C0G caps have no significant capacitance change with voltage?

[TimFox]

I have used C0G SMT capacitors up to 100 nF in critical analog circuits (integrators), but they were in large packages that scare people who think 1206 is large.
I often use 10 nF TH capacitors in regular audio circuits.
Note that NP0 is a military specification for the same material as C0G commercial specification.
If a vendor makes a part that does not meet that spec, he shall not call it C0G.

[Zero999]

They're very stable in all circunstamces but also have low capacitances, so rarely useful except for tuned circuits.

Not really accurate.  Tuned circuits are not the only type where C0G caps are useful.  And "low" is a subjective term.  It means nothing really.  I know people who regularly use 1 pF caps.  Yeah, really.

I think he meant low energy density. Film capacitors are generally a better choice for values over 10nF or so.

[DavidAlfa]

I mean, they're 10-20x more expensive than X7Rs, unless for very specific RF uses, why use them?
For decoupling, value is not critical, use a value large enough that will be good enough at any degrade level.
Of course in any tuned LC/ RC circuit stability is a must, to me it seems like almost the only scenario to justify using C0Gs.
Though I'm only at the average level and might learn something 

[TimFox]

In critical analog circuits, a voltage-variable capacitance will distort the signal, when there is a substantial signal voltage across the capacitor.  This can include coupling as well as filtering.
For bypassing the power on a digital chip, just make sure you have sufficient capacitance at the applied DC voltage.

[coppice]

I mean, they're 10-20x more expensive than X7Rs, unless for very specific RF uses, why use them?
For decoupling, value is not critical, use a value large enough that will be good enough at any degrade level.
Of course in any tuned LC/ RC circuit stability is a must, to me it seems like almost the only scenario to justify using C0Gs.
Though I'm only at the average level and might learn something 

Anywhere a capacitor is used as a source of capacitance (e.g. filters), rather than bulk energy storage (e.g. decoupling) a non-linear capacitor is a source of distortion.

[DavidAlfa]

Yeah that's also what I was refering to.
Tuned circuits include filters
I checked them last week out of curiosity when buying at lcsc, I think the largest C0G fast 1nF or so.

[TimFox]

Mouser has 100 nF 100 V C0G capacitors in 1210 and 1206 packages, along with higher voltage (larger package) and lower voltages.

[Nominal Animal]

Well, Murata GRM3195C1H104JA05D is a 50V 100nF ±5% C0G in 1206 that costs 0.167€ per cap in lots of ten at Mouser, with plenty in stock as of 2023-07-28.  Murata says its capacitance varies by less than 0.3% = 0.3nF between 25°C and 100°C, with operating range -55°C to 125°C, and that its capacitance change with respect to bias DC voltage is neglible (Murata product details page).

Note that NP0 is a military specification for the same material as C0G commercial specification.

Not exactly: NP0 and C0G encode the same capacitance temperature coefficient α using different notations.  The US military prefers the former, but the latter is what US industry came up with (specifically, Electronics Industries Alliance standard EIA-RS-198).

For example:
 P100 = M7G (α = 100±30 ppm/K)
 NP0 = C0G (α = 0±30 ppm/K)
 N750 = U2J (α = -750±120 ppm/K)

EIA uses a weird encoding of letters and numbers, whereas the non-EIA name is simply Nppm, NP0, or Pppm for ppm ppm/K, with tolerance ±30 ppm/K for |ppm| ≤ 100, ±60 ppm/K for 100 ≤ |ppm| ≤ 500, ±120 ppm/K for 500 ≤ |ppm| < 1000, and ±250 ppm/K for |ppm| ≥ 1000 (if I've understood the naming correctly).  Very straightforward and easy to remember.

Neither of these namings consider anything other than the temperature dependence of the capacitance.

[TimFox]

That is what I wrote about NP0 and C0G.  EIA is a commercial spec.
In the military notation, there are also controlled temperature co-efficient dielectrics such as N750 (-750 ppm/K), which are sometimes seen in non-temperature contexts due to their higher dielectric constant.

