EEVblog Electronics Community Forum

EEVblog => EEVblog Specific => Topic started by: EEVblog on June 07, 2014, 10:48:46 am

Title: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 07, 2014, 10:48:46 am
Dave explains, shows, and measures a potentially big trap with using high value ceramic capacitors.
If your 10uF capacitor really 10uF in your circuit?
Those humble X7R caps you think are a "stable" dielectric? think again...

Class II and above ceramic capacitors can vary their capacitance drastically with DC bias voltage level and also the applied AC voltage.
Links:
http://www.murataamericas.com/murata/murata.nsf/promo_dcbias.pdf (http://www.murataamericas.com/murata/murata.nsf/promo_dcbias.pdf)
http://www.avx.com/docs/masterpubs/mccc.pdf (http://www.avx.com/docs/masterpubs/mccc.pdf)
http://www.ece.ucdavis.edu/vcl/asap/asap_v1/docs/X7R_C.pdf (http://www.ece.ucdavis.edu/vcl/asap/asap_v1/docs/X7R_C.pdf)
http://psearch.murata.com/capacitor/product/GRM21BR60J106KE19%23.html (http://psearch.murata.com/capacitor/product/GRM21BR60J106KE19%23.html)
http://www.murata.com/products/design_support/simsurfing/outline.html (http://www.murata.com/products/design_support/simsurfing/outline.html)
http://www.avx.com/SpiApps (http://www.avx.com/SpiApps)
Capacitor Tutorial - Ceramics & Impedance
https://www.youtube.com/watch?v=TDDoi70cxw0 (https://www.youtube.com/watch?v=TDDoi70cxw0)
The piezoelectric effect demonstrated:
https://www.youtube.com/watch?v=KFCRB4d991E (https://www.youtube.com/watch?v=KFCRB4d991E)

EEVblog #626 - Ceramic Capacitor Voltage Dependency (https://www.youtube.com/watch?v=2MQyQUkwmMk#ws)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: rs20 on June 07, 2014, 12:22:34 pm
Wow, thanks for this video -- I never would have suspected this. Makes me wonder whether I should be overspeccing my capacitances, or my capacitor voltages more (e.g., using 12V caps instead of 6.3V caps for 5V rails) as I presume the capacitance dropoff would be smaller if you reach a smaller percentage of rated voltage in general (let's just ignore the counterexample in the video with the -90% dropoff on the 100V caps).

Also, I never thought that every little decoupling cap I had could be used in a pinch as a varicap/varactor. This could make the most confusing, obfuscated schematic for a VCO ever...
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: VK5RC on June 07, 2014, 12:26:00 pm
A great topic, Thanks
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: mmilejski on June 07, 2014, 12:35:46 pm
Aren't some of the letter codes in the table for tolerance reversed?
I have a bag of Y5V capacitors and it says that they are -20% +80% and not +20% -80%.
And googling it gives various results:
This is similar to the one Dave used: http://m.eet.com/media/1174806/table1.jpg (http://m.eet.com/media/1174806/table1.jpg)
and this one seems to be right: http://2.bp.blogspot.com/_c8EartLqdVs/TMmr-6Wn0WI/AAAAAAAAAkc/8A00KvYYR0g/s1600/Tolerance+code.bmp (http://2.bp.blogspot.com/_c8EartLqdVs/TMmr-6Wn0WI/AAAAAAAAAkc/8A00KvYYR0g/s1600/Tolerance+code.bmp)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: nathanpc on June 07, 2014, 02:14:07 pm
This is incredible, I never thought a DC bias could affect the capacitance that much. I'll probably do some experiments with some capacitors myself too.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: opablo on June 07, 2014, 02:33:22 pm
Is it possible that the errors come from the fact that the func gen could not be able to provide enough current to make the math work ?

Can you re-test one of the worst cases using a high current power supply and the dso triggering a single shot ?

You can see in your dso that the initial voltage is not being able to stay solid at the begining of the charge curve.


(I'm not an EE... just a hobbist... so if I'm wrong just point me wrong in a friendly way)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: AndreasF on June 07, 2014, 02:56:21 pm
I'm not sure if I'm missing the point completely, but I don't understand why this is any way surprising, in fact I would be very surprised if a capacitor didn't behave this way. Is this really unique to ceramic caps???

Given a bias voltage (i.e. a constant voltage difference between the two pins/"plates" of the capacitor) I would think that the capacitor already holds a certain charge*. The total amount of charge it is able to hold depends on the (total) voltage differential (in a non-linear relationship). Therefore increasing the voltage differential from 0 to 1V will add a lot more charge than changing the voltage differential from 5V to 6V, which will only be able to add a smaller amount of charge thus making the time constant shorter. So, yes, in a sense the capacitance changes, but mostly because some of its capacitance is already "used up" by the bias voltage.


*This can easily be demonstrated: as soon as the bias voltage is removed, the cap will discharge.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: T3sl4co1l on June 07, 2014, 03:38:42 pm
I'm not sure if I'm missing the point completely, but I don't understand why this is any way surprising, in fact I would be very surprised if a capacitor didn't behave this way. Is this really unique to ceramic caps???

Yes.  Ferroelectric materials saturate in the same way that ferromagnetic materials do.  Which is better appreciated among transformer designs, but inductors frequently miss the mark on that fact.

Quote
Given a bias voltage (i.e. a constant voltage difference between the two pins/"plates" of the capacitor) I would think that the capacitor already holds a certain charge*. The total amount of charge it is able to hold depends on the (total) voltage differential (in a non-linear relationship).

By definition, if capacitance is constant, the charge is directly proportional to (linear with) voltage, Q = V*C.

If C varies with V, then Q = V * C(V), which is kind of odd.  You can imagine then trying to solve the differential equation for charging a capacitor through a resistor, but having to substitute some nonlinear function into it...

Quote
*This can easily be demonstrated: as soon as the bias voltage is removed, the cap will discharge.

Leakage is not an intrinsic property of capacitance.  It is an inevitable result of imperfect materials and their physics, but nothing at all to do with charge or capacitance or bias.

The cells in an EPROM or Flash contain very high quality capacitors (high purity SiO2 glass dielectric), which leak something on the order of single electrons per month.

Tim
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: DJ on June 07, 2014, 03:43:57 pm
Aren't some of the letter codes in the table for tolerance reversed?
I have a bag of Y5V capacitors and it says that they are -20% +80% and not +20% -80%.
And googling it gives various results:
This is similar to the one Dave used: http://m.eet.com/media/1174806/table1.jpg (http://m.eet.com/media/1174806/table1.jpg)
and this one seems to be right: http://2.bp.blogspot.com/_c8EartLqdVs/TMmr-6Wn0WI/AAAAAAAAAkc/8A00KvYYR0g/s1600/Tolerance+code.bmp (http://2.bp.blogspot.com/_c8EartLqdVs/TMmr-6Wn0WI/AAAAAAAAAkc/8A00KvYYR0g/s1600/Tolerance+code.bmp)

Y5V's are +20/-80

www.avx.com/docs/catalogs/cy5v.pdf (http://www.avx.com/docs/catalogs/cy5v.pdf)

I avoid them and do not keep any on hand - don't want to inadvertently use one when prototyping a ckt




btw, there is also a voltage coefficient of resistance that was pronounced in old carbon composition resistors.

here's a couple articles

http://www.barthelectronics.com/pdf_files/Application%20note%201%20Voltage%20Coefficient%20Products_Pulse%20Page.pdf (http://www.barthelectronics.com/pdf_files/Application%20note%201%20Voltage%20Coefficient%20Products_Pulse%20Page.pdf)

www.vishaypg.com/doc?49997 (http://www.vishaypg.com/doc?49997)








Caps have all sorts of interesting characteristics,  dielectric absorption etc. Much analog fun.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: T3sl4co1l on June 07, 2014, 03:47:11 pm
Thanks Dave!  It is an important lesson that needs to be said more often!

