Big film caps and noise

I'd brought this up in responses to other threads... I thought I'd document it here seperately with a hard measurement. The ubiquitous advice with respect to caps is 'film is always better-- if you can afford it'. That's not actually true; you'll probably not be able to measure the THD contribution from just about any cap, ceramic and tantalum excepted. Also, huge film caps pick up noise. Case in point below.

I've been constructing a small fleet of gainclones for some self-powered monitors. These are intended to be refined, thoroughly debugged 'milking what you can from a great little chip'. I'll post a schematic if AndrewT demands, but for now suffice to say it is a pretty standard little gainclone with a Panasonic FC blocking cap for Ci and an Orange Drop .22uF input coupling cap (big dipped polypropelyne film/foil).

The initial prototype fired up and with the inputs shorted produced the following self-noise graph:

withcap.jpg


[The vertical scale is dBV mesaured at the driver. The gray line is the measurement, the red line is the sampler self-noise reading]

Good-- but not great. That's just audible in a quiet room.

The foil cap turns out to be coupling hum and hash noise coming from the PSU. Although shileding the input cap helps, eliminating it entirely or using an electrolytic yields:

nocap.jpg


...much better. That's the output of 30dB of amplification from a chipamp with +/-35V rails. The only difference is avoiding using a obscenely large film cap.

Presented for nothing more than illustrative/documentation purposes.
 

sharpi31

Member
2005-10-20 12:57 pm
I find these results very interesting.

I (like many others) have found teflon caps to have the cleanest HF reproduction of any type of cap I've tried. But teflon caps are huge compared to other types... hmmm...

All the teflon caps I've tried have been russian military types with a metal case. Perhaps the case helps to shield the cap against noise...?

Either way, non-shielded teflon caps also seem to be liked for their clean HF reproduction (not that I've ever listened to one).

Please don't think I'm questioning your results - I'm genuinely intrigued :)
 
sharpi31 said:

I (like many others) have found teflon caps to have the cleanest HF reproduction of any type of cap I've tried. But teflon caps are huge compared to other types... hmmm...

All the teflon caps I've tried have been russian military types with a metal case. Perhaps the case helps to shield the cap against noise...?

A metal can provides shielding if the can is grounded to the chassis, yes.

The problem with claims like 'teflon has the cleanest HF' and 'film is more transparent than electrolytic' is that in small signal coupling applications, I can measure zero-- and I mean zero-- performance difference between them up to 50kHz. The tools can tell the capacitors apart when measuring to performance extremes that have nothing to do with audio (amps of current flow, nonlinearity at 200V, impedence differences at radio frequencies), but in an audio signal coupling application the measurement tools say there is no difference whatsoever (measurement depth of 20 bits).

Of course, at high power or other applications there are real differences. But for small signals, all the film and electrolytic caps are ridiculously overqualified.
 

sharpi31

Member
2005-10-20 12:57 pm
But for small signals, all the film and electrolytic caps are ridiculously overqualified.

Again, I have no grounds to doubt your measurements but I have always considered it incredibly easy to hear the differences between different coupling capacitors. Judging from the prevalence of threads discussing these differences, it seems I am not alone.

Xiphmont: Ignoring your measurements, do you actually hear a difference between different types of cap?
 
Re: What's the software?

gmikol said:
Slightly OT, but...

I don't recognize the look of the software. Judging by the UI, I'd say it's a Linux-based application?

What is it?

--Greg

Yes, it's a program I wrote years ago when I was doing alot of speaker array building and testing for some local theater groups. I never finished it (just got it good enough to work for what I needed right then). I've been working on it again recently along with a suggested schematic to plug into the soundcard to use along with it. I'll post a thread about it when I get a little farther along.

The source is at http://svn.xiph.org/trunk/spectrum

The version there runs now but is incomplete (that's what I used to make the graph)


AndrewT said:
what does this mean?

I mean that most people are using 600V capacitor capable of full amps of RMS ripple with HF characteristics suitable for use into tens or hundreds of megahertz for audio coupling, an application that doesn't scratch a tenth of a tenth of a tenth of a percent of the capacitor's rated/intended capabilities. The nonlinearities you see caps exhibiting out toward their rated limits are damned near undetectable in an audio application. 'Soakage' is about as real a phenomenon in these caps as 'skin effect' in speaker cables. It exists, but you'll never measure it in audio with even the most sensitive instruments.

