Is it the asymmetry or the pulse which makes it more visible? Wouldn't other CR (e.g. lead-lag networks, Zobel) in the circuit have a similar effect but with a much shorter time constant? I'm still struggling to get a grip on this. If it is a linear phenomenon (as many people say - wrongly?) then it won't distort but it could disturb LF frequency response by, in effect, remembering the average of the previous few seconds waveform. If it is non-linear then where does the non-linearity reside?
Let's say I put 10V on a cap, then disconnect and short it for 10 sec. Then measure the voltage. Say it peaks at 0.5V due to DA. Repeat but using 20V. Will it peak at 1.0V? Will the time constant be the same?
Let's say I put 10V on a cap, then disconnect and short it for 10 sec. Then measure the voltage. Say it peaks at 0.5V due to DA. Repeat but using 20V. Will it peak at 1.0V? Will the time constant be the same?
John,
AP2 can do multitone. I do use this occasionally. Very instructive.
I once recommended this to a rather famous designer for a company that at the time had F1 tie ins.
He was not amused about what the multitone tests told him about his gear. I generally am quite happy what they tell about my stuff, now that I have an AP2...
Ciao T
AP2 can do multitone. I do use this occasionally. Very instructive.
I once recommended this to a rather famous designer for a company that at the time had F1 tie ins.
He was not amused about what the multitone tests told him about his gear. I generally am quite happy what they tell about my stuff, now that I have an AP2...
Ciao T
Walt Jung and I published a paper about 25 years ago that showed our results. Please go to his website to get it, if you do not have it. Scott Wurcer, at the time, did a SPICE analysis predicting the same results.
Using AC to test is like charging a potential battery with AC, not much happens. However, charge it with DC for a time and later you can tap this 'stored' energy.
I am somewhat surprised that these questions even need to be asked.
Using AC to test is like charging a potential battery with AC, not much happens. However, charge it with DC for a time and later you can tap this 'stored' energy.
I am somewhat surprised that these questions even need to be asked.
Last edited:
If I don't know, I ask. If I want to discover what someone else knows, I ask.
I assume you are referring to this article. It says that DA "will often be essentially linear error". It also says that DA can be non-linear in caps such as high-k ceramic (which are themselves non-linear anyway). The plots do not appear to separate any non-linear residue from the dominant linear residue. The plots may look frightening, but as they stand there is no clear evidence of nonlinearity. For this you would have to show that the output includes frequency components not present in the input, or that the output cannot be nulled by any linear reference arm. Now you may have done this, but as far as I can see the article doesn't say so.
Is there any evidence that DA can be significantly more non-linear than the basic capacitor dielectric response? (This is a genuine question; I don't know the answer. I thought that is what we do on discussion forums).
I assume you are referring to this article. It says that DA "will often be essentially linear error". It also says that DA can be non-linear in caps such as high-k ceramic (which are themselves non-linear anyway). The plots do not appear to separate any non-linear residue from the dominant linear residue. The plots may look frightening, but as they stand there is no clear evidence of nonlinearity. For this you would have to show that the output includes frequency components not present in the input, or that the output cannot be nulled by any linear reference arm. Now you may have done this, but as far as I can see the article doesn't say so.
Is there any evidence that DA can be significantly more non-linear than the basic capacitor dielectric response? (This is a genuine question; I don't know the answer. I thought that is what we do on discussion forums).
DA is primarily a LINEAR distortion like phase shift, but it will show itself as an interesting departure from the ideal path predicted by an ideal capacitor.
The most important thing that the differential subtraction test shows is that high DA caps do not follow the same time path as more ideal caps. This is why we now use more ideal, low DA caps, because they are a more ideal cap doing what we ask the cap to do, and closer to the mathematics that we use to predict the cap's performance initially.
DA does not sound exactly like distortion, but more like adding 'reverb', that can actually make some circuits more 'forgiving' to what is processed through them, perhaps making marginal fidelity into acceptable fidelity. However, it is not, high fidelity.
The most important thing that the differential subtraction test shows is that high DA caps do not follow the same time path as more ideal caps. This is why we now use more ideal, low DA caps, because they are a more ideal cap doing what we ask the cap to do, and closer to the mathematics that we use to predict the cap's performance initially.
DA does not sound exactly like distortion, but more like adding 'reverb', that can actually make some circuits more 'forgiving' to what is processed through them, perhaps making marginal fidelity into acceptable fidelity. However, it is not, high fidelity.
OK, we seem to have gone in a circle but I'm glad my original understanding of DA has been confirmed. The main (continuous sine wave) effect of DA would be to modify LF phase shift? It will add a tail to a pulse, which looks bad in the time domain but is simply the effect of a linear filter.