A good rule of thumb to choosing ceramic capacitors:
- Don't pick anything worse than X7R if you can help it,
- Look for minimum double the voltage rating you need, preferably triple,
- Design your circuit to tolerate twice the capacitor's rated tolerance, to account for manufacturing, temperature, voltage and aging (+/-30% or so for X7R?).

I did this plot some time ago, because I felt like it:

(http://seventransistorlabs.com/Images/Z5U_CV_Plot.png)

- Random "203Z" ceramic disc from the junk box, in series with a much larger cap; the common node was biased with a large resistor (>100k?) to the voltages shown.
- Capacitance was measured by resonating with a known inductor, I think it was around 100kHz (varying, of course) and 1V rms.
- The capacitor was sitting around for at least ten years, so likely had aged since its original salvage.  I annealed it by heating each leg with the soldering iron (set for 350C) for 10 seconds, then measured again.

Would it be possible to do one of these videos for magnetic components as well?  The effect is, after all, exactly the same -- ferroelectricity and ferromagnetism are two sides of the same electromagnetic coin!

My pet peeve: ferrite beads.  I ask you: what's the point of a 10A ferrite bead when it saturates at 100mA?  Nothing, that's what!  Even if you dig through their database programs (Kemet, TDK, etc. all have their things), you rarely find bias data for ferrite beads.  If you need DC filtering, don't look at ferrite beads, look at the "inductor" listing -- they make multilayer ferrite chip inductors, outwardly identical to ferrite beads, that are just as cheap and plentiful, but actually perform correctly.

Tim
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Tek_TDS220 on June 07, 2014, 03:57:24 pm
It is scandalous that this is not mentioned on the datasheets, since it affects a large fraction of all applications.  I guess that using a type 2 ceramic cap as a DC blocker for audio applications would be awful.  How much distortion do you get by passing a 50 Hz sine wave (2 V p/p or so) through a ceramic cap?  I don't have the tools to measure distortion, but it might be enough to see on a scope.  I'll try it.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: NiHaoMike on June 07, 2014, 04:18:12 pm
My pet peeve: ferrite beads.  I ask you: what's the point of a 10A ferrite bead when it saturates at 100mA?  Nothing, that's what!  Even if you dig through their database programs (Kemet, TDK, etc. all have their things), you rarely find bias data for ferrite beads.  If you need DC filtering, don't look at ferrite beads, look at the "inductor" listing -- they make multilayer ferrite chip inductors, outwardly identical to ferrite beads, that are just as cheap and plentiful, but actually perform correctly.
Beware that inductors are designed for low losses while chokes are designed to be lossy. Using an inductor in an application that calls for a choke can cause unexpected ringing.

I wonder if anyone has used the voltage coefficient of ceramic capacitors to an advantage. They're basically big varactors...
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Alana on June 07, 2014, 04:27:14 pm
Big Thx, i think one of my projects failed due to exacly this "feature".
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: mariush on June 07, 2014, 04:38:04 pm
A bit offtopic

Dave, when you said at around 5:30  "Come with me" i almost said out loud "Dave, you watched too much Cosmos" :D


Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: gxti on June 07, 2014, 04:40:30 pm
All other things equal, higher voltage rating caps won't necessarily have better DC bias characteristics. Sometimes the 12V cap is just as bad as the 6.3V cap. X7R is better than Y5V for sure, but that isn't always enough either.

The best predictor for bias tolerance is the physical size of the capacitor. I stay away from 1uF caps smaller than 0805, and 10uF smaller than 1206.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: peter.mitchell on June 07, 2014, 05:13:41 pm
I find a lot of comments in this thread misleading.

The effect of DC bias on capacitance is significant, however the effect of a capacitors "voltage rating" on the effect of DC bias is not significant, eg, a 50v 0.1u capacitor will have a similar drop of capacitance to a 6.3v 0.1u capacitor if all else is the same.
The thing that most closely correlated to the effect of DC bias on capacitance is physical dimensions of the capacitor, eg, a 0402 will lose more capacitance than a 0805 and a 0805 will lose more than a 1210.

So as a general rule of thumb, get the physically largest capacitor you can that is practicable for your electrical specification needs.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: DutchGert on June 07, 2014, 05:28:59 pm
Another problem is that even if u are aware of the problem 99% of the LDO and Switcher Datasheets and Appnotes just mention a let's say "stable with 2.2uF". I mean, is that a 2.2uF rated cap or do the mean a 'true' 2.2uF at the regulated output voltage......? I tend to design according to the latter but u end up using a lot more board space or more expensive caps just because u want to be sure and whack in a 10uF cap at a double or more voltage rate.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Wytnucls on June 07, 2014, 06:06:47 pm
Here is the physical process, as explained by muRata.

The mechanism of DC bias characteristic:
In the high dielectric constant capacitor type of monolithic ceramic capacitors, at present mainly BaTiO3 (barium titanate) is used as a principal component of high dielectric.
BaTiO3 has a perovskite shaped crystal structure and above the Curie temperature it becomes a cubic shape with Ba2+ ions to the vertices, O2- ion to face center and Ti4+ ion in a body centered position.
At the Curie temperature (approx 125°C) or more, it has a cubic crystal structure, and below the Curie temperature and within an ambient temperature range, one axis (axis C) stretches and other the axes shrink and turn to a tetragonal crystal structure.
In this case, polarization occurs as a result of the unit shift of axially elongated Ti4+ ion crystal. This polarization occurs without applying an external electric field or pressure, and is known as "spontaneous polarization." As explained above, a characteristic that has a spontaneous polarization and a property of changing orientation of spontaneous polarization by an external electric field to reverse is called "Ferro electricity."
The reversal of the spontaneous polarization per unit volume is equivalent to relative permittivity. Relative permittivity is observed as a capacitance.
Without a DC voltage, spontaneous polarization can happen freely. However, when a DC voltage is externally applied, spontaneous polarization is tied to the direction of the electric field in the dielectric, and independent reversal of spontaneous polarization is inhibited. As a result, the capacitance becomes lower than before applying the bias.
This is a mechanism of decrease in the capacitance after applying DC voltage.


http://www.murata.com/products/capacitor/design/faq/mlcc/property/05_more.html (http://www.murata.com/products/capacitor/design/faq/mlcc/property/05_more.html)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Stonent on June 07, 2014, 06:20:38 pm
A bit offtopic

Dave, when you said at around 5:30  "Come with me" i almost said out loud "Dave, you watched too much Cosmos" :D

Well he did name his kid after Carl Sagan.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Skimask on June 07, 2014, 06:59:23 pm
A bit offtopic

Dave, when you said at around 5:30  "Come with me" i almost said out loud "Dave, you watched too much Cosmos" :D

Well he did name his kid after Carl Sagan.

Next kid to be named Neil?  Maybe Tyson?

(P.S.  For all practical purposes, I can't stand what N.D.T. did to Cosmos)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 07, 2014, 07:11:58 pm
Great episode!
Dave -- you're correct ... not a lot of experienced EEs know about this issue ... you sure know how to pick good topics ...so, how were you (Dave) first made aware of this issue? I.e., from prev. in-the-field experience, word-of-mouth, discussions/requests on this blog?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: mpep on June 07, 2014, 07:39:21 pm
Fascinating topic. Not something I've ever been made aware off!
Good point about LDOs by DutchGert.
How should you design for this effect? The measured (rated) capacitance, or use a higher rated capacitor and use the de-rated value to get to the 'correct' value?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: retrolefty on June 07, 2014, 08:47:37 pm
Here is the physical process, as explained by muRata.

