Selecting Capacitors

Note: This is a rewrite of a message I wrote to the Parallax Basic Stamp Mailing List, but it seemed generally interesting. This was all off the top of my head, so it may or may not be very complete.

There are many factors that measure the quality of a cap. Selecting a capacitor means maximizing the factors you need while still keeping the other factors reasonable. What are the factors? My list looks like this:

bulletPrecision - How precise the value is (+/- 20% is not uncommon).
bulletStability - How much does the value change with time & temp.
bulletLeakage - All real-world dielectrics have some resistance (perhaps in the hundreds of megohms) and therefore real capacitors leak some DC current.
bulletBreakdown voltage - Dielectrics break down at particular voltages.
bulletCost - $$$.
bulletPhysical Size - 1 Farad ceramic capacitors are as big as the Titanic (just kidding; there are no 1 Farad ceramic capacitors). However, some of the new "super caps" can be 1F or larger. These are mostly useful for memory backup applications.

Leakage and other resistance (and impedance) effects leads to lots of other measurements including Q, power factor, and Effective Series Resistance (ESR). These are all useful ways to measure what resistance is doing to your capacitor. An
ideal capacitor has no resistance and therefore no leakage and a phase angle of 90 degrees. Real caps have some series inductance and resistance (from the leads), and some shunt resistance from dielectric leakage. That means
the phase angle will not be 90 degrees (which will degrade the power factor) and the DC current will not be zero.

For many circuits this isn't a big deal because the values are inconsequential. But at microwave, or micropower levels they may be important.

So what are the practical ramifications:

bulletCeramics - Usually el cheapo. Usually low stability and precision. In general, physically smaller ceramics will be more temperature sensitive than a larger unit of the same uF value (at the same voltage). There are two ceramics used in manufacture. One has a + temp shift and the other a -. They are mixed in NPO ceramics to provide some measure of temperature stability. (For those who must know the materials are magnesium titanate and calcium titanate).
bulletElectrolytic - El cheapo. Same as ceramics except they have much smaller physical size for a given value. Usually polarized.
bulletTantalum - Better characteristics than electrolytic but still small for high capacity values. Polarized.
bulletPoly film - These use polyester or polypropylene as a dielectric (this has mostly replaced paper capacitors). In general, slightly better characteristics than common ceramics. Usually very low leakage currents. Sometimes the conducting part is metallized right onto the film and they are called metallized film capacitors.
bulletMica/Silver Mica - These use mica plates as the dielectric and are known for being temperature stable. Usually large physically.
bulletPolystyrene, Teflon - Very temperature stable. Polystyrene breaks down, however, at high temps (say >80C)
bulletGlass, Air, Oil - You don't see these much any more except for HV work or around big motors and such.

Of course, there are exceptions. Some electrolytics have better designs than others. There are ceramics built to be stable over certain temp ranges (known as the temp coefficient). Others have leads designed to minimize inductance.

So, how to proceed? You need a capacitor of a given value. You can ask yourself:

1) How stable does it need to be?

2) How large can it be?

3) How much leakage can I tolerate?

4) How much temperature sensitivity can I stand?

5) How much will I pay?

Take a filter capacitor for a power supply. We don't care exactly what value the cap is. We just want a big one. That's why electrolytics are popular in this application. They tend to put the most capacity in the smallest container at the lowest cost. When I designed very small instruments for the DoD, we would often use tantalum which can be quite small, but costs much more.

Bypass capacitors are another place where the exact value isn't that important. You don't need big capacity values, so ceramic is very popular here.

If you are building a tuned circuit (say an LC tank) or an oscillator, you want a capacitor with tight tolerances and low temp sensitivity. Polystyrene is good here if you can operate without heat over 80C (which is actually pretty hot). Silver Mica is very good, if you can afford the space for a bulky capacitor. A carefully selected ceramic with good temperature coefficient at the temperature of interest will save money and space, but will probably never be quite as good as polystyrene, teflon, or silver mica.

So you can see this is a bit of an art. You have to get the best performance where it counts while still keeping the other factors reasonable. You can usually replace a "bad" capacitor with a "better" capacitor. For example, you could easily replace a ceramic capacitor with a poly film capacitor, all other things being equal. You can always use a capacitor with a higher working voltage than the one you are replacing too. However, you usually don't want to trade "down". Don't replace a silver mica with a garden-variety ceramic, for example. Watch out for polystyrene if you are in a high temp environment (automobiles, for example). Don't put polarized caps of any kind where a non-polar is called for. Going the other way should be OK, since I don't know of a rational designer who would take advantage of the polarizing effect (I could be wrong, though -- it would be possible).

Some other links: