EEVblog #1037 - Solving Ceramic Capacitor Cracking

[EEVblog]

Dave demonstrates various practical PCB and component selection solutions and techniques to prevent flex stress cracking of Multi Layer Ceramic Capacitors (MLCC) on PCB's and preventing your product from catching on fire!
Open Mode capacitors.
Flexible Endcap Soft Termination Capacitors
Lead Frame Capacitors
Series Capacitors

TDK Solutions for Ceramic Capacitor Cracking
https://product.tdk.com/info/en/products/capacitor/ceramic/mlcc/technote/solution/mlcc02/index.html#anc01

Soft Termination:
https://product.tdk.com/info/en/catalog/datasheets/mlcc_commercial_soft_en.pdf

Open Mode: https://product.tdk.com/info/en/catalog/datasheets/mlcc_commercial_openmode_en.pdf

AVX Flexiterm:
http://www.avx.com/products/ceramic-capacitors/surface-mount/automotive-mlcc-with-flexiterm/

Kemet Flex Termination
http://www.kemet.com/Lists/ApplicationBriefs/Attachments/58/Flexible%20Termination%20Technology%20(FT-CAP).pdf

[T3sl4co1l]

It's too bad the leadframe parts are so horrendously expensive.

Also, I wonder what they use, they're clearly soldered but is it lead (RoHS by exemption -- high temperature solder?) or some other alloy?

Tim

[joeqsmith]

Shown looking at the cross section of a ceramic.  You can see the layers.  Also shown is looking at the resistance of each layer. 

[EEVblog]

It's too bad the leadframe parts are so horrendously expensive.
Also, I wonder what they use, they're clearly soldered but is it lead (RoHS by exemption -- high temperature solder?) or some other alloy?

You would presume it's not solder because they would just fall apart in the reflow oven. So would be interesting to find out what.

[trophosphere]

It's too bad the leadframe parts are so horrendously expensive.

Also, I wonder what they use, they're clearly soldered but is it lead (RoHS by exemption -- high temperature solder?) or some other alloy?

Tim

I know in at least the deep earth exploration field for high temperature solder that one would use gold based alloys such as Au-Si and Au-Ge. They have melting points in excess of 350C.

[tszaboo]

It's too bad the leadframe parts are so horrendously expensive.

Also, I wonder what they use, they're clearly soldered but is it lead (RoHS by exemption -- high temperature solder?) or some other alloy?

Tim

I know in at least the deep earth exploration field for high temperature solder that one would use gold based alloys such as Au-Si and Au-Ge. They have melting points in excess of 350C.

I don't see, how reflowing the solder between the leadframe and the cap would cause any issues. The same way solder will align components, the same way, solder will keep the capacitor together. Surface tension.

And yes, the leadframe types are really expensive. One of the solution is to just use through hole capacitor, because those are pretty much the same benefits (maybe higher ESL) but they dont have this inflated price.

Also, this is more of an issue, if you have a large board, and really stressed areas. For small boards and simple stressed areas, like an USB connector this is not an issue. For 0603 and even 0805 this is rarely an issue.

Just be aware, where it is needed. And dont start placing these capacitors on the next arduino compatible boards please.

[unclebob]

Hi Dave, you mentioned that the leadframe type caps can help with microphonics. I'd love to see a video about to which extent you can silence a shrieking power supply with these caps.

[chris_leyson]

The Kemet KPS stacked caps will go through a standard lead free reflow profile without any problems. https://content.kemet.com/datasheets/KEM_C1020_X7R_KPS_SMD.pdf page 9. If you are prototyping with these parts don't apply heat for too long otherwise they will fall apart 

[T3sl4co1l]

I don't see, how reflowing the solder between the leadframe and the cap would cause any issues. The same way solder will align components, the same way, solder will keep the capacitor together. Surface tension.

We're not talking 0603s here.  2220 (and larger) chips will slump to the ground and fall apart.  Try it   (Well... if you can find any... I'd gladly try, but I don't have any. )

Huh, would be really interesting if they were spot welded or something like that, and just coated with solder for connectivity.  I wonder if end caps can tolerate spot welding?...

And yes, the leadframe types are really expensive. One of the solution is to just use through hole capacitor, because those are pretty much the same benefits (maybe higher ESL) but they dont have this inflated price.

Well... maybe.

