De-coupling capacitor on a power regulator

Hi,

If I look at the suggested circuit diagram that comes on the datasheet for a basic voltage regulator there's frequently a big capacitor (say 100 micro farads) next to a smaller capacitor (0.22 micro farads).

I know the smaller capacitor is there for de-coupling, but how does this fit in with classic circuit theory that says that two capacitors in parallel have a equivalent capacitance equal to their sum - in this case 100.22 micro farads.

100.22 micro farads is pretty damn close to 100 micro farads so why bother with the smaller capacitor? It makes little or no difference to the overall capacitance

Bigger caps take care of the bigger, slower frequency noise effects. Little caps take care of the smaller, higher frequency noise effects. Little caps (0.1uF, or 0.22uF) next to an IC also provide the energy for high speed switching current surges within a chip.

Thanks Crossroads - but I kinda knew that. What I'm really getting at is that from the circuit theory perspective the tiny little capacitor next to a bigger one has no electrical effect. Is this one of those 'practice ain't the same as theory' moments?

I think this is more like the sum of the components contribution isn't the same as contribution of each individual component. It is more a function of time than of some steady state condition.

A hypothetical perfect capacitor has impedance Xc=1/wC so that as w increases, Xc should decrease in proportion. In the real world this formula does not give you a totally accurate picture, and Xc does not follow this law precisely. Different capacitor types and sizes depart to a different degree, so need to be selected for suitability.

Try reading up on 'capacitor ESR'. This may help answer your question.

I'm sorry if I'm chatting nonsense here. I'm just learning this capacitor noise filter thing myself at the moment too, but...

You say the caps are in series... are they polarized? If they're not polarized then perhaps that means the first cap (lets say that's the 100uF) allows all of the high frequencies through to the smaller cap which then filters out all of the REALLY low frequencies and sends the resulting waveform to ground. Which if it's a mains regulator would possibly be anything outside of say a 220 - 250 volts bandwidth (in the UK).

Then... it sends all of the frequencies which it didn't allow to ground (the clean signal) back through the lager cap (being as it's none polarized) and to the output.

Like I say this could be wrong. Sorry if it is.

Fulliautomatix: but how does this fit in with classic circuit theory that says that two capacitors in parallel have a equivalent capacitance equal to their sum

theobsoletemovement: You say the caps are in series...

ahh right... I think i get it... sorry i thought you said in series.

But yeah, if they are in parallel... in this instance the capacitor ratings do not get added together as it is down to the instantaneous voltage difference (transient speed... or rather the lack of it) of the noise to depict how much the capacitor discharges. So when the the low frequency noise transients become slower the low-pass cap begins to allow flow but the hi-pass one does not necessarily (unless the high frequency noise transients also become slower at that exact moment). Either way... all that gets through to the rest of the circuit is what is inside the band which it creates.

Am i close?

Is this one of those 'practice ain't the same as theory' moments?

No, mind you it is one of those "your theory is far far too simplified" moment.

Either way... all that gets through to the rest of the circuit is what is inside the band which it creates. Am i close?

About half a solar system away.

Quite simply any capacitor has associated with it a stray inductance, the bigger the capacitor the larger the inductance. When the frequency is such that the inductive reactance equals the capacitave reactance it stops looking like a capacitor. Any frequency higher than this and it looks electrically like an inductor and so stops smoothing or shorting out high frequencies. Their values do add up but the absolute value is not the most important parameter here. What is important is the frequency response of the system.

This point is known as the self resonance point. The smaller the capacitor the higher is the self resonant point. Hence you use two capacitor, one for low frequencies and the other for high.

See the graphs here http://www.thebox.myzen.co.uk/Tutorial/De-coupling.html

theobsoletemovement: I'm sorry if I'm chatting nonsense here.

Apology accepted.

theobsoletemovement: ahh right... I think i get it... sorry i thought you said in series.

But yeah, if they are in parallel... in this instance the capacitor ratings do not get added together as it is down to the instantaneous voltage difference (transient speed... or rather the lack of it) of the noise to depict how much the capacitor discharges. So when the the low frequency noise transients become slower the low-pass cap begins to allow flow but the hi-pass one does not necessarily (unless the high frequency noise transients also become slower at that exact moment). Either way... all that gets through to the rest of the circuit is what is inside the band which it creates.

Am i close?

The dominant effect is series inductance inside the high value capacitor and from its leads, nothing to do with capacitance per se. The small value cap is surface mount direct on the device to minimize series inductance. At logic speeds series inductance is a huge effect.