Tantalum, Ceramic or Aluminum capacitors

[new299]

I'm trying to understand how to choose between ceramic, tantalum and aluminum capacitors.

My current understanding is that for low values (and voltages) I use ceramics. For mid-range values (up to ~470uF, and < 16v?) I can use tantalum or aluminum. Above this aluminum capacitors are the only practical option. Is this approximately correct?

So, my question is how to I choose between tantalum and aluminum capacitors or tantalum and higher valued ceramic capacitors (e.g. 10uF). When does ESR matter? For example is it a consideration for decoupling capacitors?

[klunkerbus]

I'd say you are approximately correct.  Also note that tantalum and aluminum will usually have a polarity to them, while ceramic does not.  Tantalum is expensive - one of the reasons you don't find large value tantalum capacitors.  ESR is important when the capacitor will be subjected to a lot of ripple current - as in power supplies. 

Here's what I apply in my hobby projects:
small value caps up to maybe 0.1uF are ceramic, mylar, polyester, or other low leakage current, non polarized materials
0.1uF to maybe 47uF will be tantalum where possible
Greater than about 47uF will be aluminum

[tautech]

Consider the 4 main uses for caps:
Coupling
Decoupling
Bulk capacitance
Smoothing.

Each type of use has differing requirements and frequencies to deal with.

You then should know which type is best to use where.

Ceramics and Tants are fast and best used to suppress spikes and therefore commonly used for Decoupling needs. Both have the advantage of small size. Tants historically have been used for bulk and decoupling and sometimes seen on a rail far from the supply.

Aluminum is the backstay of most other common electronic use, smoothing and bulk capacitance along with low ESR needs.

[new299]

Thanks for your reply. So it sounds like for 10uF I'm free to choose between tantalum and ceramic. For 100uF capacitors I can choose between tantalum or aluminum.

Based on cost I'd go with 10uF ceramic and 100uF aluminum. I guess other selection criteria is  ESR (this is a consideration in capacitor switching inverters for example?) and reliability?, Tantalum caps also seem to be lower profile which is handy in a project I'm working on.

[new299]

Ceramics and Tants are fast and best used to suppress spikes and therefore commonly used for Decoupling needs.

When you say fast, the "speed" of a capacitor is a function of its capacitance right? Or is there something else going on?

[fivefish]

Dont forget cost difference.  (prices @ Mouser.com)

10uf 50V tantalum, radial, leaded - $3.26 to $15.81 per piece
10uf 50V electrolytic, radial, leaded - $0.08 to $0.97 per piece

* price goes down at higher quantity

[mariush]

Ceramic capacitors go lower in capacitance with voltage ...  a 10uF ceramic capacitor rated for maximum 16v, won't be 10uF in a circuit that uses 12v, it would probably be 4-8uF ... see http://www.maximintegrated.com/en/app-notes/index.mvp/id/5527    You may need to go with 10uF 50v or 100v rated ceramic which may be more expensive than a polymer or tantalum capacitor.

Also, some ceramic capacitors have too low ESR, for example some linear regulators require a capacitor with a minimum 0.01 ohm ESR and maximum 1 ohm, and some ceramic capacitors are better than this... that's why you'd often see tantalum capacitors mentioned in datasheets.

In keeping up with this example, 10uF 16v electrolytic could have too high ESR (above 1 ohm) depending on brand, series, what it's optimized for so that's why you'd often see electrolytic capacitors rated for much higher voltages than needed (35v, 50v) as larger electrolytic capacitors have better ESR.

With most ICs you need some 0.01uF and/or 0.1uF capacitors for decoupling and some ICs also require a bit of "bulk" capacitance at the input, it's common to see values of 10-100uF in datasheets.

Tantalum capacitors have low ESR but not very low like ceramic capacitors but they don't like overvoltage, you usually want to see less than 70% of the rated voltage at any time. So if you use a 16v rated tantalum for a 12v circuit, you really have to know your circuit and be sure you won't see more than 12v ever, even for milliseconds.

[T3sl4co1l]

Ceramics and Tants are fast and best used to suppress spikes and therefore commonly used for Decoupling needs.

When you say fast, the "speed" of a capacitor is a function of its capacitance right? Or is there something else going on?

