Interesting article.
I miss Wireless World a lot.
My boss used to buy them then when read passed on to me.
I spotted a "readers circuit" in about 1985 for a "soft limiter" which simulated valve type distortion. I built one and improved it a bit and used it with my electric guitar for many years. It worked really well.
Thanks for sharing this article! It directly addresses the question I asked in my opening post regarding the capacitor ("C3" in this article) in the feedback path.
My question was prompted by the results shown in Cyril Bateman's earlier "Capacitor Sounds" series of articles. In that series, he showed that a signal as small as 0.1Vac across a bad polar capacitor can cause atrocious distortions (e.g., -60dB 2nd harmonic from a tantalum cap). When a polar cap is put in the feedback loop, even if the cap value is large enough such that the -3dB frequency of the low-pass filter formed by the cap and the R before it is at say 1Hz, the AC voltage across the cap can still be significant at 20Hz, and I did not feel comfortable to accept the assertion that the effect of the cap distortion on the quality of the amp is de minimis.
Moreover, I was also wondering how the distortions might be associated with or caused by the amount of current flowing through the cap, not just by the voltage across the cap. This is because the current at frequencies way above the -3B frequency is not diminished by the Z of the cap, and should approach a level equaling the NFB voltage divided by the R. This means that the effect of the cap distortion might not be limited only to low or ultra low frequencies (i.e., "unclean at any frequency"), if the cap distortion also depends on the current. I am glad that Cyril Bateman also talked about this issue in this article.
Last edited:
These are good points. Although I would think that even if the current is appreciably, if the frequency is high enough, the voltage across the cap resulting from that current is still very small and gets smaller with frequency. And it's the voltage that determines the distortion, I think.
Anyway, there's much more from Cyril on my website. I have been in contact with Cyril before he passed away and he has allowed me to post the articles, with agreement by Electronics World.
Cyril Bateman's Capacitor Sound articles | Linear Audio NL
Jan
Anyway, there's much more from Cyril on my website. I have been in contact with Cyril before he passed away and he has allowed me to post the articles, with agreement by Electronics World.
Cyril Bateman's Capacitor Sound articles | Linear Audio NL
Jan
-60dB (i.e. 0.1%) is "atrocious distortion"? This with 0.1V AC across the cap - much higher than you would get from an appropriate value cap. Let's do some sums:
Assume signal, and feedback, voltage is 500mV. Assume the lowest relevant frequency, and assume that the cap sets an LF rolloff 10x lower. Hence we have around 50mV across the cap. Assuming a polar cap then we can assume mainly 2nd order distortion. -60dB at 100mV means -66dB at 50mV, which is then -86dB with respect to 500mV. This feedback network distortion sets the distortion for the amp. So we have -86dB i.e. 0.005%. This is what the numbers say, using a cap which many would regard as poor - you certainly would not use this cap in a passive speaker crossover.
Didn't someone famous say that without numbers our knowldge is of a meagre kind?
Assume signal, and feedback, voltage is 500mV. Assume the lowest relevant frequency, and assume that the cap sets an LF rolloff 10x lower. Hence we have around 50mV across the cap. Assuming a polar cap then we can assume mainly 2nd order distortion. -60dB at 100mV means -66dB at 50mV, which is then -86dB with respect to 500mV. This feedback network distortion sets the distortion for the amp. So we have -86dB i.e. 0.005%. This is what the numbers say, using a cap which many would regard as poor - you certainly would not use this cap in a passive speaker crossover.
Didn't someone famous say that without numbers our knowldge is of a meagre kind?
That would be the wrong conclusion. The right conclusion is that the cap quality matters, but not nearly as much as some people believe. I would not use a high permittivity ceramic because they are very nonlinear. I might use an electrolytic, but to play safe I might use a bipolar electrolytic even though I would not expect that to make much difference.
So -60dB is not "atrocious" enough for you?
I know you are trying to make the point that whether a component is good or bad depends on the particular context in which it is used - "horses for courses." That is surely a valid point, but I am not sure if your particular example is convincing enough.
You'd not be proud of designing an amp that has -86dB of distortion at 500mVac input (so ~ 12.5W assuming a gain of 20 into an 8ohm load), would you? I'm more impressed by something like the Benchmark AHB2 amp, which was measured by AudioScienceReview to have a distortion of 0.00016% (or -116dB) at 185W. It is a pity that I cannot justify buying a $3000 amp, so I am still playing with cheap Chinese amp kits.I enjoy playing with numbers too.
Without talking about numbers it would be nearly impossible to have meaningful scientific or engineering discussions.Unfortunately steady state signals like sine waves are not representative of music signals.
Linsley-Hood wrote an article about capacitor distortion in 1993. In this he looked at series loss in electrolytic dc blocking capacitors in the nfb decoupling arm to earth - describing the parasitic resistance in series with the capacitance as a long way from being an ideal resistor.
