Another definitive advice on capacitor distortion.

[abbey road d enfer]

Recently saw this article, that apparently claims to close the lid on the subject of capacitor distortion.
https://www.we-online.com/en/suppor...t-of-harmonic-distortions&activity_id=4326949
What are your thoughts?
I really contest the methodology...

 

[zamproject]

Interesting

Maybe I read it to fast, but I also have some interrogations ....

-still don't get what is -intensity- in all the graph ?!?
-How the author merge THD electrical measurement with dB SPL and hearing sensitivity ?
- the statement:

As discussed in section 02 the hearing threshold for
distortions is about 7%

not expert in ear physiology nor psychoacoustic, but this number seem way to high ?!?
did someone have a source for this ?

-Study is done at single fundamental frequency where ear sensitivity curve follow harmonic damping/distribution curve.
What about a test tone just one octave below max ear sensitivity (let say 3kHz tone for a f2 @ 6k...)

-What happen if you add a consequent DC to the ac signal (which is in first place the usual reason to put a capacitor in audio path...)
-What about a complex audio program, full spectrum, with all intermodulation, complex frequency summing, etc...
-What about real life design where you easily find 10 capacitors in the signal path.

In other hand I never had issue nor complained about capacitor distortion when doing my mixing job...

 

[KrIVIUM2323]

not expert in ear physiology nor psychoacoustic, but this number seem way to high ?!?
did someone have a source for this ?

Well it's from a different field (loudspeakers) but still... Earl Geddes and Lydia Lee work seems relevant ( they come to conclusions up to 20db in some cases):

http://www.gedlee.com/Papers/papers.aspx

http://www.gedlee.com/Papers/The Perception of Distortion.pdf

 

[JohnRoberts]

Less is better.....

JR

 

[k brown]

Less is better.....

JR

In general, I would agree - but perceptual masking is a very real thing. Below a certain threshold, things simply are not heard; any lower is not perceived.

 

[JohnRoberts]

In general, I would agree - but perceptual masking is a very real thing. Below a certain threshold, things simply are not heard; any lower is not perceived.

I promised myself to not argue with people on the internet about what they say they can hear.

I am very familiar with perceptual masking as it applies to the design of dynamics (I killed a lot of brain cells designing companding noise reductions).

I discovered personally many years ago, that I could measure distortions that I could not hear. I strive for lower distortion when practical. For one example of practical considerations, back when I designed the sine wave oscillator inside the Loftech TS-1 audio test set. I could have made the sine wave distortion lower, but the settling time would increase and this settling time would be more annoying to the customers than the distortion. I was dealing with low tenths of one percent distorion. Audible in close A/B listening tests, but not annoying in general use.

JR

 

[merlin]

Ugh, what a badly written paper. TLDR: He concludes electrolytic cap distortion exists, but is so small that you can't hear it if you believe his 7% 'threshold' figure. Nothing new here.

still don't get what is -intensity- in all the graph ?!?

Magnitude of the harmonic relative to the fundamental. Someone with an engineering doctorate should be able to label graphs better!

 

[abbey road d enfer]

We all know that electrolytic capacitor distortion is most audible at LF, for at least two reasons:
one is that distortion increases with voltage, which is coincident with the increase of voltage loss across a coupling capacitors, the other is that very low frequencies are not efficient at masking low-mid harmonics.
IMO measuring capacitors at 500Hz is not representative of the conditions where their distortion is most audible.
The author does not consider the effects of cascading dozens of CR circuits as is the case in a typical audio chain.
I think he is more concerned with proving that their products are exempt of defects.

 

[merlin]

I think he is more concerned with proving that their products are exempt of defects.

+1.

 

[jp8]

In my Audio Analyzer, there are three elcaps in the audio path when doing a loop-back measurement. Harmonics measure below -106dB in a loop-back test. IMHO, elcap harmonic distortion is about the last thing to worry about. I'm really surprised about the high THD levels of the WE elcaps, btw.

Jan

 

[abbey road d enfer]

In my Audio Analyzer, there are three elcaps in the audio path when doing a loop-back measurement. Harmonics measure below -106dB in a loop-back test. IMHO, elcap harmonic distortion is about the last thing to worry about.

Capacitor distortion is frequency and level dependant. It can't be quantified with a single figure. In many cases, sound reasoning allows reducing distortion to an acceptable, even neglectable level.

I'm really surprised about the high THD levels of the WE elcaps, btw

That is due to the measurement method, which is completely uncorrelated with what happens in reality.
In his historic works, Ciryl Bateman had to push his measurements way beyond normal conditions, in order to "magnify" the results.

 

[JohnRoberts]

In my Audio Analyzer, there are three elcaps in the audio path when doing a loop-back measurement. Harmonics measure below -106dB in a loop-back test. IMHO, elcap harmonic distortion is about the last thing to worry about. I'm really surprised about the high THD levels of the WE elcaps, btw.

Jan

I haven't told this story recently. Back in the 70s when I was trying to extend the resolution of my budget test bench I bought a old used Singer spectrum analyzer for $700 (nowadays free from a smart phone app). I fed the product output from my Heathkit distortion analyzer into the spectrum analyzer input. This gave me 40 dB more measurement resolution floor, in theory. In practice I could see the THD caused by the heathkit circuitry. By dropping my THD analyzer nominal 0db level -10dB I could make clean measurements again with 30 dB more resolution, not bad for my budget test bench.

