Knowing about components and their characteristics is important. Here's a list of a few things you need to know about capacitors:

• they have series inductance and resistance and parallel resistance
• initial tolerance ranges from quite good to really broad; if you need 0.1% tolerance you are in trouble
• some ceramic capacitors decline in value sitting on the shelf, then return temporarily to nominal after being heated to soldering temperature
• electrolytic capacitors "deform" if left for a long time without applied voltage
• electrolytic capacitors "wear out"
• electrolytic capacitors can be "poisoned" with chlorinated PCB cleaning solvents
  
• some capacitors are very stable with temperature, some have terrible stability, some are deliberately made to vary with temperature
  
• some caps have a large voltage coefficient of capacitance
• some caps have very low dielectric absorption, some have large da
• some caps handle fast large amplitude current pulses well, some fail miserably
• some fail short-circuit if subjected to a single high-magnitude fast-rising current event
  
• some dielectrics melt at low temperature, so they can never be used in surface mount capacitors
  
• some capacitors are quite microphonic

The list goes on.

 

I've bought some supercaps that were like "500F +80% -20%".

+80% / -20% is a common tolerance for many low-cost electrolytics, not just supercaps.

ak

 

Here is something they don't teach you in electronics class.

They did in at least two of my classes, but that was back in the early 70's.

It is common practice to put two or three decoupling capacitors in parallel across the supply and ground pins on an IC.
These could be a 0.01μF, 0.1μF and 1μF capacitors.
Theory of capacitances in parallel tells you that they add. Hence the net capacitance is 1.11μF.
However a 1μF cap with 20% tolerance negates having the additional lower value capacitors.
What gives?

(Assuming that is not a rhetorical question) the answer has nothing to do with value tolerance. The reason for multiple capacitors of different sizes is that when the caps have different constructions, each contributes to an overall capacitive effect that is better than any one of the individual parts. Common scaling is two orders of magnitude, so something like a 1 nF silvered mica, 100 nF ceramic, and 10 uF aluminum electrolytic in parallel gets you the superior high frequency performance and very high self resonant frequency of the smaller caps with the electrolytic's much lower impedance at low frequencies. This creates a very wide frequency range of low impedance decoupling.

ak

 

I think the advantage is that each cap has it's own bandpass.

 

The whole business of paralleling multiple values of capacitors has changed significantly with surface mount devices. It is still common to use a "bulk" cap in parallel with a high-frequency cap, but it is much less common to use multiple values of HF caps in parallel. The reason is simply that the compact leadless structure of SMT capacitors removes much of the parasitic inductance that was unavoidable with physically large leaded capacitors. A 100 nF cap can perform as well as a 1 nF at high frequency, in terms of decoupling. SMT also often allows getting the capacitors closer to the required place, though the necessity of vias can sometimes thwart intentions.

If you want to see really horrible capacitor tolerance, look at typical high dielectric constant types in surface mount. The initial tolerance and temperature coefficient of capacitance are poor to awful and most have such a horrendous voltage coefficient of capacitance that you can be down to 20% of nominal just due to that if you operate anywhere near rated voltage. The initial tolerance isn't usually an issue for decoupling or power supply filtering, but the voltage coefficient certainly is.

Inductors used in switchmode converters are another component class where large variation often must be considered. Cores materials such as powdered iron, molybdenum permalloy powder and "sendust" type all decline in permeability with DC bias. The amount of "swing" depends on the specific formulation used, and can actually be useful (you can get cores that are a composite of ferrite and powdered iron and are designed to have large swing), but a reduction in inductance of 30% isn't unusual. Permeability can increase with AC excitation, so it all gets very "interesting." I have yet to see a switcher simulation at AAC that includes inductor swing.

 

The whole business of paralleling multiple values of capacitors has changed significantly with surface mount devices. It is still common to use a "bulk" cap in parallel with a high-frequency cap, but it is much less common to use multiple values of HF caps in parallel. The reason is simply that the compact leadless structure of SMT capacitors removes much of the parasitic inductance that was unavoidable with physically large leaded capacitors.

Agree. I was going to edit that to the end of my post, but you beat me to it.

ak