Generally the best number of caps in a circuit is the right number as suggested by the circuit architecture. People using more or fewer than this are usually creating trouble for themselves, although they might not always realise this. Not being a faradaphobe I don't try to eliminate capacitors, but I may try to eliminate unnecessary phase shifts and put the LF rolloff in the best place (usually the input).
A slightly different way (though basically equivalent) to look at it is that the looking-in impedance is the parallel combo of the cathode resistor with (RL + rp)/(1+mu). This again shows the usefulness of CCS loading of the plate.
Of course, when vague claims about the sound changes are made which are unsupported by data, one has two possibilities- poor engineering or a refusal to comprehend how the ear-brain works. We've focused on the former, but one does have to consider the latter as a likely possibility in some circumstances, especially when amplified (as it were) by sociological factors.
Well, Steve does explain how he performed the tests, and I agree the scope photos are not so great, but he does discuss his results in general. You should be able to evalute whether his approach is a valid one.
Glad to see you admit that the numbers might not capture everything ... now there's progress.
Lots of people here dismissing the evidence of people with far more kudos (Steve Bench) for their work in innovative audio. Brave.
Shoog
Lots of people here dismissing the evidence of people with far more kudos (Steve Bench) for their work in innovative audio. Brave.
Shoog
Yes. Reason why CCS can sound good #19: it lets you calculate the cathode bypass cap using the popular (but wrong) formula which everyone quotes, and still get the right answer! Of course, this does require a CCS which is good down to subsonic frequencies (preferably DC) so some 'gyrator-like' designs will not do. Oops, I'm being boring and introducing engineering again.SY said:
Kudos does not count; evidence does. SB may have seen hysteresis, but he presents no evidence for this. The most likely explanation for an X-Y trace opening in the middle is phase shift; hysteresis would require non-linearity at the ends too. He may have seen this, but his pictures don't show it.
The design is always the most important..
Component choice after ensuring the design functions..
The difficult bit to understand is why two polypropylenes seem to be different..
Meridian DAC's also show a difference with a cap change..not value just type.
You don't need to spend megga money..I hate to say it but Obbligatos are quite good..Some Mundorf..Silmic II are not expensive parts..Caerafine are not expensive either..Not compared to conductive plastic pots for Volume etc.. Resistor types..this is still engineering..I'll admit its difficult to quantify..what works well in what position with what topology..
Regards
M. Gregg
Component choice after ensuring the design functions..
The difficult bit to understand is why two polypropylenes seem to be different..
Meridian DAC's also show a difference with a cap change..not value just type.
You don't need to spend megga money..I hate to say it but Obbligatos are quite good..Some Mundorf..Silmic II are not expensive parts..Caerafine are not expensive either..Not compared to conductive plastic pots for Volume etc.. Resistor types..this is still engineering..I'll admit its difficult to quantify..what works well in what position with what topology..
Regards
M. Gregg
Last edited:
A really good point and a real problem when selecting best sounding parts.
When possible, you can, sometime, subjectvely distinguish between hidden or added detail comparing the part with no part at all (part replaced by jumper or removed according the case).
Maybe / maybe not [re: getting caps out of signal loops] ... Its challenging technically to "get them out" - and kind of fun to see the different amplifier stage designs that do. Yet, at least in Tube work ... they're kind of a necessary evil.
As to [DF96]'s notion that "could just be harmless phase shift" ... yes, I imagine so. As you pointed out, the oscilloscope is a lousy instrument to do this kind of thing, especially since the nominal range of acoustically meaningful analog is at least 80 dB (40 dBV, or 1:10,000). Oscilloscopes are at best able to show a few tenths-of-a-percent (1:500 typical). Much better would be to use a reference A->D converter to do the measurements, then subtract in digital domain the response from the theoretically-perfect capacitor-resistor-integrator response, and scale it to actually show the nonlinearities.
And I again think we should remember [DF96's] observation: that many capacitors are manufactured with really high intolerances: +50%/-20% is typical for power-supply caps. (since there is relatively few issues with having "too much" capacitance) At the low end, phase shift becomes an interesting side-effect.
GoatGuy
As to [DF96]'s notion that "could just be harmless phase shift" ... yes, I imagine so. As you pointed out, the oscilloscope is a lousy instrument to do this kind of thing, especially since the nominal range of acoustically meaningful analog is at least 80 dB (40 dBV, or 1:10,000). Oscilloscopes are at best able to show a few tenths-of-a-percent (1:500 typical). Much better would be to use a reference A->D converter to do the measurements, then subtract in digital domain the response from the theoretically-perfect capacitor-resistor-integrator response, and scale it to actually show the nonlinearities.
And I again think we should remember [DF96's] observation: that many capacitors are manufactured with really high intolerances: +50%/-20% is typical for power-supply caps. (since there is relatively few issues with having "too much" capacitance) At the low end, phase shift becomes an interesting side-effect.
GoatGuy
To distill down that above point, you are saying that because one hears something but cannot quantifiably measure it, then it doesn't exist, or is a product of "bad design", whatever that is. There are alot of things about our senses that are not quantifiable. I am not sure we have a good way to quantitively measure taste and smell, and yet there are people with developed sensory capabilites and the ability to perform consistent and repeatable qualitative assessments. Not to say they cannot confuse themselves under certain conditions. And don't forget we are talking about attempting to measure an electrical analog of an acoustical event that has to be transformed back to an acoustical event. Do we even know what we are attempting to measure or quantify when a very complex musical signal interacts with a circuit. At what point does the extremely low level information in that signal that may be what I am trying to hear as "different" is itself masked or distorted by the circuit, even a "well designed" circuit. What if that information is below the resolution of the measurement device? Can we correlate between measured results and audible results at all levels?
Some people have extra-ordinary senses. They can demonstrate them fairly reliably under test conditions. A few audiophiles may fit into this category; I suspect most do not. Evidence for this:
- 'obvious night and day differences' disappear under test conditions - were they ever really present?
- known distortions favoured by those who use FX boxes (e.g. nonlinearity, frequency response shaping) also seem to be favoured in a weaker form by some audiophiles - e.g. 'tube buffers', other demonstrably poor circuits yet receiving glowing reviews.
A circuit does not know that it is being fed a very complex musical signal. It just sees a voltage, which it is expected to amplify. All we need to do is ensure that it does this as perfectly as reasonably possible. I simply do not accept that circuits with clear design faults can do this better than circuits without such faults, yet this is what people seem to claim. We don't know everything, but we do know some things. I would be more convinced if people expressed audible preferences between different but equally good circuits, suggesting that there are further levels of detail to be explored, rather than preferring known poor circuits over known good ones (suggesting a preference for error).
- 'obvious night and day differences' disappear under test conditions - were they ever really present?
- known distortions favoured by those who use FX boxes (e.g. nonlinearity, frequency response shaping) also seem to be favoured in a weaker form by some audiophiles - e.g. 'tube buffers', other demonstrably poor circuits yet receiving glowing reviews.
A circuit does not know that it is being fed a very complex musical signal. It just sees a voltage, which it is expected to amplify. All we need to do is ensure that it does this as perfectly as reasonably possible. I simply do not accept that circuits with clear design faults can do this better than circuits without such faults, yet this is what people seem to claim. We don't know everything, but we do know some things. I would be more convinced if people expressed audible preferences between different but equally good circuits, suggesting that there are further levels of detail to be explored, rather than preferring known poor circuits over known good ones (suggesting a preference for error).
I have experience with this.
JENSEN copper foil Alu tube capacitor 0.47uf 630V | eBay <= to me these sound awful, I notice they are also being sold as "guitar caps" which makes perfect sense to me.
After using MKP cap types that is where I am staying......or transformer coupling next maybe....🙄
Unfortunately [dmcgown], there's another quad of factors at play, which renders this whole topic fundamentally unresolvable.
(1) - pride of knowing
(2) - pride of authorship
(3) - pride of abilities
(4) - lack of quantitative, sufficient-resolution data acquisition
Pride of Knowing - pros and amateurs (esp.) "grow into" knowing the reasons, supposed and real, for different aspects of something so particular as "sound-of-capacitors". We collect ever more information, some substantiating, some refuting the surprisingly large bowl of interlinked theories and counter-theories. Much of the argumentum ... comes from the "pride of knowing", in particular, defending one's personal framework of knowing. Mental swordplay. Sharpening up the opus of theory into a stance, a statement. Festooning one's arguments with acidic adjectives, "totally", "utterly", "absolutely", "incredible (more/less/different)", "remarkable", and all their opposites.
