How does an X7R capacitor fail when exposed to a long term over-voltage?
I have a selection of leaded (through hole) X7R capacitors in various values from 10nF to 220nF. All are rated at either 25V or 50V
Yesterday I put a 220nF 50V X7R on test at 63.2Vdc with a 1k0 current monitoring resistor in the feed.
Today the current monitoring resistor is still showing a Vdrop of 0.0mVdc
The cap bead is still cold.
I know from datasheets that these ceramic capacitors lose capacitance as the voltage is increased. But I don't have any 63V data for 50V capacitors.
How do X7R cermic capacitors fail?
Do they explode, or go open circuit, or go short circuit, or capacitance simply goes down to near zero nF?
I ask because I want to increase the supply voltage for a +-50Vdc amplifier to +-59Vdc. It is currently fitted with these 50V caps as supply rail decoupling.
I have a selection of leaded (through hole) X7R capacitors in various values from 10nF to 220nF. All are rated at either 25V or 50V
Yesterday I put a 220nF 50V X7R on test at 63.2Vdc with a 1k0 current monitoring resistor in the feed.
Today the current monitoring resistor is still showing a Vdrop of 0.0mVdc
The cap bead is still cold.
I know from datasheets that these ceramic capacitors lose capacitance as the voltage is increased. But I don't have any 63V data for 50V capacitors.
How do X7R cermic capacitors fail?
Do they explode, or go open circuit, or go short circuit, or capacitance simply goes down to near zero nF?
I ask because I want to increase the supply voltage for a +-50Vdc amplifier to +-59Vdc. It is currently fitted with these 50V caps as supply rail decoupling.
Like most of solid state devices (understood in the broadest definition, ie not only semiconductors) they fail short, or at least low impedance.
As far as I know, there are two main mechanisms of failure: either a small initial defect present in the bulk of the material grows until it reaches catastrophic proportions, or a slow process of electrode metal migration/electrolysis happens, eventually causing a short somewhere.
You can eliminate the likelihood of the first type of defect by screening the capacitors at a much higher voltage than the one you are going to subject it to: 3 times Vworking for example, in your case ~180V.
This will eliminate short term failures, because the breakdown susceptibility vs applied voltage is exponential, but is no guarantee against the slower mechanisms (which can happen even below the official Vnom).
That said, if you use capacitors from a reputable manufacturer far from its temperature limits, 60V for a 50V cap seems safe enough, especially if you HV-screen them first.
Note that 63V X7R capacitors are easily available....
As far as I know, there are two main mechanisms of failure: either a small initial defect present in the bulk of the material grows until it reaches catastrophic proportions, or a slow process of electrode metal migration/electrolysis happens, eventually causing a short somewhere.
You can eliminate the likelihood of the first type of defect by screening the capacitors at a much higher voltage than the one you are going to subject it to: 3 times Vworking for example, in your case ~180V.
This will eliminate short term failures, because the breakdown susceptibility vs applied voltage is exponential, but is no guarantee against the slower mechanisms (which can happen even below the official Vnom).
That said, if you use capacitors from a reputable manufacturer far from its temperature limits, 60V for a 50V cap seems safe enough, especially if you HV-screen them first.
Note that 63V X7R capacitors are easily available....
Thanks.
I'll jack up my test voltage to >>100Vdc and see if I get a catastrophic failure.
So far >24Hrs @ 63.2Vdc and still near zero current.
I don't order one component at a time.
I'll have to wait until I need lots of other components and then I could order eight 100V X7R (or more likely 50off)
I'll jack up my test voltage to >>100Vdc and see if I get a catastrophic failure.
So far >24Hrs @ 63.2Vdc and still near zero current.
I don't order one component at a time.
I'll have to wait until I need lots of other components and then I could order eight 100V X7R (or more likely 50off)
The cap might handle 500 to 1500 volts without failure. A long time ago in the development lab, if we were hi-potting components and had a little time to waste at the end of the day, we would hi-pot different bits and pieces to see what it took to break them. With many low voltage caps it took 1500V to fail them. Same with low voltage multi-conductor cables (similar to Cat3). But I don't remember any details.
Lead stress causes cracking and later failure, due to moisture leakage.
This seems much more common than voltage related failures, in ceramic caps.
