In the context of ceramic capacitors (often also called MLCCs, for Multi-Layer Ceramic Capacitor) the term “tolerance” refers to deviations in device capacitance from the nominal value that are caused by variability in the manufacturing process, and this one factor alone. It is measured under strictly defined test conditions that are specifically designed to EXCLUDE the influence of other factors on the measured capacitance of a given device.
It is a common misunderstanding to interpret “tolerance” in an all-inclusive sense, and expect a measured value to fall within the stated tolerance range of the nominal value, regardless of measurement conditions. Many products are stable enough in their characteristics that this distinction is easily overlooked; ceramic capacitors are not among them. (Those based on class 1 dielectrics being an exception; see this post for an explanation of dielectric classes.)
Unfortunately, many ceramic caps based on class 2 and 3 dielectrics are so horribly, awfully, wretchedly unstable in their capacitance characteristics that casual measurement will often show a capacitance value outside the stated ‘tolerance’ from the nominal value. This does not mean that the products are defective or out of spec, it’s simply the nature and character of the technology.
Three important factors that affect the observed capacitance of a ceramic capacitor can be remembered as the 3 T’s: Test signal, Temperature, and Time.
Test signal: The image below, excerpted from the CL31A106KAHNNNE characteristics sheet, is an example of the effect that DC bias and the amplitude of an AC test signal can have on observed capacitance. Simply varying the amplitude of the AC test signal used can, in this case, result in a shift in observed capacitance of anywhere from -15% to +10%. The effect of DC bias is even worse; at only a quarter of rated voltage, the capacitance of the device is reduced by half. This can come as a very rude surprise to those who assume that the listed +/-10% tolerance figure is all-inclusive.
Note that variabilities of this magnitude are quite common, and the part chosen as an example is not appreciably better or worse in this regard than comparable products. The manufacturer of this product has, however, done a commendable job of making the characteristics of specific devices easily accessible in the form of a characterization sheet.
Temperature: The image below shows example capacitance variations with temperature for a few capacitor series with different temperature characteristics. With the exception of the Class 1 dielectric (C0G) devices, a 10% variation is about the minimum one should expect for applications that exhibit substantial changes in operating temperature. Note that this illustrates the effects of temperature only: other factors will tend to increase observed capacitance variations even further. Again, this is likely to come as an unwelcome surprise to those assuming that the listed “tolerance” is an all-encompassing figure.
Time: Finally, ceramic capacitors (class 1 dielectrics again excepted) are subject to an aging phenomenon, whereby they slowly lose capacitance with time, measured from the last occasion on which they were heated to a material-specific temperature, which happens during initial manufacturing and possibly also during assembly. The magnitude of the effect is material-dependent, cumulative, and often as much as a few percent per decade-hour. This article describes the effect in greater detail. Depending on the circumstances it can either add to or mask the effects of other influences on observed capacitance, and in any event is capable by itself of causing devices to appear to be out of tolerance when “tolerance” is incorrectly understood as an all-encompassing term.
“Tolerance” describes only 1 factor that influences the observed value of a ceramic capacitor, NOT all of them combined.
These other factors can have effects that are much larger than the specified tolerance value. This does not constitute a defect or failure to meet product specifications.
The test signal used, device temperature, and time since the last de-aging of a device influence the observed capacitance of a ceramic capacitor. All of these must match the manufacturer’s test conditions in order to make a capacitance measurement that is valid for determining compliance with the manufacturer’s specifications.
The capacitance measurement feature found on many multimeters is not a valid tool for determining whether or not a product meets the manufacturer’s specifications.
If you need a known, stable capacitance, don’t even consider ceramic capacitors other than those based on class 1 dielectrics (C0G or NP0 being the most common).