I often see schematics of basic buck converters. The majority of these schematics use a P-MOSFET as high side switch.

Why is this design preferred over a low side, N-MOSFET buck converter?


simulate this circuit – Schematic created using CircuitLab

Would this design work? The load is a 12 V fan, 3.6 W. The MOSFET would be driven directly from the microcontroller in PWM, 5 V logic, 32 kHz carrier frequency.

  • \$\begingroup\$ If I turn off the mosfet, how can the load get 15V? What do you mean by flipped? \$\endgroup\$
    – Francesco
    Sep 21, 2017 at 10:33
  • \$\begingroup\$ I've never seen such a buck converter. I'm not 100% sure that it won't work, it seems like the ground is switched by the mosfet and the diode, while the input remains connected to the output. I would actually use nmos instead of pmos, because then you have fixed potential at the source. The advantage of this circuit with respect to the classical is that if it works you are spearing the boostrap circuit. \$\endgroup\$
    – judoka_acl
    Sep 21, 2017 at 10:37
  • \$\begingroup\$ I dont see any flaw in this circuit, except that load is not referenced to the ground but on positive, instead. \$\endgroup\$ Sep 21, 2017 at 10:39
  • 1
    \$\begingroup\$ You really shouldn't accept a answer so quickly. Now you may never know what other people might have said. It's best to wait 24 hours at least. That way everyone around the world has a waking day to answer. \$\endgroup\$ Sep 21, 2017 at 11:34
  • 2
    \$\begingroup\$ @Marko: The flaw is the high dV/dt of the load's common mode voltage. While the load itself will be fine, this is never going to pass the radiated emissions test. Not even close. \$\endgroup\$ Sep 21, 2017 at 11:36

4 Answers 4


Yes, that works.

The advantage is that the low side switch is easier to control, since its input is ground-referenced.

The downside is that the load is not ground referenced. If you are sure you have a floating load, then this is a very valid thing to do. I usually drive solenoids with a similar circuit, for example. This topology also works if the load is the primary of the transformer in a flyback switcher.

For example, here is a snippet of a schematic I'm currently working on. The product is a piece of industrial equipment.

Note that you can flip the inductor and the load. That fixes one side of the load to the positive supply, which reduces the common mode voltage swing of the load.

In this case, I added a deliberate inductor, even though the load is sufficiently inductive to smooth out the individual pulses. The reason for L6 and C30 is to filter the voltage swings on the SolValve- wire. Without these two components, that wire would carry the full switching pulses. That would cause a lot of RF emissions.

Note the Schottky diode to catch the flyback current pulses. Schottky diodes are good for this as long as the voltage isn't too high. 24 V is well within the range where a Schottky diode makes sense.

You might wonder why I'm worried about pulses when the solenoid being driven is rated for 24 V, and that's also the available supply voltage. I could just turn on Q6 to turn on the solenoid valve. However, that takes a lot of power. I plan to turn on Q6 for about 500 ms to initially activate the solenoid, then fall back to a lower average current by using PWM. The PWM duty cycle will be chosen to ensure the holding current thru the solenoid, as opposed to the initial activation current. Many relays and solenoids are specified to require less current (or voltage) to keep them activated than it takes to initially activate them.

The main advantage of this topology is how it's easy to control the low side switch. In this case, the VALVE signal is coming directly from a 0 to 3.3 V microcontroller digital output. This particular FET is rated for 37 mΩ maximum on-resistance with 2.5 V gate drive. At 285 mA, it will only dissipate 3 mW. That's not enough for you to notice the temperature increase by touching it with a finger.

  • \$\begingroup\$ thank you Olin. I guess that by reducing common mode voltage you aim at reducing EMI, correct? \$\endgroup\$
    – Francesco
    Sep 21, 2017 at 11:29
  • \$\begingroup\$ @Fran: Yes. --- \$\endgroup\$ Sep 21, 2017 at 11:30
  • \$\begingroup\$ ok I got it. Now I was thinking in radiated emissions. You mentioned that, for obvious reasons, the design as-is would fail the test. Could you please explain the difference, In terms of EMI, between your circuit and the one I posted? I am interested to understand if is a matter of voltage floating or current path also? \$\endgroup\$
    – Francesco
    Sep 21, 2017 at 11:53
  • \$\begingroup\$ + 1 for your nice answer. At least you have closed mouth to those that were saying that this concept of buck can't work. \$\endgroup\$ Sep 21, 2017 at 14:08
  • \$\begingroup\$ This reminds me of the basic microcontroller based relay driver circuit. In that circuit, we put a free-wheeling diode across the inductive relay coil. \$\endgroup\$ Mar 20, 2019 at 7:26

I second Olin's advice, so here's your shopping list:

At 300mA IRFR3910 is overkill, since you have 2 fans, pick any cheap SO-8 dual FET with low Qg for low gate drive current. Add gate resistor to slow down the edges to avoid EMI. At 32kHz you don't need 10ns switching time, 1µs will be just fine.

Use a schottky diode for fast recovery (also less EMI in discontinuous mode).

Shielded inductor.

For low output voltages you can also reduce the frequency... but you're unlikely to use very low output voltages, as the fan wouldn't start.

  • \$\begingroup\$ Thank you. The reason we want to go at 32 khz is to eliminate any possible audible noise that could be generated by the switching \$\endgroup\$
    – Francesco
    Sep 21, 2017 at 12:05
  • \$\begingroup\$ I was thinking to drive the 2 fans in parallel. \$\endgroup\$
    – Francesco
    Sep 21, 2017 at 12:21

What about the feedback to regulate the output?As the load is not ground referenced a direct divider from output will not work.You will have to use optocoupler to shift the reference to ground.

  • 1
    \$\begingroup\$ Welcome to SX.EE. This is not a full answer, more a clarification. in this case, please use the comment functionality instead of answer. \$\endgroup\$
    – gommer
    Mar 8, 2018 at 12:01

I've designed buck converts with this exact configuration and it works very well. The main difference, application wise, is that you have to run two wires to your load. Aside from that, it's easier to control, calculations and performance are the same - even better - because it's easier / cheaper to find low resistance N-channel MOSFETs.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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