MLCC X5R capacitor degradation

[CosteC]

I got interesting issue.
There is 0805 4.7 uF 25 V X5R capacitor working in half-bridge bootstrap circuit. Capacitor is charged from 12 V, so well below max rating.
It looses capacitance slowly with number of operations - operations happen very slowly - one HB swing every couple of seconds.
After some time (thousands operations) capacitance drops to 3.5 uF or so.

Temperatures are almost room, 25-30'C. No mechanical shocks.

Any ideas what is mechanism of degradation and capacitance decrease?

[helius]

Class II dielectrics like X5R experience capacitance loss with DC bias.
You would have better results with a film capacitor or C0G ceramic.

I found this description: https://passive-components.eu/high-cv-mlcc-dc-bias-and-ageing-capacitance-loss-explained-2/

[mzzj]

High K dielectrics drift slowly down in capacitance. Heating up(soldering) temporarily resets the drift so fresly soldered X5R measures higher than 1 month old or 1 year old.
My guess would be that if you resolder the bad capacitor it ”works” again.

[CosteC]

High K dielectrics drift slowly down in capacitance. Heating up(soldering) temporarily resets the drift so fresly soldered X5R measures higher than 1 month old or 1 year old.
My guess would be that if you resolder the bad capacitor it ”works” again.

Not much. Capacitor measured in circuit (tracks cut to avoid thermal stuff) looses capacitance after some thousands of cycles. Does not seem to re-gain it "overnight". Happens in product and happens from "new" capacitors from Farnell (same type, who knows if same batch).

BTW you claim X5T slowly looses capacitance? With or without bias voltage? Any paper about it?

[floobydust]

MLCC degradation is actually quite bad, to the point equipment fails unexpectedly. There is a need for some standards on this, as Continental lobbied several MLCC manufacturers but I don't know where that left off. It also depends on the cap manufacturer and they did not say who the worst offender is.
Here is one curve showing it drops within HOURS- which allows you to make fake claims on a datasheet. And is your basic engineering nightmare why the circuit worked and no longer does.

[CosteC]

MLCC degradation is actually quite bad, to the point equipment fails unexpectedly. There is a need for some standards on this, as Continental lobbied several MLCC manufacturers but I don't know where that left off. It also depends on the cap manufacturer and they did not say who the worst offender is.
Here is one curve showing it drops within HOURS- which allows you to make fake claims on a datasheet. And is your basic engineering nightmare why the circuit worked and no longer does.

Do you have original/full paper? I think DC bias is not my case, but would like to read more about it. Maybe I am wrong.

[Siwastaja]

mzzj's "claim" is like claiming water is wet, it's not like some extraordinary claim requiring proof to be delivered to you. MLCC aging is obvious to nearly everyone here; yes, it can be pretty bad, and some old assumptions like "X7R is not that bad, only Y2V sucks" are not true. Just do your own homework and use Google, you will find as many papers and appnotes as you want.

[KE5FX]

MLCC aging is obvious to nearly everyone here;

There is absolutely nothing "obvious" about capacitance degradation at timescales on the order of hours.  What textbook(s) discuss this phenomenon?

[Siwastaja]

Textbooks? You are looking at completely wrong type of source. Textbooks are great for fundamental physics or methodology (e.g. circuit analysis), but for up-to-date practical information about components, you need to read application notes, reports, websites, even discussion forums such as this one.

The fact it can happen within hours is not obvious to everyone, but ageing itself, in timescale of months, is extremely well documented. What difference does it make? If it ages faster, it's only better as the problem will be revealed sooner.

[tooki]

High K dielectrics drift slowly down in capacitance. Heating up(soldering) temporarily resets the drift so fresly soldered X5R measures higher than 1 month old or 1 year old.
My guess would be that if you resolder the bad capacitor it ”works” again.

Not much. Capacitor measured in circuit (tracks cut to avoid thermal stuff) looses capacitance after some thousands of cycles. Does not seem to re-gain it "overnight". Happens in product and happens from "new" capacitors from Farnell (same type, who knows if same batch).

BTW you claim X5T slowly looses capacitance? With or without bias voltage? Any paper about it?

Can you share the capacitor manufacturer and part number?

[Vovk_Z]

Any ideas what is mechanism of degradation and capacitance decrease?

I don't have ideas, I just may recommend to use large caps, e.g. 1206 or even 1210. I gues 4u7 1210 don't degrade at all (comperativly).

[floobydust]

MLCC degradation is actually quite bad, to the point equipment fails unexpectedly. There is a need for some standards on this, as Continental lobbied several MLCC manufacturers but I don't know where that left off. It also depends on the cap manufacturer and they did not say who the worst offender is.
Here is one curve showing it drops within HOURS- which allows you to make fake claims on a datasheet. And is your basic engineering nightmare why the circuit worked and no longer does.

Do you have original/full paper? I think DC bias is not my case, but would like to read more about it. Maybe I am wrong.

