High capacitance MLCC ESR values and current carrying capabilities

[Yansi]

Hello,

why is that only very seldom manufacturers publish ESR values and maximum RMS currents for MLCC capacitors?

I have a 40 A buck converter design, that requires up to 20A RMS current through the input capacitors. That is a lot of electrolytics, so I thought I'd use a ceramic.

But I have hard times finding any data about ESR and current carrying capabilities of these suckers.

So for example, a typical 10uF/50V/X7R cap in 1206 package, what may be the ballpark of ESR and max RMS current?  From my little research, the ESR may be in the 10-15 mR neighborhood. So for 20 A rms current I'd need quite a bunch of those for them to sustain the ripple continuously without melting. 

[Zeyneb]

I agree but Taiyo Yuden is an exception.

Check here:
https://www.yuden.co.jp/or/product/

[uer166]

I use https://ksim3.kemet.com/capacitor-simulation and similar vendor cap simulators, they give not only ESR vs frequency curves, but also temp rise vs. RMS current vs. Frequency.

[Phoenix]

I use https://ksim3.kemet.com/capacitor-simulation and similar vendor cap simulators, they give not only ESR vs frequency curves, but also temp rise vs. RMS current vs. Frequency.

I second KSIM. Heaps of data/curves available for their parts. I've used it to spec AC coupling capacitors for an LLC - I expect you will be better off with a 1210 package for current handling.

[Yansi]

Awwwww! That KSIM is lovely!

Now if I'd know how different properties may have caps from different manufacturers. For example, Samsung caps are decent cheap.

But back to my original issue of the 20 A rms current.  Great to see ceramic caps can handle tons ... i mean amperes of current!
To get any sensible voltage ripple on the input, seems I'd need about 100uF to get below 1V of ripple.  That would be a LOT of ceramic caps, considering the capacitance decrease with operating voltage applied.  According to  KSIM, the decrease of capacitance may be -80! Damn..

//EDIT: Seems also Wurth has the "Red Expert" cap sim tool, although not as comprehensive as KEMET has.

[T3sl4co1l]

Simple, keep shopping until you find parts that specify Irms.

Also, I take it this is single phase?  Why not phase interleave?  Ripple goes down to a fraction, at expense to overall complexity; but when you're stuck between a truckload of caps or a couple more transistors (and a controller that can support it), it's a good deal!

I made such a decision on this design,



It's a dual buck module, with the right side being a moderately high voltage input (above SELV, but below rectified mains) and 20A output.  The overall form factor was important, as was the design and BOM budget (so, no highly optimized GaN and all that).  Also it's adjustable CC/CV with digital control, which I didn't see any likely candidates among controller ICs so it's a full custom design anyway (but one that I've more or less used before, so again, not a lot of design time).  Anyway, the capacitance requirement alone would've blown the module spec (particularly height) if I went single channel, but as you can see, five and four little electrolytics is all that's needed for cool and comfortable operation by going to just dual phase.

Tim

[Yansi]

Of course 2 or more phases will be much much better. But the design time will go through the roof, as I do not have any experience with any of the available controller ICs, so there will be a steep learning curve.

And in case of complicating the buck things, the whole thing does not make much sense at all, as that is supposed to be a "bolt on hotfix" to bring a 34V unregulated main supply rail down to 24V. So rather then complicating the buck side with interleaved operation and such, redesign of the whole main power supply would have more sense. (400 to 600V input range). (.. and that is of course, what I have already recommended).

[T3sl4co1l]

Ah. Literally on the other side of that decision then!

LM25119 might be attractive -- I've used it before, seems to be pretty straightforward.  But yeah, if you don't have the time or risk to try it out, brute force it is.

Tim

[Siwastaja]

TDK also has the data. They have a decent interactive web page for each capacitor.

MLCC ripple rating is seldom a problem (I have never hit such a case, but I imagine it's possible), but when you check for the actual C under DC bias, you can double-check it has enough ripple current rating for the job as well.

