What is the impact of package on SMD capacitor

Question

Most of electronic component exist in different package for different reason, it can either be power rating, reachable value, power dissipation... As far as I understood, the size of resistor principally impact the power rating.

So, a 1kΩ resistor in a 0805 package will have a different power rating than a 1kΩ resistor in a 0603 package.

If the package has an importance in the selection of resistor, is it the same for MLCC capacitor?

Answer 1

You are correct that for resistors power rating is one of the big factors affecting size, but it is not the only one. Voltage rating is also important to be aware of.

If you are running a circuit at 100V for example, you wouldn't use an 0402 resistor because the breakdown voltage of 0402 resistors is generally much lower than this (i.e. it will short through if you put too high a voltage across it). The larger the package, typically the larger the voltage rating.

---

In the case of capacitance there are several reasons for choosing a larger package over a smaller one. For one larger packages typically allow for higher capacitance because there is more physical space - you couldn't for example get a decent 10uF cap in an 0402 package.

If we assume you compare two capacitors of the same value (e.g. two 100nF caps), the larger ones will again typically have a higher rated working voltage. This is beneficial for two reasons. The first is obvious, and that is if you need a higher working voltage for your circuit - you wouldn't choose an 0402 rated at 10V if you need to run a circuit at 25V. The second is more subtle and I'll come to it in a moment.

A third reason is that there are different dielectrics. X7R is typically the best performer in terms of stability and has better DC performance. X5R is less good in that regard. Generally X7R capacitors are physically larger for the same voltage/capacitance ratings than X5R capacitors.

Furthermore, if you put an MLCC in a circuit charged to a DC level such as with a decoupling cap, you actually want to pick a rated voltage much higher than your working voltage. The reason for this is that the rated capacitance of an MLCC is heavily dependent on DC voltage.

A 10V MLCC might have a capacitance 50% lower than rated when running at 5V DC whereas a 25V MLCC might be only 10% lower than rated for the same running voltage. You can for example get a 100nF 0201 6.3V MLCC, but if you try to use that to decouple a 5V power line, you might find that the actual capacitance is only 10nF or less! As such, if you have the space, you typically want to go with a larger package with a higher rated voltage if you have the space

Answer 2

Tom Carpenter's comments about derating for DC bias demonstrate a very common misconception. While it's true that a higher-voltage cap might have slightly less capacitance loss with DC bias than another of the same value in the same package, it is package size that predominantly determines capacitance degradation with DC bias. A 1206 will see much less loss than an 0603, all other things being equal. Just spend some time with Murata's simsurfing website to see for yourself.

Answer 3

A few points:

• Tom Carpenter mentions the curves in some datasheets showing capacitance drop as DC is applied. First, these curves are not guarantees. Second, manufacturers will change the ceramic recipes periodically, for numerous reasons, which will change that curve. So for example, what once was a stable DC-DC converter, is now an oscillator. The only way to get around this is to use an AEC Q200 rated part. Then the manufacturers are required to notify you if they change anything about that part. Another option is use a 50V part instead of a 10V part, and hope and pray.
• Another important, poorly spec'd, and often ignored component of a MLCC is ESL. I'll leave researching planar inductance as a exercise for the reader, but the short version is the wider the package, the lower its ESL. For example, if you have a 10uF cap in 1206 and 1210, the 1210 will have lower ESL, therefore a higher SRF.
• Parts smaller than 0603 are significantly more prone to tombstoning during reflow.
• Parts larger than about 1812 are prone to cracking, due to flex in the PCBA either due to depanelization or application stresses.

Answer 4

According to Murata, package also affect decoupling. It's common practice to add several caps in parallel to filter various frequencies. Their study demonstrates that it's not really the case since the larger cap does moreless the same job as all the other combined. Except if you use packages of size decreasing with the capacitance: For example: 1uF in 0804, 0.1uF in 0603 and 0.01uF in 0402.download pdf here

Answer 5

You also must PLAN for heat removal. Each square of copper foil is 70 degree Centigrade per watt flowing across the foil. Thus a 100 mil by 1,000 mil copper path has 70*10 = 700 degree Centigrade per watt of heat flowing. At some point (I did some Finite Element sims using a SPICE grid of resistors, and 2cm was the answer), most heat exits that trace and flows THROUGH the epoxy-fiberglass to the underlying planes.

Get thee to a SPICE program, and lay down some 1ohm resistors horizontally, and some 200ohm resistors vertically, in a grid. Like this

^{simulate this circuit – Schematic created using CircuitLab}