Sound-relevant effects in decoupling caps for analog nodes/measurement methods

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The idea for this thread was born from the discussion on ftorres' excellent thread in the digital forum on decoupling caps for digital ICs.

While it is almost universally accepted that electrolytics, ceramics and some plastic film caps make poor coupling and analog filter caps and it is at least partially clear what effects are detrimental to sound quality in those application, the situation is no so clear for decoupling caps on analog nodes such as supply rails of power amps, voltage reference for discrete constant current source or reference decoupling pin of a DAC.

Many believers advocate a certain line of caps in this application but fail to explain the theory behind their reasoning or preference.

Clearly, the caps should have a very low impedance to ensure a good decoupling. However, this is not necessarily true in all cases. If the impedance between the rectifier diodes is low enough, the ripple current will increase if more caps are added or low-ESR caps are used. This will put a certain strain on transformer, diodes and caps which might in some cases degrade the sound. Also, if the ground layout is poor, it might induce 100 Hz noise plus whatever harmonics are on the mains.

Then there is microphonics. DAC and CCS references would be most susceptible to this.

Finally, there is the question of how the caps react to a sudden current demand. Assuming the voltage does not sag too much, capacitance vs. voltage is not relevant. It boils down to the question of how smoothly the current is delivered, if it triggers strange effects in the electrolytics and if it starts sound waves within the cap. A CCS reference will not have a modulated current draw. A DAC reference, if there is no further internal buffer, may see a capacitor charging current which would come in short spikes at some 100 kHz to MHz frequency. This may or may not be dependent on the audio amplitude that must be converted.

As a starter, I shamelessly copied one of my own posts from the digital forum. It deals mainly with ceramics, but the questions raised are also relevant to electrolytics.

microphonics and nonlinear properties


Originally posted by ftorres
Last thing, just to relaunch the debate : What about the piezo-electric properties of Y5V
compared to X7R ? Ceramic layers are so thin in these caps that vibrations may induce some
unwanted noise. So we will have to damp them. Any suggestion for a damping material have
nice HF dielectric properties ?


That question would almost merit a new thread. Obviously, other factors than ESR, ESL and DA may influence the audio quality of a cap.
The most obvious is microphonics, as pointed out by you. My guess is that a dielectric with a high dielectric constant would be more susceptible. But it would also depend on the mechanical properties of both the dielectric
material and the overall design of the cap.

Most sensitive application in my eyes is not rail decoupling but reference decoupling, i.e. voltage reference in a DAC or zener diode in discrete current sources.

Then there is the question how a cap reacts to voltage drops and current draw. ESR, ESL, DA are, after all, just linear measurement methods, i.e. they are done with a huge DC offset and a small AC test signal.

Capacitance vs. voltage is specified but measurement is carried out again only by changing the DC offset from measurement to measurement. How is linearity with voltage drop spikes such as may occur when a DAC charges its internal capacitors? Again, this is not only a question of the dielectric properties alone but also of the mechanical properties as the electrodes may "move" when voltages change or currents flow.

Listenting tests might help but might only produce new myths. We should try to come up with tests for nonlinear capacitor properties....
low frequence effects in the electrolyt -> red herring?

another self-quote from digital:
"Now for something that confuses me: BC components (formerly Philips Passive) type 136 low impedance cap specs give ESR at 120 Hz and impedance at 100 kHz. For 1000uF/25V, ESR is 0.22R and impedance is 0.034R.

To my understanding ESR (equivalent series resistance) is a real impedance. A HF complex impedance can be no smaller than the real series impedance. Strange, uh?"

I took a closer look at the BC data sheet and I am beginning to get the idea although the data sheet does not explain the measurement method.

At 100 Hz there seems to be a procedure to measure and compute the resistive or real (i.e. no phase shift contribution) part of the impedance, i.e. get rid of the 1/(i*omega*C) contribution of the capacitance.

At 100 kHz they seem to report the amplitude of the complex impedance of the series connection of the capacitance and a parasitic inductance and resistance, i.e. Z_data sheet = real(Z(100 kHz))=real(R + 1/(i*omega*C) + i*omega*L).

R(100 Hz) is roughly a factor of four higher than real(Z(100 kHz)), showing a slight dependence on the rated voltage of the cap.

Looking at the plot of ESR(f) on page 13 of the 136-series data sheet (download e.g. from and assuming ESR=R in my equation, there seems to be an additional resisitve contribution at 100 Hz that deminishes with increasing frequency until R becomes constant in the low single digit kHz range.

