EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: jheatac on August 14, 2024, 03:37:26 pm
-
Newbie question.
Small SMD Capacitors...
First - None seem to be marked to indicate what they should be.
Second - Is there any way to visually tell if the are ceramic, film, or other?
and lastly - Do they fail very often? It isn't convenient to pull it just to test like it is with through hole. It seems that I rarely get a meaningful measurement in circuit. Is there a good way to indicate if it could be a problem?
Thank you!
-
SMD resistors are often labeled, but capacitors only rarely.
Presumably due to temperatures involved in manufacture.
SMD film capacitors are not as common as ceramic.
All bets are off on the ceramic dielectric for an unmarked unit not in its packaging.
If you have an LCR meter, you can get a good idea of the type by measuring its "Q" at 1 kHz:
C0G is very high, maybe 1000.
X7R is much lower, maybe 100.
SMD film capacitors have lower Q than the best TH film capacitors, e.g. polypropylene around 1000.
In-circuit measurement, by definition, includes everything else in the circuit that really changes the ESR or Q of the naked part.
-
and lastly - Do they fail very often? It isn't convenient to pull it just to test like it is with through hole. It seems that I rarely get a meaningful measurement in circuit.
Something I never* do is pull parts simply to test them. Learn how the circuit works, take signal measurements when it's powered up, and draw your conclusions from that. The problem with pulling parts, testing, and then putting them back, is you can create new problems or can temporarily cover up a problem:
Sometimes the heat and mechanical stress from removal will temporarily "fix" the part, but it'll fail again later.
If you're not that skilled with desoldering, you start damaging the PCB or parts themselves.
I've had to post mortem PCBs where someone else pulled parts and then put them back incorrectly, wrong replacements, solder bridges, etc. Then it had multiple faults.
As a general rule, you should not put old parts back. It can get expensive replacing with all new parts and you'll never really know if you fixed it or what the actual problem was by just swapping things out en masse.
* never say never, but it's so rare that I do it, that it may as well be never. ;)
-
First - None seem to be marked to indicate what they should be.
Second - Is there any way to visually tell if the are ceramic, film, or other?
and lastly - Do they fail very often? It isn't convenient to pull it just to test like it is with through hole. It seems that I rarely get a meaningful measurement in circuit. Is there a good way to indicate if it could be a problem?
1. Very, very early SMD ceramic caps were marked. But they dropped this long ago. No idea why, but I wish they hadn’t!
2. If it’s a little tan or light brown block, it’s almost certainly a ceramic cap. SMD film caps (which you encounter extremely rarely in comparison) are little silvery blocks of plastic. SMD tantalum caps are in black or yellow epoxy, looking quite similar to some diode types. Electrolytics and polymer electrolytics are little aluminum cans.
3. Ceramic caps rarely fail unless exposed to significant overvoltage or if they crack due to excessive board flexing. Film caps are similarly robust. Tantalums are fairly durable too, but do not tolerate overvoltage. Electrolytics… well I sssume you already know how they are.
-
Since the subject of capacitor survivability has come up, I'd like to check two things I've never been fully sure of
Just how vulnerable are ceramic caps to over-voltage? Say you want an SMD ceramic towards the high end of capacitances, 22uF or 47uF say, you don't tend to find them rated >16V. So is 16V always good for a 12V circuit, what about a 12V circuit with some chance of brief higher spiking, or does one need something like a >=20V rated cap? And will MLCC's then die in a serious and permanent fashion from a brief over-voltage event, or a succession of them cumulatively, by say a few volts or a 10% or 20% above the rated voltage, or would they merely misbehave for the duration of the spike?
And is it specifically real flexing which mechanically damages MLCCs, or can any vibrations do it? If the board is well anchored with screws in such a way as to make sure that in the event of vibrations the whole board vibrates fairly evenly together, and you've got no locations with particularly heavy components which could make those bits of the board act have a higher inertia during vibrations and create a difference in the amount of acceleration different locations on the board would feel, does this damage SMD ceramic caps? Like, if you have a board with physically big, and high capacitance because this means the layers inside will be closer together and more vulnerable, SMD ceramic caps, and you attach it to equipment like a motor of some kind or a milling spindle or a lightweight highly vibrating drill... is this vibration going to wreck SMD caps, over time even if you arrange the board's mounting points so as to reduce obvious opportunities for flexing? Does any vibration cause some flexing, or is it only the kind of flexing where you can noticably detect parts of the board changing location RELATIVE TO other parts of the same board that you get mechanical risks to SMD MLCCs?
Thanks
P.S. jheatac: most cap failures are going to end up either with the cap shorting to become a low resistance (or near zero resistance) between its terminals, or by it ending up having very little capacitance at all. I've never heard of caps having "failures" that change there value by just a bit but leave them still acting as caps. In some circuits caps might be expected to change there value by some amount as they age or depending on the temperature of the environment they're operating in, you get some caps sold with tolerances of +80%-20% which experience such scenarios, circuits using these should usualyl have been designed to accomodate this. SMD can be easier to remove and resolder than through hole, heat both ends by getting an iron tip in contact with them, then go in with tweezers to move the part, no need to struggle to clean solder out from a hole before replacing, but cap failures will probably be findable just with in-circuit resistance testing with a sensitive milli-ohm meter. Needing actual measurement of capacitances is going to be far more likely to be needed if you're trying to reverse engineer a device back to a schematic than if you're trying to repair something.
-
Just how vulnerable are ceramic caps to over-voltage?