Even ceramic caps, which are avoided for some good reasons, show barely detectable THD contributions when used to couple microamps of line-level voltage. The noise coupling and microphonic characteristics of many caps contribute way more to signal degrdation than any other imperfection in the cap. I can easily measure a ceramic cap picking up room noise around it (acting as a condensor mic).

And no, I don't hear a difference between the caps. Honestly, I've never gone looking to hear a difference. I already know my ears are not as sensitive as my measuring tools, and if the tools can't detect a difference, I'm fully confident any difference I hear will be imagined. Placebo effect is strong and unavoidable, even in people who are aware of it. It is human nature to find patterns and detect differences where none exist.
 
Got to agree

I used to use huge film caps for coupling. The US made teflon caps in metal tubes were big faves. The surplus teflons seemed better than the approved boutique films caps. Now use many varients, small electrolytics such as Silmics and BG, 50 volt polycarbonate film, and the lowly Epco mylar.
To my ears, the smaller the coupling cap the better. Decoupling is the same. I use a lot of 20 cent ECR 1% polycarbonates in recent builds. They have mid 80's date shift codes.
I used to throw all steel leaded caps in the trash too. Cannot honestly hear the non linearities of coated steel lead component versus a copper lead. Short lead length may help here.
A better design and layout is much more important than the brand of caps used to build.


George
 
Re: Re: What's the software?

xiphmont said:

<snip> I already know my ears are not as sensitive as my measuring tools, and if the tools can't detect a difference, I'm fully confident any difference I hear will be imagined. Placebo effect is strong and unavoidable, even in people who are aware of it. It is human nature to find patterns and detect differences where none exist.
What makes you think here's something wrong with your ears? In an anechoic chamber, human ears can hear the noise of air molecules bumping into one another (about 0dB) and can tolerate 130dB for short bursts. Measuring equipment with a 130 dB dynamic range is beyond the average engineering lab.

You assume that the quality that makes one type of capacitor sound better than another is THD. There's little evidence to support this assumption, but if true, many have measured significant distortion differences in various capacitor types. See this thread in this forum http://www.diyaudio.com/forums/showthread.php?threadid=5107]http://www.diyaudio.com/forums/showthread.php?threadid=5107
 
Re: Re: Re: What's the software?

hermanv said:
What makes you think here's something wrong with your ears? In an anechoic chamber, human ears can hear the noise of air molecules bumping into one another (about 0dB) and can tolerate 130dB for short bursts. Measuring equipment with a 130 dB dynamic range is beyond the average engineering lab.

There is nothing wrong with my ears. I'm an audio engineer and I have my ears tested yearly. My hearing is excellent.

Human hearing has a dynamic range of 120+dB, but a narrowband signal depth of only about 30-35dB. Therefore my equipment which has a narrowband depth of >130dB/Hz is easily superior to my ears.

hermanv said:

You assume that the quality that makes one type of capacitor sound better than another is THD. There's little evidence to support this assumption, but if true, many have measured significant distortion differences in various capacitor types. See this thread in this forum http://www.diyaudio.com/forums/showthread.php?threadid=5107

Any time invariant nonlinearity or imperfection in a capacitor will show us as harmonic distortion. That's what harmonic distortion is. I've also demonstrated above non-time invariant noise, and how the so called 'audiophile capacitors' may be sabotaging an amp's noise performance.

The link you posted supports everything I said. I said approximately 'ceramic capacitors are not designed for audio, and yet add almost unmeasurable THD'. Despite the fact that ceramic is the single worst kind of capacitor you could choose for audio, the article measures non-negligable THD only in some of the lowest grade (X7R/Z5U) ceramics, and even then it is orders of magnitude under the audible threshold (regardless, I'd avoid these caps myself-- far far better is available for mere pennies and it is correct that imperfections stack up through the signal chain). The C0G ceramic cap measures a -125dB third and that's it. That's just short of undetectable in the lab and not even a bat could hear it.

The article goes on to say "1 uF film/foil polypropylene gives unmeasurable distortion". I'm saying electrolytic is exactly the same, even ones not made for audio.

