I'm setting up a test of what I think is the worst case ceramic I uses which is an SMD 22µF X5R 10V 20% in a 0603 package.
The first thing that I did was measure the capacitance after unpacking the new component. The meter showed 16.9 µF. That sounds like about 23% low to me and would suggest that it is not thermally formed. Then I soldered it for testing and measured it again. Now it reads 20.8 µF. I am guessing, but it would appear that the manufacturer, Samsung in this case, assumes that the reflow process will form it. It's possible that it could come up even more if I heated it that much, but I simply used a soldering iron at 580F very briefly.
Now I need to set up the test which will approximate a real life AC coupling circuit that I would design with this part.
The first thing that I did was measure the capacitance after unpacking the new component. The meter showed 16.9 µF. That sounds like about 23% low to me and would suggest that it is not thermally formed. Then I soldered it for testing and measured it again. Now it reads 20.8 µF. I am guessing, but it would appear that the manufacturer, Samsung in this case, assumes that the reflow process will form it. It's possible that it could come up even more if I heated it that much, but I simply used a soldering iron at 580F very briefly.
Now I need to set up the test which will approximate a real life AC coupling circuit that I would design with this part.
I think that it's been made clear enough here that all of these capacitor types are perfectly acceptable depending on the situation. If 30 years is a concern maybe use the best of whatever and go with it. It's just about good sensible engineering and what the requirements are. If something is intended to be cheap and have a short life, use whatever gets the job done at the right price.
And if you have never found a faulty coupling capacitor you simply have much less experience than some of us. They do exist. Perhaps not to the extent that power supply filters do, but indeed they exist.
And if you have never found a faulty coupling capacitor you simply have much less experience than some of us. They do exist. Perhaps not to the extent that power supply filters do, but indeed they exist.
Exactly, like class 2 ceramics only being good for decouling purposes with proper designing in mind. Sorry, not gonna let you push your bad info any longer
For situations where I need small form factor, I've become a huge fan of acrylic caps like ECPU1C105MA5 (voltage limit permitting). If I need more than 1uF, I simply stack/parallel them. They aren't as sensitive to hand soldering as some would have you believe, or maybe I'm just really good at it. 😁
Hows that? 10uf 16V rubycon mu costs the same as 10 1uf 16v ecpu (mouser). Minus the time you need to stack them, costs even less Agreed some lower values may be a tiny bit more expensive comparing to stacking, but still acounting for time spent, i'd say it's pretty much equal, not to mention more clean.
Agreed some lower values may be a tiny bit more expensive comparing to stacking, but still acounting for time spent, i'd say it's pretty much equal, not to mention more clean.
1pc 10uF 16V is $8. 10pc ECPU1C105MA5 cost me less than $5 (I bought hundreds years ago). Plus, buying ECPU1C105MA5 in bulk and having them on hand allows me to create any value I need. I'm not saying it's ideal for every situation/project/person, but it works for me. I can see maybe needing a single larger value at some point. I needed 22uF recently and just used MUSE because I had them.
Whole 'nother story if you have a stash with early cheaper prices, but current difference is 0.30€. Which to me is definitely not worth the hassle of stacking them 😅 But yeah, in gemeral what works is fine. I personally like to calculate my time in as well.
PS. Also have a stash of 0.1uf ecpu, full little compartment 😅 Amazing cheap opamp decouple.
PS. Also have a stash of 0.1uf ecpu, full little compartment 😅 Amazing cheap opamp decouple.
Well, partially what prompted the thread in the first place, McIntosh 20 year Rubycon - both fail and huge drift across so many of them, when it wouldn't be necessarily expected. Let's also be honest, electrolytic constitute a tinny sound in comparison to anything else.
I'm still investigating this cermaic capacitor thing when I can find time because it's quite the interesting and irresistable rabbit hole.
