TPN19008QM, 1 A	81.2 V	80 V
TPN19008QM, 60 mA	47.8 V	80 V
IRFS7530, 1 A	48.8 V	60 V

Current and frequency move the line

Where both FETs lose the same

Fig. 20e holds for one motor current and one PWM frequency. Set the two losses equal and solve for the frequency where neither FET wins:

f_even=I²(R₂−R₁)E₁−E₂

=(10 A)² × 17.6 mΩ18.5 µJ=95 kHz

That is 10 A with 1 A of gate current. Above that frequency the TPN19008QM loses less, and below it the IRFS7530.

Break-even frequency

One line for each motor current, on log scales. Above its line the TPN19008QM loses less at that current, below it the IRFS7530. The dot is 60 mA at 30 kHz, the turn-on step and PWM rate moteus-c1 uses.

The numbers

PWM frequency where both FETs lose the same, at each motor current.
Gate current	2 A	5 A	10 A	15 A	20 A
60 mA	1.3 kHz	3.2 kHz	6.3 kHz	9.3 kHz	12 kHz
80 mA	1.7 kHz	4.2 kHz	8.3 kHz	12 kHz	16 kHz
120 mA	2.5 kHz	6.3 kHz	12 kHz	19 kHz	25 kHz
140 mA	3.0 kHz	7.3 kHz	14 kHz	22 kHz	29 kHz
170 mA	3.6 kHz	8.8 kHz	18 kHz	26 kHz	35 kHz
190 mA	4.0 kHz	9.9 kHz	20 kHz	29 kHz	39 kHz
260 mA	5.3 kHz	14 kHz	27 kHz	40 kHz	53 kHz
330 mA	6.5 kHz	17 kHz	34 kHz	51 kHz	67 kHz
370 mA	7.0 kHz	19 kHz	38 kHz	57 kHz	75 kHz
440 mA	7.9 kHz	23 kHz	45 kHz	67 kHz	89 kHz
570 mA	9.2 kHz	29 kHz	58 kHz	86 kHz	114 kHz
680 mA	10 kHz	35 kHz	71 kHz	106 kHz	135 kHz
820 mA	11 kHz	40 kHz	82 kHz	123 kHz	162 kHz
1 A	12 kHz	45 kHz	95 kHz	141 kHz	187 kHz

396

Break-even frequency

More current favors less resistance

Field marks. Resistance loss grows with the square of the current, switching loss roughly in step with it. The line climbs with current. With 1 A of gate current it sits at 12 kHz for 2 A and 187 kHz for 20 A.

Look-alike. One crossover gate current for every drive. Fig. 20e’s 290 mA holds at 10 A and 30 kHz. At 30 kHz the 5 A line crosses at 590 mA, and the 20 A line at 150 mA.

Range. At 140 mA and below, the 15 mΩ FET loses less at 30 kHz at every current here, from 2 to 20 A.

Habitat. moteus-c1 turns its gates on at 60 mA and runs at 30 kHz, above every line. At 60 mA the highest, for 20 A, is 12 kHz. Its 20 mA turn-off only lowers the lines. At 10 A it moves the line from 6.3 kHz down to 2.4 kHz.

397

Choosing a FET

Two FETs, one 48 V leg

The answer turns on how hard the driver pushes the gate and how often the leg switches. The key sorts them.

Key to FET choice

Which FET for your drive?

1aThe driver turns the gate on with 140 mA or lessGo to 2
1bWith moreGo to 3
2aThe PWM runs at 30 kHz or fasterThe one with less gate charge
2bSlowerGo to 3
3aThe PWM runs above your motor current’s line in Fig. 20gThe one with less gate charge
3bBelow itThe one with lower Rds(on)

For these two FETs on a 48 V leg, at 2 to 20 A. Another pair has its own lines, from P=I²R+Ef. moteus-c1 turns its gates on at 60 mA and runs at 30 kHz, so it ends at 2a. Its 20 mA turn-off widens the gap. The TPN19008QM’s leg loses 3.19 W, a pair of IRFS7530 23.9 W.

Pick the FET yourself

Every gate current and motor current

Choose a gate current and a motor current. Against PWM frequency each FET’s loss starts flat at its resistance loss, then climbs as its switching energy is paid more often. Both scales are logarithmic.

Where the two lines cross is the break-even frequency of Fig. 20g.

