I'm building a kind of CAN bus IO module with an ESP32. This module will control the taillights of my electric conversion project.

What is a robust (reverse polarity protection, overvoltage protection etc) way of stepping down the 11-14V to 3.3V?

I'm planning on using JLCPCB's assembly service. I was looking at a few LDO's, but I honestly have no idea what type I need, in addition to the capacitors required, and what they're for. I am running 6 relays with NUD3160 drivers, drawing 900mA maximum (maximum for the NUD3160) when powered on.

TL;DR, I need a very robust way of stepping down automotive 12V to 3.3V for a microcontroller

  • 1
    \$\begingroup\$ Have you looked at reference designs from semiconductor companies? If you need stable output during cold-cranking you may need additional complexity. \$\endgroup\$ Commented Nov 13, 2022 at 18:57
  • \$\begingroup\$ Like I said, I want as robust of a system as possible. Eventhough there is no cold cranking in an electric vehicle haha. But the biggest problem is that I have no idea where to start, some say a switching IC, some say an LDO, maybe something else? \$\endgroup\$
    – Lolimpol
    Commented Nov 13, 2022 at 19:06
  • \$\begingroup\$ LOL, good point. Start with limiting the maximum transients (negative and positive) and then you know what the SMPS must withstand. For low power, maybe not much more than a diode and a TVS. There are complex approaches such as the LM74700-Q1 (stocked by LCSC, btw, but probably overkill). \$\endgroup\$ Commented Nov 13, 2022 at 19:14
  • \$\begingroup\$ I have no clue what transients are, let me google. But I hear the word overkill and this sounds great to me! (BTW I am in my 3d year of a bachelor electrical engineering... my school is bad lol) \$\endgroup\$
    – Lolimpol
    Commented Nov 13, 2022 at 19:20
  • \$\begingroup\$ Just use a regulator that is listed for use in "Automotive Applications". For best efficiency a switcher type is best but not absolutely required. \$\endgroup\$
    – Nedd
    Commented Nov 13, 2022 at 19:33

3 Answers 3


You absolutely need a switching regulator for this. 12V to 3.3V is a drop of 8.7V, and you should expect to see about 250mA max draw on an ESP32. That puts the power dissipation in a linear regulator at around 2.2W. In reality, automotive 12V varies a lot, and it isn't unusual to see 14-15V while the vehicle's engine is powered on. That increases the worst-case voltage drop to 11.7V, which corresponds to a power dissipation of nearly 3W.

The other issue you'll likely run into is the ripple and noise on the line. Automotive supplies are extremely noisy. As such, you'll want to ensure that you have enough ripple rejection to keep that noise off your ESP32's supply.

I've previously designed a board for a quite similar purpose to yours. I used an XL1509-5.0 buck converter to get down to 5V, then an AMS1117-3.3V LDO to drop from 5V to 3.3V.

The XL1509's datasheet provides a reference design and tells you all the parts you need. The bill of materials for my buck converter design was:

  • Input capacitors: 2x RVT1V221M0810 220uF Electrolytic (C3340), 1x CL31A106KBHNNNE 10uF 50V MLCC (C13585)
  • Output capacitors: 1x RVT1V221M0810 220uF Electrolytic (C3340), 1x CL21A106KAYNNNE 10uF 25V MLCC (C15850)
  • SS34 Schottky diode (C8678)
  • SMDRI127-680MT inductor (C9907)

The numbers in brackets are the LCSC part numbers.

Here's my schematic:

enter image description here

enter image description here

The parts with the red crosses are do-not-populate, so just ignore them. You also don't need D1.

The output of the XL1509 goes straight into the input of the AMS1117-3.3V LDO, with a CL21A106KAYNNNE 10uF capacitor between the output and ground. The AMS1117 has very good ripple rejection and is very cheap.

The ESP32 was decoupled with 1uF and 100nF MLCCs very close to the VCC pins.

For reverse polarity protection, I'd recommend just using a Schottky diode. Your current demands are small so you don't need anything fancy like an ideal-diode MOSFET.

For overvoltage protection, add a 20V Zener diode clamp over the input. The XL1509 can take up to 40V, so the clamp will kick in way before.

You might also want to consider a TVS diode for ESD protection.

