Which Capacitors for Audio?

[ct.waveform]

I don't think anybody denies the fact that loudspeaker cables interact with the amp and speaker. However, in most cases, the differences are way outside the domain of perception.

I'm afraid I really couldn't disagree more. I could hear the differences back in the early '80s, which is somewhere around when Monster Cable came out in the US and people here in the UK started using either 4mm2 RS 56 Strand or Naim NACA 4 cable (also 4mm2 x-section). Remember that before this the usual speaker cable (considered good enough because people thought it made no difference) was either mains cable (and probably not even 1.5mm2) or bell wire. In the interim period I have tried out hundreds of cables and built a few dozen myself, hoping to get something better than the then prevailing best available. I have been able to hear the differences nearly every time (some cables are quite similar) and my own efforts have never come anywhere close to being as good until very recently. I doubt there's a single loudspeaker manufacturer who believes that speaker cables make no difference.

Even the DCR of the cable is not such a significant factor, considering 75% of the speaker impedance is resistive and inductive losses in the voice-coil. It may matter in terms of interaction with passive x-overs though.

There's a difference in both active and passive systems but for frequency response variations due to DCR, passive systems are obviously worse. It's becoming more significant these days as the trend is to lower nominal impedance. 6 ohms is more or less the norm today with 4 ohms nominal being common. And the IEC spec for a 4 ohm speaker is that it doesn't fall below 3.2 ohms. Bob Cordell covers all this in his book and his graph of what you get at the speaker terminals, especially due to the impedance mismatch, is a horror story. And that graph of the voltage the speaker sees doesn't cover what the extra resistance does to the Q of the fundamental resonance of the bass unit. Unfortunately I don't have it to hand or I'd post it here.

If standing waves in the MHz range affect the amp's stability, it shows a very poor design. One can't determine a concept on the basis of a lemon.

Just the capacitance alone. 5 meters of cable at about 100pF per foot. Of course it affects the phase margin and amplifier designers will take this into account. Incidentally, not all amplifiers have an output inductor, with the trend there being to make them smaller anyway. Nor is the usual Boucherot cell any kind of impedance match so RF just makes a beeline for the input stage, with its life made easier if there's a small C in parallel with the feedback resistor. Incidentally you can make this better by putting a series CR across the speaker terminals - something like 1nF+220R - and that's audible too, though not usually tonally.

About 40 years ago, Sony had an amp with a rise-time of 0.2us IIRC. That's a BW above 1MHz. You couldn't use just any cable with it. Some may have concluded that the cables that worked with this amp were superior. Others (including me) thought the amp was simply inadequate.

A BW above 1MHz is just inviting trouble, but it does make my point. Funnily enough some people are boasting about their amps getting halfway there today. They obviously don't understand that the dimensions of the instruments and of the microphones mean there's a very low probablity of there being any signal on the recording at even 20kHz, so I don't know what they expect to "reproduce".

 

[ct.waveform]

Wiper resistance is often substantial and nonlinear. The exceptions are wire wound and cermet track controls.

Yup. And have you seen the price of cermet pots and presets these days?

 

[abbey road d enfer]

I can't find the post with your noise calculations but it's in my email Inbox. This is on the rising 1/f part of the noise graph. The corner frequencies where the rise begins may not be dead on but it's not going to make much difference. I'm using 1Hz noise x Sqrt( ln (corner f/Aperture reciprocal)) for the rms total.

You may note that I have edited my post. At first I had not figured out you integrated the 3dB/octave curve.

I thought it was going to come out much higher, though it would have to be a hell of a lot more before it becomes significant in an rss sum. Even integrating it visually you'd imagine it was higher, but then these tricky sods move where it intersects the y-axis so you can't see at a glance whether it's bad or not. But, as I said, the surprise here for me was that in low noise op amps the 1/f noise also comes down. I don't know if I have just chosen two LN amps where that happens but, if that's what happens in general, why does anyone ever bother about it? People do write about it as though it's important, but it really seems not to be unless you have a very narrow bandwidth.

It's important in medical, vibration and seismologic applications, not really in audio. Some devices that have their 1/f transition (corner) frequency spilling into LF/low-mids should simply not be used for sensitive audio.
Much progress has been made in this domain since the heroic days of 5534 and TL072.

 

[ct.waveform]

It's important in medical, vibration and seismologic applications, not really in audio.

Thank you. As I was typing I was scratching my head as to what the use case could be. That explains why all the guys from the big semiconductor companies keep mentioning it.

