Step down converter LM2596 with voltage spike on output when powering up, destroys subsequent circuits

Question

I use this step down converter with input 30V DC and output set to exactly 5V. At the output I connected an ESP8266 (WEMOS D1 mini, has 3.3v regulator on its board) and an MP3 player module (DY-SF5W.) One thing is, I experience sparking on the input side of the step down converter module whenever applying power to it. But more concerning, I destroyed my load at the output (the MP3 player module DY-SF5W.)

I set the output to exactly 5V and connected the ESP8266 and the mp3 module to the output of the step down module. After powering up the step down converter 2 or three times (30V DC at the input), the mp3 module was destroyed. The microcontroller on the mp3 board smoked and you could see it glowing. The voltage was definitely set to 5V, and the ESP8266 was still working ok. I then connected another MP3 player board/same model, which first worked ok, but again after 2-3 times powering up the input of step down converter, again the uC of MP3 player board smoked exactly as the first one. The ESP8266 again survived.

I then monitored the output of the step down converter closely with oscilloscope, all looked fine, but after setting time base to 25ns, I could see that at the moment when connecting the input voltage to the step down regulator, its output VCC spikes to plus 18V and down to minus 10 volts, but only 25-50 ns in length, and then back to 5V. The question is, is it possible that this very short Vcc change is dangerous to certain parts, and why I do not find more results when searching google for damages due to that? Even with a 4700uF cap at the output of the step down, I sometimes get those spikes. Another step down converter module of same type shows the same spikes. How can I securely get rid of them?

Answer 1

The spike problem may be due to a "lag" in feedback (resistor-dividing) circuit. It means there's a delay between feedback voltage and output voltage. If you have added a capacitor in FB (Feedback) pin, remove it. If the problem still exists, add a small cap (in nF range) between FB (feedback) pin and output voltage.

Answer 2

When you connect your power supply, a spark happens (due to the charging of buck converter input capacitors.) If you will check on the oscilloscope this voltage goes up to 300 - 500 volts for a fraction of a second. It causes the destruction of your LM2596 buck converter.

To protect, change your circuit to the following:

^{simulate this circuit – Schematic created using CircuitLab}

Use a 2 watt resistor at least. I use this technique in my E-Bike when powering aux at 48V and it works.

Answer 3

after setting time base to 25ns, I could see that at the moment when connecting the input voltage to the step down regulator, it's output VCC spikes to plus 18V and down to minus 10 Volt, but only 25-50 ns in length

If there are decoupling capacitors on that rail voltage, it is very suspicious to measure a 50ns spike that goes 13V over the output voltage. I bet that's common mode noise due to the spark and inrush current being displayed by the scope. See this answer for details.

Note these cheap "LM2596" modules are usually fake and counterfeit. Sometimes they don't even bother to fake the National logo and font right, and the chip isn't even a fake LM2596 since it runs at the wrong frequency. Probably some other chip being relabeled. Eventually the low quality, high-ESR output cap will die and output voltage ripple will go out of control, but that shouldn't happen when it's new.

Solution: replace counterfeit DC-DC with good brand, see this mouser search link for example. You can find canned DC-DCs for a few $.

Now your ESP8266 survives but your MP3 player dies. The fact there is some 30V around hints that's a power supply for an audio amplifier that is connected to the MP3 player. Presumably this signal connection also has a ground.

So there is another possibility: a ground loop. You say there is a spark when connecting LM2596 power input, which means you do it by hand with wires. If the power connection is established before the ground connection, or the power ground connection gets loose, then current will find its way through the signal ground between the MP3 player and the power amp. Depending on the path the current takes on the board this may damage your device, or cause it to latch up and burn. So you should make sure ground is not interrupted, check for loose breadboard or connector contacts, that kind of thing.

Answer 4

Are you using the LM2596-Adjustable or the LM2596-5V version? The measurement might be wrong or the curent changes. This regulator supplies as much current as you request. I have had problems burning strips of LEDs, because there was no current limitation.

Also these chinese devices are not very good.

Answer 5

Seems like a surge/transient, ESD problem or maybe your supply to the step down converter gives a overshoot during stepload or just isn't stable.

Anyway a proper varistor or serveral ones should clamp the over voltage if done correctly. I would also advise you to put TVS diodes on the output, if the component is disconnectable(as it seems they are) to avoid ESD.

Answer 6

I put a resistor of 22 Ohm in series to the input of the step down, which removed the biggest spikes at the output, and also the sparks when connecting input power to the step down. now the spike is max 7 volts for 50ns, and no more negative voltages at output of step down. maybe I use an additional voltage limiting Z Diode circuit at the output to prevent those very short over voltages. Will try that. Regarding the ground path remarks, the mp3 board was not connected to other grounds, it simply had a speaker connected to the amp output. In fact there were only three wires to the mp3 board: GND, VCC, serialTX from ESP8266.

Thanks to all for the valuable remarks