The "speed" is essentially the time constant C * ESR (or the cutoff frequency proportional to 1/x of this).  It's largely a matter of technology and construction, not capacitance (you can make a bigger capacitor that should be slower, but it's going to have lower ESR, so it stays the same speed).

The time constant of batteries is in the seconds to hours range; supercapacitors, seconds to fractions of a second; electrolytics, milliseconds to microseconds; and most others (ceramic, film, etc.) microseconds and below.

This determines the range of frequencies these components are best suited to bypassing.  Lower frequencies (longer times) are most efficient for a given type, while moderate frequencies can be filtered effectively, if not necessarily with low losses.  Example: you might use an electrolytic to bypass some digital logic, where the frequencies are high but the average current is low; but not for a switching regulator, where both frequency and current are high, and power dissipation would be excessive.  Sometimes, the loss is a helpful addition; an excellent use of electrolytic or tantalum capacitors is in parallel with smaller ceramic or film capacitors, which have quite low ESR which can resonate with circuit inductances.

It's noteworthy that the time constant of tantalum capacitors is not much less than most electrolytic capacitors.  But, the response is far more stable, and much simpler.  Electrolytics have all sorts of transmission line and ionic diffusion effects (a result of their wound construction and electrolytic nature), and ESR is vastly higher at low temperatures; the impedance measured in circuit (C, ESR and ESL) vary in all sorts of subtle ways.  Tantalums are basically ESR (with a small tempco), with ESL limited to that due to lead length and basically nothing else.  (The ESR is due to the "electrolyte", a conductive solid conformed around the tantalum pellet.  It's more conductive than electrolyte, and the conductive path is direct.)

Tim

[that_guy]

Tantalums are reputed to fail explosively if you exceed their voltage rating even slightly. For that reason I only consider them on the output side of a voltage regulator and never on the input side.

[Howardlong]

For low ESR applications, you might also want to take a look at aluminium polymer as opposed to bog standard non-solid aluminium electrolytics in around the 47uF to 1000uF range.

[T3sl4co1l]

Yes, aluminum polymer are largely equivalent to film capacitors (in terms of energy density, ESR, and even self-healing), except with large, practical values at low voltage ratings.

Tim

[Cliff Matthews]

Aluminum vs. Polymer over-voltage video:
It would be nice to know if a failed polymer/aluminum cap can be easily identified on a PCB.

[TimFox]

Also, please note that "ceramic" covers a wide range of materials and properties.
NP0/C0G has the best electrical performance (high Q and excellent linearity), but is available in relatively low values due to the low dielectric constant of the material.  People who disdain "ceramics" usually don't mean NP0/C0G.
X7R has mediocre electrical performance (medium Q and some drop in capacitance with voltage).  X7R and X5R are similar, and are often used in place of polarized capacitors at relatively large values.
Be careful of Z5U and similar types, whose capacitance falls badly with applied voltage.  They are small, however, at high values.

[suicidaleggroll]

Also keep in mind that you need to voltage de-rate the crap out of tantalum capacitors.  With aluminum and ceramic you can do maybe 50% derating on the voltage, but with tantalum you have to do 2-300% in many cases.  So if you're planning on using it to decouple a 30V rail, you would want at least 45-50V for aluminum or ceramic, but you want closer to 90V for tantalum.  This jacks the price of tantalums WAY up if you need to decouple any more than about 10V with a moderate amount of capacitance.

[TimFox]

What is your reference for such a huge derating factor (2x or 3x) on tantalums?
I looked at the data sheet for Kemet T491 industrial SMT tantalums, listed in their catalog as "standard tantalums".
They specify:  2000 hrs at 85C at rated voltage  and  2000 hrs at 125C at (2/3 rated voltage, x1.5 derate).
Surge rating: at (1.32 x rated voltage), 85C or at (1.2 x rated voltage), 125C for 1000 cycles.
The specified result is +/- 10% of C for the 2000 hrs tests, with ESR within original spec and DC leakage within 1.25x original.
For the surge test, the result is +/- 5% of C and  ESR and leakage within original spec.

[suicidaleggroll]

Just search google for "tantalum derating", you'll find hundreds of references.  Here's just one from Kemet:
http://www.kemet.com/Lists/filestore/Derating%20Guidelings%20for%20Tantalum%202011%20(3).pdf

Last table in the paper has the recommended voltage ratings for standard tantalum and polymer tantalum.

