They're used for coupling in a growing number of applications as it's becoming understood that they are actually ideal in many applications and SMD in particular. The are not just cheap, they are small and 100% reliable when you realize that the strange parameter shift is not damage with age but rather a characteristic change. It's more like dealing with a large tolerance that isn't fixed. Dealing with this is a matter of good engineering.
The distortion issue is also a matter of design, engineering, and making the right choice for the given circuit. C0G/NP0 is readily available up to about 470n before the price starts going out of site and that covers quite a lot of ground with filters. Alarm over the X7R and even X5R is unwarranted. At this point I'm not seeing any mysteries here.
The distortion issue is also a matter of design, engineering, and making the right choice for the given circuit. C0G/NP0 is readily available up to about 470n before the price starts going out of site and that covers quite a lot of ground with filters. Alarm over the X7R and even X5R is unwarranted. At this point I'm not seeing any mysteries here.
I read somewhere (but I don't remember where) that the specified capacitance usually applies 1000 hours after the last time the capacitor was heated above the Curie temperature. The test rejection limits are increased a bit to get the right capacitance after 1000 hours (at 0 V DC, 25 degrees Celsius and the specified AC test voltage and frequency).
If you have to use class-2 capacitors, you could consider trying something like this.
The AC voltage across the upper capacitor drops with 40 dB/decade rather than 20 dB/decade, which may help to keep the signal small without needing very large time constants. A disadvantage is that you get a subsonic response bump.
If you have to use class-2 capacitors, you could consider trying something like this.
The AC voltage across the upper capacitor drops with 40 dB/decade rather than 20 dB/decade, which may help to keep the signal small without needing very large time constants. A disadvantage is that you get a subsonic response bump.
What I was trying to describe was that it all must be in context and with details. An unqualified distortion residual means nothing. It's nowhere near enough information to be scientific. This subject is not at all that simple but it is certainly understandable.
I also tried to point out that if you understand the specific circuit, the bounds of the component's nature, and the circuit works, there is no problem at all. In other circuits it may be a disaster. I'm trying to describe what good science and good engineering looks like. I'm not seeing any here at the moment and I only want to help.
Components should not be discarded based on superstition. The EDN article does make most of this clear, but it does fall short regarding the solution. It doesn't describe what a good engineer does to deal with putting such components to use while understanding their limitations and also understanding that sometimes they are not limitations at all.
And indeed this applies to the electrolytic scenario as well. It would be perfectly fine to use an electrolytic, but I am simply trying to make the point that none of it is so on/off. It all has to do with the specific circuit/application and design goals. You may have all the space you need, no budget constraints and you could use aluminum foil and window panes to construct a giant capacitor for the trill of it. When it's a hobby, crazy things are allowed.
Another scenario is if you are trying to engineer something that has requirements and constraints defined by whom you work for. It's all about context and competence. And where the hobby scenario is concerned, fun.
Everyone has different priorities regarding their hobby and I just happen to enjoy the pursuit of being a better engineer while having fun with old stereos and guitar stuff.
I also tried to point out that if you understand the specific circuit, the bounds of the component's nature, and the circuit works, there is no problem at all. In other circuits it may be a disaster. I'm trying to describe what good science and good engineering looks like. I'm not seeing any here at the moment and I only want to help.
Components should not be discarded based on superstition. The EDN article does make most of this clear, but it does fall short regarding the solution. It doesn't describe what a good engineer does to deal with putting such components to use while understanding their limitations and also understanding that sometimes they are not limitations at all.
And indeed this applies to the electrolytic scenario as well. It would be perfectly fine to use an electrolytic, but I am simply trying to make the point that none of it is so on/off. It all has to do with the specific circuit/application and design goals. You may have all the space you need, no budget constraints and you could use aluminum foil and window panes to construct a giant capacitor for the trill of it. When it's a hobby, crazy things are allowed.
Another scenario is if you are trying to engineer something that has requirements and constraints defined by whom you work for. It's all about context and competence. And where the hobby scenario is concerned, fun.
Everyone has different priorities regarding their hobby and I just happen to enjoy the pursuit of being a better engineer while having fun with old stereos and guitar stuff.
I think the best coupling capacitor is no capacitor at all. Use a DC servo instead [ https://imrad.com.ua/userdata/modul.../banner/cy-xxi-v-7-karta-naprug-jpg6TT9F0.jpg ]
Well you can read the residual amplitude as being about 45dB below the signal, that's compared to perhaps 100dB for electrolytic in same situation, i.e. very gross. You don't need to do detailed measurements for very gross! Anyone with a distortion meter will see similar results, but the must important detail is the nature of the distortion is not just 2nd order or 3rd order, but includes something more like cross-over or hysteresis distortion.
I try to avoid electrolytics because of their reliability problems, but ceramic capacitors aren't 100% reliable either. SMD MLCCs in particular can crack due to thermal expansion relative to the PCB. In audio applications this might happen if one is near a heatsink or other hot component.
Do you force a sinusoidal current through the capacitor and measure the voltage across it?
Well, they can be. The problem described here would depend on the failure to recognize these thermal issues. Good thermal design inculdes anticipating such stresses and assuring that they have no such consequences.
One over the last decade. It was a capacitor from the 1970's that I had reused in a DIY circuit in 1987. It failed a couple of years ago.
Failed power supply capacitors are more common, especially with switched-mode power supplies.
Last edited:
I have found more than I expected to over the years while doing repairs. All of them were at least 30 years old. Not long ago I repaired a Japanese 7 band EQ that had horrible distortion in one channel and bad distortion in the other. Two 2.2 µF electrolytic coupling capacitors were the sole problem. After replacing just the two offending parts it worked nicely and then I replaced the remain electrolytics as a routine recap.
It certainly happens. They do dry up and become usesless even when the duty is light but.... I have also been very surprised when measuring the leakage on some 50+ year old electrolytics to find them like new. With electrolylics there's a finite operating life and both time and thermal cycles play into it. Failing seals is another matter.
What I have not seen is a 5 year old electrolytic leaking badly or worse unless the equipment went up in smoke or at least was severly damaged.