Like this one, it's probably just a regular 1812-ish chip, with leads tacked on and dipped in epoxy:
https://www.digikey.com/product-detail/en/murata-electronics-north-america/RCER71H106MWM1H03A/490-7518-3-ND/4277142
so it's not really extreme priced.  But these things,
https://www.digikey.com/product-detail/en/avx-corporation/SE045C106KAR/478-5612-ND/2060743
Are they one huge chip?  Multiple stacked (side by side)?  Or these,
https://www.digikey.com/product-detail/en/united-chemi-con/KTD101B226M90A0B00/565-4652-ND/6189636
(oh hey, I didn't even realize UCC made ceramics), where their diagram shows a single chip, but who knows, that's just a representation anyway.  And then there's this,
https://www.digikey.com/product-detail/en/avx-corporation/ST205C107MAN10/478-5068-ND/1913137
but it's pretty clear what that is.

There's also TDK's potted Ceralink modules, which are probably somewhere inbetween.

(These were selections from a simple search on DK, ceramic capacitors, through hole, equal or greater 10uF 50V.)

Also, this is more of an issue, if you have a large board, and really stressed areas. For small boards and simple stressed areas, like an USB connector this is not an issue. For 0603 and even 0805 this is rarely an issue.

Just be aware, where it is needed. And dont start placing these capacitors on the next arduino compatible boards please.

Huh... what do you mean by "simple stressed areas"?  Taking that literally, I can't think of a good reason why the stresses around a connector should be simple; they'll have complex curvature, up, down and around...

I don't think that's what you meant, though, really that the situation itself is 'simple', like, who cares about USB, right?  But then, what's so simple about that -- if you don't need a capacitor there, then why put it there at all?   (Which is a valid and important question for most USB connector situations, I might add!)

Anyway, smaller caps I think are less vulnerable on account of their being smaller than the board thickness, therefore the board beneath them looks much stiffer than otherwise.  Though they aren't as elastic or strong either (smaller chip, less elastic stretchy range -- which is to say, same bulk elasticity), so it's not clear if that should work in our favor, or wash out.

Ah, think of it this way: a relatively large cap will see torque (bending) from its electrodes, as well as stretching due to the board bending.  Whereas a very small cap sees only the stretching of the board surface.  This does not save smaller caps in any way (the percent stretch of board surface is independent of scale, so an 01005 is just as prone to failure as an 0402 on, say, a 3mm thick board), but disadvantages large caps (an 1812 on 1mm board will see significant bending as well as tensile strain).  You'll see the difference in the fracture shape, where bending tears the corners off the chip, whereas stretching either shears the pads off the chip, or cracks it more vertically (I think?!).

Tim

[David Hess]

It's too bad the leadframe parts are so horrendously expensive.
Also, I wonder what they use, they're clearly soldered but is it lead (RoHS by exemption -- high temperature solder?) or some other alloy?

You would presume it's not solder because they would just fall apart in the reflow oven. So would be interesting to find out what.

It can still be solder but with a higher melting temperature.

For boards which have to make more than one pass through the soldering process, a higher melting point solder can be used on the first pass so those parts will not desolder on the second pass.

[free_electron]

Anything larger than an 0603 must be kept away from mechanical stress.

Here are the traps for beginnning players:

- mounting close to any plug-in connector or screw terminal connector is a no-no. if they need to be close they must be placed perpendicular to the flex point. ( if mounted between two adjacent pins they should be installed with their long side facing the pins. )

O | O  is ok
O --- O is NOT ok.

so that cap under those screw terminals is a big mistake. too close and wrong orientation.

- mounting close to board edge is a no-no , especially if the boards are v-scored and need to be broken apart. Again capacitors should be placed perpendicular to edge

bad :
------------ edge   
     --- cap

good :
------------ edge
      | cap

- ceramic capacitors on high-powered unprotected rails should be placed two in series , and under a 90 degree angle from each other. that way if one fails short the other one prevents a DC path. Definition of such a rail is a power rail that is unfused or not current limited.

(+) ---||---||--- (-)

placement :
  --- Cap 1
| Cap 2

- Keep capacitors away from mounting holes. the board flexes under vibration . mount perpendicular to hole

O| : not ok
O - : ok


- use soft electrode caps.
- use thru-hole caps as the legs can dampen the stress.

[HwAoRrDk]

Hmm, after watching this video I'm now nervous about a PCB design of mine where I'm using a 1206 10uF MLCC across the primary power input rails.

I originally wanted to use an electrolytic, but in the end physical constraints dictated it needed to be a surface-mount component. I'm in two minds as to whether I should make more of an effort to change it.