[TimFox]

Thanks for the link.
It is interesting to see how much less derating is recommended for the "polymer" type, compared with the original MnO2 type.  The polymer type also typically has better ESR.

[max666]

Aluminum vs. Polymer over-voltage video:
It would be nice to know if a failed polymer/aluminum cap can be easily identified on a PCB.

Did that look like Polymer tend to fail with a dead short? If so, would that now be a good or a bad thing.
I'm a little disappointed the video didn't show all of the Polymer caps destroyed. Now we only have a partial sample.

[Zero999]

Did that look like Polymer tend to fail with a dead short? If so, would that now be a good or a bad thing.
I'm a little disappointed the video didn't show all of the Polymer caps destroyed. Now we only have a partial sample.

Lots of capacitors can fail short circuit. Ceramic capacitors do this too.

[Cliff Matthews]

It would be nice to know if a failed polymer/aluminum cap can be easily identified on a PCB.

Did that look like Polymer tend to fail with a dead short? If so, would that now be a good or a bad thing.
I'm a little disappointed the video didn't show all of the Polymer caps destroyed. Now we only have a partial sample.

It may have shorted..  pressure seemed to push down and eject the lead, tilting the whole cap sideways.

Must be the reason why Dave gets all excited to see hot-snot pin-down possible flying (shorting) caps?

[SeanB]

Any capacitor with a rated voltage below 10V is going to need derating. Tantalum typically is derated more because there is no real method on solid devices to repair defects in the oxide layer, so anything that makes a low resistance leakage will cause it to blow up if there is enough supply current. Thus the typical use of a 25V unit on a 5V rail instead of a 10V unit. Wet slug units will self heal, though they turn into small guns if you severely overload them. I did use some 10uF 35V units to replace the output caps in a PC power supply ( only ones I had to hand late one night and I wanted it working bad) and they were quite happy as the high current filter along with a regular 85C electrolytic. It would not work with only the electrolytic, but ran for a few years with the 40uF extra decoupling.

[Howardlong]

Ceramics also crack, there is little tensile strength in them.

Aluminium polymer are very different to bog standard aluminium electrolytics in many ways, including ESR, voltage derating and that they won't dry out. Probably they should be considered as a completely different type of cap, despite having "Aluminium" in them. Be wary of putting Al polymers in the same bucket as Al bog standard.

[Monkeh]

It would be nice to know if a failed polymer/aluminum cap can be easily identified on a PCB.

Did that look like Polymer tend to fail with a dead short? If so, would that now be a good or a bad thing.
I'm a little disappointed the video didn't show all of the Polymer caps destroyed. Now we only have a partial sample.

It may have shorted..  pressure seemed to push down and eject the lead, tilting the whole cap sideways.

Must be the reason why Dave gets all excited to see hot-snot pin-down possible flying (shorting) caps?

It looks more like they went and cut the lead because it was a dead short.

[klunkerbus]

Shorted tantalum caps - oh man, what a flashback.  Many years ago, I was the project lead over a team responsible for developing four large circuit boards (VXI size, maybe 11 x 18 inches?) that were to replace the functionality of two racks of equipment used for command and control of geosynchronous satellites.  Our whole team was watching the day we powered up the first prototype board for the first time.  The only VXI chassis we had at the time had no current limit provision on any outputs.  Within a fraction of a second powering up the board, the power entry corner of the board started glowing red and the amount of glowing red kept expanding. 

What we found is that the SMD tantalum cap on the + or - 15V input had been installed backwards, and had shorted out when it failed.  The board designer hadn't realized the autorouter used a very narrow inner trace for this power input, and the internal trace to the capacitor commenced to vaporize when the cap shorted out. When the trace vaporized, it charred the dielectric between copper planes in a way that the +5V plane (driven by a 70A supply in the chassis) became conductive to ground.  It wasn't a direct short, but enough to generate a lot of heat.  Thankfully we did kill power within the first second or so. 

It would have taken an unacceptable amount of schedule and cost budget to replace that prototype.  As project lead, I took on attempting salvaging the board myself. I spent two days picking away at the board, peeling back copper layers and cleaning out charred dielectric until the DC resistance between 5V and ground increased to what we thought was acceptable.  We continued to use that prototype for a long time after that.  For years, I showed many engineers fresh out of school that board as an example of how things can go really bad, really fast.