Würth Elektronik’s new Application Note ANP125 publishes the results of a study “The acoustic effects of harmonic distortion of aluminum electrolytic capacitors” on harmonic distortion of electrolytic capacitors. The result: Capacitors don’t cause any appreciable signal distortion.
The discussion is ongoing in the audio technology world about what circuit elements affect the sound quality of amplifiers. The app note by Würth Elektronik provides empirical evidence to add to this discussion and answers questions that most audio engineers have.
Application Note ANP125 is the result of international research collaboration between R&D teams at production sites in Asia and the Würth Elektronik Competence Center in Berlin. The text begins with an introduction to human hearing and psychoacoustics and goes on to examine harmonic distortions in capacitors. Furthermore, results from model calculations are presented in order to check the plausibility of the measured results. The measurements show no appreciable distortion of signals caused by capacitors.
Material variations also tested
Dr. René Kalbitz, Product Manager in the Capacitors & Resistors Division at Würth Elektronik eiSos and author of the study, explains:
- The investigations indicate that material variations have a negligible influence on distortions, and these are below the hearing threshold. Electrolytic capacitors do not add any appreciable harmonics to the fundamental frequencies in signal transmission, so, to a good approximation, they can be considered as linear components. It is likely that other voltage-independent capacitor types and passive components, as a rule, generate similarly low distortion amplitudes compared to the audibility threshold. Consequently, the choice of non-linear components such as operational amplifiers and diodes has a greater distortion impact on the audio quality of the amplifier, i.e., the overall distortion characteristics, than the choice of electrolytic capacitor.
ANP125
The discussion is ongoing in the audio technology world about what circuit elements affect the sound quality of amplifiers. The app note by Würth Elektronik provides empirical evidence to add to this discussion and answers questions that most audio engineers have.
Application Note ANP125 is the result of international research collaboration between R&D teams at production sites in Asia and the Würth Elektronik Competence Center in Berlin. The text begins with an introduction to human hearing and psychoacoustics and goes on to examine harmonic distortions in capacitors. Furthermore, results from model calculations are presented in order to check the plausibility of the measured results. The measurements show no appreciable distortion of signals caused by capacitors.
Material variations also tested
Dr. René Kalbitz, Product Manager in the Capacitors & Resistors Division at Würth Elektronik eiSos and author of the study, explains:
- The investigations indicate that material variations have a negligible influence on distortions, and these are below the hearing threshold. Electrolytic capacitors do not add any appreciable harmonics to the fundamental frequencies in signal transmission, so, to a good approximation, they can be considered as linear components. It is likely that other voltage-independent capacitor types and passive components, as a rule, generate similarly low distortion amplitudes compared to the audibility threshold. Consequently, the choice of non-linear components such as operational amplifiers and diodes has a greater distortion impact on the audio quality of the amplifier, i.e., the overall distortion characteristics, than the choice of electrolytic capacitor.
ANP125
That application note has been discussed here previously. It was found to be wanting, to put it briefly.
Also IIRC, the application note may be somewhat in conflict with something Doug Self previously published regarding electrolytic cap distortion.
Moreover, to discuss cap HD in isolation would seem to tacitly imply that HD is the only thing about cap sound that could potentially be audibly disagreeable. Not necessarily so IME.
Also IIRC, the application note may be somewhat in conflict with something Doug Self previously published regarding electrolytic cap distortion.
Moreover, to discuss cap HD in isolation would seem to tacitly imply that HD is the only thing about cap sound that could potentially be audibly disagreeable. Not necessarily so IME.
Last edited:
470uF is a whopping big value for coupling or servo - didn't Sam and Scott use 100nF for their bridge approach to measurement? I am a bit chagrined that their Linear Audio article wasn't mentioned in the footnotes.
With respect to the use of electrolytic caps, I believe that Nelson wrote "Just get over it" or words to that effect.
With respect to the use of electrolytic caps, I believe that Nelson wrote "Just get over it" or words to that effect.
I haven't been able to measure any detrimental effect of modern electrolytic caps either. That said, I tend to use the Nichicon UES-series. The bright green ones. They're quite good.
Same can be said about ceramic capacitors. A physically large C0G/NP0 capacitor can measure better than a fancy polypropylene film cap.
Tom
Same can be said about ceramic capacitors. A physically large C0G/NP0 capacitor can measure better than a fancy polypropylene film cap.
Tom
I had not seen the Wurth white paper and thank you for sharing.
I have always found Steve Bench's analysis on capacitor linearity and distortion to be accurate and fair enough for audio purposes. It can be found at the link below.
https://diyaudioprojects.com/mirror/members.aol.com/sbench102/caps.html
In general he confirmed that some capacitors like paper in oil and metal film have more linear transfer characteristics than electrolytic, tantalum, or mica.
