Bizarre behavior grounding constant current regulator SW pin

[amaschas]

Hey all, I've run into a really weird behavior with a constant current LED driver circuit I'm testing on a bread. Basically what is happening is that the circuit is only outputting a fraction of the power that it should, but I discovered accidentally that if I touch the grounding clip of my oscilloscope probe (even with the oscilloscope off!!) to the SW net of the circuit, it suddenly starts working properly. I see the correct current output, and I also see the expected pulses from the SW pin at the correct duty cycle.

The solution I am designing for requires a 28V input and drives 8x LEDs with a Vf of 3.15 for a total of 25.4V/0.36A output, but when I breadboard out the circuit, I get 18V @ 0.01A. I'm using this IC https://www.ti.com/lit/ds/symlink/lm3402.pdf, and this is the schematic of my breadboarded circuit:



I believe that the issue I'm having with the circuit is noise related, because when I probe the SW net (without the grounding clip connected) I see this on my oscilloscope.



However, when I attach the ground clip I see the waveform I expect with the appropriate duty cycle.



From what I understand, all that is occurring is that the SW pin is being connected to earth ground through the oscilloscope grounding clip. I actually sacrificed an IC to see what would happen if I simply grounded the pin, and I got the expected magic smoke. What exactly is happening when I connect the grounding clip?

I've added photos of the breadboarded circuit with and without the IC mounted to see if there's a mistake I'm not seeing.

[jmelson]

You really can't build these circuits on solderless breadboards.  They run at very high frequencies.  Your caps CIN1 and CF1 need to be very close to the IC pins.
Jon

[amaschas]

You really can't build these circuits on solderless breadboards.  They run at very high frequencies.  Your caps CIN1 and CF1 need to be very close to the IC pins.
Jon

So I'm not really trying to build this thing for any long-term purpose, I'm just attempting to work out some confusion I had from the design math to make sure that my input/output power will actually work with this IC. I know that using a breadboard to lay out a SMPS is going to result in a lot of noise, though I have successfully tested out other switching ICs on breadboards before. There is noise, and you would never use them for anything permanent, but it works to figure out if you've chosen the right passives and stuff like that.

What is really puzzling to me here is that connecting earth ground to the SW node of the circuit suddenly makes everything work correctly. I've actually just connected the earth ground of my power supply to the SW node and it starts outputting 360mA as designed.