Some devices that have their 1/f transition (corner) frequency spilling into LF/low-mids should simply not be used for sensitive audio.

That's kind of what of thought was going to happen here. Same sort of principle as on a distortion curve; as you lower the noise floor it just means it starts rising from a lower frequency.

Still, it's lovely to have something I can safely ignore.

 

[abbey road d enfer]

Bob Cordell covers all this in his book and his graph of what you get at the speaker terminals, especially due to the impedance mismatch, is a horror story.

I don't know if these are your words or Cordell's, but speaking of impedance mismatch in an amp-cable-speaker connection is ludicrous. The amp's impedance is typically a fraction of ohm, the cable is probably a few dozen ohms and the speaker is 4-16 ohms. No way anything will be "matched". I don't think any kind of standing waves can develop, though, due to the cable being much shorter than the highest frequency wavelength.

And that graph of the voltage the speaker sees doesn't cover what the extra resistance does to the Q of the fundamental resonance of the bass unit.

Maybe because it doesn't change much. A decent cable may have a resistance of about 0.1 ohm, which is nothing compared to the DCR of the voice-coil. The change in Q caused by temperature variations is 2 orders of magnitude higher.

Just the capacitance alone. 5 meters of cable at about 100pF per foot. Of course it affects the phase margin and amplifier designers will take this into account.

1500pF is 1 kohm at 100kHz. An amp that is not capable of dealing with it is poorly designed. You can't conclude that the cable is defective, it's the amp that is.

Incidentally, not all amplifiers have an output inductor, with the trend there being to make them smaller anyway. Nor is the usual Boucherot cell any kind of impedance match so RF just makes a beeline for the input stage, with its life made easier if there's a small C in parallel with the feedback resistor. Incidentally you can make this better by putting a series CR across the speaker terminals - something like 1nF+220R - and that's audible too, though not usually tonally.

I seriuosly doubt a "1nF+220R" network across the speaker changes anything perceptible; I don't think it's even reliably measurable.

 

[Ricardus]

I'm afraid I really couldn't disagree more. I could hear the differences back in the early '80s, which is somewhere around when Monster Cable came out in the US

Subjective listening tests? Not double blind?

 

[ct.waveform]

The amp's impedance is typically a fraction of ohm, the cable is probably a few dozen ohms and the speaker is 4-16 ohms. No way anything will be "matched".

That's the point, nothing is matched, but it's unquestionably acting as an aerial and the out of band mess was there in Cordell's measurements. I worked out the characteristic impedance at the time the book came out, which is over 10 years ago, and though there was quite a margin as to what it could be as I didn't have 100m of the cable to measure, I seem to remember it came out somewhere around 220R, hence the 1nF +220R. I was going to do a dip switch attachment to try different values of C and R, as it might have been as low as 150 ohms, but I never got round to it. Also I couldn't hear any difference between different values I tried but I could hear a difference between with and without. I still use a CR across speaker terminals to this day.

I don't think any kind of standing waves can develop, though, due to the cable being much shorter than the highest frequency wavelength.

I wouldn't imagine they can, but this is not really an area I know very much about. I can't remember what frequency Cordell's graph went up to, or the left hand scale, or even the kit he did it on (which might have given me an idea of what it was) but the out of band mess was alarming. It was a mountain range! I don't think I had a System One back then but I guess I could repeat what he did, if even if only partially. The CR was his suggestion but he didn't give any indicative values - I guess for proprietary reasons. He is a consultant after all. It really is a shame I don't have his book here.

Maybe because it doesn't change much. A decent cable may have a resistance of about 0.1 ohm, which is nothing compared to the DCR of the voice-coil.

Cable may be 0.1 ohm nowadays but it didn't used to be until people started taking cable seriously! Quite a lot of things to do with the Q have been ignored in the past and several still are today. I've been looking at this quite recently and my estimate is that quite a large number of speakers that were sized for, say, a Q of 0.707 actually have a Q closer to 1. (This turned out to be exactly the measured Q of a pair of Isobariks I have, with the resonance at a not especially impressive 45Hz, which is some long way from the specced free air resonance of 25Hz at a Q of 0.37!) The thing is that once you start putting resistances between the driver and the amplifier it doesn't just raise the Q but that resistance also means nothing to the 100 ohm peak of the resonance so you get a bump there in the voltage as the sensitivity of the system comes down.

The change in Q caused by temperature variations is 2 orders of magnitude higher.

