I just read Morgan Jones amplifier book. Page 218 to 222 talk about different type of capacitors. He explained about the "d" of the cap which caused by the polar molecules of the dielectrics. That the loss is between upper audio frequency to low RF. It is being regard as a series resistance that cause loss at the frequency band.
My question is for tube circuits. The signal path of tube circuit are usually high impedance circuits. Input impedance usually in >10kohm. What difference does it make if there is a few ohms in series with the cap in the signal path? Why is it important to get caps with low "d" for signal path that a few ohm loss is nothing?
I can see it's important for filter caps, but not for signal caps.
I can see this maybe a little more important in SS amp as the circuit impedance are lower than tube circuits.
I design guitar amps and effect pedals, I actually make it a point to use ceramic caps if all possible. I really don't hear the different in guitar amps.
My question is for tube circuits. The signal path of tube circuit are usually high impedance circuits. Input impedance usually in >10kohm. What difference does it make if there is a few ohms in series with the cap in the signal path? Why is it important to get caps with low "d" for signal path that a few ohm loss is nothing?
I can see it's important for filter caps, but not for signal caps.
I can see this maybe a little more important in SS amp as the circuit impedance are lower than tube circuits.
I design guitar amps and effect pedals, I actually make it a point to use ceramic caps if all possible. I really don't hear the different in guitar amps.
Last edited:
High-k ceramic caps are fine for use a coupling capacitors in wireless sets and guitar amps, and anywhere else where high-fidelity is not a design aim. For hi-fi they will introduce some distortion due to their non-linear capacitance. Low-k ceramics are fine as they are linear but they tend to be huge if you need serious amounts of capacitance.
Almost any film cap has sufficiently low losses for audio use. The issue is dielectric linearity - and again most are plenty good enough. An exception is polyester/Mylar which is best used only in non-critical points as it is slightly nonlinear.
Almost any film cap has sufficiently low losses for audio use. The issue is dielectric linearity - and again most are plenty good enough. An exception is polyester/Mylar which is best used only in non-critical points as it is slightly nonlinear.
Why does dielectric linearity matter in a coupling capacitor that is supposed to have hardly any signal voltage across it?
It doesn't matter, except for the cap which sets the LF rolloff. In some designs this cap is the one which feeds the output stage grid, so signal levels are quite high. Interestingly, some people using some designs can hear that polyester caps sound slightly different but they wrongly ascribe this difference to superiority rather than minor inferiority.
I read the article posted by woody. The testing was done by RC network with R=1K. BUT in tube amp, grid leak resistor for small tubes are usually 1M. That is 1000 times larger than 1K!!! The effect is going to be a lot smaller.
I do see problem with coupling caps, yes the voltage across the cap is very small compare to the grid leak resistor of the following stage at the pass band as it supposed to be. The problem lies with the upper and lower pass band frequency when the impedance of the coupling cap is high. eg, for 0.1uF coupling cap. At 20Hz, the reactance is about 80Kohm. That is not like a short circuit with a 220K grid leak resistor of the power tube.
I don't see there's a problem in the more critical higher frequency, reactance of the coupling cap is much lower at higher frequency, effect of change of capacitance at high frequency is really not that important.
I do see problem with coupling caps, yes the voltage across the cap is very small compare to the grid leak resistor of the following stage at the pass band as it supposed to be. The problem lies with the upper and lower pass band frequency when the impedance of the coupling cap is high. eg, for 0.1uF coupling cap. At 20Hz, the reactance is about 80Kohm. That is not like a short circuit with a 220K grid leak resistor of the power tube.
I don't see there's a problem in the more critical higher frequency, reactance of the coupling cap is much lower at higher frequency, effect of change of capacitance at high frequency is really not that important.
Last edited:
For many years, people kept using caps that deviated from optimum in at least two ways: non-linear distortion in large amounts like most ceramics and tantalum caps., or high in Dielectric Absorption like Ceramic, electrolytic, or even Mylar.
We measured this and documented it decades ago. Light loading only reduces DA, but it does not completely eliminate it. Just because some people think that virtually all caps sound OK is not necessarily so. My clients use the best caps that they can find. Recently Silmic caps that are silk based electrolytics look like a good compromise, when you need larger values for coupling caps, yet have no space available.