I think it was this one: http://electrical-integrity.com/Paper_download_files/DCE11_200_slides.pdf

"Five MLCC manufacturers (AVX Corporation, Kemet Electronics, Murata Electronics Europe, Samsung Electro-Mechanic, Wurth Elektronik) and Continental Automotive as the end automotive user representative at the round table open discussion." HIGH CV MLCC DC/AC BIAS AGEING CAPACITANCE LOSS EXPLAINED
Summary of the outcome: "Open letter to AECQ-200 committee about MLCC class II Capacitance DC BIAS issues”

Continental is now managing “DC-Bias Aging” and training their engineers on MLCC values. The meeting did not yield any cap manufacturer improved ratings. Kemet AVX fellow was same guy Dave interviewed on X film caps failing. In the end, they have to qualify each capacitor: "Capacitance versus DC-Bias Ageing data measured at room temp., for 24h, at three points: 20%, 40% and 60% of rated voltage."

Class-II MLCC DC-Bias Aging Issues in Automotive Applications was 2021 update ok to read as well

edit: had wrong company for Mr. Demcko

[CosteC]

Any ideas what is mechanism of degradation and capacitance decrease?

I don't have ideas, I just may recommend to use large caps, e.g. 1206 or even 1210. I gues 4u7 1210 don't degrade at all (comperativly).

I am not so optimistic as I do not understand mechanism... also if I would have space for 1206 I could put tantallum or other there.
I suppose I need to make test couple of hypothesis...

KEMET... Exact number I need to check tomorrow.

[bdunham7]

BTW you claim X5T slowly looses capacitance? With or without bias voltage? Any paper about it?

https://www.digikey.com/Site/Global/Layouts/DownloadPdf.ashx?pdfUrl=91D84559BB484E21BD284CC0B41A4783

[Vovk_Z]

I am not so optimistic as I do not understand mechanism...

IDK mechanism too, but, e.g. if a problem is a current-related (pulsating current) then large size still must help, because of a large surface arrea and a large volume to dissipate a power.

[helius]

Kemet was same guy Dave interviewed on X film caps failing.

I re-watched that video yesterday, it was rather an engineer at AVX. (The failed capacitor was from neither company, but a Taiwan cheapie branded "Suntan".)

[helius]

I am not so optimistic as I do not understand mechanism...

IDK mechanism too, but, e.g. if a problem is a current-related (pulsating current) then large size still must help, because of a large surface arrea and a large volume to dissipate a power.

The mechanism for "aging" is rearrangement of the crystalline structure of the barium titanate. The stable structures are different above and below the Curie point, so the rearrangement follows an exponential decay pattern starting at the last time it was soldered.
The mechanism for capacitance loss with DC bias is that some of the barium titanate dipoles become "blocked" or unable to rotate with the AC field. This is proportional to the DC electric field, so it is a worse problem when the capacitor is smaller.
These are both elucidated in the link I gave in Reply #1.

[exe]

Good manufacturers provide either provide curves DC voltage vs capacitance, or simulators, like murata (https://ds.murata.co.jp/simsurfing/mlcc.html?lcid=en-us).

Considering it was 0805, going down from 4.7uF to 3.5uF at 12V is not that bad, imo. Could be worse (see screenshot of random 16V cap). Going with physically bigger mlcc helps with DC bias. Note that this degradation is nothing to do with voltage rating. All MLCCs are affected by this, especially the high-capacity ones.

As of ageing without DC bias, Idk about that, but won't be surprised if this happens, but slower.

[DavidAlfa]

This has completely changed my perception of ceramic caps.
-80% at 12V bias?? Holy crap!

[ejeffrey]

This has completely changed my perception of ceramic caps.
-80% at 12V bias?? Holy crap!

Yes.  And it isn't solved by derating the voltage.  A 35 V and 16 V capacitor of the same dielectric and volume will usually have the same effective capacitance at 12V.  If a MLCC manufacturer doesn't provide capacitance vs voltage plots you should probably just ignore them. 

The way to find the right parts is to use parametric sort to filter by thickness.  In a given footprint, the only way to get more capacitance at working voltage is to add thickness.  This will help you filter out all the capacitors that are not worth looking at.

[mawyatt]

Folks, please do a little research before placing all ceramic caps in the same category.

It's well known that C0G/NP0 ceramic material is highly stable with temperature, voltage and age. In fact this ceramic is so good that it's utilized in the high end high resolution DMMs as the main integrating capacitor in the multi-slope intergrating ADCs

However, here's an example of the typical cheap ceramic thru-hole disc that's commonly utilized for power supply decoupling. This capacitor is obviously made with a high K dielectric and shows very poor stability with applied DC voltage and temperature.

BTW these plots were produced by some new Python routines for use with LCR meters for frequency, voltage and temperature sweeping and plotting, you can find details here:

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

Best,

[T3sl4co1l]

This has completely changed my perception of ceramic caps.
-80% at 12V bias?? Holy crap!

