ESP32 Forum

So I recently came across this design challenge while building the designs for a product where I have to power an ESP32-S3 module from the 16V RMS AC output of a transformer. So far what I can tell is that its rectified with a shottkey FBR, and somewhat smoothed with some large 4700uF caps, but there's also the fact that the device consume huge amounts of power in small bursts.

My initial design was too use an LM2596-5V alongside an AMS1117-3.3 for a relatively cheap yet somewhat ripple free power source. What's Espressif's thought on scenarios like this? Do they recommend a single fast buck converter, or a buck+ldo combo? If so what chips are recommended?

It may be beneficial to point out that the design has 10 PWM outputs (5xMCPWM + 5xLEDC) and heavily relies on BLE connectivity to a phone no more than 2m away...

Sounds good. In general, HF ripple is the thing to take care of as it affects radio sensitivity. A buck+ldo combo rarely is a bad choice, but given you're talking about short-range BLE, I'd also happily put a switch mode converter by itself here.

Personally i'd just stick one of these in there for your 5v reg. Ya, its a bit more in cost than that old dog of a LM2596, but by the time you add the cost of the diode, inductor, caps its going to add up. Plus this is an isolated converter so your totally isolated from the power side of things.

SPBW03F-05
https://www.digikey.com/en/products/det ... 05/7707003

Follow that up with your 3.3v Reg and you'll be good to go.

Ya, its a bit overkill, but these are my go to power supply's.

The combination of a switched DC/DC converter and a low drop linear regulator is golden, in my experience. The DC/DC brings the voltage down to a "manageable" level, and does that very efficiently. From there (usually 5V) you can smooth out most of the introduced RF noise by the linear regulator, which doesn't have to do handle "large" voltages now (which would only make it dissipate a lot of power). Dependent on whether I need 5 V on the board by itself, I usually set the DC/DC converter to even lower a voltage, say 4.2 V, just the amount the LDO needs to operate.

Also be aware that DC/DC converters have a maximum input voltage. And that if you use longer feeding lines, the DC/DC circuitry may induce a voltage (on the input side) that is significantly higher than the supplied voltage, for a very short time, when powering up. This may be sufficient to kill the DC/DC control chip. Been there, done that I'd advise to add a TVS diode on the input side of the DC/DC converter, a few volts higher that the expected voltage. TVSes are dead cheap. For those taking power from a battery that's nonsense of course, but for your setup, I'd call it smart

And as others have learned: The ESPs tend to produce fast current transients of a few hundred mA when using Bluetooth and/or WiFi. Make sure you have enough smoothing capacitance between the buck converter and the ESP.

I usually set the DC/DC converter to even lower a voltage, say 4.2 V, just the amount the LDO needs to operate.

Doesn't that kind-of defeat the purpose of the LDO? Any transient dip in voltage from the DC/DC would pass straight through the linear regulator.

I usually set the DC/DC converter to even lower a voltage, say 4.2 V, just the amount the LDO needs to operate.

Doesn't that kind-of defeat the purpose of the LDO? Any transient dip in voltage from the DC/DC would pass straight through the linear regulator.

The LDO isn't there to fix voltage dips, the DC/DC converter (equipped with sufficient capacitors) already does that The LDO is there to smooth out RF switching noise.

I am not completely sure it's really required to remove all RF noise when feeding a ESP32, so the LDO could go in theory, but I often need a 5V rail anyway.

For those that didn't already know:

Don't use "just" electrolytic capacitors in places where the "signal" is varying (DC with ripple) at somewhat higher frequencies than 50/60 Hz. This often the case. If you suspect RF ripple, add a small solid state (e.g. ceramic) capacitor, in parallel with the electrolytic capacitor. Mostly 100nf is sufficient. This will short circuit the RF ripple before it can erode the electrolytic capacitor.

(At least) half of any noise/ripple will be negative w.r.t. the set target voltage, so the "headroom" needs to be at least as large as the expected noise amplitude (plus the minumum LDO drop). But generally I agree: Use the efficient DC/DC for as much of delta-U as possible and minimize the drop over the LDO.

I have grown an aversion against linear converters since the days (way back...) I had to use 7805's to derive 5V from a 15V power supply (this was required, couldn't simply use another power supply here), for a harddisk. Which of course it couldn't do, 10V * 1A = 10W dissipation . So we added multiple big power resistors, which felt even worse to do

I am really interested in experiences of people feeding ESP32's (or dev boards, which is not really the same, of course) from just a (modern) DC/DC converter. I can imagine it works perfectly fine. The power dissipation in a linear regulator doesn't feel right to me, I'd rather avoid that.

A little background: most of my ESP's are powered by 19V laptop power supplies, which are small and can be very power efficient. So I really need to use DC/DC converters.