1. read the datasheetAlan0354 said:
or
2. check any colour coding or marking on it
or
3. use experience - if it is physically small but has large value or large voltage rating then it is high-K
Almost all ceramic caps used for coupling in valve audio circuits will be high-K. Anything over a few hundred pF is likely to be high-K.
I am still looking at Jung/Marsh's article. There is something that IMHO quite miss leading. As you can see, the experiment for ceramic caps in Fig.4 is using R=1K and C=0.1uF for the graph. But if you calculate the reactance of a 0.1uF is j80Kohm at 20Hz, j8Kohm at 200Hz, j800ohm at 2KHz. The reactance is very significant compare to 1K resistance. There is a significant voltage dividing ratio in this low frequency region, the reactance DOMINATES the divider network!!! Of cause you see significant distortion that looks scary!!! But if you plot the same graph using 220K resistor, you can use the way I calculate in post #15 to find the divider ratio. You'll see the distortion is down to very low, way more than 220 times lower as you have to use complex number division, it's not the same as simple real number division.
I don't know the % of capacitor change with frequency, but lets just for a moment assuming that the % change of capacitance is the same between 20Hz to 2KHz ( my guess is going to be close). If you use the graph in Fig.4, if grid leak is 220K, you scale the frequency down where you read 2KHz on the graph as 10Hz. OR better 4KHz on the graph as 20Hz!!!! So the valve amp is operating totally out of the high distortion area of the graph even at 20Hz. From reading the graph, it's only about 0.03% at 20Hz!!!
And remember, if you use a bigger cap like 0.22uF or higher, you push the distortion even lower.
Please comment on what I wrote so far. I am not commenting anything on whether you can hear the difference, I only look at the data available and use math to calculate the distortion.
I don't know the % of capacitor change with frequency, but lets just for a moment assuming that the % change of capacitance is the same between 20Hz to 2KHz ( my guess is going to be close). If you use the graph in Fig.4, if grid leak is 220K, you scale the frequency down where you read 2KHz on the graph as 10Hz. OR better 4KHz on the graph as 20Hz!!!! So the valve amp is operating totally out of the high distortion area of the graph even at 20Hz. From reading the graph, it's only about 0.03% at 20Hz!!!
And remember, if you use a bigger cap like 0.22uF or higher, you push the distortion even lower.
Please comment on what I wrote so far. I am not commenting anything on whether you can hear the difference, I only look at the data available and use math to calculate the distortion.
Last edited:
This is the well known practice of how to make a coupling cap work: one can use a nasty dielectric in a coupling cap as long as the time constant is very oversized, and thus the signal voltage across the cap is very low. Most of these errors are voltage related, so if the voltage across the cap is reduced, by increasing the capacitance of the 'sloppy' cap, then the errors are also reduced.
The problem arises when one wants to use a cap for in-band filtering, where there will be significant signal voltage across the cap. In this case, you can't just make the cap bigger to make it more linear, because you actually want a specific time constant.
From my memory of the Jung article(s), this was the point of their ceramic high pass example. Sometimes, you actually want a high pass filter that actually chews on some LF in the signal, so there will be signal voltage across the cap, and you can't just oversize it to make it good. An example would be an 80Hz high pass filter inside of a mixing console designed for live sound - the caps that make up that filter _will_ have signal voltage across them - that's the whole point.
The problem arises when one wants to use a cap for in-band filtering, where there will be significant signal voltage across the cap. In this case, you can't just make the cap bigger to make it more linear, because you actually want a specific time constant.
From my memory of the Jung article(s), this was the point of their ceramic high pass example. Sometimes, you actually want a high pass filter that actually chews on some LF in the signal, so there will be signal voltage across the cap, and you can't just oversize it to make it good. An example would be an 80Hz high pass filter inside of a mixing console designed for live sound - the caps that make up that filter _will_ have signal voltage across them - that's the whole point.
I am mainly talking about DC blocking caps here. I agree that you have to be very careful in using caps for tone stack.
The problem I have is people keep insisting in using the best cap in power amp which mainly serves as DC blocking caps for interstage coupling. Yes, I agree the feedback capacitor has to be of good quality in the NFB loop. People should not just blanket saying avoid ceramic cap at all cost etc. Before people making this statement, they need to specify under what condition you avoid ceramic caps!!! JMHO.