I have always found Steve Bench's analysis on capacitor linearity and distortion to be accurate and fair enough for audio purposes. It can be found at the link below.
https://diyaudioprojects.com/mirror/members.aol.com/sbench102/caps.html
In general he confirmed that some capacitors like paper in oil and metal film have more linear transfer characteristics than electrolytic, tantalum, or mica.
What were you trying to measure, and under what test conditions? Only HD? DA (linear distortions can be audible and unwanted in some cases)? Microphonics? Anything else?
IIRC, Doug Self found that electrolytic distortion depended on the AC voltage across the cap. If less than 60 some odd mV then no measurable HD. That could be one reason for using a big 470uf cap.
Also, if you ever listen to bunch of different electrolytic coupling caps from different manufacturers in different uf values, they sure don't all sound the same if used for DAC Vref filtering. They also tend to change over a period of time, sometimes not settling fully for a few weeks. There are ways to know that such effects are not imaginary too. Build two DUTs and age the caps different lengths of time before listening. If you don't hear any difference, please check your ground loops and or get rid of SMPS, Class-D amps, ferrites in the wrong places, etc., until differences finally become audible. Your system will sound better afterwards for the cleaning up efforts too. BTW, in the foregoing context when I say "you" I don't mean any specific person. Just suggestions for those readers who don't hear differences.
EDIT: BTW, I agree the little green bipolar electrolytic caps are pretty good.
Last edited:
I look at many performance parameters, including: THD, THD+N, multi-tone IMD, transient response, etc. See the Performance Graphs tab on the product page of any of my power amps if you're interested.
I'm not interested in another subjective vs objective thread. If you believe so strongly that you can tell a difference then set up a double-blind experiment. If your participants can pick out a particular cap 19 times out of 20 (or 39 times out of 40 if we're talking a two-tailed test with a p of 0.05) I'll believe there is a difference.
I'm aware of Doug Self's writings. I've been following them since his articles in EW/WW back in the 1990s. I'm also aware of dielectric absorption, voltage coefficient, etc. As for the microphonics: That's a solved problem. You just have to choose the right dielectric.
Tom
I'm not interested in another subjective vs objective thread. If you believe so strongly that you can tell a difference then set up a double-blind experiment. If your participants can pick out a particular cap 19 times out of 20 (or 39 times out of 40 if we're talking a two-tailed test with a p of 0.05) I'll believe there is a difference.
I'm aware of Doug Self's writings. I've been following them since his articles in EW/WW back in the 1990s. I'm also aware of dielectric absorption, voltage coefficient, etc. As for the microphonics: That's a solved problem. You just have to choose the right dielectric.
Tom
Tom,
I don't want to get into a public argument with you, because I think you are serious guy trying to do good work. However, what you said is, "I haven't been able to measure any detrimental effect of modern electrolytic caps either." At the same time you say you are aware of DA, but not that you measure for it. Thus I don't think you have been trying to measure as much as you seem to suggest in terms of "detrimental effects". Its as though you feel there is no threshold of audibility for group delay.
Given that there is in fact a threshold of audibility for group delay, and that DA can be modeled as an RC ladder network, then it seems that in some cases DA could have some detrimental audible effect (which IME it sometimes does). Thus maybe you shouldn't make a general claim about what you can measure if all you are measuring is for is one thing, HD. Surely you can measure for more than that. That's all.
Mark
I don't want to get into a public argument with you, because I think you are serious guy trying to do good work. However, what you said is, "I haven't been able to measure any detrimental effect of modern electrolytic caps either." At the same time you say you are aware of DA, but not that you measure for it. Thus I don't think you have been trying to measure as much as you seem to suggest in terms of "detrimental effects". Its as though you feel there is no threshold of audibility for group delay.
Given that there is in fact a threshold of audibility for group delay, and that DA can be modeled as an RC ladder network, then it seems that in some cases DA could have some detrimental audible effect (which IME it sometimes does). Thus maybe you shouldn't make a general claim about what you can measure if all you are measuring is for is one thing, HD. Surely you can measure for more than that. That's all.
Mark
Last edited:
I find the Würth paper a bit biased.
First is the odd choice of 470uF. While such large values are a good choice in the feedback circuit, DC blocking caps at the input of audio circuits are rather in the range of 47uF. Therefore it would have been worthwhile to also investigate a 47uF capacitor.
Since the capacitance investigated is that large, the conclusion is not really representative.
While distortion is well measurable, the authors conclude that it is not audible. This is why they explained what is audible and what is not at the beginning of the paper. I even find there was significant distortion measured (more than I expected), or maybe I'm just misinterpreting figure 8?
