Nowhere that i know of, either. I agree with SW - hard to get excited.
However, if you use non-sine waves, what would you find.... maybe that back to back friction fit contact diode that cancels under sine wave tests. [ no i havent done that test on resistors.] Such tests need to be done before we can say fer shur there is nothing audible.
Last edited:
But if they both are caused by the same properties, in such case they can correlate.
I can't see this link being mentioned before, MEASURING CAPACITOR DISTORTION, has some reasonable info ...
I just noted the updates to that material, here: capacitor distortion and part-2.
Frank
I just noted the updates to that material, here: capacitor distortion and part-2.
Frank
Last edited:
Indeed it is feasible in general. It's much more tricky, as the order of operations is important. Klippel demonstrated some processing that backed the (nonlinear) distortion out of loudspeakers, and referred to the technique as a Mirror Filter.
I often wonder how many people really know what linearity means. Such a key concept, so powerful, and although an idealization, still extremely important and useful. Indeed, although every real physical system exhibits nonlinear behavior at some level, to deal with it we are not usually really much further along than to analyze it in little bits, "piecewise linear".
I worked with a guy at UCLA who was hazy on the subject. He had been a night assistant at Lick Observatory and worked with an instrument known as the Robinson-Wampler Image Tube Image Dissector Scanner, which for obvious reasons was dubbed the Superscanner for short. The device had virtually no offset "fixed-pattern" signal (often called, casually, "fixed-pattern noise"), so to calibrate the sensitivity per picture element it was only necessary to illuminate it with a flat field, and use the result to normalize other data. When Dan got to UCLA I explained how data needed to be reduced from the photodiode array spectrometer, which had both non-uniform dark current and a quite substantial amount of fixed-pattern signal, in addition to a pixel-dependent response function.
So it was necessary to subtract a reference scan from a data scan before additional operations, typically division by a flat field with the fixed pattern noise similarly removed. The man just could not grasp why this order was important, or even the need for subtractions at all, even when I used simple numerical examples. It was very frustrating, with him at one point saying in essence, Just do it --- that's what we did at Lick.
Yes, you will have necessary correlation at some fundamental level. Capacitor cannot be separated from capacitor dielectric.
Just to make mattrs a bit MORE fun or confusing for listeners and causes of thier complaints is the following observations from decades ago -->
Most metalized film caps use aluminum for the plate material... this thin plate will quickly oxidize if air gets into it. Then you have aluminum oxide and aluminum. Not a great conductor. Then the ends of the plate - and this is same with foil plates are strayed with molten aluminum droplets to quickly and cheaply attach all the plates together. Then the lead (tinned copper or steel usually) is laid against the sprayed aluminum and sprayed again to attach the lead to the plates. To save metal, the sprayed ends are often masked off so only a strip of the plate ends get attached. Often the lead is barely attached and can be lifted off by your fingers. So, we have contact issues.... issues which may only be an issue with high currents passing thru its connection/contact. But ove time higher esr, lower total C and possible real distortion at certain volatge thresholds.
One dumb stunt done for audiophiles is to put stranded leads on the cap... this is a path way between strands for air to get in and oxidize the plates and contacts.
Most caps are Not hermetically sealed... unless MIL spec'ed to be. The quick test for oxidation , I used, was to measure the capacitance.... it often goes down.... ditto when metalized caps which have had too many over voltages applied and plate area is reduced/burnt off. With very high Z circuits -- like fet inputs and mic cartridges/capsules, the surface leakage of the input cap and pcb et al is very important..... if you touch with your fingers, the cap body, the leakage will not be what you think it should be for the film material used. Ditto high MegOhm value resistors or the super high Victoreen resistors in the GigOhm range. They must be absolutley clean... can be useless for you app by sloppy handing in packaging or in recieving or your own dumb self. Many caps do not use the same film on the outside as the cap is made of... usually it is mylar wrap with the values printed on it. This too has leakage greateer than many better cap films. Electrolytic caps are really poorly made.... air ruitinely gets in and drys everything out but also oxides the connnections... the foil is often only loosely crimped in the first place to the leads. Oxidation over time causes an increae in many undersirable effects. I have suggested many times that all one need to do is replace the caps with new ones of the same brand and get an improvement in pecieved performance.