It's nearly impossible to kill MLCCs with slight overvoltages, voltage rating is more an arbitrary value by the manufacturer, one of the least useful numbers. Of course, using them beyond ratings causes strange looks and your explanations might be questioned, so that's a good reason not to deliberately overvolt them, but applying a 25V pulse on a 16V rated MLCC won't be a reason for failure.
Though, in reality, often you want to derate voltage significantly, for the voltage-biased capacitance loss effect. But there is no good rule-of-thumb percentage for that, you need to check the curves and operate at a voltage which still delivers you the capacitance you need.
But, especially with sizes larger than 0805, mechanical damage is the real risk. Any amount of board flex easily kills these capacitors, so does careless soldering with way too much heat, into a poorly designed footprints (i.e., no thermal reliefs, direct connection to planes).
-
It's nearly impossible to kill MLCCs with slight overvoltages,
It was also my opinion until last year.
We started to use MLCCs when it was tantal crisis due to the earthquake in Japan in 2002. I have read than that unlike for electrolytic you need not to make 2x voltage reserve. In ceramic capacitors datasheet I sow that they are tested at 2.5 times the rated voltage. So capacitor that withstands 25V in test is sold as 10V rated voltage capacitor.
We didn't forced contract manufacturer to use specific capacitors. When size, dielectric, capacitance and rated voltage were as specified he could use capacitor.
But last year we had a serie our devices broken what costed us a lot.
At 24V supply input there were 33V 600W transil then 1A diode and then 100n/50V capacitor. It happened it be shorted.
Then there were LM317 set to 20V at output (used as overvoltage protection for a transients that potentially can make 40V pulse at 33V transil). At these 20V where I didn't expected any overvoltage I used 10u/25V ceramic capacitors and for 15 years had no problem with it. But many of them failed last year.
In local distributor I found that for any capacitance/voltage combination (I am speaking not about nF but uF) when you select to get cheapest capacitors you get one manufacturer (don't want to use name, as I'm not 100% sure of my conclusions) and in datasheet I didn't found information about capacitors being tested at voltage being 2.5 times rated voltage. I have also confirmed with contract manufacturer that they used these capacitors.
I just suppose that some manager decided that he can win at market by giving up this 2.5x voltage test.
So now I agreed that contract manufacturer can use any capacitor except from that manufacturer. I also consider to increase by 2x all capacitors voltage specifications as who knows if because of competition other manufacturers will not give up these 2.5x test also. As during years ceramic capacitors got cheaper we can pay for higher voltage specification just for sure.
-
Maximum capacitance of MLCC's has been going up for a while. This can only mean very thin insulation layers. Apparently reliability is going down now and we're aproaching limits of what can be done with these things.
-
But last year we had a serie our devices broken what costed us a lot.
At 24V supply input there were 33V 600W transil then 1A diode and then 100n/50V capacitor. It happened it be shorted.
Then there were LM317 set to 20V at output (used as overvoltage protection for a transients that potentially can make 40V pulse at 33V transil). At these 20V where I didn't expected any overvoltage I used 10u/25V ceramic capacitors and for 15 years had no problem with it. But many of them failed last year.
And what sort of failure analysis led you to believe this had anything to do with voltage?
I mean, derating by 2x advice is completely novel; no one does that. On the other hand, mechanical failure is very well known, I have witnessed that, too, and there are many well known mitigations such as placing two capacitors in series, or using soft-terminated variants. What makes you think it was not mechanical failure or just poor quality of that manufacturer? I mean, mechanical failure can be even soldering related, doesn't need to be actual board flex.
-
Maximum capacitance of MLCC's has been going up for a while. This can only mean very thin insulation layers. Apparently reliability is going down now and we're aproaching limits of what can be done with these things.
The problem is that even very small cracking causes the layers to touch. Thin ceramic layer, when intact, can take large voltages.
-
And what sort of failure analysis led you to believe this had anything to do with voltage?
I mean, derating by 2x advice is completely novel; no one does that. On the other hand, mechanical failure is very well known,
I just didn't know anything about mechanical failures so (now I start thinking that wrongly) didn't even taken them into account.
Only one information can suggest voltage. One of devices powered by 12V worked (in such case at these capacitors there is about 9V) but when powered by 24V (in that case there is about 20V at these capacitors) failed at once. But this is no 100% credible information.
MLCC manufacturer of previously used capacitors makes 1206 10u/6V3...10u/25V all as thick but used this time (partially because of accessibility problems) 10u/25V were 2 times thinner.
Do you know of any correlation between being polarized close to rated voltage and greater susceptibility to mechanical failure?
-
Do you know of any correlation between being polarized close to rated voltage and greater susceptibility to mechanical failure?
It has been said that higher voltage increases the risk of small mechanical failure causing shorting, but since the mechanism is unclear to me I would be careful from drawing conclusions from that. In any case, even in such cases, the root issue is the mechanical failure. Voltage close (or slightly exceeding) rated voltage, AFAIK, never causes a failure, and mechanical damage often/possibly causes a failure, where higher voltage increases the risk of dramatic consequences.
The most dramatic MLCC shorting I saw was Murata X7R 1210 50V at mere 10V applied to it. The capacitor became so hot it glowed bright yellow; by luck I happened to be present when it happened, because it happened at power-up on equipment which has sat some weeks unused. The crack was there for months!
Often one can see tips like "use soft termination when low-impedance voltage source e.g. battery pack is connected to the capacitor directly", and to me this seems odd. It is like trying to avoid conditions where failure is dramatic. But in my opinion mechanical failure should be avoided in all cases, even if it's just one capacitor acting as a 100ohm resistor, across a nice current-limited and fused supply, causing increased power consumption and possibly lack of bypassing at that site. This is why I'm always very careful with board flex and footprint design with anything larger than 0805.