As Rod Serling used to say on The Twilight Zone, "Offered for your consideration"... are two capacitors. One, a 3 uF polypropylene, with 0.02% dielectric absorption, and a dissipation factor of about 0.0013. The other, a 3 uF polyester (Mylar), with 0.17% dielectric absorption, eight times worse, and a dissipation factor of 0.019, more than ten times worse.
To make the values exactly equal, and this is very important, a small polyester was paralleled with the large polyester. The above measurements included this.
One end of each cap was connected to a GR 1390B noise source, and the other end of each cap terminated to ground with a 1% metal film 20K resistor. The common lead of the caps was driven with 2 VRMS of random noise covering the entire audio band.
A high gain differential amplifier (Tek 1A7A) compared the signals across each 20K resistor. If there was any difference at all between the signals passed by these two caps, it would show up. The results? The residual common mode of the amplifier with both probes connected to the same cap was about 100uVRMS, plus a trace of 60hz pickup.
Measuring between the caps gave the same 100uVRMS signal, plus the same trace of 60hz pickup.
That means that any difference between the two caps due to DA and DF was less than 0.005%. Considering the CMRR of the amplifier, a lot less. It also means that any difference from unknown mechanisms is also less than that, because the test doesn't care where the difference comes from. IMO, this test is pretty representative of how coupling caps are used.
Now, maybe there's some condition of non-linear source or load impedance where a larger difference would show up (I gotta give people something to hold on to), but logic says if the signals are the same, the sound has to be the same. IMO, if less than 0.005% doesn't qualify as "the same", and you can hear it, caps are the least of your worries.
If you want to see larger differences, just mismatch the cap values a bit. The low frequency difference will skyrocket simply because the RC constant is slightly different. Very small mismatches shouldn't affect the sound, but it's a variable that has to be considered when caps appear to sound different. Another way to see larger differences is to test against an electrolytic or a ceramic!
Ok, did I listen to them? Not yet- too much junk on the test bench from running the test. But I will. The difference is that now I don't expect to hear any difference, unlike the frame of mind I'd be in if I had great expectations due to the DA and DF differences.
No doubt some will consider this test bogus and useless- my flameproof underwear is on, so give me your logic as to why.
To make the values exactly equal, and this is very important, a small polyester was paralleled with the large polyester. The above measurements included this.
One end of each cap was connected to a GR 1390B noise source, and the other end of each cap terminated to ground with a 1% metal film 20K resistor. The common lead of the caps was driven with 2 VRMS of random noise covering the entire audio band.
A high gain differential amplifier (Tek 1A7A) compared the signals across each 20K resistor. If there was any difference at all between the signals passed by these two caps, it would show up. The results? The residual common mode of the amplifier with both probes connected to the same cap was about 100uVRMS, plus a trace of 60hz pickup.
Measuring between the caps gave the same 100uVRMS signal, plus the same trace of 60hz pickup.
That means that any difference between the two caps due to DA and DF was less than 0.005%. Considering the CMRR of the amplifier, a lot less. It also means that any difference from unknown mechanisms is also less than that, because the test doesn't care where the difference comes from. IMO, this test is pretty representative of how coupling caps are used.
Now, maybe there's some condition of non-linear source or load impedance where a larger difference would show up (I gotta give people something to hold on to), but logic says if the signals are the same, the sound has to be the same. IMO, if less than 0.005% doesn't qualify as "the same", and you can hear it, caps are the least of your worries.
If you want to see larger differences, just mismatch the cap values a bit. The low frequency difference will skyrocket simply because the RC constant is slightly different. Very small mismatches shouldn't affect the sound, but it's a variable that has to be considered when caps appear to sound different. Another way to see larger differences is to test against an electrolytic or a ceramic!
Ok, did I listen to them? Not yet- too much junk on the test bench from running the test. But I will. The difference is that now I don't expect to hear any difference, unlike the frame of mind I'd be in if I had great expectations due to the DA and DF differences.
No doubt some will consider this test bogus and useless- my flameproof underwear is on, so give me your logic as to why.
Cool test. I like the idea of the differential signal measurement. I proposed doing a similar test on speaker cabales in another thread.
