[FAQ] TPS63020: Phase margin improvement method

Hello Hasegawa,

Please be aware of the following recommendation from the datasheet: The recommended nominal output capacitors are three times 22 μF.

You are using 3 times 10uF which is lower than the recommendation causing most probable the low phase margin.

In addition, it is possible that the used MLCCs do not even get close to the value written on the datasheet. Please check out the so-called DC Bias of the capacitors. MLCCs loose capacitance when used at a DC level. For example, if you are using 6.3V rated capacitors at an output voltage of 5.4V they could get down to 20% of their capacitance or even lower.

Some more detailed information about phase margin measurements:

Phase margin is needed because of the following reasons:
1. Time for the control loop to get back into default regulation value after a disturbance (at 45 deg the load transient response shows one undershoot one overshoot (or vice versa) and the default value is reached fastest according to regulation theory).
2. To keep the system stable even if the components change over time.

The second point is the one that is more unknown and therefore in depth knowledge of the system is necessary to know how much phase margin is ok for the system lifetime.
As the amount of different components on a converter (with all switching components integrated) is low, experience says that 30 deg is in such systems very often good enough over lifetime.

To distinguish if the stability in your system is good enough, please have a look at the worst case load transient response in your system. If this is ok and still has some margin to the system requirements, it should be a stronger decision point than phase margin. Normally small signal disturbances are causing less problems in systems than large signal disturbances.
Especially when you measure phase margin with 5.4V input voltage and 5.43V output voltage, TPS63020 is permanently changing from buck mode to boost mode and back, because you are in the region where the converter changes modes. Therefore you have during a small signal measurement a large signal change, so the measurement result is not very accurate.

So I think for your system it is important to do worst case load transient measurements. Worst case can be on the following points:
mode transition boundaries (buck to buck-boost and buck-boost to boost)
highest and lowest VIN (often the lowest VIN is worse)
highest load transient with fastest slew rate
lowest temperature (because the converter normally reacts faster at low temperature)