I suppose the next question is whether it is this effect which is apparently degrading sound or whether DA magnitude is acting as a useful proxy for some as yet undiscovered dielectric problem.
I suppose the next question is whether it is this effect which is apparently degrading sound or whether DA magnitude is acting as a useful proxy for some as yet undiscovered dielectric problem.
DA itself is not a bad thing, but since capacitors don't have inside ideal resistors and capacitors like on an equivalent schematics diagram, it signifies presence of non-linear effects hard to measure. Also, even when capacitors are ideal coupling of stages with non-linear input and output impedances causes phase non-linearities.
Whatever, the fact is that caps store energy from the signal. This CHANGES the signal in a way that is usually not calculated into the time constants (for example, for a phono EQ) and quibbling over 1Hz or 1% value accuracy might be made moot by the DA inaccuracies in the cap itself. If one would only look at the results of what Walt and I published, and I personally made all the tests, these selected as representative out of hundreds of measurements made by me, as well as additional measurements made by Walt Jung, you 'might' get a 'feeling' for the inaccuracies with the passage of time, especially with PHOTO 6, a more representative example of normal coupling capacitor loading (50Kohm driven by a 10uf cap). IF you look, you will note a NEGATIVE ERROR while the pulse is being applied, and then a POSITIVE ERROR when the pulse is removed and only a 50 ohm source is there. This implies REDUCTION of the peak signal, and ADDITION of extra energy as the cap releases its stored DA energy. Check it out!
John, it's clear that one can measure DA effects in an appropriate test circuit (e.g., the one that Scott developed which you used or in some of the timing circuits discussed by Pease in his book). The more interesting question is, if DA is an issue in AF circuits, you should similarly be able to measure the effects with audio signals. Do you know if anyone has done so?
Sorry, I was unclear; that'll teach me to type slower. What I mean is, if you take an amplifier with low DA coupling or EQ caps, and compare it to the same amp with high DA coupling or EQ caps using music or test signals or whatever, what differences in the overall amplifier performance can you measure?
Scott,
You can use RC Circuits to produce a fixed delay above cutoff (of course there will be the little problem of amplitude nonlinearity). Even better, you can use a piece of glass as delay line...
There is little end to ingenuity.
Ciao T
You can use RC Circuits to produce a fixed delay above cutoff (of course there will be the little problem of amplitude nonlinearity). Even better, you can use a piece of glass as delay line...
There is little end to ingenuity.
Ciao T
Hi Sy,
That would be an unexpected but very welcome turn of events.
Well, you can measure differences that are signal dependent, of course. What else?
Actually, as an observation, the documented non-linearities in capacitors are insufficiently explained by DA. But who gives a flying french coitus anyway?
Ciao T
That would be an unexpected but very welcome turn of events.
Well, you can measure differences that are signal dependent, of course. What else?
Actually, as an observation, the documented non-linearities in capacitors are insufficiently explained by DA. But who gives a flying french coitus anyway?
Ciao T
SECAM TV's did that. The idea of "memory" as in an echo of the signal coming out of a capacitor at a later time is not accurate.
Please explain "fixed delay above cutoff" you lost me. AFAIK there is no solution to an RC mesh that yields a traveling wave.
I really don't know. I just use 'accurate' caps for EQ and direct couple everything else. Pointless for me to do otherwise.
I once lost a listening test with HK back in 1978, and I pinned down that the 'problem' was a 10uf 50V polycarbonate cap that even Mark Levinson could not afford to put into the JC-2.
For some reason, the 2uF 200V polycarbonate from SE&I worked OK, but this larger, more expensive cap, from the same manufacturer, added an unpleasant factor to the sound. However, my competition in this HK preamp trial had LOTS of DA in the caps used and it sounded sort of 'extra smooth' or 'forgiving'. Neither preamp was ever made by HK, in any case. I consider 'forgivingness', an error as well, and I do not design it into my circuits, if I can help it.
I once lost a listening test with HK back in 1978, and I pinned down that the 'problem' was a 10uf 50V polycarbonate cap that even Mark Levinson could not afford to put into the JC-2.
For some reason, the 2uF 200V polycarbonate from SE&I worked OK, but this larger, more expensive cap, from the same manufacturer, added an unpleasant factor to the sound. However, my competition in this HK preamp trial had LOTS of DA in the caps used and it sounded sort of 'extra smooth' or 'forgiving'. Neither preamp was ever made by HK, in any case. I consider 'forgivingness', an error as well, and I do not design it into my circuits, if I can help it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.