The mechanism of DC bias characteristic:
In the high dielectric constant capacitor type of monolithic ceramic capacitors, at present mainly BaTiO3 (barium titanate) is used as a principal component of high dielectric.
BaTiO3 has a perovskite shaped crystal structure and above the Curie temperature it becomes a cubic shape with Ba2+ ions to the vertices, O2- ion to face center and Ti4+ ion in a body centered position.
At the Curie temperature (approx 125°C) or more, it has a cubic crystal structure, and below the Curie temperature and within an ambient temperature range, one axis (axis C) stretches and other the axes shrink and turn to a tetragonal crystal structure.
In this case, polarization occurs as a result of the unit shift of axially elongated Ti4+ ion crystal. This polarization occurs without applying an external electric field or pressure, and is known as "spontaneous polarization." As explained above, a characteristic that has a spontaneous polarization and a property of changing orientation of spontaneous polarization by an external electric field to reverse is called "Ferro electricity."
The reversal of the spontaneous polarization per unit volume is equivalent to relative permittivity. Relative permittivity is observed as a capacitance.
Without a DC voltage, spontaneous polarization can happen freely. However, when a DC voltage is externally applied, spontaneous polarization is tied to the direction of the electric field in the dielectric, and independent reversal of spontaneous polarization is inhibited. As a result, the capacitance becomes lower than before applying the bias.
This is a mechanism of decrease in the capacitance after applying DC voltage.


http://www.murata.com/products/capacitor/design/faq/mlcc/property/05_more.html (http://www.murata.com/products/capacitor/design/faq/mlcc/property/05_more.html)

 Dear God, please don't let the 'golden ear' crowd hear about this. They're sure to claim it's audible.  ;)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: WarSim on June 07, 2014, 08:55:01 pm
These fundamental Friday videos are beginning to worry me. 
There has not been a single one that taught me something new.  I am not an engineer.  I do not have a degree.  I have been employed in a junior engineers positions, and performed engineering duties in the past.  I wanted to use these videos to fill in the gaps because I do not have the formal degree. 
What do the people that made the electronic devices in my environment actually know.  Is it safe to to turn commercial electronics on anymore?  I know I have many things to learn in the engineering realm.  After all I am only a lowly Tecnologist.  Well that is what the idiot engineers I have met have told me.  :)
I will continue to watch these wonderful videos and wait for an introduction to something new to learn. 


Sent from my iPad using Tapatalk
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: VK5RC on June 07, 2014, 09:30:30 pm
A interesting quote from muRata, to me it may imply that capacitance in one polarity may be different than the other with DC bias.
I might need to have a look myself.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: T3sl4co1l on June 07, 2014, 09:37:23 pm
Indeed, there can be a frozen-in potential (electrets), and there can be polarization through various other physics (piezoelectricity, thermoelectricity, etc.).

Tim
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 07, 2014, 09:57:17 pm
Another problem is that even if u are aware of the problem 99% of the LDO and Switcher Datasheets and Appnotes just mention a let's say "stable with 2.2uF". I mean, is that a 2.2uF rated cap or do the mean a 'true' 2.2uF at the regulated output voltage......?

That''s the thing, you don't actually know. It's a suck it and see thing. Although if they recommend a 2.2uF ceramic, then you'd have to think they have tried a real 2.2uF ceramic. But what type? What size? Good app notes will have actual part numbers.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 07, 2014, 10:00:32 pm
These fundamental Friday videos are beginning to worry me. 
There has not been a single one that taught me something new.

Well, the title says it all, fundamental. If you know your fundamentals then there is nothing to learn. That's just the way it is. Sometimes I do lesser known ones like this, but if you know those too oh well  :-//
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: bitwelder on June 07, 2014, 10:11:22 pm
So all of this does *NOT* apply to Class I capacitors?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: JoannaK on June 07, 2014, 10:13:47 pm
These fundamental Friday videos are beginning to worry me. 
There has not been a single one that taught me something new.

Well, the title says it all, fundamental. If you know your fundamentals then there is nothing to learn. That's just the way it is. Sometimes I do lesser known ones like this, but if you know those too oh well  :-//

It's never possible t oplease everyone..  I liked this one, cause I have not used those particular caps and thus seen need to learn about this effect.  :-+

Afterall. There is no real component that would match the ideal (simplified) idea of the parts. Sometimes it comes and bites designer, sometimes not.

Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 07, 2014, 10:19:17 pm
so, how were you (Dave) first made aware of this issue? I.e., from prev. in-the-field experience, word-of-mouth, discussions/requests on this blog?

Not sure exactly when, but I first learned about non-linearity and effects like this in ceramic hydrophone sensors (they are essentially just physically large ceramic capacitors and have similar issues) at work in about 1995. And I've seen the odd effect in circuits a few times since then.
Don't recall really knowing about it before that. Of course that was pre-internet when this stuff was not easy to look up or stumble upon.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 07, 2014, 10:21:11 pm
Afterall. There is no real component that would match the ideal (simplified) idea of the parts. Sometimes it comes and bites designer, sometimes not.

Yes, you could certainly go your whole career without running across this issue in practice. Depends on your "luck".
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: eV1Te on June 07, 2014, 10:46:52 pm
I knew that ceramics were non-linear, but i would never have expected them to have this big of a non-linearity!

So why are ceramics used so often as decoupling in power rails, since that is the application where they behave the worst (small ripple voltage, large DC bias)? Wouldn't some film capacitors have better actual capacitance in this case, or are they to "slow"?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: c4757p on June 07, 2014, 11:00:03 pm
eV1Te - Two reasons: Ceramics perform better at high frequency, and ceramics are cheaper. If you want 1 µF at full bias, it's usually cheaper just to use 2.2 - 4.7 µF nominal in a ceramic than to use a film cap.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: DJ on June 07, 2014, 11:16:27 pm

<snip>

 Of course that was pre-internet when this stuff was not easy to look up or stumble upon.

Very true. When I started back in the late 70's, we each had large libraries of data books and application notes.

I assembled (and still have) binders specific to various passive components.

Covers a whole wall of my lab.

Many interesting things found, often lost and then rediscovered.  Even going back to my old Philbrick apps book has yielded insight.

I donated many of my books to a local university.  Wonder if they were ever used or even survive.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: BUkitoo on June 07, 2014, 11:33:32 pm
Very good video!
Suprisignly I learn about this capacitor behavior just a year ago, reading some papers. That I have checked some datasheets and I have found all the data reagarding the capacitance variation with DC voltage (the good brands, of course).

I would like to mention that I think Dave`s wrong when he says that a square waveform from 0 to 5v do not add a DC bias to the capacitor.  Of course the DC bias voltage is the mean value of that. It depends on the duty cycle, low and high level voltages. I think that the capacitance test should be done with a AC voltage with no mean value and take that capacitance value as the refernce point, and then add DC bias and check the difference.

bye!

Franco
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: W7NGA on June 08, 2014, 12:19:16 am
> I would like to mention that I think Dave`s wrong when he says that a square waveform from 0 to 5v do not add a DC bias to the capacitor.

Joseph Fourier agrees ...

dan W7NGA
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: rs20 on June 08, 2014, 12:37:24 am
I would like to mention that I think Dave`s wrong when he says that a square waveform from 0 to 5v do not add a DC bias to the capacitor.  Of course the DC bias voltage is the mean value of that. It depends on the duty cycle, low and high level voltages. I think that the capacitance test should be done with a AC voltage with no mean value and take that capacitance value as the refernce point, and then add DC bias and check the difference.

bye!

Franco

Correct criticism; wrong solution though, I suspect. "Of course the DC bias voltage is the mean value of that"? What does this mean in engineering terms, that the capacitor is omnisciently predicting the future and thinking to itself "I think Dave is going to charge me to 5V again, so I'd better divide that by 2 and use my 2.5V capacitance..."?

