Measuring uF range MLCCs

[Rooster Cogburn]

I noticed that all the MLCC capacitors I have in the 1uF - 100uF range measure consistently significantly lower (15-30%) than their spec. Since they all measure lower and they all cluster together close to the lower value, I figure it's not just a large tolerance or a bad measurement but some other phenomena. I remember reading that measuring high value MLCCs can be tricky, like here:

https://forum.digikey.com/t/testing-high-capacitance-mlcc-s/34

So is what I'm seeing to be expected? I've tried measuring with a Peak ESR70 Gold and a Fluke DMM.

Thanks!

[wraper]

What you see is aging, heating them above curie point will reset the process.

[Rooster Cogburn]

Now that's interesting, I had no idea that's a thing. But you're absolutely right, I just soldered a few to a perfboard and now they measure significantly above their rating. But they'll be back to the lower value in a few months/years anyway. So, how should I deal with this? Just buy MLCCs with higher capacity than what I would've bought if I had gotten electrolytics? Also, is there any truth to the measurement difficulty claims or should I expect accurate readings from my DMM/ESR meter?

[T3sl4co1l]

Ah, since you're one of the daily 10k, https://xkcd.com/1053/ let's also add the other effects:
- C(V) -- capacitance varies with bias voltage, rated value is not what you will get under load.  Check characteristics sheet; if they don't provide chars, keep shopping.
- Rated voltage: irrelevant to C(V) curve.  Some are -20% at rating, others -90%.  Only means a safe operating value, guaranteed not to break down (fail shorted).
- Aging: C decreases as roughly 1/log(time since annealed).  Which is a very slow process, if you don't need centuries/millenia of product lifetime anyway.
- Cracking, of course.  Soft termination and other mitigating types may be of interest, for high-rel/high vibe applications.  Not usually an issue, but do avoid high stress locations, like near mounting screws, etc.
- High-K types are worse on just about all these factors; generally avoid Z5U and such, with X5R being acceptable and X7R preferred, especially for equipment that will run on the hot side.
- C0G (and related type 1 dielectrics) is an essentially ideal capacitor.  Even dielectric absorption is low enough for many sensitive analog applications (precision S&H, etc.), and ESR is extremely low.  This is somewhat a concern, in that, it can be difficult to even wire them in parallel without getting stray resonances between them! Whether this affects your circuit's performance, or function, or EMC, obviously depends, but it's at least interesting that this can even be a concern.  So, just pay attention to PDN (power distribution network) impedances and peaks more often.  X7R etc. are somewhat susceptible (say when wired with trace lengths rather than pours, or in adjacent stacks), films vary, and electrolytics generally don't (ESR dominates at most any would-be resonant frequency).

Related topic: ferrite beads.  Both materials are "ferroic", type 2s being ferroelectric, beads being ferromagnetic.  It should be no surprise then, these two components have similar characteristics.  Namely, the L(I) (or Z(I)) of ferromagnetic parts is a matter of concern, and particularly low with ferrite beads, which can saturate (say 30% reduction in |Z|) in the 10mA-4A range (varies with Z rating and chip size).  Rated current, then, is only the safe operating current -- a thermal rating, irrelevant to the Z characteristic.  FYI, Laird publishes characteristic curves on almost their whole product line; others rarely do.

Tim

[Rooster Cogburn]

Yeah, sorry, I'm just a complete amateur when it comes to this stuff, no question about it.

I've been looking at the higher capacity Class 2 MLCCs as replacement for electrolytic caps during repair. I knew about a few of the downsides like capacity loss with DC bias and being more susceptible to cracking due to obvious mechanical differences. Aging was a new one for me :/

I had a look at this article and found it helpful:

https://www.doeeet.com/content/eee-components/passives/high-cv-mlcc-dc-bias-and-ageing-capacitance-loss-explained/

Especially the 'Summary of MLCC class II capacitance loss behaviour in operating conditions' section.

From what I understand the MLCC will drop like 5% for each 'time decade'. So that's 20% in a little over a year? That doesn't seem so slow. Sure, it won't go down much after that, but that still means I basically need to buy 20% higher capacity to account for this.