Last edited:
The "voltage squish" of the X7R dielectric is a factor to consider with high operating voltages, but thankfully, a number of vendors have simulator tools on their sites to allow you to bias a cap up to some random voltage and then generate a SPICE model at that operating point, as well as datasheet curves for the component at that operating voltage. So, you can see how they will function at a specific bias point pretty easily. Murata has a nice web tool for this that they call 'simsurfing', and IIRC Kemet has a downloadable tool as well.
At higher voltages, this effect may cause issues with your circuit, and different size caps will "squish" differently, so it's worthwhile to go through the various models and see how they fare at your operating voltage.
MLCCs in general tend to not fail from over-voltage much at all. C0G MLCCs were tested by NASA and the worst failure was to arc over at something like 500% or 1000% of their faceplate voltage rating. X7R might be different, and the "voltage squish" is going to be a more practical problem, but still, arcing over and failing short is not a common failure mode with modern, quality MLCCs. As was stated above, cracks due to mechanical flexure that induce a shorting failure are more common. You can get 'flex-term' caps that have slightly flexible terminations to minimize this. As far as i can tell, the flex-term caps don't seem to measure any better or worse than standard models, so if you need high reliability, getting an automotive rated flex term cap might be wise.
I'm referring mainly to SMD MLCCs, but a lot of this applies to leaded MLCCs, since they're the same basic chip with leads added, which is then dipped in encapsulant. I'd recommend using SMDs, but if you're working with a through-hole layout, then buying the chip with pre-made leads is the best answer.
Another random data point: I tested a bunch of 1uF 100V X7R 3216 chips recently, and almost all of them behaved really nicely at 17V bias. I was testing their impedance and their overall distortion when used as output caps with a linear regulator chip, driving the regulator output with 3mA of 2kHz through a Kelvin resistor probe. Despite the X7R voltage coefficient 'sloppiness', the distortion was entirely due to the regulator and its gain bandwidth. They also didn't seem to behave much differently than a 50V part, but still, 100V parts are easily available and are not much bigger than lower voltage parts.
They work well, there's a ton of 100V parts, and they will all squish up differently with voltage, so that's the most important thing to check IMHO.
At higher voltages, this effect may cause issues with your circuit, and different size caps will "squish" differently, so it's worthwhile to go through the various models and see how they fare at your operating voltage.
MLCCs in general tend to not fail from over-voltage much at all. C0G MLCCs were tested by NASA and the worst failure was to arc over at something like 500% or 1000% of their faceplate voltage rating. X7R might be different, and the "voltage squish" is going to be a more practical problem, but still, arcing over and failing short is not a common failure mode with modern, quality MLCCs. As was stated above, cracks due to mechanical flexure that induce a shorting failure are more common. You can get 'flex-term' caps that have slightly flexible terminations to minimize this. As far as i can tell, the flex-term caps don't seem to measure any better or worse than standard models, so if you need high reliability, getting an automotive rated flex term cap might be wise.
I'm referring mainly to SMD MLCCs, but a lot of this applies to leaded MLCCs, since they're the same basic chip with leads added, which is then dipped in encapsulant. I'd recommend using SMDs, but if you're working with a through-hole layout, then buying the chip with pre-made leads is the best answer.
Another random data point: I tested a bunch of 1uF 100V X7R 3216 chips recently, and almost all of them behaved really nicely at 17V bias. I was testing their impedance and their overall distortion when used as output caps with a linear regulator chip, driving the regulator output with 3mA of 2kHz through a Kelvin resistor probe. Despite the X7R voltage coefficient 'sloppiness', the distortion was entirely due to the regulator and its gain bandwidth. They also didn't seem to behave much differently than a 50V part, but still, 100V parts are easily available and are not much bigger than lower voltage parts.
They work well, there's a ton of 100V parts, and they will all squish up differently with voltage, so that's the most important thing to check IMHO.
I have seen high value multilayer ceraimic X7R caps fail at well under 2* rating voltage. Short circuit failure.
With a drive impedance of 1 kΩ, you'll probably not get enough current flow to cause a destructive arc.
As others have pointed out, the voltage coefficient of X7R is pretty high. It's common that you only get 20-25 % of the rated capacitance at the full rated voltage.
Tom
- Status
- Not open for further replies.