I decided to dig through my cermaics to find the worst one. The winner is a 22 µF, 10V, X5R, 0603. Almost worst case and a great one to demonstrate the extremes. I was happy to see that I was thinking correctly that the distortion residual is not at all a spray of harmonics but almost entirely clean 2nd and 3rd harmonics in some proportion that is not always predictable but always looks like that of a tube or transistor. This happens because it's a voltage coefficient modulating the capacitance in a very nice way that ends up logarithmic ind the end. If there are any strange higher harmonics, it happens at very low distortion levels that are beyond the possibility of being audible. The higher order harmonics are due to other quirks of the capacitor and I have an example here where the voltage coefficent distortion so dominates that others are lost down near the noise floor of the system. I'm only going down to 0.0004% here and beyond that I would need to build equimpment.
The next thing I noted is that this low order distortion can be very large and very proportional to frequency. The capacitor I'm using to force the extreme reached 6%! This was at 10 Hz. At 1 kHz under the same conditions it measured 0.174 %. Still bad for Hi Fi but remember that this is a deliberate worst case and it's REALLY worst. The loading for these tests are 1K Ohm which is pretty heavy for a small signal circuit like this and some op-amps would be challenged. I'm using a 5532 inverting amplifier for drive.
All of this said, the phenomenon is very predictable and consistent and this could work for coupling by using enough capacitance to bring the voltage drop across it down and high enough load impedance to not instigate the distortion. This however would be ridiculous in most applications because it's horrible and would require huge amounts of capacitance and depend on very light loading. It's purely acedemic and an argument for using electrolytics. This isn't true for all cases, but this one is a perfect example where it's obvious.
One of the things I find interesting about this is that the voltage rating of the capacitor is so critical. Notice I used a 10 Volt capacitor and I'm driving it with a peak amplitude of nearly 10 volts to bring my signal to 100% of that. Not a normal thing in a circuit. I'm doing everything I can to get a worst case example for reference to show that blanket judgements of parts based on their construction doesn't work and is not good engineering.
If the voltage of this part were instead 100 Volts, the distortion for many parts and manufacturers would drop by nearly 80%. This seems quite crazy but that's how X5R behaves and why of course why it's out of the question in many cases. It's just as interesting that X7R is drastically better and brings some of these parts into the range of being usefull without significant distortion. That will be my next step and I will at some point bring some graphics into this.
The worst case example makes specacularly large distortion at very low frequencies but very low order and least audible distortion. It's in no way usable for Hi Fi signal paths but maybe X7R is. I hope to demonstrate that. I'm also hoping to still "know" that this works for coupling. If I'm wrong, that will be reported as well
I decided to dig through my cermaics to find the worst one. The winner is a 22 µF, 10V, X5R, 0603. Almost worst case and a great one to demonstrate the extremes. I was happy to see that I was thinking correctly that the distortion residual is not at all a spray of harmonics but almost entirely clean 2nd and 3rd harmonics in some proportion that is not always predictable but always looks like that of a tube or transistor. This happens because it's a voltage coefficient modulating the capacitance in a very nice way that ends up logarithmic ind the end. If there are any strange higher harmonics, it happens at very low distortion levels that are beyond the possibility of being audible. The higher order harmonics are due to other quirks of the capacitor and I have an example here where the voltage coefficent distortion so dominates that others are lost down near the noise floor of the system. I'm only going down to 0.0004% here and beyond that I would need to build equimpment.
The next thing I noted is that this low order distortion can be very large and very proportional to frequency. The capacitor I'm using to force the extreme reached 6%! This was at 10 Hz. At 1 kHz under the same conditions it measured 0.174 %. Still bad for Hi Fi but remember that this is a deliberate worst case and it's REALLY worst. The loading for these tests are 1K Ohm which is pretty heavy for a small signal circuit like this and some op-amps would be challenged. I'm using a 5532 inverting amplifier for drive.
All of this said, the phenomenon is very predictable and consistent and this could work for coupling by using enough capacitance to bring the voltage drop across it down and high enough load impedance to not instigate the distortion. This however would be ridiculous in most applications because it's horrible and would require huge amounts of capacitance and depend on very light loading. It's purely acedemic and an argument for using electrolytics. This isn't true for all cases, but this one is a perfect example where it's obvious.