[T3sl4co1l]

Yup, pretty much. And if you need accurate values, you're looking at the wrong thing anyway.  Fortunately, power supply applications are often more like +/-2x tolerance, so a modest reduction due to C(V) and aging is tolerable.

If you need accurate and stable values, polymers to some extent, and film if possible, are the way to go.  Or C0G in smaller values or higher voltages, or tantalum in larger values (preferably avoid, part cost + conflict minerals).

Tim

[wraper]

There are many issues with such replacements. If you do that for LDO or buck converter, they may start oscillate due to too low ESR. If you put them into audio path, now you have microphonic effect and likely distortion due to voltage dependence of capacitance. God forbid you put them into any active filters. Not to say replacement may need lower or higher capacitance rating. If it's say buck converter operating at very high frequency, you may need lower capacitance because ESR and ripple current rating is what's important there, not bulk capacitance. Or higher because for particular circuit the amount of bulk capacitance is important, but it drops due to voltage bias, temperature and ageing.
Polymer capacitors are much more adequate as drop-in replacements for electrolytics, however they shouldn't be put anywhere where low leakage current is needed. And ESR may be too low for some things too.

[Rooster Cogburn]

I was just surprised you said about aging 'Which is a very slow process, if you don't need centuries/millenia of product lifetime anyway.' when there's a significant drop in the first year or two. After that it pretty much stops for all practical purposes, though.

I heard that voltage regulators need some ESR and that MLCC can vibrate and susceptible to vibration. How much of an issue is the latter, really?

Polymer/Film caps are a bit exotic and expensive, but are Tantalum a good drop-in replacement for electrolytics and can we expect a longer lifetime?

Interesting that you shouldn't use them in audio circuits, I would've expected that to be fine since we're talking about AC here.

[wraper]

Polymer/Film caps are a bit exotic and expensive.

Huh? Polymer and film capacitors are entirely different things. Neither they are that expensive or exotic. Modern computer hardware is loaded with them. They do not come in large voltage ratings though.

[Vovk_Z]

I've been looking at the higher capacity Class 2 MLCCs as replacement for electrolytic caps during repair.

Sometimes ceramic caps and electrolytic caps are interchangeable but not always. So, you have to consider it twice where you can do such replacement and where it is more safe use just new electrolytic caps. Ceramics MLCC have much smaller ESR than electrolytic caps so it may lead to excessive ringing, e.g. somewhere in power lines.

[Rooster Cogburn]

Polymer/Film caps are a bit exotic and expensive.

Huh? Polymer and film capacitors are entirely different things. Neither they are that expensive or exotic. Modern computer hardware is loaded with them. They do not come in large voltage ratings though.

Sorry, I shouldn't have mixed those two together.

I've been looking at the higher capacity Class 2 MLCCs as replacement for electrolytic caps during repair.

Sometimes ceramic caps and electrolytic caps are interchangeable but not always. So, you have to consider it twice where you can do such replacement and where it is more safe use just new electrolytic caps. Ceramics MLCC have much smaller ESR than electrolytic caps so it may lead to excessive ringing, e.g. somewhere in power lines.

Yes, exactly, I'm also concerned about the capacity drop with DC bias and I obviously just learned that they also might drop like ~20% capacitance 18 months after I soldered them in. Makes me think if maybe fixing something with reflowing solder joints really just fixed it because the aging process on the MLCCs has been reset

[T3sl4co1l]

MnO2 tantalum aren't recommended for power supplies as they are prone to ignition when used on low-impedance circuits.

Polymers are, I think, whatever -- anything's fair game.  Mind the recommended derating levels.
https://www.vishay.com/docs/40246/tantcapsvoltderautoindapp.pdf
https://www.avx.com/docs/techinfo/Tantalum-NiobiumCapacitors/PolymerGuidelines.pdf

Tim

[Zero999]

I heard a story about a company having problems with a batch of ceramic capacitors oscillating in a regulator circuit. They measured the values and found they were over 25% above the specified value. After discussing it with the manufacture, it turned out it was because their own fault, as they were baking them for a long time to drive out the moisture.