Somehow it appears to me as if the study was designed to show desired results.
As a side note, figure 3 is not a good model of an electrolytic capacitor. A much more accurate model is suggested by Cornell Dubilier (Presented at IEEE Industry Applications Society Conference, Oct 17, 2002), which includes dielectric loss.
There have been several independent measurements before (like the very detailed and excellent work of Bateman) I tend to trust more.
Another short, but good investigation by @PMA : https://pmacura.cz/capdist.htm
Conclusion was that electrolytic capacitors with low capacitance value may show significant distortion if the voltage across the capacitor is high.
The easiest way to reduce the voltage across the capacitor is to increase the capacitance. This is what the Würth engineers did: Investigate a non-issue and conclude there is no issue.
Sorry for being a bit harsh, but for me as a non-scientist, this does not meet my scientific standards.
First is the odd choice of 470uF. While such large values are a good choice in the feedback circuit, DC blocking caps at the input of audio circuits are rather in the range of 47uF. Therefore it would have been worthwhile to also investigate a 47uF capacitor.
Since the capacitance investigated is that large, the conclusion is not really representative.
While distortion is well measurable, the authors conclude that it is not audible. This is why they explained what is audible and what is not at the beginning of the paper. I even find there was significant distortion measured (more than I expected), or maybe I'm just misinterpreting figure 8?
Somehow it appears to me as if the study was designed to show desired results.
As a side note, figure 3 is not a good model of an electrolytic capacitor. A much more accurate model is suggested by Cornell Dubilier (Presented at IEEE Industry Applications Society Conference, Oct 17, 2002), which includes dielectric loss.
There have been several independent measurements before (like the very detailed and excellent work of Bateman) I tend to trust more.
Another short, but good investigation by @PMA : https://pmacura.cz/capdist.htm
Conclusion was that electrolytic capacitors with low capacitance value may show significant distortion if the voltage across the capacitor is high.
The easiest way to reduce the voltage across the capacitor is to increase the capacitance. This is what the Würth engineers did: Investigate a non-issue and conclude there is no issue.
Sorry for being a bit harsh, but for me as a non-scientist, this does not meet my scientific standards.
Capacitance artifacts seems might becoming the dominant contributor to sonic differences if one believes harmonic distortions below the threshold of human hearing at perhaps > -120dB isn't reasonably arguable any more. It seems this only leaves linear distortions.
If taken to the most minuscule scale whereupon an electron represents signal, these units of signal migrate into a dielectric material of depth dependant upon the signal field strength across the material. If the field is altered sinusoidally this suggests sinusoidal movement into and out of the junction between the conductor and the dielectric material. Each electron can be viewed as a recording of the fundamental being played back in their own time in re-entry to a low impedance conducting stream. The difficulty is in measuring electron units of signal being time delayed as time smeared from normal signals resulting from pure capacitance. It seems that DA can be more critical to both understand and measure.
If taken to the most minuscule scale whereupon an electron represents signal, these units of signal migrate into a dielectric material of depth dependant upon the signal field strength across the material. If the field is altered sinusoidally this suggests sinusoidal movement into and out of the junction between the conductor and the dielectric material. Each electron can be viewed as a recording of the fundamental being played back in their own time in re-entry to a low impedance conducting stream. The difficulty is in measuring electron units of signal being time delayed as time smeared from normal signals resulting from pure capacitance. It seems that DA can be more critical to both understand and measure.
Oh the almighty "I can hear it and you can not" argument with no real proof rises its ugly head again.
Why am I not surprised?
Never ever any DBT involved of course.
I admit I cannot hear any difference but that may be because I don't use electrolytic caps for DAC Vref filtering on any of my DACs
AKM and Rohm use electrolytics in parallel with small bypass caps for Vref on their eval boards. In an old app note, ESS used an opamp with 10uf - 47uf electrolytic on the output for Vref.
Some people may be laughing hysterically at that, but could be the last laugh will be on them (at least for dacs intended for listening use).
Some people may be laughing hysterically at that, but could be the last laugh will be on them (at least for dacs intended for listening use).
Not having them can have detrimental effect. It would mean you hear more of the sound of the error amplifier in your LDO. Sometimes the sound of the caps is preferable. LDOs can sound bad despite how they look on an FFT. A nice looking FFT just means what they are doing is not PSS distortion.
Also the reason other people can't hear some things may simply be because the are in the wrong half of the bell curve. IIRC, are you not the one who claims all your dacs sound the same when level matched, except you like an old TDA1541?
Also the reason other people can't hear some things may simply be because the are in the wrong half of the bell curve. IIRC, are you not the one who claims all your dacs sound the same when level matched, except you like an old TDA1541?
Last edited: