Capacitor Distortion

The July 2002 edition of Electronics World/Wireless World has an article entitled 'Capacitor Sound ?' describing a test apparatus for measuring capacitor induced distortion, by Cyril Bateman.

"As a long term capacitor designer and measurement engineer, I believe that any truly audible differences must be understandable and measurable.
Understanding should be in terms of capacitor constructions.
Measurements may however require a change in measuring techniques."

"In order to develop suitable test methods, I have measured large numbers of capacitors of many types.
From these measurements, I have determined the distortion differences between capacitor constructions."

"What I did not expect to find - and I find this rather disturbing - is that within a small batch of capacitors, some exhibit abnormally higher distortions.
These anomalous capacitors typically exhibit some ten times greated distortion than others taped on the same card strip."

His approach is to build a very low distortion (below 1 ppm) 1kHz sinewave oscillator, send that through the on test capacitor and measure the harmonic residuals.

The results are that a 10 nF X7R ceramic capacitor gives -80 dB 2nd harmonic, -60 dB 3rd , -105 db 4th, -85 dB 5th, -110 6th, -100 7th.

10 nF COG ceramic gives -125 dB 3rd, and not measurable other harmonics.

1 uF film/foil polypropylene gives unmeasurable distortion (< -130 dB).

The one and only
Joined 2001
Paid Member
This is pretty much to be expected. Ceramic and Tantalum
caps have no business in the signal path.

What is more interesting is that even lowly aluminum
electrolytics measure much better than any semiconductor,
tube or transformer.

It is my opinion that the role of capacitor is overstated
as a source of distortion and sonic degradation.

capacitor distortion will depend on the load impedance and will be biggest around the corner frequency where capacitor and load have the same impedance. A low pass might have lower distortion than a high pass because the harmonics get attenuated. What was the configuration those caps were measured in?

There is a link to the Jung/Marsh article that first explained the load dependence in my thread on sound effects in electrolytics in the electronics and parts forum. Contrary to the load explanation in the first part of their article, they claim that lower load impedance is better sonically because it gets rid of dielectric absorption which they claim to have heard also in PP caps.

Oh, and I have a thread in the loudspeaker forum on distortion in ferrite and iron core coils. That seems to be largely unexplored territory still.

Eric from the other side
Snowball Effect

Hi Nelson, isn't the audible effect of the combined distortions of series stages additive or multiplicative according to the nature of the distortions of each of the amplification stages (ie harmonic and/or imd distorting stages) and causing an avalance effect that could cause low levels of (coupling capacitor) distortion early in the chain to become more distinctly audible at the end of the chain ?.

Hi Eric

If you have any form of nonlinearity in an amplifier stage (or whatever) this will tend to generate intermodulation products between all frequencies present.

When our now distorted signal approaches another nonlinear stage there will again be IMD products generated between the original signal components, between the original signal components and the "old" IMD products, and between the "old" IMD products as well.

The more stages, the more the sound will be "muddied" with "information" that wasn't there originally.

This is at least what I am convinced of.



P.S. I also read the article in EW&WW and I am impatiently awaiting the second part.
Hi all,

Have any time I advertise to all diyaudio about this series of articles, and I say that hurry to buy EW&WW.

Is a set of six articles, and I have the opportunity of see a sinopsis of all the work.

Really I think that are the definitive work about capacitors. All the people interesting in this topic need read.

Happy days,

Raúl Couto
NAIM AUDIO uses lots of Tantal pearls IN the signal path of the preamps and Xovers for many years now - with a lot of success, I would say. Electrolytics should be DC biased properly, and the relieability problem can probably be subdued with the right supplier.

But, although I diagree Nelson in that respect, he is defintely right in respect that the role of the semiconductors concerning sound degradation.

I tried several "cheap" and small electrolytics soundwise in the last days and was astonished how good they can sound - sometimes even better alone than bypassed with foil-caps !

Various experiments I have done with capacitors have demonstrated that very complex phenomenon can occur under particular conditions. Dielectric abnormal behaviour is know long since (1850 ?) and has been studied in the 1940's.

The simple measurement of harmonic distorsion under sinus voltage is irrelevant. As written by Mr Pass, these measures shows simply that semiconductors and transformers are more "distortionous".

Extended listening tests prove that sound degradation is clearly audible, even when measured THD is as low as -120 dB, which is not the case with transformers, by exemple. Triode amplifier with interstage transformer coupling sound better than capacitor coupling ones, although measured THD is the same, or higher.

In film capacitors, I suspect that surface acoustic wawes can travel along or across the wrapped (rolled ?) film, causing some sort of reverberation like a (bad quality) delay line. Such phenomenon cann't occur in paper-oil capacitors, because of damping caused by oil viscosity. (This can be an explanation for the various mixtures sold (very expensive) to improve (?) sound quality by simply painting or coating capacitors, or even other devices).

Another surprising thing is the response of the capacitor to the acoustic pressures : audible sound quality difference is noticed when passive filters are moved from inside to outside of a loudspeaker cabinet...

Regards, P.Lacombe.
Voltage generated by cap like condenser mic.

P.Lacombe said:
Another surprising thing is the response of the capacitor to the acoustic pressures : audible sound quality difference is noticed when passive filters are moved from inside to outside of a loudspeaker cabinet...

Regards, P.Lacombe.

Just now I grabbed a rectangular light blue Philips 100n 250vac MKT-P film X-cap (the first cap I saw) and put it across the scope and squeezed it between my thumb and forefinger and I could get about 10mV out of it that would last for several seconds if I maintained the pressure. Going by the contact area of my finger/thumb and the cap I estimate the pressure to be about 50 psi. If the air pressure in side the speaker boxis anyting like that then there is a lot of noise happening so would a couple of millivolts on top of several tens of speaker volts be audible? Maybe...