[T3sl4co1l]

MLCC ripple rating is seldom a problem (I have never hit such a case, but I imagine it's possible), but when you check for the actual C under DC bias, you can double-check it has enough ripple current rating for the job as well.

Heh, I've tested it a little, for curiosity's sake -- like abusing a 0.47uF 0603 at 100V and, I forget what frequency was it 500kHz, or lower?  Anyway, self-heating obviously is a problem, as is DC bias.  In a resonant test, the frequency shifts dramatically, which can be used to control power.  The thermal time constant makes that a bit hard to use, though.

Losses generally go down with DC bias, though I don't have any numbers as to how much.  Analogous to inductors where saturated core effectively becomes airgap.

Probably, under heavy load, they'll eventually fail due to thermal cycling and stress -- the ceramic isn't very conductive, so it takes less power to reach the same thermal gradient, or peak core temperature, as say a chip resistor (made of Al2O3, fairly conductive in comparison).

A reasonably safe assumption, I think, is taking ESR or tan delta, and figuring half the power dissipation of a comparable sized, ordinary rated (not high power or pulse) chip resistor.  So, a 1206, 1/8W or less, say.

Tim

[Yansi]

Losses generally go down with DC bias, though I don't have any numbers as to how much.  Analogous to inductors where saturated core effectively becomes airgap.

Thats exactly what I've observed in the KSIM. Was quite surprised at first, but it makes sense as you write it. Also was quite surprised, that loading a 50V rated MLCC with 35V can decrease the capacity down to 20% of the nominal (was looking at some 10uF 50V X7R 1210 part part in KSIM I think)

Some times, looking at the %capacitance vs. %voltage graphs, I remembered that at about 50% rated voltage, you get about 60 or what % of capacitance. But it seems it is very wrong to assume for all the caps. Never thought even the X7R dielectric could be as bad.

Thanks for the tip on LM25119, will have a look. I was looking at LM5143 once, but there were few things I didn't like on that IC.

//EDIT: LM25119 just seems to be little bit simpler LM5143. Unfortunately, 25119 has 42V max input voltage, what I'd say is not enough to safely accept 34V unregulated rail, that can potentially go close to 40V.

//EDIT2: With the 2phase interleaved buck, the input cap ripple current seems to be about 10A max, nothing four 470uF 50V Nichicon UHW caps wouldn't handle, likely without flinching, as the ripple current rating is at the 105°C, isn't it? At lower temperature, it could handle much more.

[ConKbot]

With an all-ceramic input, you have to mind your upstream supply as there is very little ESR to help de-q the whole thing. Even if you are using ceramics to eat a bulk of the ripple, Id still suggest some electrolytics in the mix to add some loss too.  Since you know what your power source is, a bit of modeling on that, and a simulation of at least the first order parasitics (ESR/ESL of caps with the capacitance knocked down appropriately, wiring inductance), can let you do a frequency sweep and look for where it may be a bit less behaved than you expected, and figure out what sort of electrolytic capacitor ESR you may want to let the ceramics handle most the ripple while letting the electrolytic ESR keep ringing mostly under control.

And yeah, once you are past 5-6v ceramics can start to be a bit underwhelming in terms of how much capacitance they retain, and it gets a bit silly past 12v.  Ive had situations where 2 caps in series (to prevent a crack from shorting a main power rail)  had the same capacitance as one by itself.

[Alti]

From my little research, the ESR may be in the 10-15 mR neighborhood. So for 20 A rms current I'd need quite a bunch of those for them to sustain the ripple continuously without melting.

With 0.25W dissipation per 1206 that is 4 capacitors in parallel.

[Siwastaja]

Also was quite surprised, that loading a 50V rated MLCC with 35V can decrease the capacity down to 20% of the nominal (was looking at some 10uF 50V X7R 1210 part part in KSIM I think)

Yes! The problem is, while everybody knows about the DC bias effect nowadays, "rules of thumbs" like "only Y5V is so bad, X7R goes just down to 50%" are blatantly false and unhelpful.