[edit: actually, I take it back... tantalum can be far worse in practice. A reverse-biased tantalum can fail unexpectedly in which case it will turn into a short circuit and probably explode. Never ever ever reverse bias a tantalum. They also have very low ESR, but will also blow if inrush current exceeds their rating.]
 
xiphmont:
I was asked to design an FM radio adjunct to a computer based home server. The radio was low cost, meant to be included as an alarm clock function. It used a single chip FM receiver, a single chip pre-amp module and a chip output stage producing 3.5 Watts/channel with 2" X 3" cheap speakers. Like most modern devices this was to be a surface mount design, ceramic capacitors were a natural, small and cheap.

Even with such a rudimentary inexpensive clock radio, the difference in sound quality between signal carrying X7R ceramic capacitors and slightly more expensive surface mount mylars was clear (pun intended). As an EE I do not fully understand all the reasons why this is true. I have learned to trust my ears, it's not that modern measuring equipment can't find differences, it's that we actually have little idea of what it is we should measure.

My problem is that you advise people to not even bother to check or listen. You seem to claim they will not, in fact can not, hear a difference.

Again and again the "there is no difference" school provides opinions that are contrary to a whole industry. Audiophiles are not universally foolish financial bumpkins. Most of us spend money for equipment that sounds better and it sounds better even to our non-audiophile acquaintances. This, in spite of quite vocal nay-sayers that simply can't seem to hear what others hear easily.

I certainly agree with the notion that if you can't hear a difference there's little need to spend any excess money. Please advise people to listen and judge for themselves before they commit to the objectivist or subjectivist camp.
 
FWIW, I've been measuring caps for quite a while, and with equipment that easily resolves dissipation and other differences between *any* type of cap. I can't hear the difference between different types of film caps, nor do differential measurements show any difference in signals passed by the various types. I do both hear and measure differences with electrolytics, and avoid them when possible. I suspect that some differences people hear are due to differing frequency response when different values are used. It seems people value Teflon, and are willing to install a smaller value than a design might call for, due to size or availability considerations. It doesn't take a big change in plain old frequency response to change what you hear, and sometimes a change in the low end will be perceived as a change in the high end- it's all relative. You can certainly shield a giant capacitor to reduce pick-up, but then you might end up with a large capacitance to ground in a place where it wasn't desired. IMO, there are no hard and fast rules- you have to consider each case separately and design accordingly.
 
hermanv said:
xiphmont:

Even with such a rudimentary inexpensive clock radio, the difference in sound quality between X7R ceramic capacitors and slightly more expensive surface mount mylars was clear (pun intended).

*shrug* People also claim night-and-day differences for silver cables.

If there was really a difference, you'd be able to measure it and ascertain why one capacitor worked properly in the design and the other didn't. Did you? I'd worry about a design in which one kind of capacitor really did sound different-- unless I knew exactly why and could definitively say the reason was harmless.

There are a couple good reasons aside from microphonics and nonlinearity to either avoid ceramics or to be careful about using them. One is their insanely high Q; despite their flaws, ceramics are in most other ways closer to the ideal capacitor than any other type. They will ring like nobody's business in circuit that is prone to oscillation or not expecting so little dissipative loss. Use a ceramic to couple a very high impedence input and you're probably asking for trouble.

That is not the capacitor's fault. That's the [potential] fault of the designer, a designer who doesn't understand the circuit, doesn't care to understand the circuit, and can't be bothered to check why unusual things are happening. Or tries, fails to understand, shrugs, and goes merrily along.

hermanv said:

Again and again the "there is no difference" school provides opinions that are contrary to a whole industry.

Religion pays really really well. It's a much bigger industry than high end audio. That doesn't make me any more likely to join a church. Audio subjectivism is nothing more than another religion that gets people all whipped up into honestly, deeply believing things that aren't true.

That's not to say you're absolutely wrong about ceramic caps of course-- I wouldn't use them in audio either.

hermanv said:
I certainly agree with the notion that if you can't hear a difference there's little need to spend any excess money. Please advise people to listen and judge for themselves before they commit to the objectivist or subjectivist camp.