[wraper]

I heard a story about a company having problems with a batch of ceramic capacitors oscillating in a regulator circuit. They measured the values and found they were over 25% above the specified value. After discussing it with the manufacture, it turned out it was because their own fault, as they were baking them for a long time to drive out the moisture.

Utter BS. Capacitance can rise above spec. However will soon drop to spec. IIRC specified value is given after 24 hour ageing or something like that. Not to say even if the issue with capacitance rise was true, it was a broken design to begin with.

[szlovak]

- C(V) -- capacitance varies with bias voltage, rated value is not what you will get under load.  Check characteristics sheet; if they don't provide chars, keep shopping.

This is the downside I learned the hard way (not entirely the hard way) about MLLC capacitors with high capacitance. It should be written in the first lines in red text of every data sheet that in general MLLC class II capacitance is a gimmick, BS, a pure marketing lie. I have just surprised 3 engineers how much capacitance drops off after reaching maximum voltage or even being in the middle of the range. We have a couple of 6.3V capacitors supplied at 3.6V, so we have 1/3 the capacitance. On top of that there is an additional -20% drop in capacitance over time.

[Siwastaja]

This is the downside I learned the hard way (not entirely the hard way) about MLLC capacitors with high capacitance. It should be written in the first lines in red text of every data sheet that in general MLLC class II capacitance is a gimmick, BS, a pure marketing lie. I have just surprised 3 engineers

In principle, I agree it would be nice to have said red text.

On the other hand, having an engineer who does not know this thing in year 2024 is alarmingly weird. It is such well known feature that I question the ability of these 3 engineers to solve any other common problem.

I mean, where does this end? Should the datasheet warn in red letters about very low ESR which can cause instability or voltage overshoot on hotplugging? Should inductor datasheets mention that effective inductance is lower than nominal at saturation current? If not, why not - it's the same thing after all.

In the end, you can't just design based on datasheet alone. A doctor cannot perform a surgery after reading a datasheet of their knife. You need to educate, read the relevant literature, appnotes, discussions... This is called expertise, and this is why we (hopefully) get paid for our job more than at McDonalds.

There are no ideal components of any type. Even choosing a resistor can be a difficult task. Capacitors and inductors even more so. And the C(V) characteristics of MLCCs really isn't some kind of hidden knowledge. Heck, even just Wikipedia, which is a general-purpose resource not an EE data bank, has a section about it: https://en.wikipedia.org/wiki/Ceramic_capacitor#Voltage_dependence_of_capacitance

[wraper]

^The problem is that type designation such as X7R is temperature related but voltage dependence is rarely talked about and usually isn't obvious from datasheets if mentioned at all. Probably because it wasn't such of a big problem until very high capacitance in a small package became widely used, so it became a dominating capacitance drop factor in supposedly quite stable dielectric type. So the problem is that you don't know what you don't know, and will have no clue until see it mentioned somewhere. I have a feeling that most people on this forum don't know about it.

[Siwastaja]

voltage dependence is rarely talked about

Maybe in 2010. I think during last decade or so we don't see any discussion about MLCCs without discussion about DC bias dependency.

So the problem is that you don't know what you don't know, and will have no clue until see it mentioned somewhere.

Yes, but isn't this the case for everything in engineering? You can't do this job without spending at least some time reading through material (books, courses, internet forums, appnotes, whatever). In 2010 not finding mentions about MLCC voltage dependency was a real issue, and old course books etc. probably still don't mention it, but you also can't do this job without some minimum level of engagement into following the news and discussions.

I have a feeling that most people on this forum don't know about it.

Maybe still in 2010-2015, I remember back then we saw a lot of threads popping up because people were surprised as they learnt something new (see the xkcd reference above). This has almost completely stopped, the very comment I'm replying to being a rare exception, like a blast from the past. I take this as, "everyone already knows about it, to the point it's not a sexy discussion starter anymore".

But in 2010 time was ripe for the mass learning effect - the phenomenon had been "well known" and "documented" as hidden information among the True Experts for at least a decade, and the information was in the stage of fast spread all over forums, coffee table discussions, blog posts, websites etc., so number of people who heard about it for the first time was high every day.

