Not All X7Rs Are Created Equal
Not All X7Rs Are Created Equal
Since my RC time-constant problem was far greater than would be explained by the specified temperature variation, I had to dig deeper. Looking at the data for capacitance variation versus applied voltage for my capacitor, I was surprised to see how much the capacitance changed with the conditions that I set. I had chosen a 16V capacitor to operate with a 12V bias. The data sheet indicated that my 4.7 μF capacitor would typically provide 1.5 μF of capacitance under these conditions! Now this explains the problem that my RC circuit was having.
The data sheet then showed that if I just increased the size of my capacitor from 0805 to 1206, the typical capacitance under these conditions would be 3.4 μF. This called for more investigation.
I found that the Murata® and TDK® websites have nifty tools that allow one to plot the variations of capacitors over different environmental conditions. I investigated 4.7 μF capacitors of various sizes and voltage ratings. Figure 1 graphs the data that I extracted from the Murata tool for several different 4.7 μF ceramic capacitors. I looked at both X5R and X7R types in package sizes from 0603 to 1812 and with voltage ratings from 6.3VDC to 25VDC.
Figure 1. This is a graphic representation of temperature variation vs. DC voltage for select 4.7 μF capacitors.
Note, first, that as the package size increases, the capacitance variation with applied DC voltage decreases, and substantially.
A second interesting point is that, within a package size and ceramic type, the voltage rating of the capacitors seem often to have no effect. I would have expected that using a 25V-rated capacitor at 12V would have less variation than a 16V-rated capacitor under the same bias. Looking at the traces for X5Rs in the 1206 package, we see that the 6.3V-rated part does indeed perform better than its siblings with higher voltage ratings.
If we had looked over a broader range of capacitors, we would have found this behavior to be common. The sample set of capacitors that I was considering do not exhibit this behavior as much as the general population of ceramic capacitors.
A third observation is that, for the same package, the X7Rs always have better temperature sensitivity than X5Rs. I do not know if this holds true universally, but it did seem so in my investigation.
Using the data from this graph, Table 2 shows how much the X7R capacitances decreased with a 12V bias.
Table 2. X7R Capacitors with a 12V Bias
We see a steady improvement as we progress to larger capacitor sizes, until we reach the 1210 size. Going beyond that size yields no improvement.
In my case, I had chosen the smallest available package for a 4.7 μF X7R because size was a concern for my project. In my ignorance I had assumed that any X7R was as effective as any other X7R—clearly, not the case. To get the proper performance for my application, I had to use a larger size package.
Choosing the Right Capacitor
I really did not want to go to a 1210 package. Fortunately, I had the freedom to increase the values of the resistors involved by about 5x and, thus, decrease the capacitance to 1.0 μF. Figure 2 graphs the voltage behavior of several 16V, 1.0 μF X7R caps versus their 4.7 μF, 16V, X7R cousins.
Figure 2. This graph shows performance of 1.0 μF vs. 4.7 μF capacitors.
The 0603 1.0 μF capacitor behaves about the same as the 0805 4.7 μF device. Both the 0805 and 1206 1.0 μF capacitors perform slightly better than the 1210 4.7 μF size. By using the 0805 1.0 μF device, I was thus able to keep the capacitor size unchanged while getting a capacitor that only dropped to about 85% of nominal and not to about 30% of nominal under bias.
But there was more to be learned. I was still confused. I had been under the impression that all X7R caps should have similar voltage coefficients since the dielectric used was the same, namely X7R. I contacted a colleague and expert on ceramic capacitors1. He explained that there are many materials that qualify as “X7R.” In fact, any material that allows a device to meet or exceed the X7R temperature characteristics, ±15% over a temperature range of -55ºC to +125ºC, can be called X7R. He also explained that there are no voltage coefficient specifications for X7R or any other types.
This is a very important point, so I will repeat it. A vendor can call a capacitor X7R (or X5R or any other type) as long as it meets the temperature coefficient specs, regardless of how bad the voltage coefficient is.
As an applications engineer, this fact simply reinforces the old maxim (pun intended) that any experienced apps engineer knows, "Read the data sheet!”
As the capacitor vendors have made smaller and smaller components, they have had to compromise on the materials used. To get the needed volumetric efficiencies in the smaller sizes, they have had to accept worse voltage coefficients. Of course, the more reputable manufacturers do their best to minimize the adverse affects of this trade-off.
Consequently, when using ceramic capacitors in small packages, or indeed any components, it is extremely important to read the data sheet. Regrettably, often the commonly available data sheets are abbreviated and will have very little of this kind of information, so you may have to request more detailed information from the manufacturer.
What about those Y5Vs that I summarily rejected? For kicks, let’s examine a common Y5V capacitor. I will not identify the vendor of this part, as it is no worse than any other vendor’s Y5V. I chose a 4.7 μF capacitor rated at 6.3V in an 0603 package and looked at the specs at 5V and +85ºC. At 5V the typical capacitance is 92.9% below nominal, or 0.33 μF. That’s right. Biasing this 6.3V-rated capacitor with 5 volts will result in a capacitance that is 14 times smaller than nominal.
At +85ºC with 0V bias the capacitance decreases by 68.14%, from 4.7 μF to 1.5 μF. Now you might expect this to reduce the capacitance under 5V bias from 0.33 μFto 0.11 μF. Fortunately, these two effects do not combine in this way. In this particular case the change in capacitance with 5V bias is worse at room temperature than at +85ºC.
To be clear, with this part under 0V bias we see the capacitance drop from 4.7 μF at room temperature to 1.5 μF at +85ºC, while under 5V bias the capacitance increases with temperature from 0.33 μF at room temperature to 0.39 μF at +85ºC. This should convince you that you really need to check component specifications carefully.