If there are differences, your test should show them. And a random noise would seem a good source. How about trying it with music? That's what it would be used for anyhow, right? A musical signal would show differences too, if there are any.
The only "flaw" in the test I can think of is that since the caps were tied together on one end, any distortion would be transmitted from one cap to the other and also reflected back to the source. Possible? I don't know - maybe not, but it's worth thinking about.
If there are differences, your test should show them. And a random noise would seem a good source. How about trying it with music? That's what it would be used for anyhow, right? A musical signal would show differences too, if there are any.
The only "flaw" in the test I can think of is that since the caps were tied together on one end, any distortion would be transmitted from one cap to the other and also reflected back to the source. Possible? I don't know - maybe not, but it's worth thinking about.
This illustrates my experiences over the years. I have been involved in a lot of blind listening tests involving caps, and what we found in every case was that polyester caps perform well for most audio applications. I know that there will be many that will flame me because of this, but this is what we have found. We assembled a group of experienced listeners, using a variety of systems (at different times, usually), and the results have always been the same. Most listeners can hear the difference between an electrolytic and a film cap, but once film caps are compared, no reliable identification could be made. We compared generic polyester and ploypropylene caps to the very expensive 'audiophile' caps, and no consistant identification could be made. For this reason, I feel that if you are using (even generic) film caps in your system, dont waste money on expensive varieties unless the rest of your system is already optimised. Put your money where it will really pay off, such as bi-amping.
Hi Hornlover,
I wouldn't recall my experience to be the same (was selling "audiophile" caps for 3 years), caps do differ in their construction, and while I agree that trying to do A/B test to identify which one is which maybe a little hard,especially at different listening sessions on differing system.
But there is identifiable audible difference, switching capacitor into circuit shows this more clearly.
Many generic cap do perform well, but the more sophisticated construction/material caps also have their use in more critical application. (more expensive is not necessarily better constructed/material ) some generic one might do well, but the manufacturer wouldn't guarantee that, they can change the dielectric to same material but lower quality ones without warning. We can not expect the metalization for 5 USD caps and 50 Cents one to be of same quality.
The problem with the so-called "audiophile" caps is that the manufacturer who really make the good ones do not want to reveal what they did to improve the caps, and do not break in ( burn in is better term here ) the caps before shipping them, even worse is that they do not tell people how to use them and in what circuit.
So the user have no idea which one did which things better, many only to be confused after their expensive purchase of "some" famous brand that in fact no better than generic ones.
Yes, it would be good to spend money on Bi-amping , but for people who already have Bi-amp-ed system, or want to improve other part of their system would benefit from better quality caps.
Speaking of Bi-amping for example, using metalized polyester inside active crossover is a very bad idea. generic one tend to develop short/noisy that can't "self-heal" in low power/voltage circuit. film and foil type is best in this case.
Now I invite people to see the differences in caps, instead of listening
, try changing your DVD player video coupling caps. Make sure the caps have been properly broken in, or it might lead to sub-optimal comparison. Yes the size of the caps in this application is big, yes it is expensive, yes the result is different.
If anybody have physically tiny metalized polyester/polypropylene in their amp zobel network, do yourself a favor and change them to more appropriate ones, this one is critical, and also that tiny resistor.
I'm not defending "audiophile" caps, I am defending "good quality caps", I only say that cap have differences, some is better quality and constructed than others, and must be used on the correct application. Vishay make caps and that's not "audiophile" brand, unfortunately that doesn't make their caps cheap
Now I wouldn't comment on resistor differences
better ask knowledgeable RF and digital people about this 
I wouldn't recall my experience to be the same (was selling "audiophile" caps for 3 years), caps do differ in their construction, and while I agree that trying to do A/B test to identify which one is which maybe a little hard,especially at different listening sessions on differing system.
But there is identifiable audible difference, switching capacitor into circuit shows this more clearly.
Many generic cap do perform well, but the more sophisticated construction/material caps also have their use in more critical application. (more expensive is not necessarily better constructed/material ) some generic one might do well, but the manufacturer wouldn't guarantee that, they can change the dielectric to same material but lower quality ones without warning. We can not expect the metalization for 5 USD caps and 50 Cents one to be of same quality.
