Can a fixed output buck have too much output capacitance?

Question

I'm working on a design with this buck regulator:

tps62130

I'm using the 5V fixed output voltage version. I'd like to have significant capacitance on the 5V output because the device can be powered through another source (separate from the buck entirely) also feeding the 5V line. I'd like to throw on something around the order of 220uF tantalum. I've read plenty of warnings about causing loop instability with too much output capacitance on buck regulators, in the datasheet and in other questions like this:

can-there-be-too-much-capacitance-on-a-buck-regulator-output

My question is since I'm using the fixed output voltage version that doesn't use the feedback pin, is it still susceptible to loop instability due to the output capacitance? The datasheet is pretty clear about recommended output inductor/capacitor selection, but the document covers all versions of the buck - including the adjustable output ones that use the feedback pin and I'm sure require the specifically tuned output caps. It's not clear if that applies to all versions of the regulator.

And if the capacitance would cause a problem, my follow-up question would be how do you deal with that situation? Is there a recommended way to isolate the two power sources so capacitance on the other source doesn't affect the buck, or any other way people have dealt with it?

Thanks for any any insight!

Answer 1

I believe, and someone can correct me if I am wrong, that a buck converter has a natural tendency to oscillate, and that proper feedback counters this, but larger LC aggravates it.

To illustrate my point, consider a buck converter as a square wave generator followed by an LC filter, followed by a load.

^{simulate this circuit – Schematic created using CircuitLab}

Here the square wave generator is "ideal". It sources or sinks current at any voltage, with 0 output impedance. There is no diode loss, etc. The frequency is 100kHz, and it the output is either 0-5 volts. Since it has a duty cycle of 50%. So it has a DC component of 2.5V. There is no feedback. Proper filtering should remove most of the AC leaving a steady 2.5V for the output.. According to the simulation it comes very close to that, except at start up, and a few "glitches".

Again, there is no feedback at this point.

Now, if we change the filter capacitor from 220nF to 2.2uF, we can see that the output is much less stable.

And, if we make the capacitor even larger, 22uF, we get this:

Adding capacitance decreases the square wave frequency ripple, but it makes the LC circuit more susceptible to longer term ringing at the LC resonant frequency.

Using feedback to adjust the duty cycle of the square wave may compensate for the tendency for the LC components to ring. But higher C creates a higher burden for the feedback loop.

As mentioned, this uses an ideal square wave generator. An actual buck converter does not have symmetrical characteristics at the high and low parts of its duty cycle. My experience, is that this compounds the problem of ringing, and one can get a sustained ring that never decays, even with a fixed duty cycle. In the 5V to 2.5 volt buck converter described, if the square wave generator is replaced by actual components like an FET and Schottky diode, the sustained ringing (never decays) may be as high as 100mV when using a fixed duty cycle.

Edit: I ran some simulations with a Schottky diode and P-channel Mosfet. I did not see the sustained ripple I have seen before. However, again, the smaller capacitor had more ripple, but better transient response. Here is circuit I used.

^{simulate this circuit}

With a 220nF capacitor, I had this output.

When the capacitance was increased to 22uF, I saw a much higher peak transient error! (But not the sustained ringing I had described).

I am wondering whether the sustained ringing that I recollect was with feedback, rather than a fixed duty cycle. Although I don't at the moment know how to reproduce that sustained ringing, it is something you may want to watch out for. Perhaps someone else has seen it, and can report.

Hope this gives you insight into some problems associated with excessive filter capacitance.

Answer 2

Here is an excerpt from the datasheet for your proposed Buck regulator. I recommend that whenever you use a regulator you take the time to read the datasheet carefully.

The external components have to fulfill the needs of the application, but also the stability criteria of the device's control loop.The TPS6213X is optimized to work within a range of external components. The LC output filter's inductance and capacitance have to be considered together, creating a double pole, responsible for the corner frequency of the converter (see Output Filter And Loop Stability). Table 4 can be used to simplify the output filter component selection. Checked cells represent combinations that are proven for stability by simulation and lab test. Further combinations should be checked for each individual application. See SLVA463 for details.

The table shows combinations up to 200uF. If you want to go higher than that, you have to simulate and/or build and test to confirm stability. Note that this particular regulator is internally compensated so you do not have the option to adjust the external compensation network.

You should be able to simulate your design using TI's web bench tools.

https://www.ti.com/design-resources/design-tools-simulation/webench-power-designer.html

I have no affiliation with TI. Nor have I had problems with them. They are one of several perfectly good options for power converters.