[T3sl4co1l]

I mean, they're 10-20x more expensive than X7Rs, unless for very specific RF uses, why use them?
For decoupling, value is not critical, use a value large enough that will be good enough at any degrade level.
Of course in any tuned LC/ RC circuit stability is a must, to me it seems like almost the only scenario to justify using C0Gs.
Though I'm only at the average level and might learn something 

I would say the bigger issue for decoupling is their losses are so goddamned low (Q > 3000 is not unheard of).  You put two in parallel, side by side, and they oscillate together.  Even larger values like 10nF (1206).  An effective PDN using just them, requires resistors EVERYWHERE to keep the Q down!

They have many redeeming qualities.  They're about as ideal a capacitor as you can find.  They're small and dense.  They have the energy density of electrolytics, when higher voltages are considered (>200V or so).  The main downsides are the low loss (when you sometimes want high!), the small values (particularly at low voltages), and the high cost.  And, the brittleness when applicable (but that's true of ceramics in general).

The low-voltage low-value thing, and the high-voltage density, go hand-in-hand: the dielectric strength is excellent, so, layers simply can't be made thin enough (100s nm?) with high yield to make low voltage high capacity parts.  Think of them as better suited to high voltages: a typical 1210 X7R runs out of steam to the tune of 10-30nF at 250V (no matter if you buy a "47nF" or "470nF" part!); meanwhile a 1210 C0G might go up to 47 or 68nF, while still boasting a 630V rating!

Note that high voltage type 2's can have much higher C(rated V)/C(0) ratios, because the same argument applies: thicker layers, and fewer of them, mean better quality dielectric, means freedom from breakdown guaranteed to higher voltages.  To reiterate: type 2 capacitors are only rated for capacitance at zero bias, and the value at rating will always be some amount less. How much, you cannot know besides looking up the characteristics.

I've been working on an LLC SMPS project for a bit and find the 10s nF 630V C0Gs (1206 and 1210) are better in both cost and PCB/layout area than film capacitors; or if not strictly on relative cost, then the absolute cost is certainly acceptable for the few parts needed in an LLC.

X7R have modest to low losses (like, low enough they can resonate together, but not nearly low enough to use for AC purposes at switching frequencies), making them efficient bypass caps for high-ripple application, and they're small and cheap enough for general purpose use.  They fall out of favor in large sizes (physically and electrically, especially at >= 100V), where film or electrolytic take over (depending on voltage and ripple requirements).

Tim

[gnuarm]

I mean, they're 10-20x more expensive than X7Rs, unless for very specific RF uses, why use them?
For decoupling, value is not critical, use a value large enough that will be good enough at any degrade level.
Of course in any tuned LC/ RC circuit stability is a must, to me it seems like almost the only scenario to justify using C0Gs.
Though I'm only at the average level and might learn something 

Power supply decoupling of sensitive analog circuits is a very good place to use C0G caps.  X7R parts can be microphonic adding noise to the supply rail. 

I made the mistake once of using a CP Clare telephony part for a POTS interface.  It had 0dB of supply rejection ratio.   That's right!  0dB!!!  This was not in the data sheet and adding larger X7R caps got rid of the supply noise which was directly coupled onto the phone line.  But it added microphonics which picked up fan noise by the bushel! 

C0G was not an option for this particular circuit, so a whopping 100 uF electrolytic (I hate using them) was shoehorned into the circuit.  Actually, I think it was added on the connector this board plugged into, as there was no space on the board for it. 

I probably could have added a small cap in series with the CODEC power with as large an C0G as would fit.  I never used another CP Clare part either.

[temperance]

I would say the bigger issue for decoupling is their losses are so goddamned low (Q > 3000 is not unheard of).  You put two in parallel, side by side, and they oscillate together.  Even larger values like 10nF (1206).  An effective PDN using just them, requires resistors EVERYWHERE to keep the Q down!