Gate current

60 mA

Motor current

Leg loss against PWM frequency

At 30 kHz the TPN19008QM loses less

IRFS7530
TPN19008QM

IRFS7530 at 30 kHz: 9.08 W
TPN19008QM at 30 kHz: 2.41 W
Same loss at: 6.3 kHz

48 V, resistance at 100 °C, twice the gate current to turn off. The dots sit at 30 kHz, moteus-c1’s rate.

The numbers

Leg loss at 30 kHz at each motor current: IRFS7530 / TPN19008QM.
Gate current	2 A	5 A	10 A	15 A	20 A
60 mA	1.74 / 0.17 W	4.43 / 0.71 W	9.08 / 2.41 W	13.92 / 5.10 W	18.92 / 8.76 W
80 mA	1.32 / 0.16 W	3.35 / 0.66 W	6.88 / 2.29 W	10.58 / 4.90 W	14.41 / 8.49 W
120 mA	0.90 / 0.14 W	2.26 / 0.60 W	4.66 / 2.17 W	7.17 / 4.71 W	9.78 / 8.22 W
140 mA	0.78 / 0.13 W	1.96 / 0.59 W	4.02 / 2.14 W	6.20 / 4.65 W	8.47 / 8.14 W
170 mA	0.65 / 0.13 W	1.63 / 0.58 W	3.36 / 2.10 W	5.18 / 4.60 W	7.10 / 8.06 W
190 mA	0.59 / 0.13 W	1.46 / 0.57 W	3.02 / 2.08 W	4.68 / 4.57 W	6.42 / 8.02 W
260 mA	0.45 / 0.12 W	1.09 / 0.55 W	2.26 / 2.04 W	3.52 / 4.50 W	4.87 / 7.93 W
330 mA	0.37 / 0.12 W	0.87 / 0.54 W	1.82 / 2.02 W	2.85 / 4.47 W	3.97 / 7.88 W
370 mA	0.35 / 0.12 W	0.79 / 0.54 W	1.64 / 2.02 W	2.59 / 4.45 W	3.62 / 7.86 W
440 mA	0.31 / 0.12 W	0.68 / 0.54 W	1.42 / 2.01 W	2.24 / 4.44 W	3.15 / 7.84 W
570 mA	0.27 / 0.11 W	0.55 / 0.53 W	1.15 / 2.00 W	1.83 / 4.42 W	2.60 / 7.81 W
680 mA	0.25 / 0.11 W	0.47 / 0.53 W	0.97 / 1.99 W	1.56 / 4.41 W	2.30 / 7.79 W
820 mA	0.23 / 0.11 W	0.42 / 0.53 W	0.86 / 1.98 W	1.40 / 4.40 W	2.02 / 7.78 W
1 A	0.22 / 0.11 W	0.38 / 0.53 W	0.77 / 1.98 W	1.27 / 4.39 W	1.84 / 7.77 W

Conditions

Board: One leg of moteus-c1 r1.2, with the loop inductances of its copper from the field runs in Note 4. Each FET pair takes the same place in it.
FETs: The IRFS7530 and the TPN19008QM, as models fitted by hand to their datasheets, within 7% of them except the small FET’s gate-drain charge, 26% high.
Driver: A DRV8323 model built from its datasheet: every step from 60 mA to 1 A to turn on, twice that to turn off, with its drive time, t_DRIVE, at its longest, 4 µs, so the big FET’s gate fills at the set current. moteus-c1’s own setting, 60 mA on and 20 mA off, is two more runs.
Load: 48 V, with 10 A held steady over each edge. 2 to 20 A in Fig. 20g and the explorer.
Resistance: Each model’s own on-resistance in the leg at 10 A, 1.04 mΩ and 13.8 mΩ. That is below the datasheets’ typicals because the driver puts 11 V on the gate, against their 10 V. Times each datasheet’s rise from 25 to 100 °C, 1.40 and 1.38, that gives 1.45 mΩ and 19.1 mΩ.
Lost: Everything both FETs dissipate over one PWM cycle beyond their resistance’s share at 25 °C: both edges, the body diode through the dead time, its recovery and the FETs’ capacitance.
Renders: Blender, one camera. Outsides from each datasheet’s package drawing.
Solver: ngspice: 142 runs of the whole cycle, 28 turn-ons for the opening and Fig. 20d, and 3 for Fig. 20f.

Simulated only. Nothing on this page has been measured on a board yet.

Limits

The IRFS7530 sits in moteus-c1’s small copper loop. Its larger package would need a larger loop, with more inductance.
81 V is the small FET model’s breakdown. From 820 mA up its ring reaches it, so the real peak would be higher.
The driver’s own power isn’t counted: 0.16 W for a pair of IRFS7530 at 30 kHz, 0.01 W for the TPN19008QM.
Whether the DRV8323’s charge pump can keep up with the big FET’s gate charge hasn’t been checked.
Both FETs are taken at 100 °C. Their real temperature depends on how they’re cooled.

Files

fsim/site.pyBoth FETs at every gate current and motor current on this page
fsim/losses.pyThe leg, the loss count and the 100 °C factors
fsim/waves.pyThe turn-ons in the opening and Fig. 20d
fsim/ringing.pyThe low-side spikes in Fig. 20f
render/site_parts.pyThe renders in Fig. 20b
gatesim/fets.pyThe fitted FET models, from Note 4

Sources

Like this note?

Rev 1 · 5 Nov 2026

Rev 15 Nov 2026First published

Each fix gets a line here, with the reader's name if they want credit.

Did it answer your question?

Want more on this?

Report a problem

What should the next note be? Ask for one