Capacitor voltage rating safety margin?

[solderfreelead]

I have a noob question. I'm building a DC-DC boost converter circuit and need a 22uF output capacitor. The maximum expected output voltage is 18V. Is it ok to use a multilayer ceramic cap rated for 25V? The 35V or 50V rated ones are significantly more expensive and I want to keep the cost down for such a simple circuit. Thanks.

[Gyro]

If you're talking about a ceramic 22uF capacitor then it is almost certainly Y5U dielectric. These suffer very significant capacitance reduction with increasing voltage.

You need to consult the datasheet and check the capacitance curve - I suspect that , by the time you have taken capacitance loss into account, the voltage rating will be high enough that it ceases to be a concern.

Alternatively you may need to specify a higher capacitance value, or multiple capacitors to achieve the actual capacitance that you require at the specified voltage. In that case, 25V is just about adequate. It will probably be better to specify 35V though to benefit from less capacitance reduction.

[Siwastaja]

The converter output voltage is regulated and shouldn't have overshoot or spikes.

MLCC caps do not require derating for safety or reliability. Especially low-voltage parts tend to withstand severe overvoltages for unlimited times (of course not recommended, though).

25V rating is definitely OK.

But, do note the DC bias effect. Choose a capacitor which has the actual capacitance vs. voltage curve available. For example, TDK parts have that info available by part number on their website.

Note that physically larger sizes tend to have better actual capacitance, but this is not automatic. Look at the actual curves. Too large, and you have too much inductance, something you want to avoid in the output of the boost converter (the cap is part of the hard switching loop). In that case, you'd better off building it by paralleling large number of small capacitors, so that you can optimize the layout - at least part of the total capacitance is very close to the switches.

Something like this may happen:
You look at a 4.7uF, 25V capacitor in, say, 0805 package. You look at the curves and find the capacitance is down to 20% rated at 18V - this is 0.94uF.
Then you look at a much cheaper 2.2uF, 25V capacitor in the same package size. Again, the curves say capacitance is down to 40% rated at 18V - this is 0.88uF.

So now you can pick the latter! I sometimes create an Excel sheet where I compare multiple different solutions (different parts, different number of caps in parallel) for the same total actual capacitance per price.

Now, the appnote / datasheet writer (assuming you got the circuit from an appnote) probably means that you need a "typical 22uF rated MLCC" (because 22uF doesn't sound like a calculation result, but an actual E series component value), and is assuming a certain DC bias effect. It would be much more useful if they told you the actual capacitance they mean. It's very likely they have calculated something like 10uF and then handwaved and tested a 22uF capacitor. That would be fairly typical.

But I bet that circuit is just stupid. I wouldn't put a 22uF MLCC in the output of a 18V boost converter. It's either going to be physically massive and expensive, or if not, then the rating won't hold water (DC bias effect would be around -90%). I would definitely use smaller caps in parallel. 4.7uF 25V and 10uF 10V, in 0805 packages, are some of my go-to sizes for low voltage converters.

[schmitt trigger]

In addition to the useful answers provided, please understand that ESR will be the main contributor (at the switching frequency) for the actual ripple voltage.
That is the reason you sometimes see several paralleled capacitors on the outputs of SMPS converters.

What Siwastaja recommends about using a spreadsheet to find the optimum combination of capacitance/ESR/cost is an excellent idea.

[solderfreelead]

But I bet that circuit is just stupid. I wouldn't put a 22uF MLCC in the output of a 18V boost converter. It's either going to be physically massive and expensive, or if not, then the rating won't hold water (DC bias effect would be around -90%). I would definitely use smaller caps in parallel. 4.7uF 25V and 10uF 10V, in 0805 packages, are some of my go-to sizes for low voltage converters.

I'm ordering parts on lcsc and apparently Samsung does seem to make 22uF 25V caps in 0805 packages. (https://lcsc.com/product-detail/Multilayer-Ceramic-Capacitors-MLCC-SMD-SMT_SAMSUNG_CL21A226MAQNNNE_22uF-226-20-25V_C45783.html)
However, I can't seem to find any characteristic curve in their 84 page long datasheet(https://datasheet.lcsc.com/szlcsc/Samsung-Electro-Mechanics-CL21A226MAQNNNE_C45783.pdf) so I have no idea what the capacitance vs voltage curve looks like.

I got the 22uF value from the datasheet of the boost converter so it's probably quite arbitrary but I don't really know how to calculate the proper value for my needs. I found a calculator on adafruit. I'm not sure if it's helpful or not. (https://learn.adafruit.com/diy-boost-calc/the-calculator)

[solderfreelead]

The boost converter ic I plan on using is the HX3608 I found on lcsc. It has a fixed switching frequency of 1.2MHz I believe. The 22uF cap is from the typical application circuit shown below.

[T3sl4co1l]

Ceramics in that value, expect to need, oh heck, freakin' huge ones (2220?) I think.  I usually see 1206-1210 sizes for 5V, let alone 18.

Ceramic voltage rating has basically nothing to do with the capacitance under bias; you have to read the datasheet for that.

Smaller ceramics with aluminum polymer, or just the latter, should be attractive.

Tim

[Siwastaja]

The boost converter ic I plan on using is the HX3608 I found on lcsc. It has a fixed switching frequency of 1.2MHz I believe. The 22uF cap is from the typical application circuit shown below.

Looks very typical. These low-voltage high-freq integrated switcher datasheet example circuits always tend to assume all-ceramic output caps, and yes, they don't specify the capacitance needed, but the "nameplate capacitance" instead, because, that's how they test their circuits as well, with real components.

The only issue becomes, unless they give you a part number, you can't know whether the capacitor they use has -50% or -90% capacitance at that DC bias. Or whether they even know that exactly.

So make sure you don't use the worst possible part, but there usually isn't need to try to reach 22uF actual capacitance either. Pick a X5R (or X7R if you plan on running at higher temperatures) from a proper manufacturer with C under DC bias curves available, and make an assumption, for example, accept a part that has -60% at your DC bias. (-60% is actually not a bad spec at all, although it sounds horrible. -80% is commonplace, even in the "better" dielectric parts such as X7R!)

In other words, the circuit is 99% likely designed and tested with actual capacitance around or less than 10uF.

For safety and reliability, look for MLCC cracking. Shortly put, avoid board flexing due to mounting screw stress, and avoid excessive heating while soldering: use proper footprints with thermal reliefs, use a modest amount of tin, solder quickly. Limiting yourself to 0805 size max makes this risk very much smaller as well.