The problem with the so-called "audiophile" caps is that the manufacturer who really make the good ones do not want to reveal what they did to improve the caps, and do not break in ( burn in is better term here ) the caps before shipping them, even worse is that they do not tell people how to use them and in what circuit.
So the user have no idea which one did which things better, many only to be confused after their expensive purchase of "some" famous brand that in fact no better than generic ones.
Yes, it would be good to spend money on Bi-amping , but for people who already have Bi-amp-ed system, or want to improve other part of their system would benefit from better quality caps.
Speaking of Bi-amping for example, using metalized polyester inside active crossover is a very bad idea. generic one tend to develop short/noisy that can't "self-heal" in low power/voltage circuit. film and foil type is best in this case.
Now I invite people to see the differences in caps, instead of listening
, try changing your DVD player video coupling caps. Make sure the caps have been properly broken in, or it might lead to sub-optimal comparison. Yes the size of the caps in this application is big, yes it is expensive, yes the result is different.If anybody have physically tiny metalized polyester/polypropylene in their amp zobel network, do yourself a favor and change them to more appropriate ones, this one is critical, and also that tiny resistor.
I'm not defending "audiophile" caps, I am defending "good quality caps", I only say that cap have differences, some is better quality and constructed than others, and must be used on the correct application. Vishay make caps and that's not "audiophile" brand, unfortunately that doesn't make their caps cheap
Now I wouldn't comment on resistor differences
better ask knowledgeable RF and digital people about this 
Well John, I'm a bit baffled. I read that article when it was first published, and have read it again. It's basically the test I'm doing. The only significant difference is that I haven't varied the impedance, though I did use various sources. There's a rule of testing that says "test it the way you use it", which is why I chose the values I did. I can certainly force a difference to be seen, but I could just as well do that by using a standard (traditional) 4-arm bridge, and looking at the residual on a scope instead of a tuned null detector once the bridge is supposedly "balanced". They never are, you know. I don't think there's any reasonable disagreement as to what's going on, just with my opinion that the differences are below any possible audible threshold if they're buried in the noise floor. Now, if the caps were used at a low impedance level, say a crossover network, the results might be different. I didn't test that, and have no opinion on the matter (though I avoid electrolytics in crossovers if at all feasible).
Regards,
Conrad
Regards,
Conrad
Conrad, your resolution is not good enough. You MUST be sure of your common mode rejection with frequency, AND you must VERY CLOSELY match the RC time constants. This is why we use an instrumentation IC op amp with 120dB common mode rejection.
If you want to see nothing at all, why bother to measure in the first place?
For the record, mylar vs polypropylene is easy. polystyrene vs polypropylene is tough.
If you want to see nothing at all, why bother to measure in the first place?
For the record, mylar vs polypropylene is easy. polystyrene vs polypropylene is tough.
John, I can force conditions where I can see a difference, but IMO, it's a red herring. (The 1A7A is a matched diff amp of excellent specs BTW, especially if you take care to adjust it properly) Both our methods of testing are fundamentally flawed anyway. D=omegaRsCs, or Rs/Xs if you prefer. We know from measurements that capacitors tend towards a semi-constant D over a wide frequency range. Not exact, but it doesn't vary all that much. Thus, Rs *must* vary with frequency. That means that these circuits can only be balanced for one frequency, and will always show a difference for caps of differing D and a non-sinusoidal waveform.
Anything will look "bad" compared to Teflon, polystyrene, or polypropylene, but no 'X-factor" has been revealed, only the difference caused by conventional capacitor parameters. And, sure, with enough sensitivity you can even see the difference between those three. Non-linearities in things like electrolytics will show up, but they also do when you look at the residual in any 4-arm bridge method. Not sure about DA, but I suspect it's not much of a factor with reasonable frequency AC signals- little soakage occurs per cycle. Thus, if I want to get down to insanely low values of D, it's easier to just use a bridge designed for the purpose, remembering that the reference cap in most bridges will be a very low DF mica, with high DA, yet it has no real influence on the results.
