1
\$\begingroup\$

I have designed a step-down converter PCB (4-layer) with MCP16301. It converts voltage from 24V to 5V.

When I connect it to power source and then connect a 100ohm resistor (50mA@5V) it works normally.

But when the load is applied before I power it up MCP16301 dies instantly and shorts Vin to Vout (Vout is 0.5V lower than Vin - probably a voltage drop across internal mosfet).

No other components are damaged. When MCP16301 is replaced it works again. I've already destroyed 3 of them.

Schematics is an exact copy from datasheet: enter image description here Components:

input cap 10uF 0805 35V

inductor 22uH

boost cap 100n 0603 50V

schottky diode 60V 1A

1n4148 diode 100V 300mA

PCB layout (gnd plane hidden): enter image description here with ground plane: enter image description here

I've been troubleshooting this for a couple of days so any ideas/solutions would be greatly appreciated.

Note: PCB was powered by connecting an alligator clip (not a proper switch) and there was a small spark. Power source is a Chinese lab power supply.

Edit:

I followed Andy aka's advice and added a zener diode (30V) but the chip still got fried.

Luckily i was able to get an oscilloscope this time. Here are the photos (because of stackexchange link limitation i can't post clickable links)

(blue - inptut voltage, yellow - output voltage; 10V/div)

There is a 2us 60V spike at the beginning (despite 30V zener diode) so I guess zener diode is not fast enough?

Edit2:

I added an electrolytic capacitor (100uF, ~0.5ohm ESR) as suggested by @Peter Smith. It now works and the voltage is also a lot cleaner (ignoring some high frequency ripple): enter image description here

\$\endgroup\$
4
  • \$\begingroup\$ The zener diode will be fast enough but it may have also become destroyed by the energy of the pulse. \$\endgroup\$
    – Andy aka
    Commented Jan 3, 2020 at 15:08
  • 1
    \$\begingroup\$ I answered a question about input inductive issues a while back (and I refer to the LTC3630 datasheet): electronics.stackexchange.com/questions/242896/… \$\endgroup\$
    – Peter Smith
    Commented Jan 3, 2020 at 15:21
  • \$\begingroup\$ I did a quick (very rough) calculation and I think the source inductance is about 2.5nH, so the R in the linked answer is something well below 1 ohm; I would start with 1 ohm in series with 39uF and record the result. \$\endgroup\$
    – Peter Smith
    Commented Jan 3, 2020 at 15:42
  • \$\begingroup\$ @PeterSmith thank you for your suggestion and linked question. Problem is now solved. \$\endgroup\$
    – nekineki
    Commented Jan 3, 2020 at 18:44

2 Answers 2

Reset to default
1
\$\begingroup\$

PCB was powered by connecting an alligator clip (not a proper switch) and there was a small spark. Power source is a Chinese lab power supply.

I've seen this problem before on switching regulators and I suspect that the answer may lie in the inductance of the power supply and cable feeding the chip. On connecting the alligator clip there is an "inrush" current to charge the input capacitor and feed current to the load. That inrush current will reduce quite rapidly and, any parasitic inductance in the supply feed wiring will produce a kick-back voltage that may take the input voltage above the absolute maximum limit of 30 volts.

When the load is not initially connected, that inrush current will be smaller.

The situation I saw it on was when an LC filter was inserted before the regulator and I was performing an output short circuit test. The test showed that the switching regulator was current limiting (as per the specification) but, on removing the output short, it fried. I correctly realized it was the input inductor and lived without it.

You might try adding a 27 volt zener diode across the regulator power input terminals.

\$\endgroup\$
1
  • \$\begingroup\$ unfortunatly zener diode didn't help. please see edit. \$\endgroup\$
    – nekineki
    Commented Jan 3, 2020 at 14:57
0
\$\begingroup\$

Could you try to place a 33pF capacitor in parallel to R3 in order to modify the compensation?

The response might be too fast for the system, so this is why there is a current spike, the response needs to be slowed down.

\$\endgroup\$
0

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