Close

Capacitors. No, it's not that simple.

A project log for D-DAQ

automotive parameter & performance monitor & logger

michael-obrienMichael O'Brien 07/12/2014 at 04:490 Comments

I'm about ready to pull my hair out. Really, OMG, this was painful... Alright, regain composure and drop curtain.

Edit: Heh, looks like Dave covered this a *just* few days ago so have a look if you've not already

So, there are these little devices that we love that store a charge called a capacitor. Not only are they handy and cool, especially aluminum electrolytic types when you force catastrophic failures, but they are highly necessary. We know they for their general characteristics of holding a charge to delay a buildup of voltage, to act like a itty bitty battery and smooth output voltages, and to decouple power from ground by doing both. They are also essential to a lot of voltage conversion circuits and blocking DC voltage. As I mentioned in the last log, I had to make some revisions due to the input capacitors being woefully inadequate. This is because the characteristic of their electrolytic, X5R, X7R, et al, has a great impact on their performance and this is what I'm writing about now.

Small pet peeve first. In MLCC-type caps the electrolytic is generally referred to by that X7R-like designation, but that isn't what the electrolytic is. This designation is a descriptor for the performance of the capacitor over it's operating temperature, which is also defined by the designation. Basically, it's a adjective, not a noun, but everyone treats it as a noun. If you get confused by the datasheets and searches, this is why.

Now, I'm dealing with a number of voltages on my mainboard simply because I have to output a several different types. By nature of simulation, I have up to 20V that I have to deal with an a regular input of 14.4V too. Things become a bit nicer with my 5V and 3.3V rails though.

If you're somewhat familiar with caps you know that they both block and let pass DC when they first charge they have to let current pass, but as the charge builds, less and less is passed. If you want a fully detailed explanation of this, feel free to check out Dave's video on the topic. It's ironic that because of this little detail, it makes perfect sense that as you add more and more DC to a capacitor that it's effective capacitance decreases and this is something I don't think most of us pay attention to. I certainly never did until yesterday. Why does it matter? Well, when you're choosing capacitors for a power applications, and I'm strictly speaking MLCC's, we have to pay attention to the voltage output.

Unlike resistors which are relatively simple devices under DC, caps become a bit crazy. We cannot just take the output voltage and scale it by 1.5x or 2x to determine the voltage rating of the capacitors you need. You have to pay attention to the electrolytic type that is used. When a datasheet says "we strongly suggest" a X7R cap, listen to them. The difference between 2 similar electrolytic types can be dramatic simply because of DC bias. Also, keep in mind that the DC Bias characteristic is affected by SMD package size too. My first clue that things were going to be complicated was this article by Maxim Integrated.

How about some examples so you can see how tricky this is, ok? Here are 2 caps that are both 1206 package and rated at 10uF. The first one is a X7R-type with a 30V rating and the other is a X5R-type with a 50V rating. Which do you think would be better?

Look closely at the Y-axis. Except for the extra "20 volts" the performance between the two is nearly identical. Okay, so this one stacks up pretty well though the only "performance" you gain with the 50V version is that your cap won't fail at at the lower voltage.

Next up is a sample between two 47uF caps, both 1206 package. The first is a X5R-type with a 25V rating, the other is a X7R-type with a 15V rating:

The difference between these two is dramatic. You might think, "eh, it's 25V and I only need it to smooth a 5V output, should be a piece of cake." At 1/5th it's DC rating, the X5R cap is derated 65% where as the X7R cap is derated only 10%. What you see above is more common difference for the values I needed.

The trouble is that the smaller your caps, the higher the likelihood that they'll have a poorer electrolytic type. Adding to the pain is that the higher the capacitance value also means higher the likelihood of a poorer electrolytic type. I've spent the better part of 12 hrs just trying to find 4 good caps in decent package sizes for D-DAQ. I even changed the design to actually use higher rated caps because their DC bias derating drops the capacitance to the value I need.

The lesson here is to not take any electrical component for granted. When it comes to MLCC caps, I suggest sticking with Murata or TDK because they both regularly publish the DC Bias graphs. The ones I pulled the screen shots from are from TDK documents. If you happen to have a TDK part number, the extended type with 4 extra digits at the end, paste it to the end of the following URL to get a webpage from them with interactive graphs including DC Bias:

http://product.tdk.com/capacitor/mlcc/detailed_information.php?lang=en&ref=us&part_no=

I so hope your endeavors are a bit less hectic. I'll release another update to D-DAQ and update the BOM shortly.

-MOB

Discussions