0
\$\begingroup\$

We are using a TPS62740 IC as DC-DC converter.

We are getting ripple at the output in the range of 10 mV to 12 mV, so we are thinking to add a pi (C-L-L) filter at the output of the DC-DC converter.

Image 1

Image 2

Image 3

Image 4

As you can see, the peak-to-peak voltage is around 12 mV, but the frequency is not stable, varying by up to 100 kHz.

The DC-DC converter switches at up to 2 MHz, V(in) = 3.6 V, we are operating with a small load (50 mA) and V(out) = 2.5 V. We want to design a pi filter for it.

Please help me out with this problem.

\$\endgroup\$
10
  • 1
    \$\begingroup\$ Convince me that it's not your oscilloscope probing technique that is giving you this ripple. Have you tried this at full load or the worst case scenario? Are you also aware that pi filters are designed for controlled input and output terminations and that what you are probably wanting is not a pi filter but an LC low pass filter with damping resistor. \$\endgroup\$
    – Andy aka
    Commented Dec 6, 2021 at 12:34
  • \$\begingroup\$ We want to Design Pi filter to reduce the noise at the output. We have connected the load at the output of the DC-DC converter, and we have tested multiple times for the reading's on output to ensure the waveforms. \$\endgroup\$ Commented Dec 6, 2021 at 12:41
  • 5
    \$\begingroup\$ OK if you want a pi filter, then what are the signal input and output resistances required. Pi filters need this to be known. What is the frequency that you want to start attenuating at? By how much do you want the filter characteristic to roll-off at (above the frequency in the previous sentence) in dB per octave or dB per decade? \$\endgroup\$
    – Andy aka
    Commented Dec 6, 2021 at 12:43
  • 1
    \$\begingroup\$ Does the load actually care about a little bit of ripple? If it's a micro, probably not, and it could even make more noise and ripple on the supply than the DC-DC. \$\endgroup\$
    – bobflux
    Commented Dec 6, 2021 at 14:01
  • 1
    \$\begingroup\$ Even a "simple" LC will give you headaches, and that's just a 2nd order. A PI filter is a 3rd order, it will be even worse. \$\endgroup\$ Commented Dec 6, 2021 at 14:27

1 Answer 1

2
\$\begingroup\$

These diagrams look like a regular PFM operation for me, especially after you've mentioned variable frequency. In my practice, there are two nice approaches besides passive filtering:

  1. If power consumption isn't critical - you may add a constant load at your output (fixed resistor) to shift the converter to the PWM mode. It will allow you to get a fixed frequency, plus often it will decrease ripple amplitude. Other part numbers often do have a dedicated "Forced PWM" pin, but with this one, I'm afraid it's the only option
  2. You may push your load point a bit higher (e.g., 2.6 V) and supply it to the 2.5V LDO with high PSRR, or even a voltage reference if the absolute value of the voltage matters (pay attention to the fact that for your output 25 mV is already 1% accuracy!). I've used such an approach for systems with capacitive touch, which are unpleasantly sensitive to such ripple, and with good LDO, it has higher efficiency compared to the first option

In case these options aren't enough and you still need a passive filter - I believe any online calculator (e.g. this one) will do the job, the number of stages will depend on your specs (how much attenuation you'll need)

\$\endgroup\$
2
  • 2
    \$\begingroup\$ Just adding that OP must think also that transient behavior will be changed for varying load, if filter not included in feedback \$\endgroup\$
    – Antonio51
    Commented Dec 6, 2021 at 14:01
  • 1
    \$\begingroup\$ Load transient will suffer, that's for sure. Regarding including an output filter in the voltage feedback - for this p/n TI doesn't provide numbers to evaluate the bode plot accurately, so there is a risk of ruining the system's stability, not worth it :) \$\endgroup\$
    – Looongcat
    Commented Dec 6, 2021 at 14:35

Not the answer you're looking for? Browse other questions tagged or ask your own question.