Also, as far as ceramics are concerned, X7R's and especially Z5U's have a great variation of capacitance with voltage. Have a look at the graph attached. I think they were 50v caps. I originally did this measurement because I wanted a snubber cap to go across a Hexfet in a SMPS that would have a high value as the fet turned off to give the greatest snubbing, but reduce it's value as the voltage rose so I wouldn't have to dump so much energy when it is discharged every cycle. But I digress...

In regard to what Capslock says, if the voltage across the capacitor is changing because it has a significant reactance compared to it's load e.g. in an active filter, then if that voltage change varies the capacitance the excess energy won't disappear, it think would have to be added/subtracted to the output voltage thereby distorting it.



  • x7r and z5u variation of capacitance with applied
    5 KB · Views: 2,327

When you experiment pressure with your fingers, it's a quasi-static pressure. But what occurs with dynamic pressure (sound) ? I have in project to experiment this with a small enclosure and suitable driver. I suppose that resonnances can magnify parasitic voltages by several order of magnitude, with appreciable ringing on short pulses. This may be audible on particular music, such percussions. Try and listen...

Regards, P.Lacombe.
Sorbothane Damped Capacitors ?

Given the mechanical forces dependence of these caps, I think we are stating the obvious.
Of course the key is to quantify it adequately.
As it boils down to it, this is (just) another souce of intermodulation.
Given that even very low levels of imd can be audible, the right mounting/isolation/damping techniques ought to help.

Re: Sorbothane Damped Capacitors ?

mrfeedback said:
Given the mechanical forces dependence of these caps, I think we are stating the obvious.
Of course the key is to quantify it adequately.
As it boils down to it, this is (just) another souce of intermodulation.
Given that even very low levels of imd can be audible, the right mounting/isolation/damping techniques ought to help.


Hola Eric,

None of the newsagent for miles around here carry E&WW, and I don't want to subscribe, so I haven't read the articles you mentioned, though I'd like to. Did Cyril only measure THD, or IMD as well? I don't think THD is as audible a problem as IMD, IME, so it'd be interesting to see how the components measure, both under static conditions, and with a two-tone burst, and looking at the response in the time domain on a spec-an and THD meter. If that test had a poor result with two tones, imagine what it must be like trying to reproduce something with an enormously complex structure like a choir.

I found this in my files, and thought you might like to see it. I cannot find the supporting documentation for it (probably lost in my PC crash last year), but I do recall it being a valid looking study.


  • ear distortion.gif
    ear distortion.gif
    9.7 KB · Views: 3,965
THD Tester

Oh boy, are you saying that my ears distort too now ?.
I always thought that my ears were 'blameless' !. ;)
Only joking - I well know that already from experience.
Recent high spl exposures and wax ear can cause extra clearly audible distortions and resonances - really infuriating/disturbing if it's bad.

This article is the first of a series and mostly describes a sub 1ppm (0.0001 %) 0.2V to 4V 600 ohm ouput sinewave generator using several 5534/2 and an SSM2018 low distortion vca to control oscillator amplitude.
The test method is the series on test capacitor going to ground via 100 ohms.
Turns out that the oscillator capacitors are important, as are the trimpot types.
Also the 5534 compensation caps (22 pF pin 5 to 8) must be polystyrene, or second choice COG ceramic, to attain low distortion amplification.
The second article will describe a low distortion, passive twin-tee notch filter and recovery amplifiers for measuring residuals.
No mention of tone burst measurements.
I just ask my local newsagent to keep a copy of EW/WW under the counter each month - no need to subscribe.
Send me a scanner and I'll email the article. ;)

Part II arrived today

--- the September issue -- at least here in NJ -- fortunately EW is wrapped in poly so the mailman doesn't read it (unlike QST).

the distortion levels he is measuring are well below the levels measured by my HP339 -- like 2 orders of magnitude.

this particular issue of EW also has an article on a DIY motional feedback speaker and a slow start for tube amps and VI control. It certainly seems worthwhile subscribing to the magazine, although stateside it isn't cheap.
Diogo said:
What about polyester capacitors, My professor told me it's the best for audio, so I try to use it the most.......

May experience doesn't tally with your Professor's. Try some polypropylene. Also take a look at Davids post here, and read the article that started it all. PPs measure better than PEs too, but it will depend on the resolution and voicing of your system whether the difference is audible, prefereable or significant to you, and there are tons of industrial PP types out there cheap, as well as better inexpensive types like the Sprague 716P 'orange drops'.

Capacitor distortion

I read this thread with considerable interest, and will try to find Cyrill's article. It occurred to me that maybe we should put some efforts in designing amp circuitry that doesn't need (or as little as possible) caps.
Obviously, we can't lose the supply caps, but other than that? If we design for a low DC offset path, there really is no need for coupling caps. We may need a cap in a DC servo, but I would expect that that would have less influence than a coupling cap that is directly in the signal path.
What else? Local decoupling, is that necessary? Anybody has tried just to disconnect local supply bypass and listen to the difference?
Decoupling of current source references, since these (should) have only DC across them, would they matter?
Just some early (06.30) Sunday morning thoughts.

Jan Didden
Joined 2002
Paid Member
don't forget compensation capacitors.

Although comp caps are usually small, they are important and most definitely IN the signal path; as miller caps for example they provide local feedback for the VAS stage. So they have to be good caps. Which thankfully is not usually an expensive problem, because they are small.