Still, an interesting test would be to find two capacitors of exactly the same value, and exactly the same DF, but of different dielectric types. Measure those against each other in your circuit, and there should be little if any difference. IMO again, whatever difference does show up will be below audibility for any decent film caps, and I'm not aware of any properly conducted listening tests that suggest otherwise (exact matched levels, blind, etc blah blah). I'm not saying there is zero measurable difference on lab equipment, just zero difference in practical application.
Going a bit OT, I once breadboarded a circuit for a research application where capacitor defects were an issue. It involved using two identical caps, one in the circuit of interest, and one in a bridge configuration that matched the operating conditions of the circuit. The bridge was nulled, and the residual amplified. The amplified residual was then injected into the circuit of interest, cancelling out the flaws of the capacitor. It was too sensitive to drift in the balance to be practical as anything other than a lab curiosity, but the concept is interesting.
Anything will look "bad" compared to Teflon, polystyrene, or polypropylene, but no 'X-factor" has been revealed, only the difference caused by conventional capacitor parameters. And, sure, with enough sensitivity you can even see the difference between those three. Non-linearities in things like electrolytics will show up, but they also do when you look at the residual in any 4-arm bridge method. Not sure about DA, but I suspect it's not much of a factor with reasonable frequency AC signals- little soakage occurs per cycle. Thus, if I want to get down to insanely low values of D, it's easier to just use a bridge designed for the purpose, remembering that the reference cap in most bridges will be a very low DF mica, with high DA, yet it has no real influence on the results.
Still, an interesting test would be to find two capacitors of exactly the same value, and exactly the same DF, but of different dielectric types. Measure those against each other in your circuit, and there should be little if any difference. IMO again, whatever difference does show up will be below audibility for any decent film caps, and I'm not aware of any properly conducted listening tests that suggest otherwise (exact matched levels, blind, etc blah blah). I'm not saying there is zero measurable difference on lab equipment, just zero difference in practical application.
Going a bit OT, I once breadboarded a circuit for a research application where capacitor defects were an issue. It involved using two identical caps, one in the circuit of interest, and one in a bridge configuration that matched the operating conditions of the circuit. The bridge was nulled, and the residual amplified. The amplified residual was then injected into the circuit of interest, cancelling out the flaws of the capacitor. It was too sensitive to drift in the balance to be practical as anything other than a lab curiosity, but the concept is interesting.
Bob Pease reported some sucess compensating DA in S/H caps:
http://www.national.com/rap/Application/0,1570,28,00.html
also:
http://www.intusoft.com/nlhtm/nl65.htm
http://www.national.com/rap/Application/0,1570,28,00.html
also:
http://www.intusoft.com/nlhtm/nl65.htm
Hi Conrad. A question about process, was the spectrum of the differential examined? Your note on the 60 Hz component implies 'yes' but it wasn't explicitly stated. If so what was revealed? With the help of solid state CCS loads tube amplification stages with a 2nd harmonic 65 dB down, 3rd well below that and nothing above are possible. If the spectral ‘difference’ between caps, assuming the worst, is a tight band of harmonics at -85 dB off the right hand side of the frequency scale (as I've measured of one poor sounding amp) it’s still arguable the coupling cap defines the sound of the circuit. I’m willing to bet a cluster of -85 dB harmonics from 5th to 15th far outweighs the sonic thumbprint of a -65 dB 2nd. The Tek 1A7A might not be sensitive enough to reveal all meaningful differences.
A suggestion for expansion, a 2 volt excitation signal doesn't come close to mimicking the full range a coupling capacitor sees in normal tube circuit duty. Another forum member describes driving capacitors in series with an eight ohm power resistor directly from the output of a power amp and typically hearing the bodies ‘sing’. This might be one non-linear effect a 2 volt noise test doesn’t capture. DC bias might reveal (or hide) another.
A final suggestion, consider high order low-pass filtering the excitation signal and examining the residual differences in the stop band. This might provide greater insight into what's happening. BTW, it's good to finally see a test in which one end of the cap isn't tied to ground, shunting the signal the test is intended to capture.
A suggestion for expansion, a 2 volt excitation signal doesn't come close to mimicking the full range a coupling capacitor sees in normal tube circuit duty. Another forum member describes driving capacitors in series with an eight ohm power resistor directly from the output of a power amp and typically hearing the bodies ‘sing’. This might be one non-linear effect a 2 volt noise test doesn’t capture. DC bias might reveal (or hide) another.
