Should bulk capacitors generally be electrolytic?

[cbc02009]

hello all,

I'm design a pcb for my capstone project, and I've realized I don't really know much about the differences between ceramic and electrolytic capacitors for this particular application. Given the same voltage rating and capacitance, is there a reason to use one over the other for bulk capacitors? How about tantalum?

Thanks!

[T3sl4co1l]

Ceramic: low ESR
Electrolytic: high ESR (relative to size)

Ceramics have the downside of tending to ring with nearby inductances.  Putting in some ESR (in the right places) helps dampen that ringing.

The other downside of electrolytics is, the ESR varies wildly with temp and age.  They can be fine, but used carelessly, they can make a poor choice that much worse.  (In that case, the wiser choice would've been a stable-ESR type like tantalum or polymer, or a low-ESR ceramic or polymer with added external resistance.)

It's all a matter of "how much?"  Some circuits are fine with a small electrolytic, or just one or a few 0.1's; others need a lot of both.

The determining factor is, how much voltage ripple can your circuit tolerate, and how much AC current does it draw?  (It's not DC current that matters, it's the change.  Say your circuit is switching rapidly between 0 and (lots)A, it's basically worst-case 100% AC*. Anything else, probably easier.)

*Well, 50%, or something near there, depending on how you define it.

Tim

[tautech]

Tim
I've always been of the understanding tantalums can serve as both bulk and local decoupling providing the necessary precautions for using them are observed.
I'd welcome your advice.

[cbc02009]

Ceramic: low ESR
Electrolytic: high ESR (relative to size)

Ceramics have the downside of tending to ring with nearby inductances.  Putting in some ESR (in the right places) helps dampen that ringing.

The other downside of electrolytics is, the ESR varies wildly with temp and age.  They can be fine, but used carelessly, they can make a poor choice that much worse.  (In that case, the wiser choice would've been a stable-ESR type like tantalum or polymer, or a low-ESR ceramic or polymer with added external resistance.)

It's all a matter of "how much?"  Some circuits are fine with a small electrolytic, or just one or a few 0.1's; others need a lot of both.

The determining factor is, how much voltage ripple can your circuit tolerate, and how much AC current does it draw?  (It's not DC current that matters, it's the change.  Say your circuit is switching rapidly between 0 and (lots)A, it's basically worst-case 100% AC*. Anything else, probably easier.)

*Well, 50%, or something near there, depending on how you define it.

Tim

Thanks for the explanation!

The circuit is a Solar MPPT controller using an atmel SAML21. Do you think it's possible that the current change from say a cloud covering the panel is enough to want a bulk capacitor on the panel input? The only other place I have them is on the outputs of the switching regulator (for the 3.3V and 5V rails).

[T3sl4co1l]

Thanks for the explanation!

The circuit is a Solar MPPT controller using an atmel SAML21. Do you think it's possible that the current change from say a cloud covering the panel is enough to want a bulk capacitor on the panel input? The only other place I have them is on the outputs of the switching regulator (for the 3.3V and 5V rails).

Ah good, we have a start.  Now the question is: how fast?

So first of all, I take it there are at least three things in this system -- you're leaving a LOT out, I suspect! -- the solar panels (DC), the MPPT converter itself (switching, with harmonics into the 10s MHz??), and some digital logic (at least an MCU, clock of 10s MHz, harmonics in the 100s?).

Which one has the highest rate of change?  How much power?  How much impedance or energy storage is required for each?

Presumably, this is charging a large energy reservoir, like a battery, or feeding the grid.  So, slow changes, like clouds, simply change the charge rate.  Rate of change (to anything the circuit cares about) is essentially zero for that.  So what does that leave?

Tim

[max_torque]

If you are getting the power to run the controller from the panel, then yes, you need to buffer the supply to the micro to include suitable "ride though" capability (or to ensure the system shuts down and boots cleanly and without glitching etc).

However, you will never buffer the main power feed to the converter as the amount of capacitance to do so would be enormous (capacitors are really bad at storing energy, as they store charge, rather than convert that charge into a different energy type (ie chemical energy in a battery)) and pointless. If there is insufficient voltage to continue to run your converter, stop running it, simples ;-)

[David Hess]

Input capacitors for non-power factor corrected off-line power supplies are sized for holdup time between cycles which results in more than enough ripple current rating even though their capacitance means using electrolytic capacitors.  In an off-line power supply with active power factor correction, this large capacitor is moved to *after* the input stage and before the switching regulator and serves the same purpose.

In a DC applications, hold up time is irrelevant except to the switching regulator controller which draws low current and can have its own low current decoupled supply.  So the input capacitor to the switching regulator needs to be sized by ripple current capability and not capacitance.  That leaves using either an large capacitance electrolytic capacitor just to get a high enough ripple current rating (1) or a lower capacitance high ripple current rating capacitor like polymer electrolytic or ceramic or film.

The good DC input designs I have seen use an aluminum electrolytic or polymer electrolytic capacitor (2) of moderate value followed by a filter inductor and then a ceramic or film decoupling capacitor.  The low ESR of the ceramic or film capacitor and the inductance divert most of the high frequency ripple current from the electrolytic input capacitor so it does not need to be nearly as large.  I assume this could be reduced to a PI filter with ceramic or film capacitors on both sides for extended life but I have not actually seen this done in practice except at much lower power levels.

(1) Output capacitors on switching power supplies have this same requirement and they are usually what fails first.

(2) Polymer electrolytic capacitors have a lower maximum voltage rating than traditional aluminum electrolytic capacitors leaving traditional aluminum electrolytic, ceramic, and film capacitors if the input DC voltage is high.