The problem I have is people keep insisting in using the best cap in power amp which mainly serves as DC blocking caps for interstage coupling. Yes, I agree the feedback capacitor has to be of good quality in the NFB loop. People should not just blanket saying avoid ceramic cap at all cost etc. Before people making this statement, they need to specify under what condition you avoid ceramic caps!!! JMHO.
You are quite right. Capacitor distortion occurs when there is significant signal voltage across the cap. If you use a big enough cap, then the cut-off frequency will be so low that there is never a significant signal voltage across it, so no distortion can occur.
The trouble with high-k ceramics is the distortion is quite high even before you get down to the cutoff frequency, typically reaching 0.01% a decade above the cut-off frequency! The distortion is also strongly temperature dependent. With polyester the distortion doesn't usually become significant until you're already a long way below the cut-off frequency (in a CR coupling filter), unless you get a particularly bad sample.
An secondary factor which I don't think has been mentioned is voltage rating. Plastic (and electrolytic) caps with higher voltage ratings usually produce less distortion than similar caps with low voltage ratings. I would guess the same applies to ceramics, though I have not tested it.
Last edited:
True, but people like to have simple answers to complex questions. Much simpler to say "ceramic caps come from the devil" than to say "calculate the corner frequency and ensure it is a decade or two below the lowest LF in the signal". When global NFB is being used it may be necessary to use smaller coupling caps (in order to avoid motorboating) so ceramic should then be avoided.Alan0354 said:
Personally, I would not use ceramic coupling caps in hi-fi but quite happy to use them in a wireless set. That is because I assume the hi-fi will use significant amounts of NFB.
Thanks, I really learn a lot in this thread. Never did understand all these until now.
On a related question. I have a SS power amp that I am experiment with. I am worry the amp blown and the output can rail and burn the speaker. I want to put a back to back electrolytic caps to block the DC for the Woofer. I already have a sub-woofer, so I don't need to have good low frequency response. The speaker is spec. 4 ohm. From calculation, if I use two back to back 20,000uF electrolytic cap, the high pass frequency is going to be 5Hz for 4 ohm speaker. Is this good enough? Can I get away with smaller cap as I already have a sub woofer?
On a related question. I have a SS power amp that I am experiment with. I am worry the amp blown and the output can rail and burn the speaker. I want to put a back to back electrolytic caps to block the DC for the Woofer. I already have a sub-woofer, so I don't need to have good low frequency response. The speaker is spec. 4 ohm. From calculation, if I use two back to back 20,000uF electrolytic cap, the high pass frequency is going to be 5Hz for 4 ohm speaker. Is this good enough? Can I get away with smaller cap as I already have a sub woofer?
bi/non polar Al electrolytics with full oxide grown on both foils measure lower distortion than the outdated "back-to-back" recommendation
Thanks, I can't find the price on any of those non polar.
I was looking on ebay, I can get 2 of 33000uF 75V for $45. Two in series will give 17000uF. According to my understanding, the higher the voltage rating, the lower the distortion. You think it is good enough.
I was looking on ebay, I can get 2 of 33000uF 75V for $45. Two in series will give 17000uF. According to my understanding, the higher the voltage rating, the lower the distortion. You think it is good enough.
I have some direct experience using an Epcos poly as opposed to a ClarityCap MR in a 10KΩ coupling application. It wasn't in a tube amp, but as a full-range input coupler to my headphone amp. Value was 4.7µF. Definitely made a difference.
At first my rationalization was "How could it make any difference? The voltage across the plates changes minimally and the currents involved are like a few hundred microamps."
Don't quite know how to make sense of it. Could the mechanical rigidity or dampening of the structure still make a difference at these levels? Polarization non-linearities? Metallurgy or the plates? Impedance of this cap favors my amp more? Scratching my head.
At first my rationalization was "How could it make any difference? The voltage across the plates changes minimally and the currents involved are like a few hundred microamps."
Don't quite know how to make sense of it. Could the mechanical rigidity or dampening of the structure still make a difference at these levels? Polarization non-linearities? Metallurgy or the plates? Impedance of this cap favors my amp more? Scratching my head.
Could very well be. I don't even have to measure them. The Epcos were 10µF and the ClarityCaps are 4.7. So there's like a 2x difference there. A better test would be to get some Epcos 4.7's and compare, as the Clarity's are going to be a bill a piece.
Seems like the things I've heard wouldn't be attributed to the difference in capacitance, like synth tones just sounding so much more "Pulled together" and coherent. But heck, I'm the first one to admit a lot of stuff in audio defies my understanding of it.
- Status
- Not open for further replies.