3V3 PSU Redesign. Noise issues

[alexxs88]

Our initial design was using the TPS560430 chip as a step-down converted from a Vin of (12-16V) down to 3V3 with the 3V3 rail looking like this (Vpp is 120mV, sorry for the quality):



Because of the chip shortage, we were forced to replace the TPS chip with the NCV3063DR2G and the following design:



Unfortunately the 3V3 rail is looking much worse, badly affecting the readings of one of the sensors:



I can see in the PSU's chip datasheet that the recommended capacitor on the output is 470uF for 800mA of current. Our design barely goes above 20mA, so the thinking here was that the 150uF should be enough. Is there any way evading this problem? Here's what we've tried so far:

Replacing the 150uF electrolytic capacitor with 3x 47uF ceramic caps stacked onto each other as a test. The performance seems to have improved, although it's still worse than before.



Have ordered 220uF ceramic capacitors for the output which should hopefully make things a bit better.

Is there any other more obvious solution? Neither of us is an expert in PSU design, so we might be missing something really obvious and simple here.

Thank you!

[T3sl4co1l]

Why zoomed out so far?  What's going on cycle-to-cycle?

BTW you can remove the offset by setting AC coupling on the vertical channel.

NCP3063 is a hysteretic controller, so it will generally be noisier than others.  It does at least run faster than the archaic MC34063, and has true current limiting, not the, whatever weird oscillator skewing hack that one had.  If you're forced to use such a controller, a good option is setting it a little high then following it up with an LDO to knock down the noisy/spiky bits.

The TPS560430 is in the traditional SOT-23-6 pinout that you should be able to find substitutes for; particularly if you're willing to add diodes, move to larger chips, etc. as the '3063 suggests.  Although, I don't have any suggestions offhand; none of the alternatives I know offhand are stocked, at Digi-Key at least.

Tim

[floobydust]

I think the low current drain 20mA is an issue, try increasing the load and seeing if the IC struggles less.

[Swake]

Indeed the low current might be the issue here:
- There is a design spreadsheet on the Onsemi website for this chip. Entering 20mA as average Iout the value for L1 jumps to 2.8mH.
- For some reason the buck example in the datasheet defines output riple from 80mA on.

Other things:
The AN calls for a 100nF ceramic C on the input and the output.
You write that Vin is 12V to 16V, your C3 is 16V, I would choose a higher voltage for that C.

[mariush]

Are you limited to that chip?

Digikey has nearly 280k of AOZ6663 regulator, which costs $0.4 if you buy 100 : https://www.digikey.com/en/products/detail/alpha-omega-semiconductor-inc/AOZ6663DI/10258159
Also, 40k of AOZ6662-01 which is just the 2A version of the above chip, $0.35 if you buy 100 : https://www.digikey.com/en/products/detail/alpha-omega-semiconductor-inc/AOZ6662DI-01/10258156

So same footprint, you could basically switch between them as needed.

The -01 versions even have a special low noise mode for when the output is set to under 4v :

The AOZ6662DI-01 features fixed frequency operation at any load when output voltage is set to be lower than 4V.
This makes it a perfect fit for low noise audio application. When output is set to be higher than 4V, Pulse Energy Mode (PEM) is automatically enabled to achieve high efficiency at standby light load. This allows flexible solution to use a single product for multiple power rails with different requirement.

Works up to 18v input (absolute max 20v), high frequency so small inductors, no diode required...

Check the datasheets, they have 12v in, 3.3v out graphs.