When it comes to electrolytics on your audio pcb's use only the 105 degree C rated ones - they are a better build and often has lower esr and esr vs temp. Other than that maybe better off with sealed tantalums for stable long life... but not necessarily better performance.
isnt this great, I can go on all day ruining it with variables until no one knows anything for sure about anything. Ouch!
Most metalized film caps use aluminum for the plate material... this thin plate will quickly oxidize if air gets into it. Then you have aluminum oxide and aluminum. Not a great conductor. Then the ends of the plate - and this is same with foil plates are strayed with molten aluminum droplets to quickly and cheaply attach all the plates together. Then the lead (tinned copper or steel usually) is laid against the sprayed aluminum and sprayed again to attach the lead to the plates. To save metal, the sprayed ends are often masked off so only a strip of the plate ends get attached. Often the lead is barely attached and can be lifted off by your fingers. So, we have contact issues.... issues which may only be an issue with high currents passing thru its connection/contact. But ove time higher esr, lower total C and possible real distortion at certain volatge thresholds.
One dumb stunt done for audiophiles is to put stranded leads on the cap... this is a path way between strands for air to get in and oxidize the plates and contacts.
Most caps are Not hermetically sealed... unless MIL spec'ed to be. The quick test for oxidation , I used, was to measure the capacitance.... it often goes down.... ditto when metalized caps which have had too many over voltages applied and plate area is reduced/burnt off. With very high Z circuits -- like fet inputs and mic cartridges/capsules, the surface leakage of the input cap and pcb et al is very important..... if you touch with your fingers, the cap body, the leakage will not be what you think it should be for the film material used. Ditto high MegOhm value resistors or the super high Victoreen resistors in the GigOhm range. They must be absolutley clean... can be useless for you app by sloppy handing in packaging or in recieving or your own dumb self. Many caps do not use the same film on the outside as the cap is made of... usually it is mylar wrap with the values printed on it. This too has leakage greateer than many better cap films. Electrolytic caps are really poorly made.... air ruitinely gets in and drys everything out but also oxides the connnections... the foil is often only loosely crimped in the first place to the leads. Oxidation over time causes an increae in many undersirable effects. I have suggested many times that all one need to do is replace the caps with new ones of the same brand and get an improvement in pecieved performance.
When it comes to electrolytics on your audio pcb's use only the 105 degree C rated ones - they are a better build and often has lower esr and esr vs temp. Other than that maybe better off with sealed tantalums for stable long life... but not necessarily better performance.
isnt this great, I can go on all day ruining it with variables until no one knows anything for sure about anything. Ouch!
That's the most general use of the term, which I find acceptable. The capacitor is doing its DA thing and the device it's supposed to be stabilizing is generating distortion attributable in part to the capacitor's DA wonkiness (technical term). Absent capacitor DA in that circumstance, the distortion thus attributed doesn't exist. That distortion, ergo, is capacitor induced.
Thanks for that info on the stranded leads, I always suspected it and have had them made with solid leads.
Yes, but that then creates confusion when people blame DA for non-linearity. Some people then create further confusion by showing a plot of the exponential time response of a linear first order filter and call it 'distortion'.Wavebourn said:
Could we perhaps start talking about 'DA-correlated' non-linear distortion? Clarity of expression can aid clarity of thought, provided people don't then abbreviate it back to 'DA distortion'.
The document that serengetiplains linked to http://www.cktsim.org/modeling/da.pdf is a very good one,
but There is no mention there to any distortion due to DA. Not even at “Conclusion”
And as jcx and Scott remarked,
the models shown are made of linear components.
Are the graphs of capacitor impedance versus freq. that scary?
As Wavebourn has said, it is the varying impedance of the “ circuit points at which a capacitor is connected to“ that should be investigated as the points of real horror.
In any case, anyone can test different capacitors on a specific part of a real circuit.