One of the things I find interesting about this is that the voltage rating of the capacitor is so critical. Notice I used a 10 Volt capacitor and I'm driving it with a peak amplitude of nearly 10 volts to bring my signal to 100% of that. Not a normal thing in a circuit. I'm doing everything I can to get a worst case example for reference to show that blanket judgements of parts based on their construction doesn't work and is not good engineering.
If the voltage of this part were instead 100 Volts, the distortion for many parts and manufacturers would drop by nearly 80%. This seems quite crazy but that's how X5R behaves and why of course why it's out of the question in many cases. It's just as interesting that X7R is drastically better and brings some of these parts into the range of being usefull without significant distortion. That will be my next step and I will at some point bring some graphics into this.
The worst case example makes specacularly large distortion at very low frequencies but very low order and least audible distortion. It's in no way usable for Hi Fi signal paths but maybe X7R is. I hope to demonstrate that. I'm also hoping to still "know" that this works for coupling. If I'm wrong, that will be reported as well
This tool is fun to play around with...
https://ds.murata.co.jp/simsurfing/...is=false&md5=3e1cb76323e80d9c0bc69fc04432f6c8
https://ds.murata.co.jp/simsurfing/...is=false&md5=3e1cb76323e80d9c0bc69fc04432f6c8
Wow! That's crazy and cool!
I'm still looking into the class 2 issue and it's far more interesting than I expected. I'm not seeing applications for them in an audio path outside of the guitar world so far. I have to find odd moments in my schedule to do more testing.
The interesting thing that I'm seeing is the lack of high order harmonics caused by the capacitor. It's all low order and huge! I started with the worst case capacitors under the worst conditions and that's likely why it's almost all 2nd or 3rd harmonic as a typical voltage coefficient based distortion is likely to do. Also at lower levels of distortion it gets more difficult to sort out other distortions and the residual can be very misleading. I am still not seeing any spray of upper harmonics even at very low distortion levels and I may need to find examples of parts with more problems than the ones I have. Upper harmonics would have to be due to more than X5R Voltage related capacitance changes resulting in logarithmic impedance changes causing low order distortion.
I also tested a guitar bypass circuit that uses my worst case 22 µF X5R for output coupling and incredibly the circuit overall stays under 0.001% THD all the way down to 20 Hz with a 2.5k Ohm load. This circuit doesn't go much beyond+4dbu and that has quite a bit to do with it. I'm certainly not using any class 2 for Hi Fi audio path circuits, but there is no doubt use for them in guitar circuits that need to be small and cheap. My investigations continue...
I'm still looking into the class 2 issue and it's far more interesting than I expected. I'm not seeing applications for them in an audio path outside of the guitar world so far. I have to find odd moments in my schedule to do more testing.
The interesting thing that I'm seeing is the lack of high order harmonics caused by the capacitor. It's all low order and huge! I started with the worst case capacitors under the worst conditions and that's likely why it's almost all 2nd or 3rd harmonic as a typical voltage coefficient based distortion is likely to do. Also at lower levels of distortion it gets more difficult to sort out other distortions and the residual can be very misleading. I am still not seeing any spray of upper harmonics even at very low distortion levels and I may need to find examples of parts with more problems than the ones I have. Upper harmonics would have to be due to more than X5R Voltage related capacitance changes resulting in logarithmic impedance changes causing low order distortion.
I also tested a guitar bypass circuit that uses my worst case 22 µF X5R for output coupling and incredibly the circuit overall stays under 0.001% THD all the way down to 20 Hz with a 2.5k Ohm load. This circuit doesn't go much beyond+4dbu and that has quite a bit to do with it. I'm certainly not using any class 2 for Hi Fi audio path circuits, but there is no doubt use for them in guitar circuits that need to be small and cheap. My investigations continue...
Last edited:
Some of these characteristics are far worse than I would have imagined. And the manufacturer daring to publish this is a good way to put it. It seems that the forthcomming nature of the data sheet is a thing of the past. I started noticing this quite a long time ago when Vbb started disappearing from transistor data sheets.
ha! Nice... lolObviously this product is promoted by the global caps lobby