Wow! I go for the 6moons reviews of this stuff, the total copper alone must cost $100's.
As usual they start right in confusing anti-imaging with anti-aliasing, but we know what they mean.
I mis-spoke slightly before, in a capacitor the displacement curent and voltage go hand in hand (I = CdV/dT) so in fact you can look at it either way, still no dV/dT no I. In my example the distorted portion of displacement current in the .1u goes through the parallel combination of .1u and 1u and hence the voltage manifestation of the distortion is attenuated (with 10,000uF VERY attenuated). We soon arrive at one of Ed's examples of -140dB distortion mattering.
Last edited:
I had to look again.
It was John Watkinson and not John Atkinson the author of the excellent book "The Art of Sound Reproduction"
Amazon.com: The Art of Sound Reproduction (9780240515120): John Watkinson: Books
George
It was John Watkinson and not John Atkinson the author of the excellent book "The Art of Sound Reproduction"
Amazon.com: The Art of Sound Reproduction (9780240515120): John Watkinson: Books
George
Wow! I go for the 6moons reviews of this stuff, the total copper alone must cost $100's.
As usual they start right in confusing anti-imaging with anti-aliasing, but we know what they mean.
I mis-spoke slightly before, in a capacitor the displacement curent and voltage go hand in hand (I = CdV/dT) so in fact you can look at it either way, still no dV/dT no I. In my example the distorted portion of displacement current in the .1u goes through the parallel combination of .1u and 1u and hence the voltage manifestation of the distortion is attenuated (with 10,000uF VERY attenuated). We soon arrive at one of Ed's examples of -140dB distortion mattering.
Very quickly the ESR and its linearity dominates in this example. If your source is 160 Ohm and the esr is 50 milliOhms you have an attenuation of around 80 dB. At higher harmonics the esl will dominate and they will increase in level. Perversely, this is where surface mount multilayer ceramics excel (if they aren't generating stuff). And at high frequencies the PSRR of the circuit deteriorates. A large film cap will have much higher inductance and may cause problems if very fast transistors are used with less than the most sophisticated layouts. Introduce digital circuits and a desire to meet FCC and you may have no alternative to the surface mount MLCC's.
I am glad you Scott clarified your statement, and I read this discussion to the end of the last page. Now we can clearly see that sources of distortions can be caused by either non-linearities of capacitors (in case of high DA it can be non-linear ESR as Demian pointed out), or non-linearities of stages that the cap couples. No miracles found.
For additional perversity note that some circuits, e.g., notoriously, low-dropout voltage regulators, have a propensity to oscillate when the output capacitor is too good, i.e., low ESR. So having a lossy part there can work better than a low-loss one. For perhaps better control of the situation one may wish to add some lumped series resistance with the lower-loss capacitors. But of course this does set a lower limit to the amount of attenuation of load current modulation.
Many years ago, I attended a seminar about a new processor chip developed by DEC (remember them?) and although most of the material was far afield from my interests then and now, I was impressed by their insistence on the use of NP0/C0G ceramic bypass capacitors for the multiple power supply lines. And big NP0 is not cheap, requiring many layers and a tightly controlled process to produce defect-free parts. I recall reading about a part for the switching power supply of the international space station, which was about the size of a small loaf of bread and managed, iirc, 270 uF, I've forgotten at what voltage
Brad
Many years ago, I attended a seminar about a new processor chip developed by DEC (remember them?) and although most of the material was far afield from my interests then and now, I was impressed by their insistence on the use of NP0/C0G ceramic bypass capacitors for the multiple power supply lines. And big NP0 is not cheap, requiring many layers and a tightly controlled process to produce defect-free parts. I recall reading about a part for the switching power supply of the international space station, which was about the size of a small loaf of bread and managed, iirc, 270 uF, I've forgotten at what voltage
Brad
That would leave fundamental on the supply, no idea what portion of ESR is non-linear it can all be characterized.
I used to have a bag of "magic" 2.2uF caps from a surplus buy at Raytheon, they simply worked when nothing else did. An even tougher test is the 8 parallel video loads on the AD8010 datasheet, record setting (my
) diff-gain and phase no special caps needed.
I really don't get a lot of this conversation, John. We handily accept that inductor performance differs (and audibly, for audio applications) with a change in the type of iron used. So, too, capacitors whose dielectric---the main determining factor of capacitor performance---has measurably different performance depending on its makeup.
