FT232R chip supply capacitors location

Question

Here I asked a question why to have the capacitors:

FT232R chip and capacitor

But now the problem of their proper setup arises i.e. how to include the excerpt at the bottom of the picture in the whole circuit? Also I have a fundamental question here. Isn't it so that when I connect capacitors in a parallel way I could simply use 1 equivalent capacitor with Ceq = C1 + C2 + C3? Thus in my case I would have only 1 capacitor with C = 10nF + 100nF + 4.7uF =~ 4.8 uF. Also why are 10nF and 4.7uF capacitors polarized and 100nF is not? I read that it's more of a technological requirement that larger capacitors ~1uF+ have to be polarized. Why then 10nF is polarized? Thanks.

The picture is from http://www.ftdichip.com/Support/Documents/DataSheets/ICs/DS_FT232R.pdf

Answer 1

The various size, style and values of capacitors shown are selected because of the advantages of each type.

Normally the smaller non-polarized capacitor will have better performance at high frequencies.

The larger valued capacitors that are polarized will have ability to work at lower frequency and smooth larger current flows at those frequencies.

Note also that one capacitor is isolated from the others via a ferrite bead so there is not a direct combination possible anyway.

So as you can see there is a disadvantage to combine all the values to a single capacitor.

Answer 2

There are several questions here which begin to answer your question e.g.Placement of decoupling capacitors

Fundamentally, capacitors are not perfect. Each value will operate most effectively to filter some frequency band of noise, and mitigate the inductance of the power supply and tracks.

Essentially they are trying to provide power to a chip which needs energy very quickly from a power supply which can't respond quickly enough, connected over wires which resist (through inductance) changing the power they supply.

Because the capacitors are supplying power very locally, rather than going across wires (aka PCB copper tracks) they reduce the amount of electrical noise emitted by the 'wires'. The effect of this is Radios, e.g. mobile phone work near the electronics.

The capacitors, especially 100nF, should be mounted as close to the chip as practical. Also, assuming you are making a PCB, try to keep power supply tracks and ground as 'wide' and short as practical to reduce resistance and inductance.

Answer 3

I could simply use 1 equivalent capacitor with Ceq = C1 + C2 + C3?

No, you cannot do that here. The various caps have different resonant/corner frequencies. It's okay to add them up when doing say a bulk filter for a power supply (e.g. use two 2200uF in parallel) or the bulk reservoir caps for a board (say several 220uF spread on the board, typically on a diagonal pattern). But you cannot "add up" 100nF + 4.7uF on a supply rail. These two serve different purpose. The larger one (uF) becomes an inductor at high frequency due to parasitics. That's why you need the smaller (nF) one in different tech. Analog's app note MT-101 discusses this in detail. It's about opamps, but applies to the supply rails of most chips. TI has a similar appnote SLOA069; here's an illustrative graph from it (but this is just the general idea regarding the effects of parasitics):

Cypress has a note AN1032 for higher speeds used in data communication. Even if you use the same tech for the caps here's what happens in calculation vs. reality when you parallel them:

That's why it's a good idea to go (at least as a start/prototype) with the values recommended in the datasheet for a particular chip. Hopefully the manufactuerer of the IC tested the effect of the combo they recommend on the real chip before putting in the datasheet.

Now, as a matter of practicality, I have combined/added the values of eletrolytics for nearby components, for example I used one 22uF eletrolytic instead of two 10uF ones for nearby opamps (which saves a bit of board space). But I left the ceramics separate for each opamp. I'm not an experienced designer and didn't do sophisticated calculations... and got away with that. I don't know how more sophisticated engineers solve this kind of problem, i.e. combining networks of caps. It's probably hard given all the variables including parasitics.