TPS54202 switching regulator shorted from VIN to GND

[amaningd]

I've been trying to replace an LM317 linear regulator with a switching regulator in a circuit that takes 24V and powers a 7.5V, 0.9A lamp. I originally had to step the 24V down to 12V with a fixed-output DC-DC regulator; the 12V then supplied the LM317.

I chose the TI TPS54202. I wired it up on a breadboard with a SOT23 breakout board according to the datasheet, trying to keep the layout as tight as possible (Schematic and breadboard layout images attached at the end; the connection to the lamp's not shown in the picture.)

I've been able to get the circuit to work briefly with three ICs I tried, but after 10-20 minutes, each IC developed a short between the VIN and GND pins. I verified this by removing the ICs from the breadboard and checking continuity. 

(edited to add: does the chip really not need a reverse-biased Schottky diode? Similar regulators I found seem to require one.)

Thanks, everyone!

[Niklas]

The TI part is a synchronous step down, which means that it uses a MOSFET to emulate the freewheeling Schottky.
The absolute maximum voltage rating is 30 V, so not much above the nominal supply voltage. Based on the picture of your setup I suspect that the parasitic from the lab board, the wires ans the long component leads, will cause overshoot and ringing. This might be enough to exceed the maximum input voltage and cause breakdown.

[amaningd]

Thanks, Niklas - that really helps! I was wondering if, well, the IC just wasn't "breadboard-able" for some reason. 

Some questions:

• I know I can place components much more tightly on a PCB and use ground planes and wide traces to to reduce trace length/impedance. Would this help reduce the overshoot?
• Re: the long supply leads that connect to the breadboard at the top right, should I expect the same issue to occur on a PCB which might be routed well but uses long supply wires (e.g. to an AC mains wall supply)?

   

[Manul]

Rule: never try to build switching converter on a breadboard.

You need PCB with proper layout. Usually datasheet gives a recommendation on that. If not, ensure that you have good, low impedance ground and try to reduce size of current loops and make them intersec as little as possible.

If space allows, use a small electrolytic capacitor at the input, which helps supress ringing. For additional input protection you might consider a TVS diode too.

[Alex Eisenhut]

Yup, this is a use case for cheap and cheerful Aliexpress all-done-for-you switcher modules.

[ArdWar]

• 500kHz of hard switching rarely pairs well with solderless breadboarding. At least do it Manhattan style, or in some case tightly laid out perfboard may suffice if pinouts allow for it
• Buck converter rarely work well with excessive parasitic inductance in the switching loop. You may alleviate that by soldering ~10uF MLCC cap directly between VIN and GND pins, but I won't hold my breath on that
• It may not looks like it for this style of packaging, but the pins and planes/traces they're connected to play a big role in heatsinking the switcher. Breakout boards often don't provide sufficient mass/surface area for effective heatsinking
• Using 28V part on 24V input is rather tight even for a decent layout. You have to make sure everything is very well designed and controlled for that.
• Don't apply the full brunt of input voltage or output load for your PSU design first power up. That's almost guaranteed effective and efficient magic smoke release mechanism. Use reduced input voltage just enough for normal operation (e.g. ~12Vin current limited to 10mA at first, then probe around to see everything is as expected)