TPS54540: Low Frequency noise at light loads

Ben,

40mV of ripple seems too high to be Eco-mode pulse skipping. You can see the expected ripple for this mode in the EVM users guide (around 10mV). My thought would be stability. How did you calculate your compensation values?

Please also share the VOUT waveform while also showing SW. I'd like to see how the switch node behaves for these VOUT waveforms.

-Sam

Hi Sam, Thanks for your reply.

I got the compensation values from the Webench. Here is the project link:

https://webench.ti.com/appinfo/webench/scripts/SDP.cgi?ID=667779377AC83207

I'm not a pro at this, so the only major thing I changed from the original suggested design, was to lower the switching frequency down closer to 1mHz. Then new compensation and inductor values were automatically calculated. 

Here is the output vs switch node at 0.3A (noisy)

and 1A (Low frequency noise gone)

Hi Sam. thanks for the suggestions.

I tried putting an electrolytic on the output earlier, but still the same problem. I left the ceramics there too though. Do you think I need to try it without the ceramics?

I also tried the putting a small capacitor across R5 earlier, and whilst it reduced the severity of this ripple by about 5mV, it didn’t do anything to stop it happening. Even with values up to 750pF.

Below is my layout for this prototype board. I have omitted part of the board that has components for LC filters which I haven't found helpful here. Bottom (green) layer is mostly 0V except for a thermal pad under the diode. Just to note, I have 2 other versions of this design that don’t exhibit this behaviour. 

Variant 1 is +15V in and +12V out on the exact same PCB design. It is stable from 0A to 4.5A

Variant 2 is +15V In and -12V out (buck-boost-inverting) on a similar layout. It is stable from 0A to 2.5A

Ben,

You can try it without the ceramics but this will increase your ripple in the non-issue condition. Give it a try and see if it helps.

The layout looks pretty good (thanks for the labels as well!).

Could you try placing a 0.1uF cap connecting VIN to GND. Place it directly over the BOOT trace and place it as close as possible to the IC. This will minimize the parasitic inductance in the loop from VIN to the cap to GND.

Another thing I'm thinking is to switch to the TPS54540B. This design may unfortunately be in a state that exhibits unexpected operation for the TPS54540. The B version is more robust and might solve this issue.

-Sam

Hi Sam,

The 0.1uF cap super close to the chip does help a little bit with noise, but doesn't fix the problem.

I will have a look out for the TPS54540B and see if this fixes the problem. 

Thanks for your advice

Ben,

Sure thing. Let me know what happens when you get parts in.

FYI these threads lock 30 days after the initial post. If you have more questions after this 30 day window, create a new thread and refer/link back to this one for context.

-Sam