AFAICT (and correct me if I'm wrong), but this DC bias effect is completely stateless. So the capacitance at time t is a function of the instantaneous voltage across the capacitor at time t, and that's all. This means that what Dave measured is a (weighted) average of the capacitance at all voltages between 0 and 3.16V. Measuring between -2.5V and 2.5V would give the average capacitance from -2.5V to 0.66V, which is arguably slightly better but hardly solves the criticism you raised.

In practice, Dave's demonstration was so completely compelling, and the effects so much stronger than any error that might be introduced this way, that it's a complete non-issue.

Pedantic, ideal solution is to use a much smaller square wave amplitude, so that the instantaneous voltage is always approximately the bias voltage. (This also gives me the interesting idea of charging a capacitor once, from 0V to maximum voltage, and seeing if these properties can be extracted from the one charging curve by looking at the gradient at each point.)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: rs20 on June 08, 2014, 12:51:40 am
It is scandalous that this is not mentioned on the datasheets, since it affects a large fraction of all applications.  I guess that using a type 2 ceramic cap as a DC blocker for audio applications would be awful.  How much distortion do you get by passing a 50 Hz sine wave (2 V p/p or so) through a ceramic cap?  I don't have the tools to measure distortion, but it might be enough to see on a scope.  I'll try it.

No, there'll be absolutely negligible distortion, if whatever's after the DC blocker has a high impedance.

If your output impedance is so low, and capacitance so small, than even an ideal capacitor would experience significant attenuation at 50 Hz, then yes, that means the capacitor isn't large enough to maintain a (near-)constant voltage across itself, and you may see interesting effects. But put it this way: you can always make a DC blocker bigger. If 1u is fine, then 10u or 100u is fine too. So if you use a 100u, and it drooped to 1u during operation, isn't that 1u still fine? The capacitor still does its job of carrying AC current.

I'm thinking about inputs to amplifiers and such here. Final output to speakers, etc, could be more of an issue. But still, distortion/harmonics are only generated where significant attenuation would occur anyway, so this is hardly a "panic and give up on all ceramics in audio applications" issue.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: David_AVD on June 08, 2014, 12:58:19 am
Using ceramics in audio filter sections of single supply applications could be an issue though?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: rs20 on June 08, 2014, 01:02:46 am
Using ceramics in audio filter sections of single supply applications could be an issue though?

Why would single-supply especially be an issue? Keep in mind, if my class 2 AC coupling cap has a nominal 10u capacitance, and it's got a 2.5V bias across it that makes it really 4.7u; what I've got there is a perfectly good, well-behaved 4.7u cap as long as the voltage across it stays at 2.5V (which is precisely the job of an AC coupling capacitor). In many cases a nominal 10u ceramic cap is much more compact and long-lasting than a 4.7u nominal electrolytic cap that'll dry out/vibrate off after x years.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: BUkitoo on June 08, 2014, 01:05:38 am
"Of course the DC bias voltage is the mean value of that"? What does this mean in engineering terms, that the capacitor is omnisciently predicting the future and thinking to itself "I think Dave is going to charge me to 5V again, so I'd better divide that by 2 and use my 2.5V capacitance..."?

So the capacitance at time t is a function of the instantaneous voltage across the capacitor at time t, and that's all. This means that what Dave measured is a (weighted) average of the capacitance

Lol, Ok. I think you are right. I just wanted to me mention that probably the method was not good enough. That is all.

Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: jesuscf on June 08, 2014, 01:34:20 am
Also, don't forget to take into account the output impedance of the function generator (Ro) for these tests.  The actual resistance of the circuit is 1k + Ro, probably 1050 ohms...
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Salas on June 08, 2014, 01:39:02 am
Those types are supposed to be fairly microphonic too? Will they display some resonance on the scope if you will hit them with a screwdriver for instance?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: T3sl4co1l on June 08, 2014, 01:46:58 am
So all of this does *NOT* apply to Class I capacitors?

Correct.  I've ran 100pF 50V C0Gs at 500V and they work just fine.

For testing I mean.  I wouldn't recommend *buying* 50V when you actually need 500V. :)

Also quite high Q (in the thousands), so they're excellent for RF and resonant applications.  About as good as silvered mica, cheaper and smaller.  Also available up to reasonable sizes (fractional uF, maybe a few up to single digit uF?), though you'll be paying for them up there.

Tim
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: David_AVD on June 08, 2014, 01:48:01 am
Using ceramics in audio filter sections of single supply applications could be an issue though?

Why would single-supply especially be an issue?

I was thinking you're more likely to have capacitors with 0V on one side and the Vcc2 (half rail) on the other.

Like the output of a volume pot (referenced to 0V) going through a coupling cap into a Vcc2 biased op-amp section.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: rs20 on June 08, 2014, 02:01:19 am
I'm going to suggest something. Whenever you look at a particular capacitor in a circuit, there's two different classes you can mostly put them in:

Bulk: Times when you need a metric truckload of capacitance, because you ideally want to hold the voltage across the cap dead-steady:
 - Power supply filtering, decoupling
 - AC signal coupling/DC blocking
 
In these situations, you need the huge capacitance because you can't just choose to make your ICs consume smaller pulses of current; and it's difficult and annoying to make the input stage of your amp have an impedance of Gigohms. Much easier to just use a huge bucketload of capacitance, without being too fussed about the precise value.

Ideal solutions: tantalum/electrolytic/class 2 ceramic.
 
Precise: Times when you need a precise, stable, predictable capacitance, but you typically don't need huge values because you control the currents that'll charge/discharge the cap. In these situations, you expect and rely on the voltage across the capacitor actually changing:
 - Timing
 - RC/RL/RLC lowpass/highpass/bandpass etc
 
In these situations, you can get away with smaller capacitances because you can increase the R's and the L's in your circuit to compensate for smaller C's (up to a point).

Ideal solutions: film/all those other low-capacitance types/class 1 ceramic.

Class 1 ceramics don't go up to nearly the same capacitance values as class 2, but that's fine, because whenever you need class 1, you can design your circuit to use those smaller values.</wild generalisation>

So as an example, a single-supply (the fact that it's single supply is of no relevance, actually) audio bandpass filtering circuit might consist of input and output buffers with class 2 capacitors interfacing to the outside world. But the internal bandpass caps should probably be class 1. Just a silly little example, but I think this sort of explains why I'm fascinated, but not the least bit worried, about this DC bias effect. None of my previous designs are adversely affected by the DC bias effect, because all of the critical, precise capacitors I had just happened to be small values, and I happened to end up with C0G/NP0 for those, even though I was completely clueless about this effect.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: rs20 on June 08, 2014, 02:01:47 am
Using ceramics in audio filter sections of single supply applications could be an issue though?

Why would single-supply especially be an issue?

I was thinking you're more likely to have capacitors with 0V on one side and the Vcc2 (half rail) on the other.

Like the output of a volume pot (referenced to 0V) going through a coupling cap into a Vcc2 biased op-amp section.

Sure, but as I've said, this doesn't lead to distortion or harmonics or anything like that, just a different value of nominal capacitance which can be easily factored into the design.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 03:20:14 am
I am baffled by some of the reactions in this topic. I can understand the you g ayers or arduino generation not knowing this, but the rest ?

This effect is limited to ceramic capacitors that use a barium titanate dielectric. Some old school through hole ceramics are different. All MLCC multilayered ceramic capacitors have this.

Dave showed dc bias dependency, mentioned temperature and frequency drift as well as case size : a 0402 is different than a 1206 all else being the same.
But, there is more : take caps with identical specs : capacitance and working voltage and class from different manufacturers and they behave totally different. It even changes inside one manufacturer depending on the series. The key is the dielectric material used.