We measured this and documented it decades ago. Light loading only reduces DA, but it does not completely eliminate it. Just because some people think that virtually all caps sound OK is not necessarily so. My clients use the best caps that they can find. Recently Silmic caps that are silk based electrolytics look like a good compromise, when you need larger values for coupling caps, yet have no space available.
Just because some people think that coupling caps do not form part of simple potential dividers does not mean that it is not so.john curl said:
DA is, I understand, well modelled by a linear array of high value resistors and capacitors so cannot contribute to distortion (but may cause a trivial change in LF frequency response). However, all these issues have been argued over many times in other threads so it might not be helpful to repeat them all here in response to Alan's 'innocent' question.
leaving the Jung/Marsh "capacitor tasting" article as the only ref is playing to the naïve subjectivist, audiophile nonsense
if you want to know a big part of why all those dielectrics, constructions of caps were there for Marsh to make up his smorgasboard, look to Ma Bell's requirements, engineer's response to objective circuit requirements and realities of materials, fabrication limitations that make for so many different types of capacitors
Bateman's Capacitor Sound series is good -even if he still doesn't "deconstruct" the use of DA to rank capacitors for audio
Pease article on Capacitor Soakage does call that into question
if you want to know a big part of why all those dielectrics, constructions of caps were there for Marsh to make up his smorgasboard, look to Ma Bell's requirements, engineer's response to objective circuit requirements and realities of materials, fabrication limitations that make for so many different types of capacitors
Bateman's Capacitor Sound series is good -even if he still doesn't "deconstruct" the use of DA to rank capacitors for audio
Pease article on Capacitor Soakage does call that into question
You mean Jung/Marsh is not universally accepted? I thought those are scientific measurements on distortion.
I use as much ceramic caps as possible in my guitar amps and guitar pedals, They sound very good. But guitar electronics is all about distortion, even the clean sound is about distorted clean. An amp without distortion sounds awful for guitar!!! Triode push pull output stage is universally considered awful sounding.
I use as much ceramic caps as possible in my guitar amps and guitar pedals, They sound very good. But guitar electronics is all about distortion, even the clean sound is about distorted clean. An amp without distortion sounds awful for guitar!!! Triode push pull output stage is universally considered awful sounding.
the "controversy" is that the implied "knowledge" of listening effects re cap type have never been shown in DBT testing
the Jung/March article has only a single blind test talking point - no data, scores, ranking system explanation or controls cited
more recent:
Capacitor "Sound" in Microphone Preamplifier DC Blocking and HPF Applications: Comparing Measurements to Listening Tests
Author:Gaskell, Robert-Eric
Affiliation:McGill University, Montreal, Quebec, Canada
AES Convention:130 (May 2011)Paper Number: 8350
Publication Date:May 13, 2011
the Jung/March article has only a single blind test talking point - no data, scores, ranking system explanation or controls cited
more recent:
Capacitor "Sound" in Microphone Preamplifier DC Blocking and HPF Applications: Comparing Measurements to Listening Tests
Author:Gaskell, Robert-Eric
Affiliation:McGill University, Montreal, Quebec, Canada
AES Convention:130 (May 2011)Paper Number: 8350
Publication Date:May 13, 2011
Thanks
They are talking about sound test with different caps in DC blocking or HPF. But is the distortion data published by Jung/Marsh valid?
From where I look at, the main problem for DC blocking is the low frequency where at 20Hz, the reactance of the cap is very high because of the limitation of how high a value of a non polar cap you can get. At higher frequency, the reactance is low that the effect is very small.
They are talking about sound test with different caps in DC blocking or HPF. But is the distortion data published by Jung/Marsh valid?
From where I look at, the main problem for DC blocking is the low frequency where at 20Hz, the reactance of the cap is very high because of the limitation of how high a value of a non polar cap you can get. At higher frequency, the reactance is low that the effect is very small.
Yes, it's the voltage divider effect. Just look at a 0.1u DC blocking cap that drive the power tubes with grid leak resistance of 220K. At 20Hz, reactance of 0.1u is 1/(2\pi f C)= 80Kohm. So the divider is 220/(j80) which is significant.
