If space and budget permit, is it always better to use a large capacitor?

Thread Starter

Obanion

Joined Nov 26, 2009
24
I'm currently working on a project where I'm designing a DC-DC converter circuit. I'm trying to make the most efficient design possible, and I've recently started to explore capacitors and their respective ESR.

I was curious. It seems that the ESR is inversely proportional to the capacitance rating e.g. a 470 uF cap has a lower ESR than that of a 22 uF of the same chemistry. I guess this is due to the function of 1/sC (in part, among other things). So in terms of actually selecting a capacitor, why not select the largest possible?

I need a minimum capacitor value of 22 uF for my input and coupling (I'm using a SEPIC topology), but I'm trying to decide why I shouldn't just go with a larger capacitor that has a lower ESR. The capacitors aren't that expensive, so it's not really a problem for me to buy larger ones. Nor is the size of the board a problem. Overall, I'm looking for efficiency.

That said, is it just better for me to buy the largest capacitors with the lowest ESRs possible?

On a side note, would you recommend that I use many smaller electrolytics in parallel, or should I try to get away with a single large capacitor?

Thanks,

Obby
 

steveb

Joined Jul 3, 2008
2,436
I'm currently working on a project where I'm designing a DC-DC converter circuit. I'm trying to make the most efficient design possible, and I've recently started to explore capacitors and their respective ESR.

I was curious. It seems that the ESR is inversely proportional to the capacitance rating e.g. a 470 uF cap has a lower ESR than that of a 22 uF of the same chemistry. I guess this is due to the function of 1/sC (in part, among other things). So in terms of actually selecting a capacitor, why not select the largest possible?

I need a minimum capacitor value of 22 uF for my input and coupling (I'm using a SEPIC topology), but I'm trying to decide why I shouldn't just go with a larger capacitor that has a lower ESR. The capacitors aren't that expensive, so it's not really a problem for me to buy larger ones. Nor is the size of the board a problem. Overall, I'm looking for efficiency.

That said, is it just better for me to buy the largest capacitors with the lowest ESRs possible?

On a side note, would you recommend that I use many smaller electrolytics in parallel, or should I try to get away with a single large capacitor?

Thanks,

Obby
I think you can use larger capacitors with the SEPIC topology with no problem. However, the transient response under feedback control will be slower, I believe.

I think putting many capacitors in parallel is not recommended because each capacitor is a series RC circuit (due to ESR, as you pointed out), not a pure capacitance.
 
Last edited:

Thread Starter

Obanion

Joined Nov 26, 2009
24
Thanks for the reply. I wasn't thinking of feedback response, so that's definitely something I'll have to consider. As far as the capacitors go, I was under the impression that the impedance would be lower overall since you have resistances in parallel, but I'm not sure if there would be any other physical phenomenon that might counter this.
 

steveb

Joined Jul 3, 2008
2,436
... As far as the capacitors go, I was under the impression that the impedance would be lower overall since you have resistances in parallel, but I'm not sure if there would be any other physical phenomenon that might counter this.
This is certainly a good question.

The complication with your idea is that the ESR in a capacitor is equivalent to an ideal capacitor in series with a resistor. This is not the same as a capacitor in parallel with a resistor.

If the latter case were true, multiple parallel capacitors would be equivalent to a single capacitor (with all capacitances added) in parallel with a single resistor (calculated with the parallel formula). In this event, your idea might hold up. Multiple capacitors in parallel could end up being preferred, depending on the properties of available devices.

However, in the former case, a parallel combination of N series RC circuits is a very high order system (Nth order to be precise). This is much more complicated than a single resistor in series with single capacitor, which is only a 1st order system. So, here you can't say that there is one effective resistance that is lower in value.

These statements of facts, in and of themselves, do not prove which method is better, and you would need to do some theoretical analysis to prove which is preferred. However, my gut feeling is that a high order system would bring up stability problems in a feedback control loop. Also, I seem to recall that the general advice given is to not use parallel capacitors, but to use only a single large value capacitor.

I'd be interested to hear other opinions and reasoning about this.
 

SgtWookie

Joined Jul 17, 2007
22,230
Well, here's my concept of "ideal" vs "real" capacitors:



Some folks attempt to model "real" capacitors as a series RLC network, with just one of each. But that ignores leakage (R3), and the lengths of the interconnects and leads or terminations of the caps, which affects L.

As you increase the value of capacitance, your ESR might go down, but I'm afraid your value of L will increase; and your leakage is bound to increase. That may or may not be an issue. Paralleling small cap(s) nearest the input to the switch should help a great deal to get rid of the highest frequency transients.

One item that could favor using several smaller caps in parallel is heat dissipation. If you have a single large cap, it might take a while for heat to propagate to the outer shell of the cap.

Anyway, just kind of thinking out loud here.
 

Attachments

thatoneguy

Joined Feb 19, 2009
6,359
Getting higher quality/low ESR caps would be a better suggestion than simply using a larger cap of the same value.

When dealing with sizes near 1uF, look for types other than Electrolytic.

Surface mount Electrolytics score a bit better as the lead length is reduced to a fraction.

For power rail stability, a combination of an electrolytic with a mylar or mica 0.1uF cap seems to work well. Spread them out on the board, one pair for each IC for bypass/decoupling/stability caps.
 
Top