The same can happen with low ESR capacitors. About 20 years ago I made a PFC with two 470µF capacitors in parallel and every time I unplugged the thing I could hear some whining. At first I thought the under voltage lockout isn't working. After some probing around, nothing could be seen. In the end I made horn from paper (you don't want to stick your years next to capacitors charged to 390V) to locate the sound. Somehow charge was being transferred between those two capacitors while the voltage was declining.

About the ceramic caps in an LLC instead of film caps. I'm doing the same thing for a 600W supply. Perhaps of interest:
https://www.psma.com/sites/default/files/uploads/tech-forums-capacitor/presentations/is94-quantifying-allowable-ac-parameters-dc-rated-ceramic-capacitors.pdf

Film capacitors able to handle 250Vpp > 10nF are almost non existent. (app notes from several manufacturers contain capacitors which are not suitable. Yes, Infineon. TI seems to prefer capacitors with lousy data sheets, probably not a match for the application) With film caps you are forced to limit your choice to a bunch of 4.7...6.8nF 1000Vdc caps in parallel if you don't want to limit your choice to one or two suppliers. I then had the idea to look for C0G capacitors, thinking they would be very expensive. Surprise, not expensive at all and they can handle over 400Vpp up 500kHz with an ESR well below 20mOhm. Try to find a film cap able to outperform a 630V 22nF 1206 C0G capacitor. The best you will find will be at least 6x18mm in size.

Did I mention the decrease in loop size using those small capacitors? An other advantage when you care about how much your supply radiates. Now the loop size is dominated by creepage/clearance requirements instead of large capacitors.

Just my two cents on C0G capacitors for anyone cramming an LLC converter into a small space.

[Kleinstein]

The usual C0G capacitor behave very well.  From the dielectric absorption / loss they are comparable to PP or PS film capacitors.  The nice thing is that C0G are also available as SMD parts, while the low loss film types are usually THT only, as they are limited in temperature. With modern MLCCs they also tend to be smaller than film types.
At least for the smaller values like < 25 nF they may even be cheaper than film types.

[T3sl4co1l]

The same can happen with low ESR capacitors. About 20 years ago I made a PFC with two 470µF capacitors in parallel and every time I unplugged the thing I could hear some whining. At first I thought the under voltage lockout isn't working. After some probing around, nothing could be seen. In the end I made horn from paper (you don't want to stick your years next to capacitors charged to 390V) to locate the sound. Somehow charge was being transferred between those two capacitors while the voltage was declining.

You can make almost any oscillate, of course; just add enough inductance.  That's somewhat of a hazard with large lumped values -- in large packages, the links between them can have relatively high inductance.

Hm. Unless those were connected with long cables somehow, it wouldn't make sense for that case to be audible -- 235uF (capacitors act in series in their loop) resonates with say 1uH (~1m apart) at 10kHz.  But with a control loop nearby, who knows, I guess?  (But I'm guessing you probed that and found nothing suspicious?  Or, you have truly magnificent hearing: 30kHz would be from a more reasonable 120nH, and it's not unheard of for some humans to hear this high.  Or, even more speculative, there could be some sort of hetrodyning, against the switching frequency let's say, but... *shrug*?!)

About the ceramic caps in an LLC instead of film caps. I'm doing the same thing for a 600W supply. Perhaps of interest:
https://www.psma.com/sites/default/files/uploads/tech-forums-capacitor/presentations/is94-quantifying-allowable-ac-parameters-dc-rated-ceramic-capacitors.pdf

Oh man, that had some slide gore... slide 5 apparently says:

Code: [Select]

Ex. Use of X7R (Class II dielectric)
• Provides Cost Savings
• Not Ideal as Series Resonant Capacitor (lack of stability)
• High ESR causes significant efficiency loss in power transition
• Quality Factor is orders of magnitude higher than C0G dielectrics
Design optimization is the important element here, but can come at a price. Therefore,
MLCC suppliers must work closely with design engineers to verify compliance with end
application requirements and ensure proper safety margin is considered while
supporting improvement efforts.

