Is ceramic capacitor package relevant?

[BorisRap]

Hello, I would like to know if there would be a difference between using two ceramic capacitors with same specs but in different packages.

Let's put an example:

I have a switching controller IC which will drive some mosfets that need fairly high peak current on switching. So I need for example, a 10uF decoupling capacitor next to the VCC pin of the chip.
If the VCC rail is 12V, I could put a 10uF 25V capacitor in size 0603, 0805, 1206 or 1210 (and more).

But does it really matter? If the voltage rating is way above the applied voltage and the capacitance is the same are there any advantages of choosing one over another?

(Electrically speaking. I don't mean stock availability, pricing, manufacturing simplicity, etc...)

Thank you very much

[strawberry]

smaller size will get worse capacitance performance
some nH will affect edge speed only above some ~5MHz
tested el-capacitors up to 1MHz is fine

[TheUnnamedNewbie]

A smaller package has a lower inductance, both the partial inductance of the package itself, but also the lowest inductance you can get when routing it. A smaller package will also allow other components around it to be routed more compactly, which again reduces the inductance of the network, and could also make routing other signals better.

[wraper]

Unless it's NP0 or similar, capacitors in smaller package in general lose capacitance with applied voltage much faster than in larger package. This is especially true when they get close to maximum possible capacitance/voltage rating for package size. So don't be surprised when there is 2uF out of 10uF left in 0402 6.3V cap at 5V.

[Vovk_Z]

A smaller package has a lower inductance, both the partial inductance of the package itself, but also the lowest inductance you can get when routing it. A smaller package will also allow other components around it to be routed more compactly, which again reduces the inductance of the network, and could also make routing other signals better.

This is true of cause, but when you need some capacitance with MLCC cap then small package is your enemy.

My experience with 4ADC 5-25 VDC DC/DC says that 4.7 uF 1210 X7R is much better than 10 uF 1206 and much-much better than 0805 22 uF. I mean 1210 4.7 uF works as it should but 0805 22uF almost doesn't help.

[Geoff-AU]

Unless it's NP0 or similar, capacitors in smaller package in general lose capacitance with applied voltage much faster than in larger package. This is especially true when they get close to maximum possible capacitance/voltage rating for package size. So don't be surprised when there is 2uF out of 10uF left in 0402 6.3V cap at 5V.

Hmm.  Given that the curves of a given dielectric+package are exactly the same across voltage ratings (eg the start of the curve X5R 0805 for 6.3V, 10V, 16V) I smell a rat.  I suspect that manufacturer is punching out exactly the same component and just labelling them with different voltage ratings.  Or, as with CPUs, they are binned after manufacturing according to leakage current or some other metric and the "happy" ones end up with a higher voltage rating than the "sad" ones.  Seems like a stretch to individually test these things though.  Bit of a moot point anyway as you'd never risk the lower voltage tolerance if your design called for more.

[mariush]

The picture is taken from the Maxim application note / blog post / tutorial here : https://pdfserv.maximintegrated.com/en/an/TUT5527.pdf

Temperature and Voltage Variation of Ceramic  Capacitors, or Why Your 4.7µF Capacitor Becomes a 0.33µF Capacitor
By: Mark Fortunato, Senior Principal Member of Technical Staff
Dec 04, 2012

Good pdf to read.

[Psi]

The small ones tend to require some more exotics materials to get the same capacity within a smaller size.
The exotic materials tend to have down sides, and odd or extreme characteristics
Like your 10V 0402 10uF cap is actually only 10uF at around 10V.   At 3.3V maybe it's only 4uF.   

[Kleinstein]

Hmm.  Given that the curves of a given dielectric+package are exactly the same across voltage ratings (eg the start of the curve X5R 0805 for 6.3V, 10V, 16V) I smell a rat.  I suspect that manufacturer is punching out exactly the same component and just labelling them with different voltage ratings.  Or, as with CPUs, they are binned after manufacturing according to leakage current or some other metric and the "happy" ones end up with a higher voltage rating than the "sad" ones.  Seems like a stretch to individually test these things though.  Bit of a moot point anyway as you'd never risk the lower voltage tolerance if your design called for more.

This can be the case, but there are also different sereies / different types or X7R that can be different. So the curves shown are only examples. This can change with manufacturer and series.
Beside the footprint, there is also the thickness. So not all 0805 are the same: some a flat and some (especially higher capacitance) can be quite thick. The thicker ones likely behave a bit more like a larger case.