The only good rule of thumb is, "if it sounds too good, it likely is", i.e., actual energy storage relates to the package size, and energy storage relates to C and applied V (not rated V).

So if the amount of C is important at all, only pick parts with ratings available. I tend to calculate $/uF actual and choose based on this. If an unrated part from otherwise reputable manufacturer, like Samsung, is available, then I have to assume it performs similarly to a similarly rated, same-volume (don't forget height!) part from a manufacturer who publishes ratings. It may be worse, but likely not many times worse. But if the price and availability is much better, then I might give it a shot.

[sandalcandal]

For high power smoothing has anyone tried TDK Ceralink caps? https://au.mouser.com/new/epcos/tdk-ceralink-flex-capacitors/
They use some fancy dielectric ceramic which actually increases capacitance with applied DC voltage bias. You actually need to derate capacitance if there isn't a DC bias. Despite high unit costs, they seem cost competitive with X7R/X7T in a similar voltage/automotive class when you account for DC bias capacitance reduction.

Edit: I've also been looking at ceramic capacitors at ~20A RMS but for use as resonant capacitors in a CLLC For the 20A current level it seems like one needs to go for multiple parallel capacitors to stay within characterised 20°C thermal rise. As other have mentioned, Kemet and TDK seem to be pretty good about making available ESR and ripple current vs. temperature rise data. Murata also has similar data available but KEMET and TDK caps seem more cost effective from what I've seen. ESR and ripple current rating can vary significantly for same dielectric, voltage and capacitance but different package so be sure to try check a few, particularly those in bigger packages.

[Yansi]

From my little research, the ESR may be in the 10-15 mR neighborhood. So for 20 A rms current I'd need quite a bunch of those for them to sustain the ripple continuously without melting.

With 0.25W dissipation per 1206 that is 4 capacitors in parallel.

Not really! Don't forget the reduction of capacitance, so in the end, the limit is not ESR/loss, but overall capacitance.

[bdunham7]

Not really! Don't forget the reduction of capacitance, so in the end, the limit is not ESR/loss, but overall capacitance.

What switching frequency are you anticipating using?

[Yansi]

Not really! Don't forget the reduction of capacitance, so in the end, the limit is not ESR/loss, but overall capacitance.

What switching frequency are you anticipating using?

Well, originally, for the sake of the simple hack-job of bringing the 34V down, I aimed for about 100kHz for simplicity, based on available core materials and mosfets (a pair of naughty hammers IRFS7530s).

But in the end it seems, solution with an interleaved buck is way "simpler", even though the component count is higher in the end.

I'll see, if the buck will be made, or they decide to change the main PSU.

[T3sl4co1l]

From my little research, the ESR may be in the 10-15 mR neighborhood. So for 20 A rms current I'd need quite a bunch of those for them to sustain the ripple continuously without melting.

With 0.25W dissipation per 1206 that is 4 capacitors in parallel.

Mind as I noted above, the maximum dissipation will be lower than for a chip resistor of the same size -- the ceramic is much less conductive.

Tim

[Yansi]

Why is it much less heat conductive? Thought ceramic material are quite good heat conductors, compared to for example silicone or mica.

[T3sl4co1l]

For high power smoothing has anyone tried TDK Ceralink caps? https://au.mouser.com/new/epcos/tdk-ceralink-flex-capacitors/
They use some fancy dielectric ceramic which actually increases capacitance with applied DC voltage bias. You actually need to derate capacitance if there isn't a DC bias. Despite high unit costs, they seem cost competitive with X7R/X7T in a similar voltage/automotive class when you account for DC bias capacitance reduction.