Mlcc also have microphone effect (piezoelectric noise)

Certain vendors have specially constructed caps that are internally asymmetrically built . As frequency increases the electrons do not go to the 'far' plates in the stack. So you can buy caps where the plates are vertically placed , or offset in the body. Take these out of the tape and reel and you better not drop em or you dont know what side is up ....

One very important thing is regulators. To meet performance a regulator requires a certain output and input capacitance. This is given in the spec. You NEED to compensate the caps for dc ! The numbers in the spec are ideal values. It is up to the designer to derate the caps.

All serious cap makers have software available that lets you find behavior.

Another problem is resonance frequency. You are making a decoupling network . Well it may not work right at the frequencies your power spikes occur !

Serious design work requires characterising the power decoupling . I use a agilent 4395a network/impedance analyser. This allows me to plot the impedance over frequency with dc bias applied. Cap  values are then inserted depending on where the impedance is out of whack.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 03:25:27 am
Another problem is that even if u are aware of the problem 99% of the LDO and Switcher Datasheets and Appnotes just mention a let's say "stable with 2.2uF". I mean, is that a 2.2uF rated cap or do the mean a 'true' 2.2uF at the regulated output voltage......?

That''s the thing, you don't actually know. It's a suck it and see thing. Although if they recommend a 2.2uF ceramic, then you'd have to think they have tried a real 2.2uF ceramic. But what type? What size? Good app notes will have actual part numbers.

App notes will have part numbers, if the apps engineer is worth his salt...
I can't speak for everyone, not even inside one company, but, in the group where i work the cap values mandated in the datasheet are IDEAL value caps. Derating is NOT taken into account. It is up to the people building the actual circuit to compensate for the behavior of whatever they use.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 03:37:37 am
One more thing...

MLCC value drifts over time. As the crsytal structure settles under dc bias the cap will exhibit drift. It takes about 5 years at dc level for the cap to become stable.
It is an exponential decay with the largest change happening the first few months in operation.

Here is the trap: heating the dielectric above its curie point resets this decay!
And the curie point is 148 degreee c for BaTi based caps.
This means : solder it and you reset the decay!
So you can forget using these as precision timing!
Worse, you got a system returning from the field after 6 months with drift. Touch it with a soldering iron and you are screwed... You reset the decay.
The one year decay is about 8%.

By the way: daves blue capacitor in the video is simply a surface mounted cap with legs spotwelded on and dipped in epoxy.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 03:42:03 am
Last thing: if you thought you heard it all : dave says in the video 'its only a resistor, the most basic of parts' ...
EEuuuuuuuuurt (big fail buzzer going off)

If you think this dc bias stuff of caps is bad, wait till you discover all the crap you can run into due to fringe effects in resistors... Noise due to sintering, iregularities in deposit, working voltage, hygroscopy, inductance in spiral cut parts, drift due to dendrite formation in voltage overstressed laser trimmed parts...  Don't get me started on that stuff...
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: WarSim on June 08, 2014, 04:18:01 am

These fundamental Friday videos are beginning to worry me. 
There has not been a single one that taught me something new.

Well, the title says it all, fundamental. If you know your fundamentals then there is nothing to learn. That's just the way it is. Sometimes I do lesser known ones like this, but if you know those too oh well  :-//

Please don't get me wrong. 
My statements are against the many egomaniac engineers I have had to work with that seeming haven't grasped the fundamentals not this video. 
I have to remind myself of the several excellent engineers that I had the pleasure of working with.  Sometimes I forget that those many are not all of them.
 
Any good engineer will learn from any source no matter what it is.  That is why they learn the fundamentals first.  So even if a lowly Technologist explains an issue with physics, an engineer should be able to verify or dispute the issue.  Not attack the source with argument based on their status. 

Oops another rant starting, I better stop now :).

Thanks for all the videos.  I will never avoid an information source like the bad examples mentioned above, just surprise me someday with something new, please. 



Sent from my iPad using Tapatalk
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: jesuscf on June 08, 2014, 04:38:25 am
I wonder how practical would it be to make a VCO using the voltage dependency of ceramic capacitors.   Something along the lines of a VCO with a varicap, but at a much lower frequency.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: WarSim on June 08, 2014, 04:47:58 am
Yes everyone listen to free electron.  These issues are present in all capacitors that are created with any type of particle suspension.  The only thing in question is the manufacturing steps taken to minimize the issues.  In general this issue is lowest in class 1 of the ceramic capacitors, not nonexistent.  So in extension, a manufacturing variance could inflate these issues even if the general rule says it's best.  Also why the capacitor manufactures with reliable and repeatable manufacturing processes are trusted the most. 
There is no such thing as a perfect capacitor, just the right one for each purpose.  Using only molecularly controlled grown crystal capacitors everywhere at $100-10,000 each would make any project cost prohibitive.  I have never seen one myself only helped make a device for the research department to develop them about a decade ago.


Sent from my iPad using Tapatalk
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 08, 2014, 06:03:14 am
Anyone know if any of the circuit-simulation software suites (SPICE, etc.) have either native or avail plug-ins for simulating these type of "ceramic distortions"?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 08, 2014, 06:36:04 am
So as a general rule of thumb, get the physically largest capacitor you can that is practicable for your electrical specification needs.
This is part of it ...
On audio DIY forums/blogs, a somewhat popular project is to mod low-cost (but otherwise high-performance Asian DAPs [aka digital audio players, like iPods] **)
Below is a stock (unmodded) PCB from a decent-sounding DAP by Chinese manuf Teclast:
(http://www39.atwiki.jp/teclast_t51?cmd=upload&act=open&pageid=30&file=T51_020.jpg)

And after a bit of experimenting -- via o'scoping, listening tests and other metric--some modders on a Russian electronics DIY audio forum made some changes that apparently improve both objective (measurable) and subjective (listening test) performance (of interest to this EEVblog topic are the cap mods. SORRY: the image below is 180-degrees opposite from above):
(https://imageshack.com/a/img189/5139/dscf5884k.jpg)

** Not all Chinese products are "Wung Hun Lo". There are quite a few high-performance brands that do not make it to USA/Europe/Australia mass-market stores, but that one can on eBay or even Amazon.

====

FYI, Russian modders' total list of mods done are below:
Replacement / differences from stock:
 
replacement negative power inverters in headphones AMP and LPF
    voltage regulators headphonesAMP and LPF with replacement filter capacitor Cbypass (is good for filtering high frequency interference)
    an additional negative regulator shoulder headphonesAMP, along with the replacement of the capacitor in the inverter on Cfly X5R 4.7uF
    replacement signal resistors located between DAC outputs and LPF
    capacitors in the LPF replacement for ceramic NP0
    replacement 10uF 1uF ceramic filters in power modulators DACs to 0,1 uF NP0
    ceramic replacement shunts opamp LPF and headphonesAMP at 0,1 uF NP0
    the signal from the low pass filter to the ear specialist now passes through the wire MC16-13(silver-plated copper) 0.05, cut lead-in track, contacts potentiometer raised and do not touch the board. Still have access to the line-out, installed protective resistors 50 Om
    removed the low quality resistors in the signal path to the headphonesAMP and line-out
    increased power amplifier to your ear +-4V
    filtering power before and after the dc-dc TD1410 (power supply for headphonesAMP)
    opamp LPF LM4562
   headphonesAMP made of LM4562+THS4222 - output up to 100mA per channel, 14mA quiescent current per channel(most portable headphones will work in Class A)
    power cap for headphonesAMP is 2x220uF Panasonic FK, LPF 2x47uF Panasonic SU(bipolar capacitors with very low distortion)
    measuring resonator 24MHz, the structure in tact CPU PLL, a generator with low jitter
    stock replacement batteries for the assembly of two LP473850 1090mAh, setting equalizer to equalize pressures between the banks
    sealed cables and sensor module aluminum tape to avoid interference on the part of the signal
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: The Electrician on June 08, 2014, 07:08:57 am
Have a look at this:

https://www.youtube.com/watch?v=5Qsmqv_O9MU (https://www.youtube.com/watch?v=5Qsmqv_O9MU)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 07:36:37 am
Anyone know if any of the circuit-simulation software suites (SPICE, etc.) have either native or avail plug-ins for simulating these type of "ceramic distortions"?