Beside the instant biasing effect, there could also be an effect of aging. Under bias the capacitance tends to go down over time.

[wraper]

This can be the case, but there are also different sereies / different types or X7R that can be different. So the curves shown are only examples. This can change with manufacturer and series.
Beside the footprint, there is also the thickness. So not all 0805 are the same: some a flat and some (especially higher capacitance) can be quite thick. The thicker ones likely behave a bit more like a larger case.

Beside the instant biasing effect, there could also be an effect of aging. Under bias the capacitance tends to go down over time.

This is only true for 0805 and above. I have not seen 0603 and below with cross section that is not square. And capacitances where you should really care about voltage capacitances are high enough that you should expect capacitor to have square cross section. 1210 and above are another thing, usually their thickness is smaller than width. But I would avoid 1206 and above unless I have no better choice. And never place them closer than 1.5 cm to any mounting holes. They are very prone to cracking. Yes there are caps with soft termination and welded tabs but it's cheaper to use a bunch of smaller caps. But you really should look for voltage charts for particular part number if capacitance matters for your design. They can be very different, there could be a few times difference for high capacitance parts. Neither I suggest to use them at all, way less reliable, and not that much more capacitance under voltage bias. Better to chose something with >5 times less than maximum possible capacitance in the package. Or use those parts for very low voltage rails, like < 1.5V

[JohanH]

Don't confuse voltage rating with capacitance drop from DC bias. It seems that manufacturers not always specify DC bias rating for type 2 dielectric MLCC. In these cases you can assume that capacitance drops quite a lot. Some manufacturers claim small impact of DC bias. At high DC voltages, you might have to consider tantalum or electrolytic caps.

[wraper]

Don't confuse voltage rating with capacitance drop from DC bias. It seems that manufacturers not always specify DC bias rating for type 2 dielectric MLCC. In these cases you can assume that capacitance drops quite a lot. Some manufacturers claim small impact of DC bias. At high DC voltages, you might have to consider tantalum or electrolytic caps.

They specify it. But you need to dig to find it as it's generally not available in datasheets, especially those for whole series and not particular part number. Often it's not publicly available and you need to request it from manufacturer. For example you need to use SimSurfing tool for Murata https://ds.murata.co.jp/simsurfing/mlcc.html?lcid=en-us&jis=false&md5=e5783a61c0acf1952ab00d812884fc87

[tooki]

This is only true for 0805 and above. I have not seen 0603 and below with cross section that is not square.

”I’ve never seen” ≠ “do not exist, ergo not applicable”.

You can go on Digi-Key and filter by package and height, and there are many 0603s and 0402s that are thinner (shorter) than they are wide. (Very few 0201s, though. I didn’t look at anything smaller than that.)

[wraper]

This is only true for 0805 and above. I have not seen 0603 and below with cross section that is not square.

”I’ve never seen” ≠ “do not exist, ergo not applicable”.

You can go on Digi-Key and filter by package and height, and there are many 0603s and 0402s that are thinner (shorter) than they are wide. (Very few 0201s, though. I didn’t look at anything smaller than that.)

Well, there are some. But most of them are special parts like X2Y caps or capacitor arrays. There are a few which are taller than their width too. Out of usual capacitors barely any is stock. But IMHO you would not use them unless need a very low profile or extreme capacitance for the size.

[Siwastaja]

The question you need to ask, what is the critical parameter you need out of this capacitor?

Do you use it to store as much energy as possible (exact amount of capacitance matters, more is better), or do you use it to respond to quick changes of current while minimizing voltage change (exact amount of inductance matters, less is better)?

If it is a 10uF part next to a gate driver driving 1nF worth of equivalent gate capacitance, then obviously it doesn't much matter if it's actually 10uF or 5uF, only ridiculously tiny part of that charge is used to deliver 1nF into the MOSFET. This is the general case with IC bypass caps - they are used for lowest possible high-frequency supply impedance, exact amount of capacitance is irrelevant.

For the same reason as placement is important, so is small package size.

Sometimes you want both minimized inductance and nontrivial amount of capacitance (say more than 1uF, so you can't just slap in any 0402/0603 and call it a day). In such cases, paralleling multiple capacitors (of exact same type/value) offer layout benefits as you can sprinkle them to conform to your other components. For similar reason, try to use at least two vias per pad when going to another layer; giving more parallel routes to the changing currents reduces inductance.

