I think they chose 470uF precisely because by oversizing it they removed the AC voltage error term across the cap and could then focus on any residual distortion mechanisms. Self discusses dis from under- value capacitance in coupling applications and references Cyril Batemans work in this regard.
I have no compunction in using oversized electrolytics for audio signal coupling if I have to. I usually use bipolar types - they are cost effective and perform well - but standard unipolar types oversized work just as well and show zero distortion on the AP.
Nelson got it 100% right: ‘So you’re worried about electrolytic coupling caps? Just get over it!’
I have no compunction in using oversized electrolytics for audio signal coupling if I have to. I usually use bipolar types - they are cost effective and perform well - but standard unipolar types oversized work just as well and show zero distortion on the AP.
Nelson got it 100% right: ‘So you’re worried about electrolytic coupling caps? Just get over it!’
Nelson got it approximately right. There is more to sound than HD. Some people would do well to get over that. Otherwise its "streetlight effect" in action.
https://en.wikipedia.org/wiki/Streetlight_effect
https://en.wikipedia.org/wiki/Streetlight_effect
You are not entitled to any proof, if any, than what anyone has provided. Not sure why you seem infatuated with DBT... why the "ugly"... and why the offence? For whom are you offended... and whom are you trying to protect? Isn't it a simple matter to assume that a claim or opinion is without DBT unless disclosed that it is? So what is your solution to what you seem suggesting as a problem? Do you believe that the general members cannot reason things out as well as you and you are protecting others from being duped?
An argument is a connected series of statements in support of a proposition or claim. The statement "I can hear it and you can not" isn't in itself an argument. Further this statement seems your interpretation of a claim made, not the claim itself. Unsubstantiated opinions can be weighted for a variety of reasons to any extent anyone pleases, so what is the problem?
Just unsubstantiated audiophile mumbo-jumbo as usual and you are as always confusing your subjective preference with objective argument. There are well-established technical grounds for avoiding electrolytic caps on DAC Vref:
- They have high inductance and do not have much to offer at high frequencies involved.
- Vref supply should be located as close as possible to load which is at DAC chip pins. Electrolytic caps with their large size do not fit well in this.
- If ultra low noise LDO supplies are used large electrolytic output capacitance offers no benefits and high ESR and inductance is very detrimental.
AKM evaluation boards have flawed layout. They are meant for testing functionality, not performance. AKM's evaluation boards even fall short of reaching the datasheet level of their DAC chips so using them as reference is not advisable. ESS reference schematics on datasheets do not have electrolytic caps on Vref (AVCC).
Your memory is failing once again. I have said that sound differences in well-measuring dacs I've tested are too small to be able to reliably make the correct pick in blind AB testing. Never have I said that I prefer TDA1541.
I enjoy Mark’s posts. Technical topic discussions like these, however, also benefit from proof. Anecdotal references and unique claims should be accompanied by proof for accountability, especially when the claims get repeated.
Scientific Method might be one reason.
Capacitors, resistors and so forth invite a lot of colloquial discussion from audiophiles, but there is never any proof brought forth to support the claims of audibility. If measurements aren’t good enough for the subjective crowd, then a well conducted DBT removes the bias of sighted evaluation.
You can quite easily parallel electrolytic caps with MLCC’s if you have to keep the total ESL down. This is a well tested and tried technique used extensively on computer boards for example. Given DAC signal edge transitions are very fast, I’d say that’s the main reason for not using electrolytics to decouple Vref - you don’t need high capacitance - what you want is low ESL. Probably best accomplished with some localised bulkl decoupling using an electrolytic and then tight local Vref decoupling using a suitable MLCC type capacitor.
There is no requirement that there exists a connected series of statements to support some conclusion as supposed proof. What changes is the weighting applied to the claim in absence of proof. If the claim is repeated the weighting increases, or decreases if rebuked with or without proof.