The problem is that people are trivially convinced of things that just aren't true every day. Audio especially is full-- brimming, teeming and overflowing-- with quacks who are not dishonest but believe to their very core they can hear things that just aren't there. The subjectivists claim the obvious superiority of silver cables, zero-feedback, etc, etc and it's all bunk. There are real considerations in choosing capacitors appropriate to a given amplifier design, but that's not what drives 99% of the people here to spring for Auricaps.

I don't object to 'listen for yourself'. I object to the chorus of 'Hey, don't believe us, listen for yourself-- oh, and by the way, the one on the left is more expensive." That's all you need to convince most people that the one on the left obviously sounds better, even if they're exactly the same part. Or don't say 'expensive', say 'silver plated for extra tonal purity-- just listen, it's obvious'. Or 'oxygen free copper'. Or 'hand matched and balanced by experienced experts with 50 years of industry experience'. And that one on the right, "well I'm not going to say anything about it, it's OK I guess. Just listen and decide for yourself if the obviously superior item on the left sounds better."

...do you see where I'm coming from? Do you understand my objection to the subjective camp?

[edit: also what Conrad said. Altering the response of a circuit through faulty design or value substitution also can't be blamed on a capacitor. I avoid electrolytics in audio circuits only because so many faulty electrolytics have made it out into the real world recently. Even good vendors like Nichicon have had problems recently.]
 
xiphmont said:

*shrug* People also claim night-and-day differences for silver cables.

If there was really a difference, you'd be able to measure it and ascertain why one capacitor worked properly in the design and the other didn't. Did you? I'd worry about a design in which one kind of capacitor really did sound different-- unless I knew exactly why and could definitively say the reason was harmless.
<snip>
That is not the capacitor's fault. That's the [potential] fault of the designer, a designer who doesn't understand the circuit, doesn't care to understand the circuit, and can't be bothered to check why unusual things are happening. Or tries, fails to understand, shrugs, and goes merrily along.
I hardly wanted to write a white paper on capacitors, I do understand capacitor measurements; dissipation factor, leakage, dielectric absorption, equivalent series resistance, piezo electric effect, thermionic effect etc. I didn't question your professional skills, just your opinion. Please do not question mine without some evidence. I have long experienced that ceramic caps sound bad, there was no need to question why something unusual was happening because it wasn't unusual.

You completely missed my point about measurements, that we don't quite know what it is we should measure. Sine waves, steps and square waves have little relation to music, using those signals to predict how something will sound has not proved particularly fruitful.
 
Assuming equally good construction, I find that my preferences in capacitor selection pretty well track dielectric absorption. I'm a little surprised that comprehensive measurements found that it was essentially the same for all the cap types mentioned.

The importance of DA is not exactly a new observation; see the original Marsh/Jung paper. A very nice test with scope photos was proposed in the 4/85 issue of Audio Amateur by Walt Jung and John Curl.

I "inherited" an unfinished modular high power amplifier design; six 1200W amplifiers in 7 RU including power supply. After dealing with the reliability and performance problems, I had the time to substitute a polyprop for a 1uF X7R. The sound went from nasal/reedy to markedly less colored/smoother. I certainly can say that the reedy sound did not meet any preconceived notion I'd ever had or heard about cap sound.

Aside from some solid Ta's, high K ceramics and some electrolytics, I've never seen an indication that THD is useful in evaluating the sonic quality of capacitors.

DA implies time delay of course, and the ear's ability to detect spuriae varies greatly with time delay.

Do try to hear the differences; bear in mind that if you do not hear the differences, it may be because of a the preconception that there are none. ;) I certainly did not expect to hear such differences, but these things became pretty clear when I got my first pair of Magnepans.

Edit: What matters is not what people say, there are trolls, flamers, and marketers on both sides; what matters is if you listen (with a quite good system) and do or don't hear differences. If not, you save yourself some money. But for you to prove the negative that others truly don't hear the differences is really not possible.
 

sharpi31

Member
2005-10-20 12:57 pm
The problem is that people are trivially convinced of things that just aren't true every day.

None more than those who are satisfied with the results of one (quite possibly flawed) 'scientific' experiment ;-)

Blind faith exists on both sides of the subjectivist/objectivist debate. Having absolute trust in your ability to design and conduct experiments (and then in your ability to interpret the results) is no different from having absolute trust in your ears.
 
hermanv said:
I didn't question your professional skills, just your opinion. Please do not question mine without some evidence.