  • \$\begingroup\$ Alright I will try to figure this out, however since it is an IO module, I'm running 6 relays with a NUD3160, it has a max draw of 150mA, 150*6= 900mA, so combined with the 250mA of the ESP, and a tiny bit for the SN65HVD230DR transceiver, I'm well into the amp ranges, will this circuit be ok with that? (Sorry for not mentioning this) \$\endgroup\$
    – Lolimpol
    Commented Nov 13, 2022 at 19:24
  • \$\begingroup\$ I can't edit, but I realised all the NUD3160's are supplied with 12v, and the ESP simply pulls the gate high, and the 150mA comes straight through the huge VSS traces. \$\endgroup\$
    – Lolimpol
    Commented Nov 13, 2022 at 20:04
  • \$\begingroup\$ Yeah, the regulator doesn't need to supply the relays so you'd be fine here. The XL1509 can deliver up to 2A with the BoM I showed, and the AMS1113-3.3V should be good up to about 700mA. \$\endgroup\$
    – Polynomial
    Commented Nov 13, 2022 at 20:12
  • \$\begingroup\$ I've added a schematic for reference. \$\endgroup\$
    – Polynomial
    Commented Nov 13, 2022 at 20:14
  • \$\begingroup\$ The reason why TVS is pretty much mandatory in automotive electronics is not so much ESD as spikes and general nastiness coming through the supply, from alternator and/or starter etc. You'll also want a big bulk cap somewhere to protect against surges. 2x200uF might be a bit low. \$\endgroup\$
    – Lundin
    Commented Nov 14, 2022 at 15:20

Here is a protection circuit that uses an ideal diode chip.

enter image description here

If you don't need to the 'ideal diode' low drop efficiency then you can replace the chip and MOSFET with a series diode. The voltage regulator should be able to withstand transients of +45V or more (I would suggest 60V as conservative) since the TVS cannot clamp perfectly (and it's not acceptable to reduce the breakdown voltage because it would no longer meet the specifications). You can pick a switching regulator chip that meets your output current requirements and can withstand worst-case input voltage. You can also go beefier on the TVS without changing the voltage- the part just gets bigger and a bit more expensive.

The actual automotive transient specs are not freely available (legally) but manufacturers of TVS diodes often reproduce the important bits in their application notes.

  • \$\begingroup\$ This looks good, I'm not so concerned with cost. The whole car project is in the 20k range, so a few extra euro's for a more robust system is allowed in my book. What are the advantages of replacing the MOSFET with a diode? I dont quite understand the low drop effiecency \$\endgroup\$
    – Lolimpol
    Commented Nov 13, 2022 at 19:29
  • \$\begingroup\$ @Lolimpol The only advantage would be price. The LM74700 is some 'deluxe' version ensuring that the MOSFET doesn't take a blow, it's a more rugged version of this common solution: electronics.stackexchange.com/questions/552381/…. As for why a MOSFET substrate diode is used instead of a plain diode: much lower voltage drop. \$\endgroup\$
    – Lundin
    Commented Nov 14, 2022 at 15:11
  • \$\begingroup\$ As for the TVS then I believe 33V is a common choice for 24V systems and in case of 12V systems we should pick something lower, around 18V. Also this one is way too weak (600W peak pulse power), I've failed industrial/automotive EMC tests when we used similar. Go with >1600W rated. \$\endgroup\$
    – Lundin
    Commented Nov 14, 2022 at 15:14
  • \$\begingroup\$ @Lundin The higher voltage spec for 12V systems comes from avoiding failure with a double-battery startup. Maybe that can be ignored for an EV. \$\endgroup\$ Commented Nov 14, 2022 at 15:40
  • \$\begingroup\$ Aah, hehe... I guess that is a valid use-case. Well, still don't pick a 600W one. The more rugged the TVS, the better. There's another rare(?) use-case scenario: clumsy mechanic shorted battery + against chassis. A rugged TVS could take that blow for a short period of time. \$\endgroup\$
    – Lundin
    Commented Nov 14, 2022 at 15:55

Since you have a decent budget purchase an automotive rated SEPIC converter. That will cover you down to maybe 3 volts or so and also do what is needed during transients, load dump, etc. I have worked with battery requirements from 6V to 24V operation normaly was required in this range for at least 60 seconds. The reason for this recommendation is your budget will not begin to cover the cost to qualify your "automotive" design. However downstream from the converter the rules change and become much less stringent.

"Load Dump" is a pseudo generic term. By that I am saying each manufacturer has there own specification of what the load dump looks like and how sever it is. It is also dependent on what other protection is in the vehicle. Some OEMs also use a protected alternator.

You can search for load dump testing you will find a lot. Shaftner is one of the manufacturers of compliance testing equipment. You can check what they offer

  • \$\begingroup\$ I can't find many of these on JLCPCB, but they sound like they can handle a lot of current, but also a lot of passive components. Also I believe load dump isn't a big issue since we don't have an alternator but a DC-DC converter. What I'm worried about is the power steering motor and brake booster pump turning on and off with their big relays. However I honestly have no clue what kind of noise/transients they create... \$\endgroup\$
    – Lolimpol
    Commented Nov 13, 2022 at 20:34
  • \$\begingroup\$ These would not be general purpose retail type items. You would probably have to get them from one of the automotive suppliers. Check with your DC/DC source. I believe the transients would fall under Load Dump. \$\endgroup\$
    – Gil
    Commented Nov 13, 2022 at 20:59

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.