But now, if a engineer who claims to do this work seriously has still not heard about this relationship, I'm very concerned. I mean, every other random beginner will know it. Do these engineers know that inductors saturate? Exactly the same thing. Do they know how to derate resistor power ratings per ambient temperature and cooling conditions? Do they know how to design in a transistor?

This being said, it is surprisingly often that, even on this forum, we see people suggesting that Vgs_th of a MOSFET is the parameter which tells us at which voltage the part is fully conducting at.

And don't get me wrong, I don't like hidden information. But we also need to be realistic here, you can't convey every piece of design information in red alarm font on the datasheet front page. T3sl4co1l listed many important points for MLCCs and there are even more. For example the tendency of cracking is at least equally serious. Appnotes are for education. If you copy-paste all important information in big red font on every datasheet, they get impossible to read.

[JohanH]

It has been mentioned repeatedly, but still isn't common knowledge. To lessen the impact of DC bias with MLCC, use as large package as possible.

[szlovak]

It is like car producer advertised in official data that their car can go 150mph but only when driving very steep road down. It's exactly the same type of passing information when we compare other types of capacitors, it simply doesn't happen that parameters worsen by that much in other parts. These capacitors almost don't exist at their maximum voltage, you get 1/5 of it's rated. It should be described as a half of its capacitance with a huge capacitance gain at low voltage and some loss at higher.

[T3sl4co1l]

where does this end? Should the datasheet warn in red letters about very low ESR which can cause instability or voltage overshoot on hotplugging? Should inductor datasheets mention that effective inductance is lower than nominal at saturation current? If not, why not - it's the same thing after all.

I mean, they do; saturation is usually a defined margin, e.g. -30%.  There's no general requirement sadly, but, often inductors plot L vs. Idc, which is nice.

I'd settle for if they put the fucking C(V) curve in the datasheet on the "Datasheet" link every time. But most of the time it's just the generic catalog with no part specifics.

Tim

[mawyatt]

The primary reason we developed this DC Bias Adapter and later Plotting routines for our LCR Meters was to investigate how various ceramic capacitors and different OEMs behaved wrt DC applied voltage (later temperature with crude setup). Somewhere within these and other post we did wrt DC bias and LCR Meters/Fixtures are some plots of high dielectric constant ceramic capacitors vs DC applied voltage (and temperature).

https://www.eevblog.com/forum/projects/bias-network-for-lcr-meter/

https://www.eevblog.com/forum/testgear/lcr-meter-plot-software/

With exception to NPO/C0G types, almost all ceramic capacitors should be considered voltage/temperature/time variable capacitors 

Best,

[T3sl4co1l]

Yes, to be clear, C0G (and AFAIK, all other type 1 ceramics?) are much better.  In fact... they're essentially ideal (the difference being tempco, which may be a desirable property or not, depending).  It's actually a bit disturbing just how good they are: for example, you'd think two 10nF 1210 chips, laid side by side (minimum distance but not touching), would simply act in parallel, but no, there is in fact a resonance between them (~40MHz), with quite low impedance, but absolutely measurable.  This could make for EMC surprises if such a loop happens to couple into wiring or fields.  You take the loss of most components for granted, but when parts have Q upwards of 3k, you start to take notice!

Tim

[tooki]

where does this end? Should the datasheet warn in red letters about very low ESR which can cause instability or voltage overshoot on hotplugging? Should inductor datasheets mention that effective inductance is lower than nominal at saturation current? If not, why not - it's the same thing after all.

I mean, they do; saturation is usually a defined margin, e.g. -30%.  There's no general requirement sadly, but, often inductors plot L vs. Idc, which is nice.

I'd settle for if they put the fucking C(V) curve in the datasheet on the "Datasheet" link every time. But most of the time it's just the generic catalog with no part specifics.

Tim

Yup.

For critical ceramic caps, I often don’t bother looking past Murata, simply because their website has a prominent, comprehensive app that shows the curves for a given capacitor. Most brands have something equivalent, but more hidden or harder to use.