Doing valid "capacitor dielectric listening tests"

[MR COFFEE]

All,
I thought I'd move this subject thread over rather than reply at the end of the "greenie capacitor - disadvantages" thread, as the subject keeps coming up around here every now and then (in a stompbox forum of all places!)

Really testing for audible capacitor differences is really, really, HARD WORK. Why? Well, for instance...

There is an interesting phenomenon in scientific testing of subtle audible differences:
In double-blind testing between experimental designs in which the subject (listener) has the control of the switch (and therefore it's timing in changing (supposed) sources), and tests in which the control of the change - or bogus change - is arbitrarily determined by some electronic timer, irregardless of what the timing duration is set at and irregardless of whether the duration is varying or fixed. This is important when determining subtle differences.

But there is LOTS MORE...

If you really want to scientifically test capacitors, you have set up an experimental design where you build two circuits, "identical" otherwise, and when you do the double-blind test, ask the subject to identify, "Which sounds "better"?".

Doing this and switching which circuit is connected to the "A" choice and which is connected to the "B" choice and retesting on more than one day is necessary to ensure you are really measuring what you THINK you are measuring. And set your criteria for real differences at 95% or so if this is supposed to be measuring something other than the bell curve.

IS it really the capacitor dielectric, or is it the sound of the relay bounce which is different between activate and deactivate? Or if you use analog switches with a fade-in or out, do the time-constants match exactly? Subtle gain differences? People CAN hear small loudness changes, which doesn't take much of a component mismatch (especially gain differences with transistors, tubes, and FETs). Frequency response differences controlled by capacitor value differences (you DID MEASURE the values and match them, right)?

Even if it appears you have found a sonic difference, you THEN have to desolder the caps and swap circuit boards (resolder the caps from "A" to "B" and vice-versa) and repeat with the same subject, including asking "Which choice sounds better?" to be sure that the "better"-sounding choice moves with the capacitors and is really because of the capacitors and NOT some other variable (especially gain differences, which are often perceived as tonal differences, NOT as loudness differences). It's a #$^@ of a lot of trouble, and...

FWIW, when I have thought I found measureable differences, it has sometimes - in fact often - been due to other variables, as in, it turns out to be that the "better-sounding capacitors" in the test circuit doesn't move like it "should" with the capacitor swapping. In low-level circuitry, I can't tell the dielectric means anything in the signal chain or the power supply bypassing (unless you have something gross like RF oscillations because an op isn't bypassed adequately: electrolytic plus small value ceramic from each supply terminal to ground and across the supplies).

HEY, what ARE we hearing as "a little bit better"...? Can you say EXPECTATION EFFECT? If I or any other human being can tell a difference, and you ask, "Which is "better-sounding", you'll get an answer. It's the human thing to do.

Sometimes even when people could discriminate A-B from A-A or B-B, they still didn't always AGREE as to which circuit was the "better-sounding" one. By this I mean they didn't always agree BOTH
1) with other listeners or
2) even express which capacitor sounded "better" consistently (with themselves) when asked on different days!!!!!!!!!!!!! (i.e., one day film sounded better, another day ceramic or electrolytic sounded better) Perhaps it correlates with humidity, temperature, or phase of the moon....  
I suppose someone could pursue it further... maybe for rap music or bluegrass a certain dielectric would appeal more to fans of that genre... :roll: :lol:

Focus your energies elsewhere folks. I have. Go by room available, derate the voltage a bit, and buy what you believe to be a durable part (well-anchored leads, looks well-made, not too pricey).
FWIW. YMMV, etc.

[Mark Hammer]

Floyd Toole, an acoustic physicist at the National Research Council here in Ottawa (although he must be retired by now) had a terrific article in the Journal of the Audio Engineering Society (or *maybe* it was the Journal of the Acoustical Society of America) some 20 years back or more on designing listening tests for loudspeakers.  If I remember correctly, the gist was that you really needed 3 points of comparison so that the test item would occur randomly in contrast to either of two other sources.  The problem is that there are both expectation and contrast effects to contend with, and simply flip-flopping between source A and source B doesn't get rid of that very well.

I have a well-worn photocopy of that paper somewhere in a binder in the basement.  I need to look it up.  It's a classic.

[Doug H]

HEY, what ARE we hearing as "a little bit better"...? Can you say EXPECTATION EFFECT? If I or any other human being can tell a difference, and you ask, "Which is "better-sounding", you'll get an answer. It's the human thing to do.
.

On the other hand, when I turn a tone control "down", I hear a less pronounced high freq response. It's not because I turned the control "down" and expected it to get dull; It's because I actually hear it. When I step on the stomp switch of my octavia, I hear this note that is an octave higher than the note I'm actually playing, as compared to my other fuzzes. It's not because I expect to hear it- it's because it is actually there.

