How do I remove this power supply noise?

Question

Above is the circuit diagram of my power supply. It converts input 12V DC into 5V DC and then 3.3V DC with linear regulator. I have highlighted it in red color in the block diagram.

So current situation is, whatever noise is there on 12V input voltage shows up on 5V and 3.3V. how do I filter this? How do I improve my design to get clean 5V and 3.3V even with a noisy 12V input.

If I use clean power supply 12v, I am getting best results. But i can't give 12V clean power supply to everyone. Customers can pick up any noisy power supply and it should work with my design.

Below I am attaching two waveforms , Yellow is 12V , Blue is 5V and Pink is 3.3V.

this is zoomed out view

Zoomed in view

If you look at the waveforms, the noise appearing on 12V is carried on to the 3.3V. I was assuming LDO will at least attenuate it more but, it is very high. On the PCB the ground is properly stitched with multiple vias from top to bottom.

How do I get rid of this, what do I have to change or add in my design to improve it?

Note: Input 12V Dc power supply used is a 220V AC to 12V DC adapter.

Answer 1

It seems to happen at harmonics of the line rate. This is showing up as common-mode noise on everything, and because it's AC potential difference between your nominally-grounded body and its local reference it's bound to cause issues with your sensor when your body is nearby.

Where is it coming from? Leakage (parasitic coupling) between the 12v DC-DC primary to the isolated secondary. All line-power supplies have it, good ones manage it better.

As an experiment, try grounding the board to safety (earth) ground to see if the noise goes away. If it does, congratulations, you have leakage. Add a common-mode filter to the 12V or get a better supply.

Another experiment: add a common-mode filter to the line-in on the 12V supply. This would break the AC loop for noise and contain it in the power supply where it belongs.

Answer 2

Good job with posting all your 'scope shots. More useful would be a frequency spectrum capture. That helps identify source of noise more easily.

In the meantime, noise pickup happens most easily through inductive pickup. If your PS wires are long(ish) and loosely strung, they will pick up more noise. Try running PS wires close together, and using twisted pair where relevant. Keep lines short as possible. Throwing capacitors at it doesn't work - as you have seen.

In fact, some of your caps might be increasing the ringing in conjunction with that L !

Answer 3

i would recommend you go to FFT mode on your scope to pin point the frequency(frequencies) of the noise. or randomly you can try using ferrite beads at the output of the buck converter and connect a capacitor of few microfarads to several microfarads(depends on the frequency) from +5V to ground after the ferrite bead.

hope this also helps Ferrite bead position