Pride of Authorship - then comes "building things", and being proud of the result. I remember building my first PieceOShyt amp, way back in high school. I had no helpers, no mentors: had to figure out the whole damned thing, from sources-of-components, to necessary design changes to substitute unavailable parts, to power supplies, chassis, knobs, connectors, grounding, etc. I was totally proud of "authoring" this device, and my opinion of its prowess was indelibly stained by the work-of-making. Only some time later did I realize it, and my glee, were misplaced.
Pride of abilities - how many times has the fruitless argument(s) here devolved into "well, my ears can hear this!". Uh... OK. I've noted it especially with people having slightly (or even significantly) impaired hearing! I had a buddy that was deaf-from-birth in one ear. Man would he rail up, down, left and sideways that he could DEFINITELY hear the difference in capacitors, and resistors and even power supply rectifiers. However, he couldn't hear you if you were on his "wrong side", at even a modestly joyful social situation. I think "pride in abilities" (or related: "Pride in Discriminative Fetishes") is amongst the most central factors that leads to endless, fruitless discussion.
Lack of hi-rez, quantitative data acquisition - Like my observation about the oscilloscope study of capacitor non-linearity, and echoing a response I wrote a few days ago citing (my theory) that much of the things of an amp that fall under [dynamics, response, resolution, sound-stage, 'effortless' and so on], are various nuances of things that are damned difficult to quantitatively measure: speaker response vs. amplifier damping; nonlinearity of gain especially near 1:20th to 1:1000th full-power input levels.
In short, I have noted an almost complete lack of "building measuring instrumentation" in order to Scientifically ferret out the supposed effects. No, I don't mean hooking up an oscilloscope to get a X:Y plot that shows a flat line, but instead, acquiring (and keeping really good care of) "reference microphones" so that A/B tests can be run - and simultaneously recorded to admittedly imperfect, but at least reproducible recording equipment for later post-test analysis.
Because, in the audio field, its ALWAYS about the ears, isn't it? "How does it sound?" and so on. Well, I for one think ... that "how it sounds" is darn important, but ultimately is sterile (and always therefore suspect) on its own. "Better" must always be relative, to "worse" - and both of them should be recorded to sufficient precision so as to be able to say what-all really was different, so that the underlying co-factors can be elucidated.
Enough. I'm sure some capacitors, at various spots in an audio-signal path, can slightly-to-markedly affect the reproduction of the audio content. I would just love to see the whole thing quantified as A:B plots. I especially would like the armies of independent sound consultants out there to fill up their ateliers less with massive quantities of now-obsolete reproduction equipment, but instead, well kept reference microphones, digitizing samplers, and a good double-blind A/B switching system to remove as much as possible "the guy" from the equation.
GoatGuy
(1) - pride of knowing
(2) - pride of authorship
(3) - pride of abilities
(4) - lack of quantitative, sufficient-resolution data acquisition
Pride of Knowing - pros and amateurs (esp.) "grow into" knowing the reasons, supposed and real, for different aspects of something so particular as "sound-of-capacitors". We collect ever more information, some substantiating, some refuting the surprisingly large bowl of interlinked theories and counter-theories. Much of the argumentum ... comes from the "pride of knowing", in particular, defending one's personal framework of knowing. Mental swordplay. Sharpening up the opus of theory into a stance, a statement. Festooning one's arguments with acidic adjectives, "totally", "utterly", "absolutely", "incredible (more/less/different)", "remarkable", and all their opposites.
Pride of Authorship - then comes "building things", and being proud of the result. I remember building my first PieceOShyt amp, way back in high school. I had no helpers, no mentors: had to figure out the whole damned thing, from sources-of-components, to necessary design changes to substitute unavailable parts, to power supplies, chassis, knobs, connectors, grounding, etc. I was totally proud of "authoring" this device, and my opinion of its prowess was indelibly stained by the work-of-making. Only some time later did I realize it, and my glee, were misplaced.
