EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: mtwieg on September 09, 2023, 04:31:17 pm
-
This is a question that has bugged me for years. Couldn't find any previous posts about it on eevblog (correct me if I'm wrong), so I thought I'd throw it in here, might get some good discussion out of it.
Multilayer ceramic capacitors (MLCCs) are inherently non-polar devices, i.e. they don't have a preferential polarity regarding applied voltage or current. However, the voltage ratings for MLCCs are nearly always given for DC voltage. Of course that doesn't mean the applied voltage is perfectly fixed, there's always some AC component (ripple, etc). But the ratings assume the applied voltage never changes sign.
But what about when the applied voltage does change sign because the AC component is larger than the DC bias? Well, there's nothing forbidding this, and many manufacturers do offer a specific guideline on this scenario: They typically suggest that the peak to peak voltage must be less than the rated DC voltage of the capacitor.
This guideline is independently suggested by several major MLCC manufacturers:
TDK: https://product.tdk.com/system/files/dam/doc/content/FAQ/20_mlcc_voltage_strength.pdf (https://product.tdk.com/system/files/dam/doc/content/FAQ/20_mlcc_voltage_strength.pdf)
Murata: https://www.murata.com/en-us/support/faqs/capacitor/ceramiccapacitor/char/0008 (https://www.murata.com/en-us/support/faqs/capacitor/ceramiccapacitor/char/0008)
Kemet: https://www.kemet.com/content/dam/kemet/lightning/documents/ec-content/KEMET-Max-AC-Rating-for-DC-Rated-MLCCs.pdf (https://www.kemet.com/content/dam/kemet/lightning/documents/ec-content/KEMET-Max-AC-Rating-for-DC-Rated-MLCCs.pdf)
So let's try applying the guideline in a few examples. Suppose we have an AC voltage source with an amplitude of 50V (100Vpp), and we want to connect it to a small MLCC in a few ways:
1. Pass it through a full bridge rectifier (with a real load so we get the classic rectified sine on the output). The capacitor voltage will swing from 0V to +50V, so the guideline says the cap should be rated for at least 50VDC.
2. Again pass it though a full bridge rectifier, but with all the diodes flipped around. Now the capacitor swings from 0V to -50V. Again, guideline says the cap should be rated for at least 50VDC.
3. Connect the AC directly to the MLCC with no rectifier. Now the cap voltage swings from -50V to +50V. The guideline says the capacitor must be rated at least 100VDC. Even though the peak and RMS voltage applied is the same.
In all three cases, the peak and RMS voltage seen by the MLCC is the same. The peak and RMS current through the MLCC is also the same! But the guideline suggests that case 3 needs twice the voltage reason.
Basically what I'm claiming is the guideline makes absolutely no sense as a general rule.
If this guideline came from just one vendor then I would probably brush it off, but seeing it from multiple large manufacturers makes it very hard to just ignore it. But the manufacturers also do not offer any explanation for the guideline.
There are some obvious hypothetical reasons, which I'll go ahead and address here:
1. AC voltage will result in large AC current, leading to heating.
Then the actual limitation is current/power dissipation, which will depend greatly on the frequency of the applied voltage. But the guideline says nothing about this.
2. MLCCs with class dielectrics aren't suited to AC-only circuits, their apparent capacitance would greatly depend on AC amplitude, create tons of distortion, etc...
Indeed, but this doesn't justify setting a conservative maximum voltage rating for AC. And since the guideline doesn't mention dielectric type, I don't think this is the reason.
3. MLCCs are not suitable for use in mains-connected circuitry (AC input filters, Y and X rated caps, etc) for safety/reliability issues.
I absolutely agree. But that's the case even if I follow the guideline. So I don't think that's what the guideline is based on.
Anyways I've probably gone on for long enough. :horse: Interested in hearing other perspectives on the matter.
-
MLCC ceramics (as other ceramic types) don't have a real accurate voltage range when they have a breakdown (unlike many polar types). They'll work some time with 4x or 10x rated voltage across (DC or DC+AC), so their breakdon voltage is totally unpredictable and fuzzy, so I don't see a clear thing to discuss.