A final suggestion, consider high order low-pass filtering the excitation signal and examining the residual differences in the stop band. This might provide greater insight into what's happening. BTW, it's good to finally see a test in which one end of the cap isn't tied to ground, shunting the signal the test is intended to capture.
Rdf- good stuff. I didn't do SA, and that would almost have to reveal more. I'm building up my test arsenal with another computer, so I should be able to do decent SA via an external sound card soon. The diff amp has an output, so I should be able to start with that. My assumption, not actually tested, is that once anything is down in the noise floor of my (listening) system, it just doesn't matter. OTOH, I'm also working at reducing both the distortion and noise floor of my system, so the bar does get raised. You mentioned our thermionic friends, which have their own requirements, but I'm pretty much concerned with solid state inputs in the 10-20kohm region. I probably won't mess with tubes again unless I fall into some deal on a prebuilt amp. I used to have modified Dyna Mark IIIs, but sold them many years ago for what seems like a song today.
Re: Re: Another cap comparison
What exactly do you think needs Correction in Linear audio coupling applications?
Bob Pease’ DA correction is based on his linear approximation of DA as applied to highly nonlinear S/H circuits where the effect is especially clear with multiplexed input signals
Using Bob's same 6 RC branch linear approximation model of the DA of a mylar cap it is easy to show that the frequency response variation from that of an ideal cap ranges from >1 mdB at 20 Hz to <1 udB at 3 KHz with Bob's 1 uF Mylar cap model and 100 KOhms of input impedance that would be common in a audio coupling applications, with ss amplifiers at lower impedance the deviation from the ideal C response is still 10X less than the DBT threshold of 0.1 dB frequency response variation at the -3dB RC response point, much less in the pass band
unless you have some new data showing ppM frequency response variations are audible, I have to conclude that the linear model of DA does not explain why different film dielectric capacitors could be audibly different in active filters or coupling applications at common impedances
I'm also curious about why people who do have decent equipment want to handicap it while trying to make sensitive measurements; "ground" is a human applied label, not a fundamental E/M concept, the label can be “moved” without affecting circuit operation
Moving the "ground" from one side of the signal generator to the other in the RC coupling circuit test usefully removes the voltage source from the measurement, reducing the CMRR requirement and allowing a much more sensitive measurement within the limitations of common test equipment - when we are interested in differences it makes sense to arrange the measurement to measure the difference - not the common part
Hartono said:jcx :
"Bob Pease reported some sucess compensating DA in S/H caps"
yes, unfortunately we can not use that method for signal coupling caps.
Cheers
Hartono.
What exactly do you think needs Correction in Linear audio coupling applications?
Bob Pease’ DA correction is based on his linear approximation of DA as applied to highly nonlinear S/H circuits where the effect is especially clear with multiplexed input signals
Using Bob's same 6 RC branch linear approximation model of the DA of a mylar cap it is easy to show that the frequency response variation from that of an ideal cap ranges from >1 mdB at 20 Hz to <1 udB at 3 KHz with Bob's 1 uF Mylar cap model and 100 KOhms of input impedance that would be common in a audio coupling applications, with ss amplifiers at lower impedance the deviation from the ideal C response is still 10X less than the DBT threshold of 0.1 dB frequency response variation at the -3dB RC response point, much less in the pass band
unless you have some new data showing ppM frequency response variations are audible, I have to conclude that the linear model of DA does not explain why different film dielectric capacitors could be audibly different in active filters or coupling applications at common impedances
rdf said:
I'm also curious about why people who do have decent equipment want to handicap it while trying to make sensitive measurements; "ground" is a human applied label, not a fundamental E/M concept, the label can be “moved” without affecting circuit operation
Moving the "ground" from one side of the signal generator to the other in the RC coupling circuit test usefully removes the voltage source from the measurement, reducing the CMRR requirement and allowing a much more sensitive measurement within the limitations of common test equipment - when we are interested in differences it makes sense to arrange the measurement to measure the difference - not the common part
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.