So some process in the electrolytic with a time constant well in the audio range inhibits the flow of current.
If I am not completely mistaken, an electrolytic is not a true dielectric in which the polarisation is only a function of the electric field. Rather, it acts like some kind of battery, i.e. there are chemical processes which store and release the charge. If this ~ms time constant was a typical time for the reaction to take place, one could understand the inhibition of the current flow. However, my reasoning is that this should also reduce the effective capacitance at low frequency.
Looking at page 12, however, capacitance (again, they don't explain measurement/computation) seems to be fairly constant up to 20 kHz where I assume the parasitic inductance begins to have a relevant contribution. There is no sign of a reduced capacitance below 1 kHz!

Can anybody make sense of the data?

Aren't effects in the Hz- 5 kHz range likely to have some audible impact?,1570,28,00.html

ESR Defined
ESR is the sum of in-phase AC resistance. It includes resistance of the dielectric, plate material, electrolytic solution, and terminal leads at a particular frequency. ESR acts like a resistor in series with a capacitor (thus the name Equivalent Series Resistance).
Microphonic effects can be quite substantial, as seen here. I've been unable to find much information on the piezoelectric qualities of different ceramic dielectrics, although NPO is supposed to be non-microphonic.

I do have some Taiyo Yuden caps that were designed for mobile equipment (cell phones, etc) and that are supposed to be less susceptable to shock-induced noise.


  • ceram_piezo.gif
    5.5 KB · Views: 420
Decoupling Caps

Hi Eric,
<I>If it pleases you,
sounding fluent
What matters measurements.......
If sound is OK for you....??</I>[Joke];)

I am sorry, back to <B><I>normal</B></I>
Routinely I was using ceramic bypass caps in all of my DIY equipment: preamp, DAC, etc.
Later I tried polyester caps; all 0.1µF in my DAC and the sound was slightly more fluent, less harsh to me.
Later I switched to polypropylene for bypassing and that was much better. I found the measurements by Francois verry interesting but a little hard to follow.
I also tried polyester caps in the powersupply; Jung like regulator but that modification gave rise to oscillations.
Those ceramic caps are certainly microphonic: If you press on them while measuring capacitance you see them change capacity!
Also they have a high temperatur coefficient. Measured them while heating with my daughters hairdryer. I believe they are called KDPU Conrad part# 453358, if this helps.


thanks for the links. On first glance, they appear to be an excellent primer on capacitors and common capacitor parasitics such as ESR, ESL, dissipation factor and dielectric absorption.

I am familiar with Walt Jungs publications on capacitors in Analog Dialogue, and this seems to be an expanded version. I have made a thick print-out for bedtime reading...

DA should be relevant in coupling and analog filter apps, but not necessarily in decouling apps.

ESR, ESL and DF are important in decoupling apps but as I tried to argue, they cannot make up the whole story. In particular, they are part of a linear analysis where we should be concerned about nonlinear effects.

Is there any passage in your links that you feel might address my rather specific topic?


Re: Gromanswespeak

mrfeedback said:
The answer of course,
is all in the conductor noise,
the noise to shunt.
or the noise to pass,
this combination,
is in the resultant.

It's all in the vibes, man,

Goodday me namesake from down under,

having suffered groman's pseudo-sophisticated blabber throughout the better days of audioasylum only to find after a couple of calm months that he has decided to get his daily dose of negative attention from diyaudio, I find your impromptu emulation of his style vexing, to say the least....
Jung/Marsh "Picking Capacitors" article

Well, I am a little behind with my bedtime literature. Did read the pickcap link provided by Harry this weekend. Seems to be an extended version of the article in Analog Dialogue which did not cover the audio implications.

Regarding my original concern about sonic effects of supply decoupling caps, there was just this: "For example, some listening tests have indicated that they can produce audible distortion when used as supply bypasses, let alone coupling!"
I dare question the universal valitdity of this as much as the claim by that various capacitor brands have their specific sonic signature no matter where inside a CD player they are used.

The rest of the article is pretty useful and straightforward, even if it treats only coupling and filter applications.

There is one thing that throws me, though. In one of the first paragraphs they explain that a highpass or DC blocking filter has the highest distortion at its corner frequency because then the resistive load is similar to the impedance of the cap which hence sees a large AC voltage. Lowering the corner frequency from 10 to 1 Hz reduces distortion at 10 Hz significantly. So far, I agree completely.

However, towards the end, in the paragraph that explains fig. 17, they recommend using a low load resistance, firstly to minimize DC error due to capacitor leakage current, and secondly to minimize dielectric absorption effects. Driving the cap from a 1 k source and loading it with 1 - 50 k, they claim to have noted a substantial sonic improvement, in mica, polyester, tantalum and polypropylene.