-
In old designs you can sometimes find where common ceramic capacitors are used as X or Y class capacitors because at the time, capacitors specifically intended for X or Y class applications did not exist. In these cases, typically the DC voltage rating of the capacitor was derated to like 25% of the peak AC voltage across the capacitor.
-
Remember that "MLCC" is a type of construction (Multi Layer Ceramic Capacitor) and a given MLCC capacitor can contain any of a wide range of dielectrics (C0G, X7R, Y5U, and Z5U are popular types) that have very different properties.
The losses (producing ESR) vary considerably between the different materials, so the self-heating will vary accordingly for a given current through the device.
Of course, the current through the capacitance depends on the voltage and frequency, and for non-C0G capacitors, the capacitance itself varies with voltage.
-
Current and heating aren't part of the particular issue you raise since you would have to mention frequency for those to have meaning. The only thing I can think of is that changes in voltage put microscopic physical stresses on a capacitor and MLCCs are probably among the most delicate. So the limiting factor seems to be the delta between the highest and lowest voltages, or between zero and the maximum voltage in the case of a signal that doesn't cross zero. After all, even those systems go to zero when they are off. So if you define that limiting factor as VDELTA(max), then in each of the four cases--DC, pulsed DC, AC+DCpk and AC (no DC or DC less than ACp-p) it matches what the manufacturer guidelines are telling you.
Exactly why that is the case from a materials science or solid-state physics perspective, IDK. But VDELTA(max) seems to be what they are specifying. They don't even make an attempt at implying an AC current rating, they just demur and tell you to not exceed the rated temperature. I wonder what sort of variability there is between samples of the same part number.
-
They'll work some time with 4x or 10x rated voltage across (DC or DC+AC), so their breakdon voltage is totally unpredictable and fuzzy, so I don't see a clear thing to discuss.
The purpose of absolute maximum ratings is, IMO, to give the designer confidence that they will never see component failures when all the ratings are observed (within reason of course). The fact that some samples can endure stress far beyond the ratings doesn't make the ratings any less clear or meaningful.
In old designs you can sometimes find where common ceramic capacitors are used as X or Y class capacitors because at the time, capacitors specifically intended for X or Y class applications did not exist. In these cases, typically the DC voltage rating of the capacitor was derated to like 25% of the peak AC voltage across the capacitor.
Right, it's not like ratings or standards are absolutely necessary for engineers to make good designs. They used the components and the information they had access to at the time.
Usually the absolute maximum ratings have some intuitive relationship to some physical process (overheating, dielectric breakdown, etc). But this guideline...?
Remember that "MLCC" is a type of construction (Multi Layer Ceramic Capacitor) and a given MLCC capacitor can contain any of a wide range of dielectrics (C0G, X7R, Y5U, and Z5U are popular types) that have very different properties.
...
Current and heating aren't part of the particular issue you raise since you would have to mention frequency for those to have meaning.
The fact that the manufacturers apply the guideline to all MLCC dielectrics equally is one of the primary reason I can't buy into it.
Exactly why that is the case from a materials science or solid-state physics perspective, IDK. But VDELTA(max) seems to be what they are specifying.
Long ago I worked with a much senior EE who explained the nuances of film capacitors and why their AC and DC ratings seem to be inconsistent (or only one rating is given). Unfortunately I forget most of the details (something about buildup of ionized material, pinholes in dielectrics...) but he was quite certain that the ratings had valid science behind them, and should be respected. But there were no guidelines as contrived as the MLCC one...
They don't even make an attempt at implying an AC current rating, they just demur and tell you to not exceed the rated temperature. I wonder what sort of variability there is between samples of the same part number.
Actually some manufacturers do specify max ripple current vs frequency on MLCCs. I know TDK lets you search by ripple current rating (for a 20C rise). Though I don't believe those numbers are to be treated as absolute maximum ratings. Just as a metric for comparing components (like most thermal-based ratings).