Hood reckoned the parasitic value likely to be voltage, temperature, polarity, and frequency dependent - and likely to be a significant fraction of the resistor in the decoupling arm to earth. He found it remarkable thus that the non- linearity of the series loss had "so small an effect" on THD of an amplifier at 1kHz and he decided to pursue IMD testing to investigate further.
Apparently what was small to Linsley-Hood was seen differently by Bateman who quantified his results with tests on electrolytic capacitors.
Hood's IMD testing was to look at the effects produced when an amplifier was driven by several simultaneous input signals using a spectrum analyser.
The aim here was to track down the spurious signals generated where real-life imperfections between one type of dc blocking capacitor and another would be more easily seen.
Hood concluded that changing the type of capacitor in that position did affect the sound quality as IMD measurements would affirm - especially if one of the input signals in the test was a square wave.
In Hood's most recent amplifiers before this article polycarbonate dc blocking capacitors were used in the nfb decoupling arm to earth. The downside there is the cost and bulk of plastic capacitors relative to the required capacitance.
A more recent alternative approach has been to use electrolytic capacitors with larger values than are strictly necessary as higher value ones have lower e.s.r. it is fairly common to use ordinary polarity types
It would be interesting to see how modern electrolytic capacitor types fare on IMD testing but who in the DIY community would have access to a spectrum analyser?
Linsley-Hood wrote an article about capacitor distortion in 1993. In this he looked at series loss in electrolytic dc blocking capacitors in the nfb decoupling arm to earth - describing the parasitic resistance in series with the capacitance as a long way from being an ideal resistor.
Hood reckoned the parasitic value likely to be voltage, temperature, polarity, and frequency dependent - and likely to be a significant fraction of the resistor in the decoupling arm to earth. He found it remarkable thus that the non- linearity of the series loss had "so small an effect" on THD of an amplifier at 1kHz and he decided to pursue IMD testing to investigate further.
Apparently what was small to Linsley-Hood was seen differently by Bateman who quantified his results with tests on electrolytic capacitors.
Hood's IMD testing was to look at the effects produced when an amplifier was driven by several simultaneous input signals using a spectrum analyser.
The aim here was to track down the spurious signals generated where real-life imperfections between one type of dc blocking capacitor and another would be more easily seen.
Hood concluded that changing the type of capacitor in that position did affect the sound quality as IMD measurements would affirm - especially if one of the input signals in the test was a square wave.
In Hood's most recent amplifiers before this article polycarbonate dc blocking capacitors were used in the nfb decoupling arm to earth. The downside there is the cost and bulk of plastic capacitors relative to the required capacitance.
A more recent alternative approach has been to use electrolytic capacitors with larger values than are strictly necessary as higher value ones have lower e.s.r. it is fairly common to use ordinary polarity types
It would be interesting to see how modern electrolytic capacitor types fare on IMD testing but who in the DIY community would have access to a spectrum analyser?
Polar electrolytics are about the worst type of capacitors, save tantalum or Z5U, etc. And, their performance (and sound) vary substantially over their lifetime and over operating conditions. What's not to dislike?
Last edited:
Nichicon Muse ES bipolar caps measured: <-120dB THD, <-140dB IMD
Has the topic gotten boring enough yet?
Has the topic gotten boring enough yet?
Clearly the price! No real substitute for that for things like filter caps. For DC blocking you can arrange the distortion to be as small as you want through sizing the thing. For the feedback network a DC servo is a viable alternative though, but more complex.
Is it really the open loop gain?
Using your circuit components, the capacitor distortion is generated by C2. It is injected into the amplifier via R28, not directly into Vfb. Won't the gain be - R35/R28? It is different from non-inverting input which we all know is 1 + R35/R28, my mistake.
Regards
Master Tibouchina.
Thanks for the comments.
Using your circuit components, the capacitor distortion is generated by C2. It is injected into the amplifier via R28, not directly into Vfb. Won't the gain be - R35/R28? It is different from non-inverting input which we all know is 1 + R35/R28, my mistake.
Regards
Master Tibouchina.
I am grateful to alayn91 for getting this research paper which shows tantalum to be a lot worse for signals than normal electro...
I repeat the link : Do Passive Components Degrade Audio Quality in Your Portable Device? - Application Note - Maxim
then further I did my own research and the distortion is minimal... seriously. The biggest problem that I found for the feedback loop is of great importance... it is related to the miniaturization of small caps, another reason for me not to buy a modern product with miniature DC stopping caps.
KCHANG should upgrade to a non-polar, for audio... no need to search further, however I doubt there will be any difference.
I repeat the link : Do Passive Components Degrade Audio Quality in Your Portable Device? - Application Note - Maxim
then further I did my own research and the distortion is minimal... seriously. The biggest problem that I found for the feedback loop is of great importance... it is related to the miniaturization of small caps, another reason for me not to buy a modern product with miniature DC stopping caps.
KCHANG should upgrade to a non-polar, for audio... no need to search further, however I doubt there will be any difference.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.