In the same way I have heard differences between film and cer caps when swapped out on the breadboard, all other things being equal and no desoldering required. Have I found some sort of "tone silver bullet" with a particular type of cap? No. Are certain caps "better" than others? No, not usually, IMO. I take it on a case-by-case basis and use what I happen to prefer at the particular time, in the particular circumstance.

Doug

[zeppenwolf]

I have a well-worn photocopy of that paper somewhere in a binder in the basement.

Beware Of The Leopard!

Eden

[saros141]

Yup, I'd agree that overall, a cap's true value accounts for way, way more than its composition; however, two caps with the exact same measured value but different dielectric construction can sound perceptibly different.  A cap meter can only tell you so much.  

I've done hours of cap-swapping, it's a fun part of the creative process.  Using those clips with the syringe-like action, I can get 'em in and out in a couple seconds.  I use very good listening equipment that I know well, and there are times I've been nonplussed when I "expected" to be wowed.  When it comes to "better", I'm with Doug: it's case-specific.  

The psychology of it shouldn't really be troubling, as far as I'm concerned.  Caps are candy.  I'm not building military or lab equipment.  (I like Vitamin Q's though)  :wink:

[bwanasonic]

On the other hand, when I turn a tone control "down", I hear a less pronounced high freq response. It's not because I turned the control "down" and expected it to get dull; It's because I actually hear it. When I step on the stomp switch of my octavia, I hear this note that is an octave higher than the note I'm actually playing, as compared to my other fuzzes. It's not because I expect to hear it- it's because it is actually there.

Yes, but the Emperor's Clothes story does not deny the existence of clothes, just questions the hype around the Emperor's outfit.  :wink:

Kerry M

[Mark Hammer]

And indeed, that is what modern psychometric techniques are largely designed to assess, not only *whether* something is different, but precisely how *much* it is different to the human perceiver.  Expectation effects can produce the appearrance of differences when they don't exist, but even when they do exist, expectations can artificially enhance or diminish them.  

Contrast effects are separate from expectations.  Here, the perceived differences would be a function of what came immediately before.  What came before might result in an actual physical adaptation (e.g., how bright an LED looks in a room just made dark after you've had illumination from a 40W lightbulb or a room with a quintet of 100W bulbs), or a perceptual/psychological adaptation from some reasonably long exposure.  This is why it is common for there to be 3 stimuli to compare instead of just two.  That way, the third "neutral" stimulus can be inserted between the two stimuli one is really interested in.  For example, when comparing tastes, researchers will have the test subjects rinse their mouth out with distilled water before they move from taste A to taste B.

Obviously, there are plenty of instances where such controls are entirely unnecessary.  Turn everything on a Blackfire to 10, and there is no possible way anyone would perceive a difference between effect and bypass signal due to "expectations".  On the other hand, when true potential differences are small, and nuance is an issue, such as the difference between the taste of 1/4 vs 3/8tsp of salt in a 10-gallon drum of distilled water, or a "transparent" compressor set for close to unity gain and very light compression vs bypass, expectation and contrast effects can play a much bigger role in what people think they hear/sense.

Essentially, any time where you're in the diminishing returns zone is cause for using fussier evaluation techniques.

In the case of caps, my sense is that any cap-type effects that do exist will almost always be sensitive to application and signal bandwidth.  I *expect* cap type to matter when it is in the signal path in a mic preamp for a device applied to acoustic signals going to a mixer.  I don't expect it to matter quite so much when it shunts distortion fizz to ground in a 10W practice amp.

[MR COFFEE]

Folks,
I have heard those differences, too, on a breadboard, but was quite embarassed when challenged to tell which was which BLIND when someone else was at the breadboard changing the caps!

That's how I got into doing a more scientific comparison. That's also how I realized that with a really good A-B comparison where I controlled the switch, I could detect pretty subtle stuff I couldn't detect reliably (95-100%) when someone else was plugging the caps into my breadboard. All of which convinced me that such sonic comparisons are VERY subtle and when asked to tell the difference in repeated trials on different days...AAARRRGGGHHH! I was personally embarrassed and several of my friends were also chagrined when they called one "better" one day and the other "better" the next. I might have lost a friend or two if I hadn't promptly told them I was equally certain I "knew" what I was hearing....

If you think your ears are "more sensitive" than both mine (hey, mine have been rock and rolled and ARE getting old  - I'm 53), AND those of several of my musician and audiophile friends, have your girlfriend do the swapping - without giving you any feedback - and have her pull one out and put the same one back in periodically, even several trials in a row, just to make it REALLY scientific - and tell her not to tell you how you are doing, just to write them down as you go. Tell her you want to see if she can really fool you as to which is which. Don't ask her between days. Then have her do it again a few times on several days over a couple of weeks. THEN look at what you "can" and "can't" hear. As I say, it was a humbling experience for me!!!!!!