Use digital signal files, arbitrary waveforms or real music as excitation signal (repeatable and well indexed). See general plan here Capacitor Distortion Part 2
Record the output of the circuit with a sound card, store the wave file, for each capacitor and signal level.
Listen to the different files.
Compare the FFTs of the files.
Digitally subtract the files, see the FFT of the subtraction.
Many thinks can be tested, practically, under real specific conditions and in less time compared to theorizing ad infinitum. Less chances to develop capacitophobia (Be nice to capacitors).
DF96
Can you please provide a link where these are explained? Thanks
RNMarsh
I am with you here (Old story with restoring activities)
The degradation with time and environmental conditions is the real issue with capacitors.
Different types, different sealing methods, different degraded behavior.
George
but There is no mention there to any distortion due to DA. Not even at “Conclusion”
And as jcx and Scott remarked,
the models shown are made of linear components.
Are the graphs of capacitor impedance versus freq. that scary?
As Wavebourn has said, it is the varying impedance of the “ circuit points at which a capacitor is connected to“ that should be investigated as the points of real horror.
In any case, anyone can test different capacitors on a specific part of a real circuit.
Use digital signal files, arbitrary waveforms or real music as excitation signal (repeatable and well indexed). See general plan here Capacitor Distortion Part 2
Record the output of the circuit with a sound card, store the wave file, for each capacitor and signal level.
Listen to the different files.
Compare the FFTs of the files.
Digitally subtract the files, see the FFT of the subtraction.
Many thinks can be tested, practically, under real specific conditions and in less time compared to theorizing ad infinitum. Less chances to develop capacitophobia (Be nice to capacitors).
DF96
Can you please provide a link where these are explained? Thanks
RNMarsh
I am with you here (Old story with restoring activities)
The degradation with time and environmental conditions is the real issue with capacitors.
Different types, different sealing methods, different degraded behavior.
George
Some actual data
I will simply quote the results from Samuel Groner's thesis.
Distortion of capacitors at +20dBu on the capacitor, 1kHz
Polysyrene LCR components FSCEX 10nF - no visible harmonics above -153.4dB noise floor (none nada zilch)
Vishay MKP1837 PP film - same
Vishay COG/NPO - same, noise floor -142.5dB (ceramic BTW, this was 1nF which I assume is why the noise floor was higher)
Dick, I'm not being a trouble maker and I'm asking nice could you please show me schematically how to put a sine wave through a C with rectifying defects at the input to get the cancellation of distortion.
I will simply quote the results from Samuel Groner's thesis.
Distortion of capacitors at +20dBu on the capacitor, 1kHz
Polysyrene LCR components FSCEX 10nF - no visible harmonics above -153.4dB noise floor (none nada zilch)
Vishay MKP1837 PP film - same
Vishay COG/NPO - same, noise floor -142.5dB (ceramic BTW, this was 1nF which I assume is why the noise floor was higher)
Dick, I'm not being a trouble maker and I'm asking nice could you please show me schematically how to put a sine wave through a C with rectifying defects at the input to get the cancellation of distortion.
Last edited:
Sorry. Assuming the capacitor and dielectric response are symmetrical then the lowest non-linearity orders possible are constant (zeroth order) and parabolic (second order). To get a linear response you would need some asymmetry, either in the construction of the dielectric or the construction of the capacitor. Even-order dielectric non-linearity means that +ve and -ve voltage peaks behave in the same way, which for a waveform means odd-order distortion. Adding a DC bias has the same effect as introducing asymmetry, so even-order distortion becomes possible. This is the same way that an unbalanced LTP can produce even-order, while a balanced LTP can only produce odd-order distortion.gpapag said:
I can't give you a link as these are my own thoughts. I worked for a while with a bunch of materials scientists, and one of the things they measured was dielectric nonlinearity. When non-poled the curves were usually symmetrical about zero electric field. You can make poled dielectrics (as in an electret microphone) but I can't think why you would do this for a capacitor. An exception would be an electrolytic cap.
- Status
- Not open for further replies.