For MLCC (multilayer ceramic capacitors) you need to be aware of the not well publicised Voltage Coefficient of Capacitance or Voltage Dependence of Capacitance. Without going into detail, for a given capacitance and package size, to ensure that you get the most capacitance at a given DC bias across the capacitor, you should choose the highest voltage rated part in your chosen package size that you can find. For example, when placed across a 3.3V supply, a 1206 packaged 100uF, 10V MLCC will generally have a higher residual capacitance than a 1206 packaged 100uF, 4V rated part. A brief introduction can be found here: [https://www.murata.com/en-eu/support/faqs/capacitor/ceramiccapacitor/char/0005](https://www.murata.com/en-eu/support/faqs/capacitor/ceramiccapacitor/char/0005)<br> <br> More: [http://www.elmac.co.uk/Voltage_coefficient.pdf](http://www.elmac.co.uk/Voltage_coefficient.pdf)<br> <br>

Thank you @andyfierman If I understand you correctly, residual capacitance is bad because it pushed the capacitor more away from its ideal; therefore, we want to minimize residual capacitance.  To do so we should select capacitor voltage rating to be as close to our actual DC voltage in our circuit.  Is that correct?

No. Exactly the opposite. Capacitance is usually quoted at zero DC bias. Residual capacitance is the percentage of the quoted zero DC bias capacitance left as the DC bias across the device is increased up to the rated voltage of the device. For an MLCC of a given capacitance, operating at a given DC bias, you should choose the highest voltage rated part in the largest package size that you can fit in your PCB. (Sorry: in my post above I forgot to say "in the biggest package you can fit on your PCB".) That will leave you with the highest residual capacitance at your required DC bias. The effect is negligible for NP0 and C0G dielectric but very apparent for X7R and X5R dielectric. Y5V and similar dielectrics are next to useless (their thermal performance is dreadful too). The effect is negligible for non-ceramic dielectric capacitors. Please study the articles I pointed you to to understand the VCC or VDC behaviour. For example, compare the residual capacitance at 5V in the DC Bias Characteristic graphs of: [https://product.tdk.com/en/search/capacitor/ceramic/mlcc/info?part_no=C3216X7R1H106K160AC](https://product.tdk.com/en/search/capacitor/ceramic/mlcc/info?part_no=C3216X7R1H106K160AC)<br> <br> and: [https://product.tdk.com/en/search/capacitor/ceramic/mlcc/info?part_no=C2012X7R0J106K125AB](https://product.tdk.com/en/search/capacitor/ceramic/mlcc/info?part_no=C2012X7R0J106K125AB)

If the supplier does not give the dielectric type then try to find it on the manufacturer's site. If you cannot establish the dielectric type then best go somewhere else. For preference only source MLCCs from suppliers or manufacturers that show the VDC or VCC curves.

Oops, where is that brown paper bag again ;)  I confused capacitor residual charge with capacitor residual capacitance.  I googled for residual capacitance but the first hit was talking about residual charge instead. <br> <br>

@Kajetan321, Easily done... Been there, bought the T-shirt! In fact I have a special drawer for them. :)

The reasons for the poor VDC/VCC behaviour of MLCCs today compared to say 30 years ago is interesting too. Basically it is an effect of electric field strength in Ferroelectric dielectric materials. Nowadays it's possible to make finer grained material and place the conductors closer together than earlier on. This means more capacitance with closer spaced electrodes so the electric field strength is higher which in turn creates greater stress in the dielectric so the grain structure and its polarisation (?) is more distorted. As the DC bias increases that  increases the effect even more and so reduces the apparent dielectric constant and so reduces the capacitance. If you compare curves carefully you can see that some caps of a given value have the same capacitance at a given DC bias even if their rated working voltage is different. That's because they probably use the same MLCC chip internally  and are just graded differently during test. For a given capacitance, increasing both the working voltage and the package size tends to invoke a bigger chip with a wider electrode spacing so the electric field strength at the same DC bias is lower so the stress on the dielectric material is lower hence less capacitance reduction. Hence the device is operating further to the right of the generic VDC/VCC curve.

I discovered this when I had to find a replacement for some caps that had gone obsolete and found that the new parts with the same working voltage, capacitance and package size had a much worse VDC/VCC curve. And people think inductors are difficult to work with...