When I worked up a switching test I could control, I did better at detecting subtle differences. I liken it to the difference between listening to a recorded guitar and playing a guitar through an amp where you "feel" the interaction of the guitar and amp at high volume. But we are talking subtle stuff here folks, aren't we!!!!

If you get different results, PLEASE share it here at Aron's.

BTW, I CAN hear the difference of quality film caps in SOME mic preamps with VERY clean recording of SOME kinds of music my ears seem to detect subtle stuff with better (like folk or Irish acoustic stuff), AND using big bucks microphones (like Neumanns), but it still isn't day and night. That's through electrostatic headphones.

And even IN the signal chain, I'm not sure if I can really tell a pure film cap from a properly-biased electrolytic with a decent-sized film cap in parallel with it. It's very close if not the same in a good discrete preamp design. I THINK the DC servo preamps that skip the big coupling caps sound better, but I might be fooling myself.

It is SOOO much work to tease out what I THINK I can hear from what I AM hearing (it's horrible to get a good experimental design with mic preamps unless they use op amps only, and they are not the really the better sonic designs IMHO).

I finally gave up trying to tell whether I was hearing capacitor construction differences or other things because, as I say, it just isn't worth it to try to prove their ISN'T an audible improvement. It costs LOTS of time and energy to build the necessary equipment to really determine WHAT I am hearing, and if I try to convince others it doesn't really matter it  just "shoots my (own) business in the foot" in the end.

I guess I got into it for a while because it bruised my ego a bit when I tried double-blind listening tests... And hey, get this, ...my discrimination went down 20% if I faced away from my wife while she swapped the capacitors :lol:  (I WASN"T "peeking" or even looking at her; and I was STILL seated directly in front of the speaker both ways). Maybe it made me nervous; I dunno.

Use just ONE speaker or really good headphones (like electrostatics) if you want to find out what you CAN hear (which is pretty amazing - it just may not be the capacitor dielectric!!!!!!!) Use TWO speakers, and if you move your head a centimeter you hear something different!!!!

FWIW, YMMV, etc.  :oops:  

[Mark Hammer]

Actually, you know what would be nice?  A self-serve blind-test program so that people can do their own testing at home and offer up suggestions based on some certainty about the audibility of whatever it is they wish to suggest.

Here's how it could work.

You wire up a pair of some sort of relay or solid state switch to your computer's parallel port.  This will not carry audio but will allow you to run cables to somewhere far enough away from the computer that you won't have to do your listening tests against a backdrop of monitor and computer fan noise.

The software initially asks you how many trials/tries you want to do and how long each tryout period should be.

The software enables the switch for one version of the signal, for some designated time, then completely kills the signal for some predetermined delay/time-out period.  Then it kicks the signal in again and lets you play.    

It does this for some number of times, and asks you to indicate via button press whether it was path A or B during each time-out period.  It does not show you a record of what you pressed until you're done.  The choice of path A or path B is made randomly by the computer, although it keeps a record of which path was selected.

At the end, you get a readout, printout of how frequently you correctly identified the signal as being path A or B.  Presumably one could insert some sort of rating into the time-out period as well.

The time-out period is important for a few reasons.  First, imposing a delay makes it harder to compare two different versions of the signal (which could be cap A vs cap B).  The delay forces you to rely on memory which will exaggerate expectation effects.  Why is this good?  Because it provides a stronger test such that if you ARE accurate, there is really something of substance going on here.  The interval is also good because if you use a relay, it will mean that even if the same signal is presented twice or more in a row, it always starts with an audible relay click, so no "Clever Hans" (look it up in Google) artifacts to subtly tip you to what it is you're hearing at the moment.

[Paul Marossy]

I still maintain that in audio circuits that use a 9V battery, such as the typical guitar effect, the type of cap you use in the audio path is pretty inconsequential. There a few, but very few exceptions to this, IMO. The only one that is fairly obvious is an electrolytic cap vs. a film, but even then, I wonder how many could really be able to consistently be able to pick out the electrolytic cap in a blind taste test.

Some people say that they can most definitely hear a difference with this cap vs. that cap. I bet in a blind taste test, most of these people wouldn't be able to consistently tell which one is which. The same goes for carbon comp resistors in a 9V circuit.

But, things do change (some) when you are in a tube amp with B+ voltages of 300-400 volts. Then the carbon comp magic comes out, and I'm sure the differences in caps become much more apparent, too.

Just my opinion, based on my own experiences. Most of the time, if I try this cap vs that cap (or opamp or transistor for that matter), I start second guessing what I am hearing, because the differences are so subtle that I wonder if I really hear anything at all. But, I do think that Mr. Coffee has a point. If you are the one building a circuit, and testing and listening to it and you believe that a film cap is the very best and all other types suck, then of course a film cap is always going to sound better, to you. But, try it out blindfolded, and let others listen to it, and I'm sure there would be a lot of surprises!