At least in the tantalum, this would have resulted in incrased LF distortion. Also, I find it hard to believe there would have been an audible effect in polypropylene. May have been common mode disto in the power amp input stage, though...
working through Harry's list:

The link deals with dielectric absorption as an effect that screws up sample and hold circuits. No relevance to power supply bypassing, but it is interesting to note that polypropylene and NPO ceramic can be better than polystyrene. Also, there seem to be significant differences between various caps of the same diele ctric material.

The article is the Jung/Marsh article discussed above. Worth reading by all means, even if not about bypassing

The file, like almost all of the hole website, is practically an ad for their product, the multicap, even if disguised as an engineering article. Apparently, the multicap is a cylindrical polypropylene film cap with concentric electrodes (rather than one long winding). It claims to have far lower ESL etc. than any other film cap on the market due to internal paralleling of multiple capacitors. I can see that it would be superior to caps with one large winding where there are simply a connecting wires touching each electrode at some point. However, stacked film caps where two sides are metallic and each single metal film is connected to its corresponding face would have the same effect. There are even single-winding caps in which each turn has a connection to the face. Where's the difference??

The article defines ESR, DF etc. Useful summary of common spec values, but no discussion why ESR would depend on frequence.

The article deals with piezoelectric effects in ceramic capacitors. No relevance to electrolytics in supply bypassing. They do not recommend the use of higly piezoelectric dielectrics in oscillator and filter applications They do give the full names of three dielectrics with a large effects, but fail to translate this into commonly known names such as X7R.

The faradnet article seems to be a reprint of a book dating back from the 1930s. It deals with the fundamentals of electrolytics, so I'll give it a more thorough reading.

I've had the same suspicion about Multicaps...I don't see at all how these are any different from more readily available stacked film caps.

Although people may dislike Wimas for various reasons, check out the datasheets on some of the FKP film and foil caps...some of these stacked film caps have low impedances into 5MHz+. I doubt Multicaps are better, and wouldn't be surprised if they are worse.
back to the red herring

i.e. my question why 100 Hz ESR is so much higher than 100 kHz impedance.

The faradnet link, in spite of being an online version of a book more than 65 years old, provided at least a clue. In an aluminum electrolytic cap, there is really a dielectric which happens to be Al2O3 and is very thin, so it is not a battery like I thought.

However, one of the electrodes is really an electrolytic solution. Maybe conduction phenomena are the reason for the frequency dependence? On the other hand, I'd expect conductivity in an electrolyte go down with increasing frequency, not up.

On a side note, there seems to be an obsolete German DIN standard for rough (rauh) and smooth (glatt) electrolytic caps. I presume this says something about the surface structure of the aluminum foil electrode. I'd expect a rough electrode to offer a higher surface and hence higher capacitance, but in reality rough caps are always bigger than smooth caps of the same rating. What are the respective properties of these electrode types?

Still puzzled,

more book reviews

The two links provided by Aggemann are on the Jensen and Aerovox variations on the four pole theme.

The idea seems to be that you let the charging current enter at one end of the windings and draw the supply current at the other end. In this way, the parasitic inductance gets used as a filter, i.e. you can save the choke before the cap. Actually, the cap works as some kind of distributed LC filter. Why this would inherently lower ESR and ESL as claimed by both manufacturers I don't see.

Re: more book reviews

capslock said:
Why this would inherently lower ESR and ESL as claimed by both manufacturers I don't see.

Because of the inherent need for using thicker foil in order withstand the current now running through the cap-foil instead been present at the terminals.

The general idea is to make choke input filters instead of pi filters as the powersupply for pure class A amplifiers.
capslock said:
Didn't think about the thicker foils. But this can be had in a low ESR 2-pole cap also, right?

Right, but rarely is, since the need is not present anymore. That was essentially what was meant with 'rough' standards in the old DIN.

Given the prices for those T-network parts, combining a (common mode) choke and high-quality conventional caps might even be cheaper.

Then you would lost the reason to use them.

A pure class A amp, especially single-ended kinds like the SOZ, normally have a transformer well capable of delivering the needed power for the circuit because of the outpower of such is a minor part of the total power consumption.

A normal pi filter powersupply have huge caps in order to filter anyway ripple currents while a single 4-pole cap of reasonable size (min 2000 yF) in a choke input filter all but removes it entirely.

Thereby eliminating the need for those big expensive caps.

Right, but rarely is, since the need is not present anymore. That was essentially what was meant with 'rough' standards in the old DIN.

Then how do modern low ESR, ESL achieve their specs if not my thicker foils and optimized geometry? Also, there seems to be a need for them, as they are becoming cheaper and more readily available all the time.

I am not sure I can follow you about the part with the smaller filter caps. In first approximation, it should not matter whether I place my 2x10 mH before or inside the cap, both would limit the peak charging current by the same amount. And I'd still need the same capacitance to get the same corner frequency.
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.