I have tried to find these rises in temperature and I haven't been able to. I seem to remember Keith Howard also tried at one time too, after me, and couldn't either. Your estimate, if it comes from what you've read, may be based on the work KEF did on the T27 and B110 (maybe in an LS3/5a, I don't know) as that's the only work I know that used a music signal (and different types of music). I mean to repeat my efforts as I have a nice idea for how to measure the temperature of the coil continuously and I need a more definitive set of figures for simulation because the change in load obviously pushes the crossover around. I'm working with a few tenths of an ohm for a normal listening level, which is essentially fixed, and then looking to see what changes in shape there are either side of that. Not 10 ohms.

1500pF is 1 kohm at 100kHz. An amp that is not capable of dealing with it is poorly designed. You can't conclude that the cable is defective, it's the amp that is.

There are plenty of conditionally stable amplifiers out there. I haven't said the cable is in any way defective, just that they change the phase margin. And if you go to a concentric cable or a Litz cable you'll get a LOT more capacitance than that.

I seriuosly doubt a "1nF+220R" network across the speaker changes anything perceptible; I don't think it's even reliably measurable.

We can see from the values that it's not inside the audio band, so there are other mechanisms at play. We know from Cordell's measurements that there are ultrasonic artefacts and he's suggesting that terminating at least one end of the cable properly will help things. I expect he's tested it too, rather than send his book out with a conjecture in it. This is only 700kHz so I expect it is measureable. One could increase the C to 2n2, which is another value I used (in fact was the one I plumped for even though I couldn't hear a difference). But the point is that these frequencies are going to jump across junction capacitances and be all over the amplifier. I'm not sure how you expect the amplifier to "deal with it". It's not a one-way street with the currents all going where we envisage them going. I'm as clueless as anyone as to what happens when you hand an amplifier something above its GBWP and slew rate (from the wrong end) but I know it's not a good idea. At the other end I put 100pF across the inputs and an RC on the way in, and that's from a shielded cable and something with grounded casework (usually). Are these then wasted components? I really don't see how one can have one but not the other.

 

[ct.waveform]

Subjective listening tests? Not double blind?

Absolutely. Double blind is useless, inefficient and unproductive. The last thing you want is for your listeners to be completely at sea. It makes them nervous, uncomfortable and self-conscious. It's a shibboleth, and is only used, if ever, by people who worship at the altar of scientism. Do you blindfold yourself at Hi Fi shows or in shops? I thought not.

 

[molke]

Absolutely. Double blind is useless, inefficient and unproductive. The last thing you want is for your listeners to be completely at sea. It makes them nervous, uncomfortable and self-conscious. It's a shibboleth, and is only used, if ever, by people who worship at the altar of scientism. Do you blindfold yourself at Hi Fi shows or in shops? I thought not.

Maybe I’m just too negative in general, but sometimes I miss a “dislike” button. You probably know that with this kind of subjective listening you’re testing all sorts of things, least of all audio performance. By the way, I would consider double blind testing to be the bare minimum and not the gold standard, but I don't go to hi-fi shows or shops either.

 

[JohnRoberts]

Absolutely. Double blind is useless, inefficient and unproductive. The last thing you want is for your listeners to be completely at sea. It makes them nervous, uncomfortable and self-conscious. It's a shibboleth, and is only used, if ever, by people who worship at the altar of scientism. Do you blindfold yourself at Hi Fi shows or in shops? I thought not.

I believe in science. Listening tests are subjective (individually specific). Only well controlled double blind listening tests carried out to realize statistical significance can bring a level of objective science to the very subjective individual results.

Properly done this is expensive and time consuming, so rarely used, but it is real science.

JR

 

[ct.waveform]

Maybe I’m just too negative in general, but sometimes I miss a “dislike” button. You probably know that with this kind of subjective listening you’re testing all sorts of things, least of all audio performance. By the way, I would consider double blind testing to be the bare minimum and not the gold standard, but I don't go to hi-fi shows or shops either.

Yeah, well, you probably believe all those double blind vaccine trials were completely above board and not manipulated at all, and that we've all had a very lucky escape from being killed by a deadly virus thanks to those brave doctors who are curiously all now a quarter of a million better off. May I suggest you won't like the majority of what I post and that we stay away from each other.

 

[JohnRoberts]

I used to write a magazine column called "audio mythology" back in the 80s (40 years ago).

Some old myths refuse to die.

JR

 

[Newmarket]

Yeah, well, you probably believe all those double blind vaccine trials were completely above board and not manipulated at all, and that we've all had a very lucky escape from being killed by a deadly virus thanks to those brave doctors who are curiously all now a quarter of a million better off. May I suggest you won't like the majority of what I post and that we stay away from each other.