Well, I did start out by posting hard data. I'm happy to post more :)

hermanv said:

I have long experienced that ceramic caps sound bad, there was no need to question why something unusual was happening because it wasn't unusual.

There's nothing more awkward than to explain to a tradesman with longstanding industry experience that not only is he doing it wrong, he's always done it wrong... This is not pointed at you (you're being rational and polite, I don't want to be rude either), I'm just getting kinda tired of several of the forum's more infamous and prolific subjectivists countering factual arguments backed by empirical data with "I've been doing this a long time, I know better than you."

hermanv said:

You completely missed my point about measurements, that we don't quite know what it is we should measure. Sine waves, steps and square waves have little relation to music, using those signals to predict how something will sound has not proved particularly fruitful.

Myth. Myth myth myth. If you throw linear superposition out the window, most of electrical and signal engineering goes with it. Signals, their combinations and their behaviors are not unknowable mysteries. We understand how signals combine and interact very well. Pure tones, square waves, pink noise, etc, have specific qualities that will stress an amplifier in ways music never will. They're *harsher* tests than music can conjure that expose flaws in glaring, easy to inspect ways.

Any circuit can be explained just about completely with appropriate effort even if a given individual may not be able to do it, Practical models may simplify things to make the math easier to wield. When a model doesn't explain something, it's easy enough to measure and when appropriate, use a more detailed model.

The claim that this is all somehow mysterious and unpredictable is just a lever to sell snake oil. There's money to be made (and egos to be built) using Fear, Uncertainty and Doubt.

I don't have to doubt. If the math says I'm right and the measurements say I'm right, I know I'm right :)

hermanv said:

None more than those who are satisfied with the results of one (quite possibly flawed) 'scientific' experiment ;-)

Blind faith exists on both sides of the subjectivist/objectivist debate. Having absolute trust in your ability to design and conduct experiments (and then in your ability to interpret the results) is no different from having absolute trust in your ears.

Quite right! The scientific method gives you a tool to argue with me: documenting a counterexample. So far, I'm the only one here to post any hard data supporting my argument. If it's easy to demonstrate my data and my argument are flawed, I invite someone here to do so. In fact, I can even tell you exactly where to start (hint: high impedence FET inputs, or high-voltage blocking applications, eg, tube amps. It's certainly possible to come up with design constraints that dictate practical considerations in capacitor choice. But I am and have been from the beginning talking about line level coupling looking into normal impedences.)

Curmudgeon: The tests did not indicate that dielectric absorbption was unmeasurable, just that it had no measurable effect on the audio. It is very easy to tell the difference between the capacitor dielectrics, but unless the coupling application is very weird, the effect is swamped into 'undetectable' by other characteristics of the circuit.
 
Actually, seriously, I would love it if we could all start throwing specific circuits and measurements around-- post your graphs! post your schematics! I say caps have no 'sound' of their own, any coloration in a given circuit is either an intentional understood compromise or just a flaw of the circuit design. The imperfections in the cap are swamped byt he imperfections of the tranducer, the rest of the amp and human hearing itself.

It also seems to me I can present a quickie listening test-- let's agree on a specific piece, and I'll run it through different coupling caps, resample it, and put the unlabelled results up somewhere everyone can listen. You won't know what you're listening to and can make some honest appraisals. Once we've all had a listen, I can reveal each source. This isn't entirely controlled and it can be gamed, but it's probably good enough to shield honest listening evaluation from suggestion effects.
 
I think Elliot talked about DA, and made a pretty convincing argument that it's not a factor. It shouldn't be hard to make up a cap with additional RC sections in series, to give any simulated DA you want. IMO, it won't affect the sound at all, since if the cap is selected properly there won't be any significant AC voltage across it (for coupling caps at least). No voltage = no DA. I've used silver mica caps in my phono preamp, then switched to polypropylene. I couldn't hear any difference, though the big flaw of silver mica is DA, in spite of its low DF, and an RIAA section is a filter, and will have AC voltages across the caps. Unlike most, I did fiddle with the values and a reverse box, so the response was within .1dB or so. Had I blindly changed caps, the response would likely have changed, then all bets are off.