JR

 

[vintageampfixer]

TBH, it reads like learned folk who've gone a long way to try and make a point and then justify it because they're somehow affiliated to Wurth.

Slight tangent (you know me) but wasn't there a recent and lively debate here expounding the virtues of kit with HF responses going way, way over 20kHz? ... so I'm astounded that no one's questioned this part!

And from that same "recent and lively" debate, I'd stand by my point that reactive components all, by their very nature, add some "distortion" through the effect of applying differing phase shifts to different frequencies which by definition change the nature of the reproduced signal when compared with the original. I wouldn't however be so bold as to try and quantify that (and I'm not affiliated with Wurth).

 

[ruffrecords]

The whole thing sounds to me like an undergraduate summer vacation project.

Cheers

Ian

 

[k brown]

TBH, it reads like learned folk who've gone a long way to try and make a point and then justify it because they're somehow affiliated to Wurth.

Slight tangent (you know me) but wasn't there a recent and lively debate here expounding the virtues of kit with HF responses going way, way over 20kHz? ... so I'm astounded that no one's questioned this part!

And from that same "recent and lively" debate, I'd stand by my point that reactive components all, by their very nature, add some "distortion" through the effect of applying differing phase shifts to different frequencies which by definition change the nature of the reproduced signal when compared with the original. I wouldn't however be so bold as to try and quantify that (and I'm not affiliated with Wurth).

View attachment 135795

The only relevent aspect of what happens above 20kHz is some things can create harmonics or other artifacts of higher frequencies that can occur in the audible range; the most recent hotly debated one is the audibity of higher sampling rates, that even though the filters are operating at frequencies above 20kHz, they can create audible distortions below that, and using higher sampling rates pushes those effects further above audibility.

 

[JohnRoberts]

The only relevent aspect of what happens above 20kHz is some things can create harmonics or other artifacts of higher frequencies that can occur in the audible range; the most recent hotly debated one is the audibity of higher sampling rates, that even though the filters are operating at frequencies above 20kHz, they can create audible distortions below that, and using higher sampling rates pushes those effects further above audibility.

IMD....

JR

 

[Voyager10]

The paper seems to be measuring decidedly average THD figures, under (what appear to be) unrealistic circuit conditions and then trying to justify that it's inaudible.

(If I understand correctly, the AC test voltage of "1.5V" is the actual ripple voltage on the capacitor, which is pretty extreme in an audio context).

A better question for them to answer would be: if I have a fancy-pants modern op-amp with its 0.00003% THD, how do I choose and apply a capacitor which doesn't entirely ruin these numbers?

 

[abbey road d enfer]

The whole thing sounds to me like an undergraduate summer vacation project.

Cheers

Ian

Well, the guy who signed the article is one Doctor René Kalbitz.

 

[vintageampfixer]

Well, the guy who signed the article is one Doctor René Kalbitz.

and he's "product manager for supercapacitors at Würth Elektronik since 2018 and oversees research and development projects in the field of capacitors"

 

[jp8]

We all know that electrolytic capacitor distortion is most audible at LF, for at least two reasons:
one is that distortion increases with voltage, which is coincident with the increase of voltage loss across a coupling capacitors, the other is that very low frequencies are not efficient at masking low-mid harmonics.
IMO measuring capacitors at 500Hz is not representative of the conditions where their distortion is most audible.
The author does not consider the effects of cascading dozens of CR circuits as is the case in a typical audio chain.
I think he is more concerned with proving that their products are exempt of defects.

Indeed, the voltage across a coupling cap increases at lower frequencies, increasing THD. The question is whether this THD rise is significant and audible.

Referring to my Audio Analyzer again, which has four (not three, as I previously said) 47uF bipolar Nichicon UEP1J470MPD coupling caps in the audio chain when looping back the signal from output to the mic preamp input. I checked the THD rise at various frequencies and output levels. The cut-off frequencies of all stages are 0.33Hz or below, so voltage across the coupling caps will be very low, even at 20Hz. At a frequency of 20 Hz and measuring at 10mV, 100mV, 1V and 5.8V (=max), THD never exceeded -90dB. So it's still in the weeds.

Things get a little different if we deliberately choose a too-low value for the coupling caps. As a test, I took two Nichicon audio grade 4.7uF elcaps, loaded by a 1k resistor, giving a cut-off frequency of 68Hz. So at 20 Hz, a significant part of the input signal will be across the elcaps. At 20Hz and 8.6 V AC RMS across the coupling caps, THD topped at -61.2 dB, mostly 3d harmonic. The 4th and 5th harmonics were at -104 and -97.5 dB respectively. Due to the increasing sensitivity of the ear to the low-mid harmonics, the 4th and 5th harmonics would probably be the first ones to be audible. Now suppose one plays this 20Hz signal through speakers at a level of 100 dB SPL, and assuming auditory masking follows the Fletcher-Munson curve, then at 80 Hz, both the 4th and 5th harmonic remain well below the hearing threshold at 80 and 100 Hz. So all in all, nothing to worry about, especially when you compare it to the THD of speakers in the LF range, which is probably orders of magnitude greater. From all this, my conclusion remains that THD from electrolytic coupling capacitors is a non-issue, unless someone can enlighten me and tell me where my reasoning goes wrong. Just take an elcap from a decent brand and choose a cut-off frequency well below the audio band and you're safe.

Jan