That must be a typo, "orders of magnitude lower" surely.

Also on pp.11-13, curious that they show a micrograph and diagram of CaZrO3, but not BaTiO3; I guess the dipole cartoon is supposed to imply domain wall movement or such, but, it doesn't really do that...

But overall, an informative and useful slide deck.  Am curious what their rationale is for "modeling" ESR, and apparently doubting their instruments (at times?).  That's not covered in the slides, but maybe there was some in the accompanying talk, or questions thereafter.

Film capacitors able to handle 250Vpp > 10nF are almost non existent. (app notes from several manufacturers contain capacitors which are not suitable. Yes, Infineon. TI seems to prefer capacitors with lousy data sheets, probably not a match for the application) With film caps you are forced to limit your choice to a bunch of 4.7...6.8nF 1000Vdc caps in parallel if you don't want to limit your choice to one or two suppliers. I then had the idea to look for C0G capacitors, thinking they would be very expensive. Surprise, not expensive at all and they can handle over 400Vpp up 500kHz with an ESR well below 20mOhm. Try to find a film cap able to outperform a 630V 22nF 1206 C0G capacitor. The best you will find will be at least 6x18mm in size.

...At frequency?  I've definitely ran hundreds of volts on induction heaters, even with commercial capacitors.  (Protip: don't use the super-compact X-type caps; they do work, surprisingly, but they get... warm... )

Conversely, I've had an application where the "DC Link" type MKPs melted.  That's some weird shit... the dielectric expands massively as it melts, carrying the electrodes with; at some point, the bottom seal (epoxy) cracks, contents squirt out, and....... they keep on operating? Sometimes?  Despite the "elephant's foot" eviscerating onto the PCB, they can actually operate that way for some time...  But eventually, enough stretch and shear occurs, electrode separation shrinks, self-healing fails, and it shorts out.

Anyway, trick is to look through high frequency and snubber types.  It's tricky, because a lot of snubbers are shit for HF losses.  CDE 935C for example, probably by way of the schoopage being whole-face (thus shielding the center by skin effect), and the generally thick build, they're good for hardly any Irms at 100kHz.  All peak, no average.  Lots of sifting through datasheets or characteristics to find V(F) or I(F) ratings (or ESR(F) and Pd ratings).

I've been impressed by, let's see, *digs through parts bin*...
Hmm B32632 (EPCOS/TDK) is now obsolete, huh...
Or what is this one, either B32672 or B32642?  Still in production, nice.
There was... B32262 maybe?? I'm sure I'm misremembering, but some 630VDC (250VAC) 0.1uF 15mm (LS) cap I used for my first big induction heating tank capacitor ever.  Rated I think 2.3A/ea at 100kHz or so?  Maybe they're still around (under the correct series).

Illinois Capacitor PPBs are nice, though on the expensive side, and if you need to obey ratings, the RMS is surprisingly not all that high.  They take abuse, at least in my limited experience; do with that what you will.

KEMET (nee Aerovox) R46, R73, etc. have some nice offerings.  Some low ESR and high density X caps, as well as pulse and resonant stuff.

I think 715P ("Orange Drop", CDE, nee Sprague) would be quite excellent if they just didn't make them with steel leads.  Dielectric losses are low and metallization seems adequate otherwise.  Expensive, probably because they're in the audio and repair boutique.

Panasonic have a number of nice dipped PP families too, though it's kind of a pain sifting through their website to find all the data.

Tim

[Zero999]

I would say the bigger issue for decoupling is their losses are so goddamned low (Q > 3000 is not unheard of).  You put two in parallel, side by side, and they oscillate together.  Even larger values like 10nF (1206).  An effective PDN using just them, requires resistors EVERYWHERE to keep the Q down!