How does one go about testing audibility of resistance and capacitance without including a multitude of resistance and capacitance mechanisms in the testing? There is a distinction between hearing something and the supposed mechanism that causes the hearing. Oftentimes the debunking relates to debunking the nature of the proposed cause not the cause itself as seeming mysterious. Creating a DBT makes an assertion that the nature of the mechanism causing the detection is known. There is no necessary "well conducted DBT".
It is considered that the nature of distortions, as linear or non-linear, is exceeding complex when dealing with dielectric materials. To think that all distortion mechanisms in dielectrics can be reasonably quantified by spectrum analysis seems far fetched, particularly if we assert that spectral components below 100DB are meaningful in DBT's, or otherwise. The question arises, if dielectric effects are far fetched at what point does it become absurd to suggest that spectral components below some point are equally absurd as meaningful in DBT.
It isn't always necessary to engage in performing DBT testing to consider potential mechanisms of distortions to mitigate them. Take for example interconnect cables with dielectric materials. As per the above it was mentioned about electron migration into dielectric materials. Electrons as signal elements migrate as a function of field strength as exiting and re-entering the conducting stream in some indeterminate fashion as a function of signal. It can be imagined if these signal electrons are diverted from the load it changes the signal at the load end. To mitigate the potential of this DA effect is to drive the cable from a zero source impedance. It seems that many DAC's have 100 Ohm resistors, or so, to mitigate capacitance loading, seeming as an invitation to DA complications. The AD811 used as an I/V in amongst these pages uses I believe 10 Ohms.
Another example of dielectric mitigation is in the use of current mode. The question arises in the use of current mode for moving coil head amplifiers. Dielectric suppression comes about by driving of the two conductors to the same potential, effectively reducing the field across the two conductors to zero as to eliminate electron movement into the dielectric. Reducing effective capacitance reduces effective DA. If one is familiar with the function of electrometers in the measurement of electron charge they often use triaxial BNC connectors. In this case the voltage sensed on the centre conductor is buffered and fed back on an inner shield, holding this at the same potential as the centre conductor. This eliminates the capacitance as seen by the inner conductor and also prevents DA.
It isn't always necessary to engage in performing DBT testing to consider potential mechanisms of distortions to mitigate them. Take for example interconnect cables with dielectric materials. As per the above it was mentioned about electron migration into dielectric materials. Electrons as signal elements migrate as a function of field strength as exiting and re-entering the conducting stream in some indeterminate fashion as a function of signal. It can be imagined if these signal electrons are diverted from the load it changes the signal at the load end. To mitigate the potential of this DA effect is to drive the cable from a zero source impedance. It seems that many DAC's have 100 Ohm resistors, or so, to mitigate capacitance loading, seeming as an invitation to DA complications. The AD811 used as an I/V in amongst these pages uses I believe 10 Ohms.
Another example of dielectric mitigation is in the use of current mode. The question arises in the use of current mode for moving coil head amplifiers. Dielectric suppression comes about by driving of the two conductors to the same potential, effectively reducing the field across the two conductors to zero as to eliminate electron movement into the dielectric. Reducing effective capacitance reduces effective DA. If one is familiar with the function of electrometers in the measurement of electron charge they often use triaxial BNC connectors. In this case the voltage sensed on the centre conductor is buffered and fed back on an inner shield, holding this at the same potential as the centre conductor. This eliminates the capacitance as seen by the inner conductor and also prevents DA.
As I've said many times as THD+N tends to zero the error and noise tends to zero and you cannot do better for fidelity, its a simple mathematical fact. Certainly there's more than just the number, but make that number small enough and you can afford to ignore other nuances because they have been reduced to complete inaudibility. Care has to be taken with the measurement bandwidth - for instance with class D (does the switching frequency count as noise for the THD+N?). Or LF cutoff (is 20Hz adequate for noise - clearly no).
And its the worst-case THD+N that counts, so you can't just test an amp from cold, but over a range of thermal histories and loads. Perhaps that's what you are getting at?
The reason for using electrolytics on Vref is because dacs draw Vref current in partially audio signal dependent way. To have enough capacitance to reproduce 10Hz well, sometimes it takes a fair amount of capacitance. Depends on the particular dac.