At 200Hz, it would be 220/j8, at 2KHz it would be 220/j0.8. Notice there is a "j" there that represents it's imaginary.
To find phase and amplitude, you need to use complex number to calculate. At 200Hz, the divider output is 220/(220+j8)=(48400-j1760)/48464=0.9987-j0.0363. This is the divider output which is in all do respect very small. Phase shift is tan^{-1} (0.0363/0.9987)=2.1 deg. Please double check my work, this is just a simple RC divider network.
This is calculate using capacitance only. If you use the data of Jung/Marsh that change of capacitance of ceramic is say 1.5%. So the result in distortion much much lower than the number above.
Call me innocent, but I am looking at the numbers and look at the weight of each distortion contribution. I am staying away from subjective listening, only on numbers.
Bottom line is IF the change of capacitance is due to frequency only, it's not that bad. BUT if capacitance change with voltage, then you are going to have even harmonics.
At 200Hz, it would be 220/j8, at 2KHz it would be 220/j0.8. Notice there is a "j" there that represents it's imaginary.
To find phase and amplitude, you need to use complex number to calculate. At 200Hz, the divider output is 220/(220+j8)=(48400-j1760)/48464=0.9987-j0.0363. This is the divider output which is in all do respect very small. Phase shift is tan^{-1} (0.0363/0.9987)=2.1 deg. Please double check my work, this is just a simple RC divider network.
This is calculate using capacitance only. If you use the data of Jung/Marsh that change of capacitance of ceramic is say 1.5%. So the result in distortion much much lower than the number above.
Call me innocent, but I am looking at the numbers and look at the weight of each distortion contribution. I am staying away from subjective listening, only on numbers.
Bottom line is IF the change of capacitance is due to frequency only, it's not that bad. BUT if capacitance change with voltage, then you are going to have even harmonics.
Last edited:
high K ceramic do have large V coefficients
wouldn't recommend them in audio signal coupling, even Mylar would be expected to be better if you can get enough uF
but even few % distortion at 20 Hz that only happens on the Hi Pass slope may not be easily audible with no tonal musical instrument having a fundamental that low, even +/- 20 Hz IMF products will be lumped in each critical band for everything higher
measured distortions, interpreted in light of circuit, signal and psychoacoustic sensitivity do indicate some cap dielectric, construction choices make sense if accuracy, "audio transparency" is your goal
its just the audiophile guru "story" caps, things like silver foil in beeswax insanity that should be avoided
wouldn't recommend them in audio signal coupling, even Mylar would be expected to be better if you can get enough uF
but even few % distortion at 20 Hz that only happens on the Hi Pass slope may not be easily audible with no tonal musical instrument having a fundamental that low, even +/- 20 Hz IMF products will be lumped in each critical band for everything higher
measured distortions, interpreted in light of circuit, signal and psychoacoustic sensitivity do indicate some cap dielectric, construction choices make sense if accuracy, "audio transparency" is your goal
its just the audiophile guru "story" caps, things like silver foil in beeswax insanity that should be avoided
Last edited:
I have to experiment that when time comes. I can tell you it does not make a damn of a difference in guitar amp that I can tell. I actually make it a point to use ceramic caps whenever possible. Common Cheap ceramic caps seems to stop at 0.022uF for 630V, so I use all ceramic caps up to 0.022uF......yes, including one used in the tone stack. To me, in guitar electronics, it's all hot air.
Even in pedals that are SS and low voltage, I use 0.01u and 0.1u ceramic bypass caps in most of the DC blocking use.
BTW, how do I tell whether the ceramic cap is high K dielectric?
I was thinking about this, why is it even a problem for DC blocking caps. In the pass band, the signal voltage across the cap is very small and doesn't change much at all!!! If the signal voltage across the cap change too much to be a problem, use a larger value cap!!!
Call me ignorant, as described in post #15. Use a larger value cap instead of paying exotic amount for a "high quality" low value cap for DC blocking, I just don't see that much of an issue even with ceramic caps. Using the calculation in post #15, if I increase the value to 0.22uF, the calculation will be good for less than 100Hz!!
Last edited:
- Status
- Not open for further replies.