The software made available by the cap makers can spit out a spice model of the cap you pick.
Every cap maker has his own secret sauce dielectrics. This isnt calculated. The cap goes through a waterfall analysis and a model is constructed.

Just like there are s parameters and ibis models for connectors, there is a model for caps. I have ran sims with these models and they are pretty damn close to measure. Og course the simulator needs accurate models for everything. As silicon maker we have the real circuit to run it on. I doubt you can do it at home...
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 08, 2014, 08:03:40 am
Have a look at this:

https://www.youtube.com/watch?v=5Qsmqv_O9MU (https://www.youtube.com/watch?v=5Qsmqv_O9MU)
Testing JUST the cap, out of ckt, may not be realistic no matter how accurate the testing routine (e.g., using correct meter(s) correctly). Of course, the cap in its permanent topology (the ckt) is not going to be as easy to model.
What TDK tests is a start but still low-hanging fruit.
In situ testing is most important ... NASA or USA military will have some answers?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 09:07:22 am
Any decent engineering kit does that. Either during design or during emc test. For mass production the circuit is checked to see there are no out of band points that could be problematic during emc test. Then circuit is optimised for bom cost by stripping unneccesary caps. If a regulator calls for 10uf then this means it needs to see 10 effective uf. Take the caps directly around it plus all other decoupling in the board and you may actually have waaaay too much. So bom cost can be tweaked by removing extraeneous parts. Deciding which ones and where takes experience and skill. Sweep the board at proper dc bias and measure both bulk capacitance and impedance. There is a reason proper lcr merers can either supply dc bias or reject it during measurement. A 4367 lcr bridge and a network/impedance analyser are required for such work.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 08, 2014, 09:26:54 am
Doug Ford reminded me of an old article of his that shows the effect of XR7 capacitor non linearity on audio distortion.
http://www.dfad.com.au/links/DFAD_PASSIVE_SURROUND.pdf (http://www.dfad.com.au/links/DFAD_PASSIVE_SURROUND.pdf)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 08, 2014, 09:34:03 am
One very important thing is regulators. To meet performance a regulator requires a certain output and input capacitance. This is given in the spec. You NEED to compensate the caps for dc ! The numbers in the spec are ideal values. It is up to the designer to derate the caps.

Not always the case.
I've seen regulator datasheets give a minimum capacitance and then proceed to specify and recommending that same value cap (including part number) in the app circuit that has no chance of meeting that minimum capacitance.
So you just don't really know whether their actual minimum value is theoretical, or practical derived using typical parts.
To be on the safe side of course you assume it's theoretical and derate the part suitably. In which case the application circuits could be in theory, wrong.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 08, 2014, 09:42:20 am
Thanks for all the videos.  I will never avoid an information source like the bad examples mentioned above, just surprise me someday with something new, please. 

Geeze, well just how a I supposed to do that?
This is probably one of the more obscure "fundamental" issues out there, and you want me to surprise you with something fundamental you don't know?  :-//
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 08, 2014, 09:45:17 am
Last thing: if you thought you heard it all : dave says in the video 'its only a resistor, the most basic of parts' ...
EEuuuuuuuuurt (big fail buzzer going off)

I was of course referring to the basic applications we were looking at here.
It's part of the story telling  ;)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 08, 2014, 09:49:48 am
So all of this does *NOT* apply to Class I capacitors?

In theory, yes it can (to a small degree). In practice, no, it's not something you generally have to worry about with a class 1 cap. Even the worst class 1 caps are an order or two better than class 2 in this area.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 08, 2014, 10:29:54 am
Another approach to addressing this issue is to base one's choice of cap type based on application, using a succinct "laundry-list" format . So ...

The voltage-dependency issue may be mostly a NON-ISSUE in the following applications:
(1)
(2)
(3)

The voltage-dependency issue is IMPORTANT in the following applications:
(1)
(2)
(3)

===

And the results of the lists above can be later turned into charts, tables, info-graphics, etc.

Yes, some topical charts and datasheet info have already been pointed out by Dave/others . But info "hidden" in the obscure sections of datasheets, and web sites are not ideal for designers (I'm sure the obscurity is partly driven by marketing and economics!)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: David_AVD on June 08, 2014, 11:26:09 am
Doug Ford reminded me of an old article of his that shows the effect of XR7 capacitor non linearity on audio distortion.
http://www.dfad.com.au/links/DFAD_PASSIVE_SURROUND.pdf (http://www.dfad.com.au/links/DFAD_PASSIVE_SURROUND.pdf)
That's a great reference and is the type of thing I was thinking about when I questioned the cap's effect in coupling and filter application.

If the capacitance changes with AC level, it's not surprising that it would result in distortion of some sort.

The filter corner frequencies could also shift depending on the overall signal level I assume.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: JackOfVA on June 08, 2014, 11:45:50 am
It's actually worse, in some respects, than Dave suggests, as the dielectric has 'memory' - probably what Robert Pease discussed in his 'dielectric soakage' notes.

A few images below from my capacitance versus voltage work.

Cap 25 & 50V Plot - Capacitance versus voltage for 25 and 50V rated X7R 1u0 1206 surface mount parts. Note the hysteresis seen in the plot - the actual capacitance value at a particular bias voltage depends on the history of the applied voltage.  This data was taken with relatively fast voltage steps, using an HP4192A with built-in bias generator.

Z5U Three Time Steps - To see if dielectric soakage explained the hysteresis, I ran a Z5u 0u1, 50V 1206 surface mount part over +/- 35V bias range, but with varying time interval between successive voltage steps. Regardless of the time between voltage steps, the same 'butterfly' pattern is apparent. There is a change in absolute capacitance but in general the same hysteresis pattern is repeated for step times of 1, 10 and 100 seconds.

Full Size - The Z5U 0u1 1206 surface mount part was exposed to alternating bias voltage of +35V and -35V for an extended time period and relatively slow intervals between bias voltage reversals and the capacitance observed.

Expanded - An expanded view of one of the cycles in the above plot. This plot is the best illustration of dielectric soakage.

CapvsV - This is an older data analysis (2001) made with a manual General Radio 1650A RLC bridge and an external bias supply, with a random selection of capacitors from my parts bin.  As might be expected, the C0G and film parts are unchanged with bias voltage, within the limits of the 1650A bridge. This data does not show hysteresis because it was taken with one continuous voltage run, slowly as the process was to adjust the bias supply, null the bridge and record the voltage, capacitance and dissipation factor. Repeat until the desired end voltage point is reached.

D versus V - Same parts and test setup as above, but this time looking at the dissipation factor. It also changes with applied voltage, and in a non-obvious fashion.

ExpFit - lastly, I tried an exponential curve fit to the 2001 data, with decent results for C versus V.


Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Bud on June 08, 2014, 03:09:24 pm

Serious design work requires characterising the power decoupling . I use a agilent 4395a network/impedance analyser. This allows me to plot the impedance over frequency with dc bias applied. Cap  values are then inserted depending on where the impedance is out of whack.

Is this done on the fully populated board or just one with only decoupling caps on the power rails in place?

Also, where is the analyzer is connected to, just to the power supply side or to each critical area of board requiring good decoupling?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 08, 2014, 03:45:13 pm
I like the earlier suggestion that an important reason for this effect may be SIZE related. So a followup by Dave/anyone, in which same value/diff.-size ceramics are compared, would be useful.