Specifically, a poled dielectric: an electret.  Some of these (the earlier series?) are poled at the factory, and must not be heated to the Curie temperature during soldering, meaning they must be hand soldered, and carefully at that.  Others, I don't know how it is they work, but they make the poling voltage/temperature such that they pick it up on first use, and these can be reflow soldered normally.

Do RTFAN, they're unusual beasts, and quite attractive in the right application!

For example, look up Google Little Box.  IIRC, the winning team used these, combined with a GaN inverter, to do what amounts to shunt PFC on the DC input -- one of the curious specs of the challenge was a particularly low input ripple, necessitating energy storage for a full inverter output line cycle.  Their solution used less space, in total, than electrolytic capacitors you'd need to do just this passively!

Edit: I've also been looking at ceramic capacitors at ~20A RMS but for use as resonant capacitors in a CLLC For the 20A current level it seems like one needs to go for multiple parallel capacitors to stay within characterised 20°C thermal rise. As other have mentioned, Kemet and TDK seem to be pretty good about making available ESR and ripple current vs. temperature rise data. Murata also has similar data available but KEMET and TDK caps seem more cost effective from what I've seen. ESR and ripple current rating can vary significantly for same dielectric, voltage and capacitance but different package so be sure to try check a few, particularly those in bigger packages.

For resonant, type 2 dielectric are no good.  You need C0G, but availability sucks in low voltages and high values.  (Incidentally, they have higher energy density than most anything else, if you're using a few hundred volts -- it might even be worthwhile putting taps on those inductors so you can make use of this!)  Next best is film, which, SMT films suck ($$) so consider PP in THT.  You can get some quite beefy pulse and snubber type caps (Illinois PPB comes to mind, or a few series by EPCOS/TDK but I don't remember what numbers).  Though again maybe not in low voltages.

The last possibility is aluminum polymer, which serves a very similar role to film caps, but at low voltages and respectively higher values -- they have similar energy density.  The tan delta (~= 1/Q) may not be enough, though (which is also to say, respect the ripple current ratings).  And they don't like much reversal, so you'd have to use anti-series pairs with DC bias.

The final option would then be: you must be doing something wrong, reconsider your network or topology; or just don't bother with resonant at low voltages.

Tim

[nuno]

If you don't need much capacitance, some polypropylene capacitors have very low ESR.

[sandalcandal]

Thanks Tim
And sorry Yansi for jacking this thread slightly.

Specifically, a poled dielectric: an electret.  Some of these (the earlier series?) are poled at the factory, and must not be heated to the Curie temperature during soldering, meaning they must be hand soldered, and carefully at that.  Others, I don't know how it is they work, but they make the poling voltage/temperature such that they pick it up on first use, and these can be reflow soldered normally.

What was the max temperature limits for those? Ceralink datasheet shows it to be rated for a pretty standard reflow soldering, 260C peak and 150C operating temperature max. They're based on a PLZT ceramic and claim to function based on antiferroelectric material property where "Permanent dipoles form antiparallel zones" https://www.tdk-electronics.tdk.com/download/1195592/1753c455d19f9c7e635942c9cfba0318/ceralink-presentation.pdf

For example, look up Google Little Box.  IIRC, the winning team used these, combined with a GaN inverter, to do what amounts to shunt PFC on the DC input -- one of the curious specs of the challenge was a particularly low input ripple, necessitating energy storage for a full inverter output line cycle.  Their solution used less space, in total, than electrolytic capacitors you'd need to do just this passively!

I remember seeing some funky input filtering in a Google Little Box entries where there was a switched LC used to do "PFC but on DC input" as you say. A potential option but I'm hoping to get away with just a high degree of multiphase (looking like 12 phases) for my final application which seems to be optimal anyway when considering available components/materials and their costs and will probably be enough of a control system challenge.

Edit: Moving my response to the resonant cap stuff to the resonant converter project thread.

[Yansi]

Thanks Tim
And sorry Yansi for jacking this thread slightly.

No problem, I've asked and got the answers I wanted, now reading and learning further, please continue.