Before I studied capacitors, I listened to them. I found in my listening that, as a general statement, and all other things being equal, teflons were better than polystyrenes, etc. Here's the hierarchy of the main capacitor types to which I listened (better first):
teflon
polystyrene
polypropylene
polycarbonate
polyester (mylar, PET)
electrolytic
Not that I needed such, but I had no measurement orientation to substantiate my personal listening observations, yet these were largely (almost overwhelmingly) corroborated by others who tend to slot capacitor sound in the hierarchy above.
Then Bateman decided to measure the things. His measurements confirmed the above hierarchy.
Notice that capacitor AC fidelity---the degree to which it approaches the theoretical ideal---generally follows its Dk (again, all other things being equal).
The more complete models of esr contain bulk R and the dielectric in series with each other and the K factor is temperature sensitive. The high K dielectrics add more to the esr than film or low K types. In the better film caps (lowest DA) the esr is limited by the bulk metal reistance which is very low.
For pwr supply bypass apps, the series Ls of the cap is very important for high freq component transients to be bypassed. The ceramic - by design - has inherantly low series L. But, as noted in many digital app notes those caps must be used as close as possible to the device (digial IC) or the series L of the pcb trace will be in series with the cap Ls and lower the effective freq range covered (the resonance of the C and Ls is the defining limit of the useful range the Cap can be effectively used as a capacitor). One reason why the MultiCap got a patent was, for its size, the Ls was much lower thus its self resonace was higher... making it more useful in filter apps over a wider freq range.. But still physical size maters in the series L and thus ceramic and small styrenes or silvered mica can do better in that app. Any claims as to hearing a change for the better in the sound is usually due to a lower Z PS over a wider bandwidth.. that might have been needed.
The reason for bypassing with smaller values is first and fundementally to extend the upper freq range which the capacitance is effective.... Once you go to a higher freq than the resonance of the Ls and C the usefulness as a cap deminishes and the circuit is that of an inductor.... not what you want from a cap. When That is the case, the sound often is cleaner and more details are easily heard. There seems to me to be new some spec limit that pwr supply (PS) Z needs to be -- something like at least 60DB below the circuit stage Z to be low enough... for high end or HD sound maybe -80 or more. Notice too that a tube circuit may not need as low a PS Z over wide freq range as SS amp would... Just speculation and a suggestion. --RNM
No magic but more info that makes using the best cap for the application easier to do and more affective.
Last edited:
It works everywhere: the higher is dynamic load resistance, the lower is current.
My friend from Israel who is crazy about capacitors uses in his RIAA correctors air variable capacitors.
Last edited:
Yes they are. Caps need to be chosen for their application. There are places where even a high-K ceramic cap is an appropriate choice (e.g. RF bypassing), and other places where it would be a very poor choice (e.g. setting LF rolloff). People who believe that certain component types are simply 'bad' will, at the very least, produce equipment which is more expensive than it needs to be and may produce equipment which has poorer performance than it could have with wiser choices.serengetiplains said:
Yes, and DA was pointed to and the correlation to listeners opinions is consistant and the same .... SO. I think it would do well for all scientific-engineering minded people to believe in what is heard First. then try to find measurments which are best at showing what is heard. THD is natoriously insensitive test and other tests which show larger problems that would be more audible need to be taken more seriously.
At this point I expect SY to pop up and point out the possible difference between what is actually heard and what is believed to have been heard. And why do people keep on banging on about THD being a poor test - please criticise people for what they believe (and say), not what more simplistic (and easier to criticise) beliefs you wish they had.RNMarsh said:
I understand what you're saying, and I agree, but the point is a bit offside what I was intending to communicate. The type and level of distortion produced by a capacitor's dielectric does not depend on the application chosen for the given capacitor. It depends only on the dielectric and goes by the name DA. Put otherwise, a capacitor's dielectric distorts inherently.
The character of that distortion has been successfully modelled as a complex impedance in the Cole-Cole parameter method. See here:
http://www.cktsim.org/modeling/da.pdf
No! It depends on the signal voltage across the cap, and the extent to which any distortion generated there can find its way back into the signal path. Both of these issues depend critically on exactly where the cap is used in the circuit. I can't understand why so many people seem unable to grasp this basic truth: distortion depends on signal level, and generally the higher the distortion order the more it depends on signal level.serengetiplains said:
My friend from Israel who is crazy about capacitors uses in his RIAA correctors air variable capacitors.
That is brilliant! (I tend to use values of 100pF and about 330pF anyway.) And the variability automatically gives some fitting to pre-RIAA curves. Brilliant!
Please post your friend's name so I can give proper credit in future. Thanks!
Chris
- Status
- Not open for further replies.