In any case, WRT the overall importance of this phenomenon ... I guess I'm still not convinced it is anything more than a marginal issue (more of a conversation piece and a fascinating scientific phenomenon).
I am involved in the wanky world of audiophile design ... and of any consumer-electronic fields, audiophilia would be the first to know about such issues (we've certainly got our share of ridiculous pre-amps and speakers with boutique, super $$ caps by Mundorf and BlackGate).
Maybe the obscurity of the "ceramic voltage dependency" issue (such as those obscure/hard-to-find datasheet graphs) is DESERVED and REALISTIC.
Convince me to the contrary, folks ;)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: JackOfVA on June 08, 2014, 04:27:56 pm
I like the earlier suggestion that an important reason for this effect may be SIZE related. So a followup by Dave/anyone, in which same value/diff.-size ceramics are compared, would be useful.

In any case, WRT the overall importance of this phenomenon ... I guess I'm still not convinced it is anything more than a marginal issue (more of a conversation piece and a fascinating scientific phenomenon).
I am involved in the wanky world of audiophile design ... and of any consumer-electronic fields, audiophilia would be the first to know about such issues (we've certainly got our share of ridiculous pre-amps and speakers with boutique, super $$ caps by Mundorf and BlackGate).
Maybe the obscurity of the "ceramic voltage dependency" issue (such as those obscure/hard-to-find datasheet graphs) is DESERVED and REALISTIC.
Convince me to the contrary, folks ;)

I've looked at intermodulation distortion at radio frequencies when signals are passed though X7R coupling capacitors and compared the 2nd and 3rd order intercepts with the same circuit but (a) with C0G parts and (b) with film capacitors.

Typically one side of the coupling capacitor is at +15V and the other side at 0 or in some circuits +20V. RF levels are around 0 dBm and the RF load impedance is 50 ohms. 

At 20 KHz, Xc (assuming for a moment the X7R is still 1uF with this DC bias on it) = about 2 ohms. Current through the blocking capacitor is on the order of 4.5 mA and the corresponding AC voltage drop across the 2 ohm blocking capacitor is 9 mV. (All RMS).

At 200 KHz, Xc is 0.2 ohms and the AC voltage drop is 0.9mV.

Analysis of this says you neglect the DC bias except in so far as it alters the capacitance value, and consider the non-linearity caused by an RF 'wiggle' of 9mV at 20 KHz or less than 1mV at 200 KHz and above.   My measurement floor is a 2nd order output intercept of +120 dBm and 3rd order output intercept around +52 dBm, limiting value is intermodulation caused in band reject filters between the device under test and the spectrum analyzer.

To date, I've not seen any significant difference between X7R 1uF 1206 SM parts and film or a bunch of parallel 0u1 C0G 1206 SM parts. Hence I continue to use X7R parts in the RF amplifiers I build and sell.  The best HF preamplifiers I build show OIP2 >= +120 dBM and OIP3 ~+50 dBm with test tones of 3 & 4 MHz, measured at 0 dBm output.

At least in this particular application, I've not found an issue with X7R coupling capacitors. I've taken a look at test tones in the medium wave band (500 & 700 KHz) and see the same thing - no change with capacitor type. It may be that lower frequencies will show degraded OIP2/3, but at the frequencies where these amplifiers are used, no degradation.

Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: w2aew on June 08, 2014, 06:26:13 pm
A nifty measurement shortcut you can use when measuring the RC time constant, particularly on an old analog scope without cursors, is to adjust the vertical scale vernier so the the waveform occupies 8 vertical divisions. Then, 1 RC time constant is when the waveform crosses the 5th division vertically.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 09:33:11 pm
One very important thing is regulators. To meet performance a regulator requires a certain output and input capacitance. This is given in the spec. You NEED to compensate the caps for dc ! The numbers in the spec are ideal values. It is up to the designer to derate the caps.

Not always the case.
I've seen regulator datasheets give a minimum capacitance and then proceed to specify and recommending that same value cap (including part number) in the app circuit that has no chance of meeting that minimum capacitance.
So you just don't really know whether their actual minimum value is theoretical, or practical derived using typical parts.
To be on the safe side of course you assume it's theoretical and derate the part suitably. In which case the application circuits could be in theory, wrong.
correct, not always the case. that is why i explicitly said 'apps engineer worth his salt' and 'some manufacturers' and 'sometimes dpeending on department' ...

it is a crapshoot.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 08, 2014, 09:41:35 pm

Serious design work requires characterising the power decoupling . I use a agilent 4395a network/impedance analyser. This allows me to plot the impedance over frequency with dc bias applied. Cap  values are then inserted depending on where the impedance is out of whack.

Is this done on the fully populated board or just one with only decoupling caps on the power rails in place?

Also, where is the analyzer is connected to, just to the power supply side or to each critical area of board requiring good decoupling?

Fully populated board , powered up and running ( devices consuming current when in operation , not when idle or twiddling their thumbs )
impedance plot is taking at every power entry of the system or system section:

power enters somewhaere at the board edge. that is to be measured.
any local regulator on the system feeds a power section. so that is where we measure again.
in essence all power rails need to be tested and there should be no 'invisible capacitors' shielded by an intermediate regulator ( or a ferrite bead ! )

you need the impedance option for the machine (option 10) and the proper testjig that lets you connect to a dc biased system i use the component testjig that allows dc biasing.

you can make this jig yourself. the schematics are out there. simply calibrate it by taking an unconnected sweep and storing that in cal memory. the 4395 then automatically compensates for all the interconnect crap so your referenc eplane is at the jig terminal.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: digital on June 09, 2014, 01:29:51 am
An informative video Dave fundamental videos invariably lead to a lot of comments on the forum so it is a win win situation for all as there is a lot of knowledge here on the forum.Thanks.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: 13hm13 on June 09, 2014, 09:06:16 am
The size-vs-effect aspect continues to intrigue me. If simply increasing the size of the same-chemistry cap decreases the voltage-dependency effect, are the longer PCB traces (and bigger solder joints) an effective trade-off?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: EEVblog on June 09, 2014, 09:27:57 am
The size-vs-effect aspect continues to intrigue me. If simply increasing the size of the same-chemistry cap decreases the voltage-dependency effect, are the longer PCB traces (and bigger solder joints) an effective trade-off?

In most cases, yes, if the voltage dependency is an issue for you.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Kjelt on June 09, 2014, 03:41:41 pm
Thanks for doing the request i made in september: https://www.eevblog.com/forum/suggestions/bridge-ciruits-for-fundamental-friday/msg288045/#msg288045 (https://www.eevblog.com/forum/suggestions/bridge-ciruits-for-fundamental-friday/msg288045/#msg288045)
I think this needs to be repeated to all generations of engineers so they won't fall in this trap.
Modern engineers depending solely on simulations can get into so much trouble.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: tchicago on June 09, 2014, 06:39:26 pm
Hi, David,

Thanks for a great video! Not so long ago, I personally was puzzled by those tiny 10uF capacitors so much that I had to go to a datasheet to figure out where is the catch and if there is any. And, yeah.. found those capacity vs voltage curves.

Speaking of the non-ideal capacitors and other components, I have an idea for an exciting topic for the next investigative video: component whine (also known inductor whine, coil noise, etc). Here is a thing: there is a strong belief this annoying noise in DC-DCs and SMPSs is caused not only by the inductors, but also by the capacitors that emit it via the piezoelectric effect.

In fact, long ago I was surprized by this noise coming from a board that did not have any inductors at all, only the bypass capacitors on the power rails on the 74xx-based TTL board. But the board had early Russian EEPROM chips, so I was not fully sure if it was coming from the chips or caps.

