I'm trying to build a buck circuit to power an ESP-2866 and neopixel LEDs for a project. I'm using the fixed 5 V version. The LEDs are powered from 5 V and there's a 3.3 V linear converter downstream that delivers power to the micro.

I've followed the datasheet parameters and also simulated the circuit with TI's Webench, but I can't get it to work. I get no output from the LM2596 and both the IC and the inductor get extremely hot. The parameters used are:

  • Input: 10 V nominal (12 V max, 9 V min)
  • Output: 3 A max (but this is just peak, the LEDs are off most of the time)

This is my schematic:

LM2596 circuit schematic

And this is the layout of my PCB (top and bottom):

Top layer

Bottom layer

Summarized BOM (all recommended by Webench design):

  • L1: Coilcraft MSS1210 AEC-Q200 33μH 20% 5.9A
  • C_IN1: 16V 39μF ESR 27mΩ
  • U5: LM2596SX-5.0/NOPB
  • D_REG: 200V 10A 900mV @ 10A DPAK Schottky
  • C_OUT1: 4.7μF ±10% 10V 1400mΩ @ 100kHz -55°C ~ +125°C CASE-A_1206 Tantalum
  • U4 and C8 are the 3.3V linear converter and its input filter cap.

(Side question: does it make any sense to have a filter cap after an output filter cap? And should I isolate the buck with a diode here?)

I've tried to have thick traces and keep them short. The LM2596 ICs, the caps and coils were from Mouser. I've tried replacing them in case I fried them soldering or testing, but nothing. I even tried a different LM2596, one from XLSEMI in case I got a batch of bad ones and still no dice.

I don't have any shorts (that I've been able to find). Pin 1 receives 12 V fine (12.3 V in fact, a bit high but within tollerance?), but the output on pin 4 is 0.3 V. I don't know what else to test or if it's just an error in my layout.

I'd appreciate any pointers or help.

  • 2
    \$\begingroup\$ Maybe the linear regulator or one of the caps has failed shorted. Check a spare blank PCB for shorts, then try removing the LDO regulator and replacing the caps. I agree with CFC that the caps seem a bit light for those currents, but I don't think that's your immediate problem (though it's possible that high spikes on the output has killed the LDO). \$\endgroup\$ Commented Apr 20, 2020 at 3:00
  • \$\begingroup\$ Is there any way to test if its the LDO? There was a minimum load on the system when I tested, about 0.2A, so the amount of current wasn't sufficient to burn it out, i think. \$\endgroup\$
    – Reven
    Commented Apr 20, 2020 at 5:03
  • \$\begingroup\$ Remove it from the board or try to measure if it is drawing the current from trace voltage drop or from it getting hot. \$\endgroup\$ Commented Apr 20, 2020 at 5:16
  • 2
    \$\begingroup\$ Replacing the linear regulator worked. I left the caps the same, as the simulator gave me those values and they were ok for my application. \$\endgroup\$
    – Reven
    Commented May 24, 2020 at 20:03
  • \$\begingroup\$ Can you tell the type of the 10A DPAK Schottky, provide (access/a hyperlink to) a datasheet? Is it a single diode package? \$\endgroup\$
    – greybeard
    Commented Jan 21, 2023 at 7:47

3 Answers 3


The input and output capacitor need to be much higher.
Input = 680 μF
Output = 220 μF

At 3 A you will certainly need a heatsink. This regulator burns a lot.

Replace 1N5824 with SK56. You can check the working design with LM2576 which is almost the same, except it works at 52 kHz so it requires a bigger coil.

enter image description here


Good start brother. There are two things wrong in the design.

1. Circuit design.

  • Use minimum 2pcs 47 μF input capacitors in parallel.
  • Use 2pcs 220 μF output capacitors in parallel.

2. PCB Design

  • The feedback pin 4 of buck converter to be connected after filtration of output capacitors. But you have connected the feedback just after the inductor.

The feedback loop got unstable, that's why it's heating up.


Beyond the other answers, your ground path isn't good.

If you follow the ground path between your input cap and output cap:

enter image description here

Since you have multi-layer and a ground plane on the bottom, add many via to improve that.

Also, the feedback should be after filtering as Deepak stated.


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.