[ Home | Staff & Contacts | DIY & Tweaks | Listening tests | HiFi Playground | Music & Books ]

Do capacitors sound different?


...and how can we test them?

[Italian version]

Manufacturer: Cicada and other capacitor boutiques
Product: Cicada & other Couture capacitors
Price: ??? depends on value, foil, voltage and order size, and for some other brands making unobtanium foil in snake oil, the bankers' bonuses look inadequate
Reviewer: Mark Wheeler - TNT UK
Investigated: 1983 - 2011

"The old fool refuses to publish tests of cables," challenge plebs, stage left, "But now he's willing to make claims for capacitor sound!"

"Indeed," replieth the old scribe, "The title question really ought to be:"

How do capacitors Sound Different?

Tom Fung, proprietor of Cicada capacitors sent me a batch of his finest new product ranges, risking ridicule and opprobrium by sending this sceptical old scribe a package of capacitors whose net worth exceeds that of whole components reviewed in these esteemed pages. Being thorough and ethical, your old scribe does not just stick a new component in place of an existing one and describe the effects as though these will be universal. All testing that has any claim to scientific rigour must make strenuous effort to reduce the effects of variables other than those under test. Hence, the inter-relatedness of neighbouring components must be considered. Component layout must also be normalised during testing. Many of us will have experienced successful breadboard prototypes that sounded dreadful when rearranged to fit a case or worse, that went into oscillation due to the alignment, inductance and length of lead-out wires.

"What is a capacitor?" enquire plebs, stage left, "and how does it differ from a condenser?"

Capacitors and condensers are the same thing and are made of two parallel conductive metal sheets separated by an insulating dielectric. The dielectric can be made of air, waxed paper, ceramic, plastics (like polypropylene or polyester or polycarbonate), PTFE (Teflon), tantalum oxide, mica or any other insulating material suitable for the design application. Hence, flexible materials dominate this list so that the two sheets can be rolled up to save space, hence the tubular shape of many capacitors. Readers might remember how conductors work with the electrons jumping from 'hole' to 'hole' in the opposite direction to current flow convention. In a metal lattice the electrons are delocalised, so there aren't really 'holes' as such, but this is an analogy. When the capacitor is 'charged' the analogy is that there is some energy in the form of extra electrons on one sheet and electron holes on the other, measured in Joules like any other energy.

In audio circuits the net utility is two-fold. The temporary energy 'storage' is used (in parallel) to smooth the rectified AC in a power supply (which without the capacitors would be 100 or 120 shifts from zero to maximum amplitude every second) sufficient to provide adequate regulation to meet the variable demands of a music signal in each amplifying stage. The reactance is essential in series to block interstage DC and in series or parallel to define bandwidth or frequency response. This is because a capacitor's resistance at falls with increasing frequency in a logarithmic relationship (hence the tightening of the parallel lines on the reactance chart). Inductors (coils of wire) have the opposite relationship, which is also very useful in audio.

The capacity of the insulator is measured as its dielectric constant (k) and varies between just over k=1 for air, through k=2 for materials like Teflon and k=6, to much more practical capacitor materials like tantalum k=27 and various ceramics ranging from k=30 to k=3000. That is why ceramic capacitors are tiny beads and Teflon capacitors are massive logs. Dielectric constants aren't even constant (rather a contradiction in terms) as they vary with voltage and temperature (the two being related by the aforementioned ESR) and also (especially annoyingly given the role in AC circuits) vary with frequency. Thus capacitors are b*gg*rs to design and specify for an application, even before the weirdness of audiophilia sets in.

[don't throw a curve]

we all should know




f in Hertz
C in Farads
L in Henries

XC is the capacitive reactance and therefore is the (theoretical) resistance at just that frequency of a (theoretical) capacitor; real capacitors have real resistance in the lead out wires and the foils because they're made of real materials that start as ore in the ground (well, their recent history does) and get mined, crushed, heated, refined and eventually form the electrical conductors in these real world capacitors.
"What is the old scribe droning on about now?," demand incredulous plebs, stage left.

The point is that the conductors in these devices are as pure as the manufacturer chooses to specify, depending on cost-benefit analysis for the target selling price. Likewise the layer of insulating dielectric between the two layers of the capacitor. Further consideration must be made in this choice for the space the capacitor is allowed to occupy, the voltage it must tolerate and the shocks and stresses it must handle. Remember, in an extreme example like the 'condensor' (old world name for capacitors) in a car distributor clumsily installed, a dented capacitor no longer has the specified thickness of dielectric between conductor layers.

Even the lumped LCR parameters might differ, according to construction type and size. How straight the lines are and at what point their rate of curve increases depends on decisions made at the component design stage and manufacturing processes and tolerances. These big first order differences are easy to specify adequately, like ESR (equivalent series resistance) which ought to be zero (no in phase resistance) but can never be zero; perhaps if the capacitor could be held close to 0Kelvin ESR might be close to zero, but the instant any current is applied the temperature would zoom up. The second order differences are plentiful and more difficult to quantify in relation to audible effect.

