Single value capacitor or two different in parallel for filtering the power supply for microcontroller

Question

I am trying to design power supply for the STM32F446 for my PCB.

I wonder if I should stick to the Datasheet of STM32F446 Page 71 and use two capacitors or be careful about it and maybe just use one 100nF. The picture from Datasheet with suggested values:

I started getting these concerns after checking some other designs, where they just used one 100nF capacitor for this micro-controller. As-well after reading this answer on StackExchange about the effect of 'anti-resonance' when using two different values capacitors. He also referenced to this question where it was elaborated in more details. In these slides they mention that 'anti-resonance' effect can be reduced by reducing ESL.

I am a bit confused by all these inputs. I think the most trusted source is definitely manufacturer Datasheet, but maybe they assume I would be using some special capacitors (for example ceramic X7R), or something else I am not taking into account. And why other designs don't follow this suggestion. It would be great if you can clarify this.

Answer 1

Manufacturer suggestions should just work.

They even mention that removing filtering capacitors might cause incorrect operation of the device.

You should do as the manufacturer says, and then worry if it does not work properly.

Answer 2

The idea in the datasheet is that you have a smaller local bypass capacitor and a larger bulk capacitor located further away (possibly near the power connection). The longer distance will contribute resistance which will damp the antiresonance peaks at least somewhat. In many cases, larger bulk capacitors will be electrolytic, which are even more resistant, although that is not the case here.

In this specific case, since the bulk capacitance value they recommend is tiny and could easily fit into a 0603 or even 0402 package, you could consider simply bypassing with a 1uF cap (or even larger to account for decrease in capacitance at bias if your package size has a significant degree of that). Using only one capacitor will prevent any antiresonance peaks. Since you will not need to use a large package to obtain the recommended bulk capacitance, you will not increase inductance (at least to any meaningful degree) by using a larger value for bypassing.

Answer 3

100 nF in parallel to 1 uF is a recommendation that holds since the 70's.

They are a low-ESR close-proximity energy source for the VDD of your microcontroller.

Without those capacitors, current will be drawn from the closest capacitor or directly from the on-board LDO or DC-DC converter which, in turn, will draw current from the external power supply.

The worst scenario, in terms of EMC compliance, is when the current is drawn from the on-board LDO or DC-DC converter because the PCB tracks are long and act as antennas.

PCB tracks are not efficient antennas but still they irradiate RF energy when fast varying currents flow into them.

Currents drawn by micro controllers contains a lot of frequencies because CMOS circuits switch fast.

It's very important that microcontrollers currents are drawn from local and close-proximity capacitors.

X7R ceramic capacitors are the best choice because the ceramic characteristics are stable over a wide range of temperature.

Every single VDD pin of your microcontroller needs at least the 100 nF ceramic capacitor.

If cost, as always, is important then share the capacitor of 1uF value.

---

Resonances

Capacitors have parasitic components inside and it's not that easy to estimate the impact that they have on the EMC signature of your board.

What people do when they have to go in an EMC lab is bringing with them a certain number of SMD ceramic capacitors and a solder iron.

If EMC problems arise, they change a couple of capacitors on the board and repeat the test.

---

By my experience I can tell you that 470 uF to 1500 uF low ESR electrolytic capacitor as an input filter of the LDO or DC-DC converter of the board solve 90% of the problems.

I love United Chemicon electrolytic capacitors because the are really low ESR capacitors and because I was able to pass the EMC compliance tests if one of my boards thanks to them.