I'd have thought you saw this already; you've been around a while!

Well, welcome to the daily 10,000, and, a good reminder that this will always be new to someone, and so always bears repeating (a thread subject that's otherwise been done to death)!

Folks, please do a little research before placing all ceramic caps in the same category.

It's well known that C0G/NP0 ceramic material is highly stable with temperature, voltage and age. In fact this ceramic is so good that it's utilized in the high end high resolution DMMs as the main integrating capacitor in the multi-slope intergrating ADCs

However, here's an example of the typical cheap ceramic thru-hole disc that's commonly utilized for power supply decoupling. This capacitor is obviously made with a high K dielectric and shows very poor stability with applied DC voltage and temperature.

BTW these plots were produced by some new Python routines for use with LCR meters for frequency, voltage and temperature sweeping and plotting, you can find details here:

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

Best,

Indeed, type 2 (X7R etc.) are also highly microphonic (piezoelectric), so are contraindicated for sensitive and low frequency (e.g. audio processing) applications, perhaps controls as well (when precision and stable controls are demanded, or when the environment is high vibration).

C0G are nearly ideal; they often have Q factors in the thousands, making them excellent for snubber and resonant applications.  They have low absorption and leakage too (which exact brands/families do, however, is a carefully guarded secret), although not quite as low absorption as polystyrene or glass did (I think?).

The main downside to C0G is cost: values under 1nF are competitive with X7R etc., but above, they diverge quickly*; and values above 100nF are hardly available at all, and quite pricey at that.  Selection doesn't improve at low voltage ratings, because the ceramic layers are just too thin -- density isn't good.  (The material has a lower K than titanates, but much higher breakdown field strength.)  On the other hand, they excel at high voltages: you can hardly find a 100nF X7R that's any good (read: >= -30% of nominal value) at 250VDC, even in 2220 size; but C0G doesn't depend on voltage at all and a, 1210 I think?, will fit such a value and voltage rating.  The energy density at higher voltages (200-600V say) in fact is competitive with electrolytics!

*If you're *really* sensitive on cost, or doing large quantities (100k+?), probably the divergence falls at an even lower point.  In the thousands, for hand-wavy pricing, it's kind of just in the noise below 1nF.

Interesting quirk: the Q is so high, and the values small to modest, that it can be kind of awkward to wire C0Gs in parallel arrays.  Even the minor stray inductance along a tightly packed array (say 6nH for a half dozen in a row over ground plane) is enough to resonate them, and you get bathtub-sloshing resonant modes, say in the 10s to 100s MHz -- which can therefore be relevant for EMI purposes, say in SMPS.

Tim

[floobydust]

It's calling out the crappy MLCC manufacturers - ones having poor voltage coefficient, hiding that data and then the aging of that parameter which the industry is sweeping under the rug. Some manufacturers have better technology.
FFS look at the graph, it infant ages in under 3 minutes 
Continental's interim solution is requiring qualification and validation from manufacturers. And what about us?

Kemet was same guy Dave interviewed on X film caps failing.

I re-watched that video yesterday, it was rather an engineer at AVX. (The failed capacitor was from neither company, but a Taiwan cheapie branded "Suntan".)

Thanks, fixed. Open Letter to AEC-Q200 Committee on MLCC DC BIAS Ageing Capacitance Loss Issues had present MLCC Manufacturers:
AVX Corporation: Mr. Ron Demcko
KEMET Electronics: Mr. Wilson Hayworth
Murata Electronics Europe: Mr. Yuki Nagoshi
Samsung Electro-Mechanics: Mr. Benjamin Blume
Würth Elektronik: Mr. Paul Le Nezet

They didn't name the dog.

[DavidAlfa]

I'd have thought you saw this already; you've been around a while!

I knew capacity rating was reduced by bias, but not that it was so extreme like -80% lol.
I'm no engineer neither I work at hw design, so yep there's still a lot to learn!

[Siwastaja]

I'd have thought you saw this already; you've been around a while!

I knew capacity rating was reduced by bias, but not that it was so extreme like -80% lol.
I'm no engineer neither I work at hw design, so yep there's still a lot to learn!

A lot of damage was made a few years ago on blog posts, youtube videos and this very forum by "helpful" people teaching around that the issue is mainly with Y5V capacitors and X7R is fine. Some rules of thumbs pulled out of thin air, such as "Y5V can drop below -80% under full DC bias but X7R only -30%" or "derating voltage helps". These are all totally false.

The best rule of thumb, by far, is simply: if the volumetric energy density is too good to be true, then it isn't. As ejeffrey writes above, if footprint size is a fixed parameter for you, then the height is relevant to look at. And you should really look at joules per volume. And calculate the joules at your DC bias voltage; the rating is meaningless. E / V = ½CU^2 / (width * length * thickness).