Hence, there are numerous possible explanations for potential sound differences between capacitors. Hence, there are also numerous possible reasons for yet more audio paranoia. Phew!

Another peg to hang some audiophilia paranoia on is dielectric absorption which is the imperfection of the charge storing described earlier. Not all the stored charge stays stored, some charge will inevitably leak or 'migrate' from the dielectric eventually. What this means to the rest of the circuit around the capacitor in question, is that its value is therefore not what it says on the tin (UK advertising catchphrase). This is extremely important in sample and hold circuit applications used in music production, for example. There are also other measurable parameters like dissipation factor, self inductance (ESL) and these have obvious effects on in-circuit behaviour and like most time and frequency domain changes are really visible on a square wave.

So how does one test capacitors?
"Obviously you compare the test capacitor with others of similar values in similar circuits," chorus plebs, stage left.
"But that involves changing other variable like solder, and the effects of the heat of an iron on a component that may have been flow soldered in the first instance," begins the old scribe.
A reactive device like a capacitor reacts with the other components in its circuit stage, for example the inductance of the resistors combined with the inductance of the capacitor lead-out wires may make a simple first order difference. The ESL of a long lead audiophile couture capacitor designed to piggy back off a circuit board might well be much higher than that of the surface mount device it replaces. More expensive components ought to behave more like a 45° straight line on a reactance chart but real world manufactured devices will never achieve this absolutely. [OK, throw a curve]

It transpires that it is much harder work comparing capacitors products than any other audio components:

Before the brickbats descend on the TNT-audio.com inbox, to some degree ALL non identical capacitors MUST perform differently. The better any capacitor is manufactured and the more more tightly it is toleranced, the less difference there will be from another equally well manufactured and toleranced capacitor. To believe that any 100V 5% capacitor will sound similar to any other 100V 5% capacitor is deluded.

Equally, to believe that all 630V paper in oil capacitors sound better than all polypropylene dielectric capacitors in all applications is...
equally deluded.

Many valve afficionados, particularly of the single ended persuasion, claim that nothing but paper-in-oil capacitors will do. They insist that clarity is optimised by sticking to one brand and never bypassing. Conversely, many push-pull bipolar transistor amplifier designers insist that mixing brands and types to minimise any particular sonic signature makes for the least coloured contribution. I know of one who spent years experimenting (before the advent of couture audiophile capacitors) and came to the conclusion that in his tests that multiple bypassing performed better than individual capacitors. Results like these confuse any capacity [groan] to draw conclusions in comparison tests.

Other valve afficionados swear by Hovland Musicaps, Jensen paper-in-oil, or Infinicaps (to name but 3 in favour on blogs) in the so called signal path and Cerafines or Black Gates in the power supply or cathode decoupling. There is no hegemony except that ridiculously expensive capacitors are believed essential in simple valve circuits to maintain any credibility among their peers. Fans of particular capacitor creeds do seem to express their views (especially on web forums) with a fervour that can only be described as religious. But, then they must justify the vast sums they have spent, like the emperor and his new clothes... or as the enlightened ones who we will eventually follow.

The delusion that we are all pursuing 'high fidelity' as though it were an overarching metanarrative, free of values but at the same time filled with virtue, is tested beyond its limit when the subject of wire and passive reactive components is raised. Each audiophile listens for those qualities they have learned to love (much as a smoker learns to love smoking then learns what brands or mixtures to love), prioritising those qualities and therefore inevitably relegating other qualities to irrelevance.

Perhaps the presence of oil or wax in some types helps to dampen internally generated vibration? This author has never seen any research to support this, but what about external vibration impingeing on the capacitor?

That sixth bullet point has particular relevance to loudspeaker crossover use. Quite apart from your old scribe's views on passive crossovers, for those who insist on staying passive, vibration works both ways in loudspeaker crossover capacitors. If mounted within the cabinet crossover capacitors are affected by the high intensity soundfield (that is how a condenser microphone works in principle) but also convert some of the signal into vibration within the component... They are affected

"Capacitors vibrating?" chorus plebs, stage left, "Did the old fool really just mention vibration generated by capacitors?"

You don't believe me? Try connecting a capacitor and resistor (in series to protect the amplifier) across a stable amplifier and playing music through it. Depending on the construction, sound may be audible emanating from the capacitor, heavily distorted, but audible nonetheless. This can only come from vibration generated in the capacitor by the audio signal. Conversely a capacitor may convert vibration into electrical energy, particularly if the cap is biased by an AC voltage like an audio signal. Ho hum.

These are not the deranged ramblings of a deluded audiophile...

"Oh yes they are!" chorus plebs, stage left, getting into the pantomime spirit.
Oh No, they're not! replies the old scribe.