-
I don't know what particularly about ceramics needs such ratings, or not, but it's conceivable with film capacitors, being constructed of layers, that the reversal causes more stress and wear, and eventual failure. Mechanisms are electrostriction, and piezoelectricity where applicable (mostly type 2's). Note that non-reversing (ripple) has less electrostrictive stress (stress ~ V^2; with bias dominant, fundamental is strongest) than reversing (2nd harmonic strongest).
We'd still expect some kind of RMS or weighted sum for a more accurate representation of things, but it may be they simply don't want to bother with such, and recommend a worst-case figure of Vpp instead.
Tim
-
The fact that the manufacturers apply the guideline to all MLCC dielectrics equally is one of the primary reason I can't buy into it...
...But there were no guidelines as contrived as the MLCC one...
So you would expect the rule to apply to ferroelectric materials but not C0G or the like? But we don't know what the basis is for the guideline. It may indeed be a bit 'simplified' but that doesn't mean it is contrived or that they all just pulled it out of thin air. I'm sure their capacitor scientists could explain and Kemet will probably respond to you if you ask them.
Here's a chart from the datasheet of a 1kV DC-Link C0G automotive rated capacitor from Kemet. They clearly are sticking to their story. I have no idea specifically what "AC voltage performance" actually means, though.
(https://www.eevblog.com/forum/beginners/ac-voltage-ratings-of-mlccs/?action=dlattach;attach=1870450;image)
-
I don't know what particularly about ceramics needs such ratings, or not, but it's conceivable with film capacitors, being constructed of layers, that the reversal causes more stress and wear, and eventual failure. Mechanisms are electrostriction, and piezoelectricity where applicable (mostly type 2's). Note that non-reversing (ripple) has less electrostrictive stress (stress ~ V^2; with bias dominant, fundamental is strongest) than reversing (2nd harmonic strongest).
See this sounds plausible, though I would expect it to only apply to class II dielectrics (AFAIK class I are never piezoelectric). And I would also expect frequency to play a large role.
but it may be they simply don't want to bother with such, and recommend a worst-case figure of Vpp instead.
Yeah I'm guessing anything beyond a crude guideline like this would have to be fairly complicated and require significant effort to validate. But if they did so, I think they would get more business too...
So you would expect the rule to apply to ferroelectric materials but not C0G or the like?
I would just expect them to be treated differently. The Vpp guideline doesn't really make sense to me for any dielectric though.
But we don't know what the basis is for the guideline. It may indeed be a bit 'simplified' but that doesn't mean it is contrived or that they all just pulled it out of thin air. I'm sure their capacitor scientists could explain and Kemet will probably respond to you if you ask them.
Years ago I did ask our Vishay FAE about this. They weren't a technical FAE, so they asked internally but could not find anyone willing to answer. Only other MLCC manufacturer we have FAE support for right now is Knowles, but I don't think they endorse this guideline anyways.
Here's a chart from the datasheet of a 1kV DC-Link C0G automotive rated capacitor from Kemet. They clearly are sticking to their story. I have no idea specifically what "AC voltage performance" actually means, though.
Right, they're apparently combining the Vpp guideline with an RMS current/temperature rise constraint (lol at the lazy curve fitting).
-
Update: I happened to have a meeting with some FAEs from Knowles/Syfer, another large MLCC manufacturer. To my knowledge, their own technical literature doesn't actually refer to the Vpp guideline (they do offer some series of MLCCs specifically rated for AC operation). Still I thought I'd ask what they thought of it, since a couple of them were formerly R&D engineers.
They immediately recognized what I was referring to, and initially made a few interesting comments. One of them suggested it only makes sense for class II dielectrics. Another suggested that sometimes a Vpp being x1.25 the rated voltage is a more appropriate rule. After a minute they apparently realized they shouldn't speculate so much, and stated that coming up with tailored AC ratings is too much effort than is justified in most cases, hence the existence of such a crude guideline. I couldn't get them to conclusively confirm whether the guideline should be applied to their MLCCs or not.
To be clear, I'm not criticizing them at all. The fact that they actually let slip a bit of speculation was more than most FAEs would allow, and I wasn't expecting anything concrete to come out of it. Nothing they said on the topic is going to impact any of my design decisions. But in the future if I have a specific application and and component in question I may approach them again for narrower advice.