Oh Dear Which doctors ? Fairly sure the ones I know are not £250 000 better off due to vaccines.

 

[ct.waveform]

I believe in science. Listening tests are subjective (individually specific). Only well controlled double blind listening tests carried out to realize statistical significance can bring a level of objective science to the very subjective individual results.

Properly done this is expensive and time consuming, so rarely used, but it is real science.

JR

I did a fair amount of Philosophy of Science back when I was at university. It's the subject that intellectually justifies the claims that science can advance knowledge, while also pointing out that it can never make a claim that something is true. In the 40 years since, I have never seen ANY scientific endeavour that looks anything like this intellectual framework. Actual science, or practical science, doesn't remotely adhere to what it claims is its intellectual justification. As an example, I will be impressed if you can find me a single audio company that regularly uses double blind listening tests to assess its products. There are plenty of companies that have tried, but admitted they either uncovered them or gave up on them because is was unfruitful. You need people who know what they are listening for and who have good ears. That comes by exposure and through experience and they are few and far between.

Incidentally, science does not require belief. Religions do, and in this context that's called scientism, which is a disease of science.

 

[Newmarket]

Maybe I’m just too negative in general, but sometimes I miss a “dislike” button. You probably know that with this kind of subjective listening you’re testing all sorts of things, least of all audio performance. By the way, I would consider double blind testing to be the bare minimum and not the gold standard, but I don't go to hi-fi shows or shops either.

tbf 'single' blind testing is useful. Not as good as double in most or all cases but a lot easier and quicker. But maybe pandemics are subjective and we are wasting our time

 

[ct.waveform]

Oh Dear Which doctors ? Fairly sure the ones I know are not £250 000 better off due to vaccines.

That's the average bonus payments from vaccine companies to doctors in the US for fulfilling their quota of vaccinated patients. We don't have the documents for the UK but it's likely to be similar.

 

[living sounds]

Maybe I’m just too negative in general, but sometimes I miss a “dislike” button. You probably know that with this kind of subjective listening you’re testing all sorts of things, least of all audio performance. By the way, I would consider double blind testing to be the bare minimum and not the gold standard, but I don't go to hi-fi shows or shops either.

Yesterday I was auditioning decoupling schemes in my console, switching back and forth between two channels (console output) as well as the pure DAC output. It happened more than once that I had the output switch set to the DAC only output and was switching channels on the console and "hearing" all the differences...

Blind testing with audio often doesn't yield the best results either, because differences can be very small, specific and your ability to judge can change drastically over - say - the course of a day. You may easily prefer the - ultimately - worse option at a given moment. I try to do mostly comparisons to a known DAC output vs signal travelling through whatever needs to be judged, and even that is often difficult.

 

[Newmarket]

That's the average bonus payments from vaccine companies to doctors in the US for fulfilling their quota of vaccinated patients. We don't have the documents for the UK but it's likely to be similar.

I see. You made it up. Shouldn't you be on Twitter with this stuff

 

[ct.waveform]

Yesterday I was auditioning decoupling schemes in my console, switching back and forth between two channels (console output) as well as the pure DAC output. It happened more than once that I had the output switch set to the DAC only output and was switching channels on the console and "hearing" all the differences...

Blind testing with audio often doesn't yield the best results either, because differences can be very small, specific and your ability to judge can change drastically over - say - the course of a day. You may easily prefer the - ultimately - worse option at a given moment. I try to do mostly comparisons to a known DAC output vs signal travelling through whatever needs to be judged, and even that is often difficult.

Yup. I too have plenty of examples of days when I have made half a dozen or more changes to something, each time keeping the change I thought was better. Then, the next day comparing the final result to what I started with and finding that it was worse than the original and not six steps better. What would have helped would be either spending more time with each iteration, preferably days, or having someone whose ears I trusted and seeing if their perceptions correlated with mine.

 

[ct.waveform]

I see. You made it up. Shouldn't you be on Twitter with this stuff

IIRC, they were CDC documents. I can't remember if I saved them but they're probably not that hard to find on the net. In this country (UK) we know officially that each doctor was paid an additional £12.50 for each patient vaccinated and that at least £50 or £60 billion was added to the NHS budget. The NAO has probably published the breakdown of where that money was spent by now, and if so it will be easily found on its website. But you prefer to call me a liar. Thanks for that.

Much more of this and I won't be on this website.