Tests by Jean Hiraga in the 70s, Walt Jung, Ben Duncan and Martin Colloms in the early 80s, yielded distinct differences which Colloms attempted to quantify as percentages. In the latter test the lowest score was earned by a fancy Rubicon Litz wired couture special that scored lower than the cheapest big aluminium electrolytics, so big money does not always buy success. In his follow up investigations Colloms identified Mechanical resonance and piezo-electric effect and measured induced vibration using a low mass accelerometer. A capacitor maker (ClarityCap, whose wares I haven't tried) commissioned Salford (near Manchester) University's Acoustics Research Centre (ARC, not to be confused with the Audio Research Corporation) to investigate the phenomenon. They identified that most capacitors (regardless of construction) of typical values to be found in audio circuits exhibited a resonance, usually between 5kHz and 25kHz. This might contribute to an explanation why what minor differences we hear between individual capacitors is usually at high frequencies (high for ears, not for electronic circuits). This was not the mickey mouse sample size of audiophile reviews, this was compiled from over 300,000 sample measurements. ARC then ran blind listening tests (n=30) to determine the audibility of these resonances and determined that 70% of the sample expressed a preference for lower resonance samples.

The foregoing tests all used a single application, a test jig. Your old scribe tried various combinations of series and parallel arrangements and also a transformer coupled isolator utilising a pair of Sowter 9111. The solder and heat question was avoided by usung screw terminals to pressure the lead out wires against a hook up wire. The terminals are cheap generic brass devices as these were all that could be obtained matching the mechanical requirements, hence their use only as clamps not as conductors. torque was approximately equalised by a quarter turn clockwise after contact. Once the various scores from the Judging System were amortised the pluses and minuses tended to be close to equal. This is what happens with big number tests, unambiguous results just do not happen. Those who crave certainty in their hobbies should look, or listen, elsewhere. Audiophilia is awash with as many opinions as there are people who have bothered to undertake thorough tests and everyone else just else tends to regurgitate them on blogs and lesser websites than this. Context does change everything, but these tests of the Cicada capacitors did support a trend among materials that I have observed before and is beginning to form into a hypothesis. You'll have to wait until part 2 to find out what it is.

"Oh he's such a tease!" observe the more camp plebs in the chorus, stage left.

[the rule]

old scribes and old readers may recognise
the slide rule reactance calculator that
clearly illustrates the log relationship

Sound Quality

The majority of the differences do occur at the top end. Rhythm, for example is mostly affected when there is significant contribution to the rhythmic structure from high frequency rich components like tambourine or riveted crash cymbals. Midrange quality, especially vocals is most affected by those recordings whose vocals have Dylanesque rasp or sibilance.

Family sounds are obvious too. Paper-in-oil from a manufacturer will share a house sound, regardless of foil material, metallised polypropylene favour 'hi-fi' qualities, while those described simply as polypropylene can be among the most neutral or worse than aluminium electrolytic. The Rel-Cap TRT Wondercap was one of the first couture capacitors to justify its cost, used extensively in Audio Research of the early 80s and also the first to convince your old scribe. Theye were used throughout as coupling caps in the little Single Ended Triode, Torlyte Cased 4W wonder described in Box Clever and major contributors to the final version's transparency and neutrality, easily as much as the Black Gate cathode bypass capacitors. At a few Euros each (just 2€ for 0.01uf) they were cost effective, but as more expensive rivals sold well, the manufacturers switched to the Infinicap brand costing several times more than the nifty Wondercap and your old scribe's cynicism returned.


In part 2 the differences, if any, will be explored and some sense of proportion will be maintained. The context sensitivity is easily as complex as cables, but some conclusions will be drawn that might help those more smitten with paranoia audiophilia.

Another case of being far from an open and shut case; indeed the case is altered with every altered context. However the trends observed in these tests might inform those readers who delve into their hardware and attempt home brewed upgrades. The findings might equally be useful to those designing from scratch.

The first conclusion is that like many audiophile conclusions, the differences are most useful as near the front end as possible. More transparency in later parts of the signal chain, merely reveals inadequacy earlier in the signal chain. In a vinyl analogue replay system the cartridge loading capacitors have most effect, then the next interstage capacitor, etc. Improving the pre-amp power supply with better (and bigger) capacitors reduces ripple and noise that would otherwise be amplified along with the signal. Hence, start at the front end if upgrading passive components.

[Go and read Part II]

Music enjoyed during this review

  • Bob Dylan: No Direction Home, Martin Scorsese soundtrack on 180g vinyl
  • Jimi Hendrix: Electric Ladyland, an album regularly in this list, both vinyl and CD today
  • Arvo Pärt: Beatus, Estonian philharmonic Chamber Choir, uncanny vocals
  • Sara K: Miles Away, Chessky classic recording
  • Little Feat: Sailin Shoes, vinyl original spinning on every turntable here since the Goldring in '75
  • The Blue Drivers: Your Mileage May Vary
  • Notorious: Radio Silence
  • Latin Quarter: Modern Times, wild sibilant vocal EQ
  • Latin Quarter: Swimming Against The tide,
  • Dave Brubeck: Time Out/Time Further Out,
  • Pink Floyd: Dark Side of the Moon, MFSL Original Master Recording gold CD
  • Gil Scott-Heron: I'm New Here

and the Manger sampler, Musik wie von einem anderen Stern
a superlative collection of fine natural sounding recordings

© Copyright 2011 Mark Wheeler - www.tnt-audio.com

[ Home | Staff & Contacts | DIY & Tweaks | Listening tests | HiFi Playground | Music & Books ]