On the one hand, I guess a 1206 isn't 'huge', so is probably at much less risk of damage; also, the board is 1.6mm and no bigger than 50x50mm, so quite stiff anyway and not subject to flex at all. There's also no screw terminals or other sources of stress forces from user interaction. But, on the other hand, if this cap failed short it would probably be fairly serious, as the power feed circuit is fused at 20A, and such current would no doubt transform the cap into a black charred mess fairly quickly.

I notice in the video where Dave is talking about soft-terminated packages, etc. that such things are commonly AEC-Q200 certified. I have seen the term before, but don't know anything about it apart from that AEC certifications apply to stuff used in the automotive industry. From a quick search, I gather this is some kind of stress test qualification for high-reliability components?

Would simply using an AEC-Q200-certified cap (regardless of whether it's specifically a soft-terminated package or whatever) mean that I would get a component that is less susceptible to stress damage? They don't seem to cost much more.

[free_electron]

Hmm, after watching this video I'm now nervous about a PCB design of mine where I'm using a 1206 10uF MLCC across the primary power input rails. :-

On the one hand, I guess a 1206 isn't 'huge', so is probably at much less risk of damage; .

1206 is problematic. 1210 is better as it is wider.

[HwAoRrDk]

Ah, I've just realised that you can get 50V 4.7uF caps in 0805. Hadn't thought before that there would be that value and voltage available in that small a size. As I have room, I am going to change the single 1206 10uF for a parallel pair of 0805 4.7uF.

[tszaboo]

Ah, I've just realised that you can get 50V 4.7uF caps in 0805. Hadn't thought before that there would be that value and voltage available in that small a size. As I have room, I am going to change the single 1206 10uF for a parallel pair of 0805 4.7uF.

And they probably have some 500nF capacitance above 30V

Tim

I really dont know, why you had to strawman my argument.

[HwAoRrDk]

And they probably have some 500nF capacitance above 30V

Is this due to the 'DC Bias Effect'? I hadn't actually thought about that as I don't fully understand it. I kind of knew that MLCCs had reduced effective capacitances at higher voltages, but I didn't think it was that bad. Or is it just a problem with smaller packages? Or affects some dielectric types more than others?

In fact, I went and looked up the graphs for the 1206 cap I have at the moment; it too has a terrible curve. At the 12V I'd be running it at, it has just under half the capacitance.

So I went back to look at 1210 (or larger) caps, as free_electron suggested, and they're better - 75% @ 12V, but they're pretty expensive.

Urgh, so many compromises...

[CNe7532294]

Thank you for this Dave. We need more awareness on things like this. It is my hope that soft terminations or something even more impressive becomes the gold standard. Sure demand gets driven up immediately and along with that comes price. However, just like side airbags and ABS we may see this in such high demand that output will try to match up to it. In other words,  it should become cheaper in the long run. After all, who wants a product that can catch on fire.?

[tszaboo]

And they probably have some 500nF capacitance above 30V

Is this due to the 'DC Bias Effect'? I hadn't actually thought about that as I don't fully understand it. I kind of knew that MLCCs had reduced effective capacitances at higher voltages, but I didn't think it was that bad. Or is it just a problem with smaller packages? Or affects some dielectric types more than others?

In fact, I went and looked up the graphs for the 1206 cap I have at the moment; it too has a terrible curve. At the 12V I'd be running it at, it has just under half the capacitance.

So I went back to look at 1210 (or larger) caps, as free_electron suggested, and they're better - 75% @ 12V, but they're pretty expensive.

Urgh, so many compromises...

Nothing prevents you from using 0805 if you want it. Just use multiple in parallel. NP0 goes up to 330nF in 0805 package 63V. Nothing prevents you from connecting even 6-10 in parallel. This way, you sure end up with higher capacitance than just one 1210.

For typical voltage rails, I went ahead and calculated the (effective capacitance)/(price) ratio for oómy preferred manufacuter. I just requested a big quote with a few dozen parts. They had effective cap/voltage curves. I dont buy 1uF+ capacitors without those. So calculate the effective capacitance, for each voltage rail, and divide that with the price. If you need, say 47uF@5V I use a different part (6x10uF 0805, 10V rating I think) than 22uF@15V. Include assembly cost.

For some designs the price of the passives can add up, and this way you can reduce it significantly. At one point I was doing a design revision, started "applying my brain", and the end result was 15% BOM cost reduction (though I was swapping opamps and analog switches also). Management was astounded, they didn't believe their eyes. End of the year, I asked for a raise they gave me 3%. So f'em, I dont work there anymore. I'm sure they will find other engineers  who are capable of this.