But I'm guessing the setup for this kind of experimentation is not going to be simple: probably need some good microphones and audio spectrum analyzers. And also a well though-through drive circuitry for the DUTs.
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: free_electron on June 09, 2014, 06:44:52 pm
BaTi based MLCC caps will emitnoise if pulsed currents are sent through them.
BaTi is a piezoelectric material.

the reverse is also true. they act as microphones. they should not be used in feedback loops of regulators if the board is going to be subject to mechanical vibration !
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: robrenz on June 09, 2014, 07:03:05 pm
BaTi has been used as the sensor for fish finders.  Driven to send a pulse into the water (speaker) and then listened to for the return signal (microphone)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: Clear as mud on June 10, 2014, 09:50:05 pm
Certain vendors have specially constructed caps that are internally asymmetrically built . As frequency increases the electrons do not go to the 'far' plates in the stack. So you can buy caps where the plates are vertically placed , or offset in the body. Take these out of the tape and reel and you better not drop em or you dont know what side is up ....

This is kind of off-topic, but about the symmetry of capacitors:  What about regular, symmetrical MLCCs?  Does mounting position matter?  The smaller values look the same right-side-up or up-side-down, and I have been assuming that it really doesn't matter which way you install them.  The larger values, such as an 0805 1uF capacitor, tend to be almost the same size both horizontally and vertically.  What happens if you accidentally solder one on with the plates vertical instead of horizontal?  Is it all right to do that?  Will it lead to less flexibility and thus more cracked capacitors or broken solder joints?
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: T3sl4co1l on June 11, 2014, 02:17:45 am
Certain vendors have specially constructed caps that are internally asymmetrically built . As frequency increases the electrons do not go to the 'far' plates in the stack. So you can buy caps where the plates are vertically placed , or offset in the body. Take these out of the tape and reel and you better not drop em or you dont know what side is up ....

This is kind of off-topic, but about the symmetry of capacitors:  What about regular, symmetrical MLCCs?  Does mounting position matter?  The smaller values look the same right-side-up or up-side-down, and I have been assuming that it really doesn't matter which way you install them.  The larger values, such as an 0805 1uF capacitor, tend to be almost the same size both horizontally and vertically.  What happens if you accidentally solder one on with the plates vertical instead of horizontal?  Is it all right to do that?  Will it lead to less flexibility and thus more cracked capacitors or broken solder joints?

This is silly:

1. The impedance, as measured, is on the graph.  Period.  Is the impedance low enough for your purposes?  Great!  You don't even need to know the capacitance, damping factor, voltage rating or anything else.  (That is, assuming the impedance curve remains true in your application.)

2. The worst possible case electrical difference would be if all the plates were flat to one side, so that the effective height above the board surface ranges from nearly zero ("face down") to nearly the chip thickness.  ...So What?  That makes a loop -- an inductor.  The loop area is maybe a half a nanohenry, for say, 0805 or thereabouts.  There is no effect on the capacitance, assuming you're talking about the frequency range where it, well, capacitates...  The only difference will occur at extremely high frequencies, where it's inductive either way.

3. All capacitors exhibit skin effect, just as all conductors do.  Indeed, radiation of any given wavelength will tend to propagate within the surface of a dielectric -- conductors are not needed at all to illustrate the skin effect.  This has applications and examples ranging from dielectric antennas (microwaves) to total internal reflection (arguably, extending up into x-rays).  This is independent of construction, limited only by the dielectric constant of the material.

Wound film types are especially notable, due to the relatively large physical construction (the effective roll-off frequency is low enough to be of significance) and the effect of schoopage (end terminal metallization) shorting across the roll (thereby forcing current away from the center at high frequencies).

A good example, from my personal experience, is CDE 940 type snubber caps.  They're rated for lots of peak current.  Beefy.  But not nearly as much RMS as you would expect from their size.  (Datasheet for reference: http://www.cde.com/catalogs/940C.pdf (http://www.cde.com/catalogs/940C.pdf) )  Take 1uF 600V for instance.  8.9A RMS, 196A peak.  Yet the voltage curve starts rolling off at only 2kHz.  (The astute reader will notice 250VAC at 2kHz draws only 3.1A; the voltage curve drops after this, but if you calculate current, you see it rises to 6.3A at 10kHz, then around 10A at 100kHz -- presumably 8.9A from the table.  The graph might show more since it says 25C ratings, not 70C.)  Now, it's not obvious from the data if the current rating rolls off at all.  It seems to climb.  In this case, the datasheet doesn't help me illustrate my point, unfortunately!

The reason I give this example from personal experience is, I've tried them before, and they get hot.  I tried using them at more like 200-400kHz, at somewhat less than rated current, but the losses were real, disproportionately higher.  They are definitely better suited to lower frequency use (under 100kHz).  Other than the hazy foreboding of data ending at 100kHz, the datasheet doesn't tell you this.

FWIW, that application ended up with Epcos metallized MKP style caps, which worked just fine -- within ratings.  Push a smidge past and they'll heat up and melt and puke and burn -- finely engineered down to a price, without a doubt.  Other types and brands were also tried, with more and less suitable results.

A better example to illustrate my point:
http://industrial.panasonic.com/www-data/pdf/ABD0000/ABD0000CE47.pdf (http://industrial.panasonic.com/www-data/pdf/ABD0000/ABD0000CE47.pdf)
Page 6, ECWFA, 250V, permissible current.  Notice how the larger uF values reach peak current at lower frequencies.

Now, regarding current ratings, let me be clear: I don't mean to assert that skin effect, or in general, eddy current effects, are limiting current capacity of large capacitors at high frequencies.  There are many possible reasons, any of which may dominate in a particular case.  (A few pages ahead, the higher voltage types show nearly flat current limit curves, at least in the smaller values.  Is skin effect a consideration for those parts?  Who knows.)  I'm no capacitor designer, so I don't know what all goes into them, I only make educated guesses.  Skin effect will vary greatly with construction, so blaming it for the unsuitability for a particular application (like the 940s in my example) is only suspicion.

TLDR: ceramic caps are hunks of ceramic with a characterized impedance over frequency.  It doesn't matter how you put them in.

Tim
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: GK on June 11, 2014, 02:55:41 am
It's actually worse, in some respects, than Dave suggests, as the dielectric has 'memory' - probably what Robert Pease discussed in his 'dielectric soakage' notes.


Dielectric absorption. "Memory" is a rather misapplied term. A capacitor exhibits a memory effect no more than a still-hot light bulb remembers that it was turned on. My earliest reference for a capacitor model for describing and estimating the effects of DA is Dow Corning 1954.

I’ve posted these before: 


(https://www.eevblog.com/forum/projects/home-brew-analog-computer-system/?action=dlattach;attach=55579;image)
(https://www.eevblog.com/forum/projects/home-brew-analog-computer-system/?action=dlattach;attach=55581;image)
(https://www.eevblog.com/forum/projects/home-brew-analog-computer-system/?action=dlattach;attach=55583;image)
(https://www.eevblog.com/forum/projects/home-brew-analog-computer-system/?action=dlattach;attach=55585;image)
(https://www.eevblog.com/forum/projects/home-brew-analog-computer-system/?action=dlattach;attach=55587;image)
Title: Re: EEVblog #626 - Ceramic Capacitor Voltage Dependency
Post by: JackOfVA on June 17, 2014, 10:20:14 am
It's actually worse, in some respects, than Dave suggests, as the dielectric has 'memory' - probably what Robert Pease discussed in his 'dielectric soakage' notes.


Dielectric absorption. "Memory" is a rather misapplied term. A capacitor exhibits a memory effect no more than a still-hot light bulb remembers that it was turned on. My earliest reference for a capacitor model for describing and estimating the effects of DA is Dow Corning 1954.


Thank you for posting the reference - well worth reading.

However, I note that Paragraph 3-9(c)(2) uses the term 'memory' as well as 'soakage.'