Once again:
Regarding noise in caps, please see attached. Other electrolytics also show 1/f noise.
Regarding fidelity, its not just about low PSS distortion and noise floor. There is also the stuff not commonly measured, which are signal-dependent noise, and linear distortions which can be particularly audible at LF.
Scott Wurcer and Sy Yaniger discussed DA as a source of distortion at length on the Blowtorch thread. In short, they concluded it is not a factor. Amplifiers designed and built by people on this site achieve measured distortion levels of low single digit ppm or better. These amplifiers use input coupling capacitors and capacitors in the compensation networks. If DA was a serious source of distortion, then surely it would show up in these examples?
Last edited:
Driving the shield is usually done to improve CMRR and maintain signal bandwidth by removing capacitive coupling to surrounding circuits. I’ve never heard of it being used to remove DA distortion mechanisms.
IME, DA is probably a factor. It results in a linear distortion which time-smears LF transients.
I'm talking about reproducing signals, not distaste! I want an amplifier that can reproduce nasty sounds exactly as nasty as they are, as well as good sounds as well as they are. Just like I like a computer monitor that reproduces colours accurately, and not rose-tinted...
By the way we've known about the mathematical truth of accurate reproduction of a signal for a much longer time than 1938. And remember the amp is only one part of the chain, the easiest part really. Best not to fail with the low-hanging fruit.
By the way we've known about the mathematical truth of accurate reproduction of a signal for a much longer time than 1938. And remember the amp is only one part of the chain, the easiest part really. Best not to fail with the low-hanging fruit.
DA implies the absorption and release of electrons as a function of time being buried to whatever depth and ultimately unburied, that by such action doesn't suggest there would be harmonic distortion. There needs to be a distinction between conventional RC type phase shift as a result of pure capacitance and fundamental frequency phase shifts created by DA. Spectrum analysis combines all alike frequencies into one combined frequency that include DA generated fundamentals.
If we consider every electron as a signal element how do we come to grips with an electron that becomes buried into a dielectric material coming out at a later time? This isn't the same as an RC network, nor can it easily be imagined as coming out distorted. We end up with buried signal elements acted upon in real time by the present signal field. DA suggests an uncorrelated mess being generated.
I glanced at the app note, I can say "at least it tries." Have a comparison with this Panasonic Aluminum Electrolytic thing from a few decades ago - most of it's actually pretty good until you get to Page 7-2, "Audio Equipment:"
https://archive.org/details/ABA0000TE7/page/n2/mode/1up
https://archive.org/details/ABA0000TE7/page/n2/mode/1up
I can also easily hear different capacitors in a signal path or different values/types of caps used for voltage rails filtering & decoupling. My kids can pick the differences as well - easily. A few of my friends are musicians - they can tell the difference as well. The best is not to use them (capacitors) in a direct signal path, if possible...
However, I can not always tell the difference in a blind, quick A/B test (which I find very tiring and pointless)... BUT, I can tell which cap is in a signal path if I give it a few days of casual listening (especially if I need to focus on work, and the music is playing in the background)... but that kind of testing takes time.
However, I can not always tell the difference in a blind, quick A/B test (which I find very tiring and pointless)... BUT, I can tell which cap is in a signal path if I give it a few days of casual listening (especially if I need to focus on work, and the music is playing in the background)... but that kind of testing takes time.
One thing about DBT test rigs, if volume trim pots are used to level match what are already very low distortion A/B devices, the pots can be the dominant distortion source in the sound (which is not to say they necessarily are, only that its a possibility). For that reason I don't use any volume pots in my system. Simple A/B testing shows that pots like the popular Alps blue add distortion that a Goldpoint stepped attenuator doesn't add. I don't even use the stepped attenuators included in my power amps, because they do more than just change the volume level. It not that they sound like added distortion, its more like HF is attenuated a little bit (at least for those particular stepped attenuators). The point is that anything in a system can be bottle neck for the sound, and finding them can be tricky.