I'd like to drive a resistive heater using an IC operating at 3.3 V. The heater may need to dissipate up to 18 W, and will be powered by a 28.8 V battery. The IC is able to generate PWM signals, but driving the heater should be as little noisy as possible. Finally, the driver shouldn't dissipate too much energy.

To prevent switching the high current flowing through the resistor, and thus EM noise emission, I thought about putting a simple RC low-pass filter before the actual driver, which would smooth its input signal.

Now, for the driver itself, using a MOSFET would be easy, however they dissipate more power than bipolar transistors, so a Darlington pair might be more appropriate. Then, the issue of the high base-emitter voltage arises, since the PWM signal will have a 3.3 V amplitude.

What is the appropriate way to do this?

  • 9
    \$\begingroup\$ FETs don’t “dissipate more power than bipolars” that statement is either blatantly false, or at the very least circuit dependent and requires more qualification. \$\endgroup\$ Feb 23, 2019 at 15:22
  • \$\begingroup\$ @EdgarBrown The resistive nature of MOSFETs leads to power losses quadratic with the current, while the fixed voltage drop of BJTs leads to linear power dissipation. Then, FETs dissipate more power than bipolars, at least asymptotically. Are the constants involved the key, or I am being wrong elsewhere ? \$\endgroup\$
    – Spirine
    Feb 23, 2019 at 15:49
  • 10
    \$\begingroup\$ “quadratic with the current” is only an issue if the quadratic curve is above the linear one, and then it’s only an issue for the static losses not the dynamic ones. A MOSFET with a 5mOhm Rds_on requires 40A to equal the static losses of a bipolar and, as dissipation is quadratic, would have 1/2 the dissipation at 20A. And this is for ideal bipolars. If your statement was true CMOS technology would not be the dominant one, precisely due to power dissipation. \$\endgroup\$ Feb 23, 2019 at 15:56
  • 3
    \$\begingroup\$ What does "as little noisy as possible" mean? Where (what circuit node) are you concerned about having noise in? How much noise is acceptable ("as little as possible" is not a useful spec)? What frequency bands of noise cause problems? What frequencies is your PWM source able to operate at? \$\endgroup\$
    – The Photon
    Feb 23, 2019 at 15:56
  • 2
    \$\begingroup\$ Suppose heat required is less than about 100 C. A RC filter ahead of a MOSfet driver results in the MOSfet(s) dissipating heat - can you mount the MOSfet(s) driver as part of the assembly that delivers heat? You can likely use a RC time constant that only filters the gates rise & fall edges of PWM from your microcontroller - how much determines the heat delivered by the MOSfet(s). \$\endgroup\$
    – glen_geek
    Feb 23, 2019 at 16:43

3 Answers 3


Resistive heater doesn't need to be turned on and off very frequently to control it. Try using a long switching period, say 1 second, while changing the "on" pulse lenght to control it's temperature. If on/off transitions still produce too much noise then try to slow them down. Thus, the power dissipated on MOSFET will be reduced significantly. You can check the whole thing using pulse generator and MOSFET. Once you determine parameters you can implement them in actual circuit.


Working on this problem you're going to encounter a number of tradeoffs and for each problem, you encounter you'll be able to solve it a couple ways introducing new problems.

In the comments people are describing efficient solutions because really big transistors are heavy and need heat sinks and airflow to cool them.

We're going to assume your heater dissipates 18W when connected to 27.8V so you have enough overhead to be able to switch it and still meet your target. This means we can find it's resistance solving P=V^2/R to find R=V^2/P = 43Ohms. If not, then you'll have to do some sort of voltage conversion to drive the heater properly.

If you're driving a resistive heater with a 43Ohm resistance at 50% with an appropriate analog drive wave your worst case power output is going to be 25% of the total from the switching element (regardless of the type) 4.5W is a lot, but you can get packages and heatsinks to handle it. If you do you can filter the gate input with a RC circuit as you've described and you'll see very little switching noise. If you do so with a MOSFET (which is a decent way to do it) it will dissipate the same amount of power as a darlington pair but the MOSFET might be less suited to dissipating it properly. Don't use a MOSFET or a BJT's rated maximum voltage and current to compute rated maximum power. It's a separate limit and is often defined by various thermal limits on the package.


I used this NPN darlington transistor to drive a resistive foil Heater at 1KHz PWM directly from FPGAs 3.3V IO with a small series resistor. The transistor has pretty high hFE and rated upto 80W: https://www.onsemi.com/pub/Collateral/TIP100-D.PDF


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.