[nctnico]

A smaller package has a lower inductance, both the partial inductance of the package itself, but also the lowest inductance you can get when routing it. A smaller package will also allow other components around it to be routed more compactly, which again reduces the inductance of the network, and could also make routing other signals better.

This is true of cause, but when you need some capacitance with MLCC cap then small package is your enemy.

My experience with 4ADC 5-25 VDC DC/DC says that 4.7 uF 1210 X7R is much better than 10 uF 1206 and much-much better than 0805 22 uF. I mean 1210 4.7 uF works as it should but 0805 22uF almost doesn't help.

Actually there is no real relationship between package and capacitance loss. There is a clear relation between cost and capacitance loss. If you look closer, you'll find that many manufacturers have better (=less loss of capacitance / voltage) capacitors in the same package size. So without comparing actual part numbers & DC bias graphs, a generic discussion about package sizes is meaningless.

[TimFox]

For Class-II dielectrics (X7R, etc.), the dielectric itself has a substantial change with applied DC voltage.
For the material in question, the change in dielectric constant is a function of the field strength in the material, which is the voltage gradient (V/m) from plate to plate.
Therefore, a physically smaller capacitor will have a higher voltage gradient for a given DC voltage, which is why the curves posted above are "worse" for the smaller packages for a given dielectric.

[wraper]

Actually there is no real relationship between package and capacitance loss. There is a clear relation between cost and capacitance loss. If you look closer, you'll find that many manufacturers have better (=less loss of capacitance / voltage) capacitors in the same package size. So without comparing actual part numbers & DC bias graphs, a generic discussion about package sizes is meaningless.

There is no clear relation in price without taking into account package size, especially at distributor pricing. But there is a clear relation in size. Good smaller size cap is not that likely to have better capacitance/voltage relation than cheaper larger size cap. Especially at top end of capacitance rating for the size. No matter how expensive it is, it will totally suck. It's just some are totally useless at larger voltage bias while other still a little bit useful.

[floobydust]

Physical size does matter, the larger package part has a lower E-field across the dielectric, giving less stress and unwanted side effects.
Although, this also depends somewhat on the technology of the MLCC capacitor manufacturer. Some do offer parts that appear to be better.
Whether this is due to bullshit specs or not I am not sure.

Auto parts manufacturer Continental got burned by MLCC's varying widely in spec and quality. They had an ECM LDO go unstable (TPS796xx input cap) despite their (internal) design rules for derating due to voltage coefficient, tolerance, temperature. The final clincher was aging, which engineers do not really know about or design for. Massive drop in the first several hours!
There was a meeting and open letter to MLCC cap manufacturers to stop hiding their bullshit on the datasheets. AVX Mr. Ron Demcko was their too, Dave interviewed him a while back.
Open Letter to AEC-Q200 Committee on MLCC DC BIAS Ageing
Update: Class-II MLCC DC-Bias Aging Issues in Automotive Applications

It's worthwhile for engineers to know the deceptions around MLCC capacitors.

[Vovk_Z]

Actually there is no real relationship between package and capacitance loss. There is a clear relation between cost and capacitance loss.

I would say wise versa. Very often, cost doesn't say anything, until something is clearly stated by the manufacturer in a datasheet. I mean we are talking about the 'usual' general use MLCC caps but not about special one types.

[tooki]

This is only true for 0805 and above. I have not seen 0603 and below with cross section that is not square.

”I’ve never seen” ≠ “do not exist, ergo not applicable”.

You can go on Digi-Key and filter by package and height, and there are many 0603s and 0402s that are thinner (shorter) than they are wide. (Very few 0201s, though. I didn’t look at anything smaller than that.)

Well, there are some. But most of them are special parts like X2Y caps or capacitor arrays. There are a few which are taller than their width too. Out of usual capacitors barely any is stock. But IMHO you would not use them unless need a very low profile or extreme capacitance for the size.

Did you even look? Just look in ceramic caps, choose either 0402 or 0603, and then filter by height. Most are just ordinary (no special features, or just "low profile"). Not talking about arrays.

I'm not saying that they in any way make up the majority of 0603 and 0402 caps, but they're not exotic, and certainly do not support your blanket statement of "This is only true for 0805 and above. I have not seen 0603 and below with cross section that is not square."