Pride of abilities - how many times has the fruitless argument(s) here devolved into "well, my ears can hear this!". Uh... OK. I've noted it especially with people having slightly (or even significantly) impaired hearing! I had a buddy that was deaf-from-birth in one ear. Man would he rail up, down, left and sideways that he could DEFINITELY hear the difference in capacitors, and resistors and even power supply rectifiers. However, he couldn't hear you if you were on his "wrong side", at even a modestly joyful social situation. I think "pride in abilities" (or related: "Pride in Discriminative Fetishes") is amongst the most central factors that leads to endless, fruitless discussion.
Lack of hi-rez, quantitative data acquisition - Like my observation about the oscilloscope study of capacitor non-linearity, and echoing a response I wrote a few days ago citing (my theory) that much of the things of an amp that fall under [dynamics, response, resolution, sound-stage, 'effortless' and so on], are various nuances of things that are damned difficult to quantitatively measure: speaker response vs. amplifier damping; nonlinearity of gain especially near 1:20th to 1:1000th full-power input levels.
In short, I have noted an almost complete lack of "building measuring instrumentation" in order to Scientifically ferret out the supposed effects. No, I don't mean hooking up an oscilloscope to get a X:Y plot that shows a flat line, but instead, acquiring (and keeping really good care of) "reference microphones" so that A/B tests can be run - and simultaneously recorded to admittedly imperfect, but at least reproducible recording equipment for later post-test analysis.
Because, in the audio field, its ALWAYS about the ears, isn't it? "How does it sound?" and so on. Well, I for one think ... that "how it sounds" is darn important, but ultimately is sterile (and always therefore suspect) on its own. "Better" must always be relative, to "worse" - and both of them should be recorded to sufficient precision so as to be able to say what-all really was different, so that the underlying co-factors can be elucidated.
Enough. I'm sure some capacitors, at various spots in an audio-signal path, can slightly-to-markedly affect the reproduction of the audio content. I would just love to see the whole thing quantified as A:B plots. I especially would like the armies of independent sound consultants out there to fill up their ateliers less with massive quantities of now-obsolete reproduction equipment, but instead, well kept reference microphones, digitizing samplers, and a good double-blind A/B switching system to remove as much as possible "the guy" from the equation.
GoatGuy
Indeed, and that's what I did for a living for many years (taste, smell, and haptics). The key words are "consistent" and "repeatable," under conditions where the sensory variable is the only cue. In the case of sound, I use the phrase "ears only."
But this is way off track here. Engineering an amplifier so that its output is audibly indistinguishable from its input using ears only is straightforward engineering, and that's what we're talking about. The analysis is likewise extremely straightforward, as you've seen here.
Try using transformers, where they are essential only, eg. in the amp output.
The magnetic hysteresis loop is saying loudly that they are not linear devices.
I do not understand the new fashion of using transformers everywhere.
Careful design and good MKP capacitors is far better. IMHO. 😉
Last edited:
I think Steve Bench did see dielectric hysteresis.
It is very common that the scale factors affect the look of the graphics.
It is also notably that Mr Bench not accustomed to scientific publications, is not clear, at least for me, nor the experimental setup neither a theoretical introduction, etc.
This is also consistent with the conclusions of this thread, if the applied voltage on the capacitor is small enough, the dielectric hysteresis curve approaches a straight line.
Evaluation of macroscopic polarization and actuation abilities of electrostrictive dipolar polymers using the microscopic Debye/Langevin formalism - Abstract - Journal of Physics D: Applied Physics - IOPscience
Attachments
Last edited:
Originally Posted by DF96
Actually, I suspect even my correction is wrong. There ought also to be a contribution to R from the mu-transformed load impedance, so you end up with something like R = Rk || ( 1/gm + RL/mu ).
A slightly different way (though basically equivalent) to look at it is that the looking-in impedance is the parallel combo of the cathode resistor with (RL + rp)/(1+mu).
You guys will probably prove to be right, but I'm puzzled: shouldn't the load impedance appear as being in parallel with rp rather than in series? I tend to think in terms of an inner "transconductance engine", and this issue may be where my little mental model breaks down, so I'm unsure.
Thanks, as always,
Chris
Last edited:
On second thought, that can't be right. Sorry. But can the external load be more important than its influence on gm?
I need to work on this.
Thanks,
Chris
I need to work on this.
Thanks,
Chris
- Status
- Not open for further replies.