The same can happen with low ESR capacitors. About 20 years ago I made a PFC with two 470µF capacitors in parallel and every time I unplugged the thing I could hear some whining. At first I thought the under voltage lockout isn't working. After some probing around, nothing could be seen. In the end I made horn from paper (you don't want to stick your years next to capacitors charged to 390V) to locate the sound. Somehow charge was being transferred between those two capacitors while the voltage was declining.

What sort of capacitors were they? 470µF 400V are usually aluminium electrolytic which is not piezoelectric.

[TimFox]

to T3sl4co1l:

"I think 715P ("Orange Drop", CDE, nee Sprague) would be quite excellent if they just didn't make them with steel leads.  Dielectric losses are low and metallization seems adequate otherwise.  Expensive, probably because they're in the audio and repair boutique."

Another SBE series 716P uses copper leads:  they are still available from CDE.  The shape of the body is somewhat squashed, compared with 715P, but the specs are the same.
https://www.cde.com/resources/catalogs/716p.pdf

These two series from Sprague are metal film and polypropylene foil, instead of the more common metallized film units that are smaller.

[temperance]

What sort of capacitors were they? 470µF 400V are usually aluminium electrolytic which is not piezoelectric.

Just regular (in that time BC components) 450V 470µF capacitors connected together on a copper plane. Maybe some relic from 20 years ago is still somewhere in box in a shelve between failed attempts at construction a class D amplifier.

What you hear is electrostriction. Of course it wasn't loud and about 10kHz. But clearly audible at a short distance.

T3sl4co1l
...At frequency?  I've definitely ran hundreds of volts on induction heaters, even with commercial capacitors.

100...250kHz, 250Vpp or 88Vrms. I've looked at those from TDK and Kemet.
For example the one you mentioned: B3265x: https://www.tdk-electronics.tdk.com/inf/20/20/db/fc_2009/MKP_B32651_658.pdf

I will stick with the C0G capacitors. It saves a lot of space minimizes loop area and they can be placed with a regular pick and place. Cost, about 10 cent for a 10nF 650V capacitors in quantities. Not bad at all. I don't mind if 8...10 pieces will be required.

The CDE 716P is similar and I would need smaller capacitors in parallel to reach my goal unless reverting to two pieces 2kVdc measuring about 32x11mm each.

Slide gore. I didn't even see that while scanning for other more relevant bits and pieces. Nonetheless, the slide contains some valuable information.

[ballsystemlord]

These might interest readers:
https://www.vishay.com/docs/28536/geninfo-cd-rfi-safcaps.pdf
https://www.raviyp.com/difference-between-x7r-x5r-x8r-z5u-y5v-x7s-c0g-capacitor-dielectrics/

[Weston]

I am surprised that no one has mentioned U2J capacitors yet. Its a (new-ish?) Class I dielectric with not great temperature coefficient compared to C0G and a somewhat lower but still high Q  (seems to be ~250 to >1000) and a higher energy density and lower cost.

Based on the product literature it seems targeted for power applications, both resonant capacitors and high voltage DC decoupling. I have seen it used in a few projects and have it designed in for a LLC converter I am currently working on.

https://www.digikey.com/en/product-highlight/k/kemet/u2j-dielectric-multilayer-ceramic-capacitors

[TimFox]

I am surprised that no one has mentioned U2J capacitors yet. Its a (new-ish?) Class I dielectric with not great temperature coefficient compared to C0G and a somewhat lower but still high Q  (seems to be ~250 to >1000) and a higher energy density and lower cost.

Based on the product literature it seems targeted for power applications, both resonant capacitors and high voltage DC decoupling. I have seen it used in a few projects and have it designed in for a LLC converter I am currently working on.

https://www.digikey.com/en/product-highlight/k/kemet/u2j-dielectric-multilayer-ceramic-capacitors

From the Kemet sheet, that dielectric seems to be the same as “N750” ceramic (military code).  I have used that dielectric in disc capacitors and ceramic trimmers where the higher dielectric constant resulted in a smaller size than NP0, and I could tolerate the higher tempco.