Hi there,
The claims I hear for audible differences existing for different electronics amount to frequency response, noise and distortion.
These distortion graphs are between the two I/V stages of a DAC :
(http://i982.photobucket.com/albums/ae301/gbyleveldt1/IC-IV_zpsb757b132.png)
And :
(http://i982.photobucket.com/albums/ae301/gbyleveldt1/DiscIVB_zps28f9f192.png)
The claim is that these DACs measure flat, but the measurements show the effects of certain filters at play. Looking at the distortion components, is this enough of a difference to cause gross audible differences? The claims I hear are that different filters used in DAC design can result in audible differences irrespective if that DAC measured perfectly flat.
What do you think?
... the claim is that these DACs measure flat, but the measurements show the effects of certain filters at play. Looking at the distortion components, is this enough of a difference to cause gross audible differences? The claims I hear are that different filters used in DAC design can result in audible differences irrespective if that DAC measured perfectly flat.
What do you think?
It is definitely not "enough of a difference to cause gross audible differences". Any amplifier manufacturer with aspirations to audiophile quality would be very pleased to attain such low THD+noise figures - in other words, the amplifier you play these DACs through will likely sound worse than the DACs.
It is definitely not "enough of a difference to cause gross audible differences".
What makes you think this?
It is definitely not "enough of a difference to cause gross audible differences".
What makes you think this?
How about that the biggest harmonic is 100 dB less than the primary signal?
Even if you filtered out the 1 khz signal, you wouldn't be able to hear the distortion unless you are gaming the volume control.
I was having a debate with a DAC designer who cited these graphs as reasons for his sonic preferences, especially the first graph. He claimed that the harmonic distortion was the reason why he preferred the sound. But you are saying that the harmonic components are so far down, 100 dB's down and therefore inaudible.
Here are the impulse response graphs, respectively :
(http://i982.photobucket.com/albums/ae301/gbyleveldt1/IMP_SHARP_zpsffebced4.png)
and :
(http://i982.photobucket.com/albums/ae301/gbyleveldt1/IMP_SLOW_zps387ac0eb.png)
Is this an example of a false positive result?
Is this an example of a false positive result?
What result? I see 2 pics with different post and pre-ringing. This proofs nothing for real audibility with a correctly working DAC. You also may try to visualize at what frequencies this ringing happens.
The claim is that the harmonic distortion components in the first graph match the sonic preference of the lister .. the claim is that the added distortion makes things sound "nicer". His claim is that science agrees with his claim, because harmonic distortion makes things sound "nice".
If you saying that because the harmonic components are 100 dB down which no human could possibly perceive under any circumstances then the audible experience was based on a false positive result or conclusion. Correct?
Just wanted to ask : can different filters possibly affect sound staging or timbre? Or are those qualities determined by other metrics?
If you saying that because the harmonic components are 100 dB down which no human could possibly perceive under any circumstances then the audible experience was based on a false positive result or conclusion. Correct?
Correct.
Just wanted to ask : can different filters possibly affect sound staging or timbre? Or are those qualities determined by other metrics?
They *can* definitely affect timbre, if the effects of the filter reach down into audible frequencies, but no properly designed filter should do that. On the other hand, there is a lot of "audiophile" gear where the rule book has been thrown out of the window and bad design is justified with voodoo.
They *can* definitely affect timbre, if the effects of the filter reach down into audible frequencies, but no properly designed filter should do that. On the other hand, there is a lot of "audiophile" gear where the rule book has been thrown out of the window and bad design is justified with voodoo.
I always assumed that perceptions of sound staging were related to SPL, the source, physical positioning of speakers and the acoustical profile of the room. How would a digital filter affect sound staging? Assuming the DAC had a flat response and harmonic components 100 dB below the musical signal, how could that possibly be?
I've heard at least 5 times now in the past two days that the filter options on DACs can *definitely* affect timbre and sound staging. The article cited in support of this can be found here :
http://www.audiostream.com/content/what-ar...ence-labs-techn (http://www.audiostream.com/content/what-are-digital-filters-and-why-are-they-requried-todays-audio-dacs-resonessence-labs-techn)
Now I'm no EE, so I'm not qualified to make objective statements concerning this, but I would like to know from those experienced in the relevant fields what their take on this issue is. Thank you very much in advance!
Sorry, just to add : looking at those graphs, the THD was 0.0013 and 0.0026 respectively. That is ridiculously - no, extraordinarily low distortion. Are there any perceptual studies that show what our threshold of audibility is to harmonic distortion?
Can and do are two different things. Personally, I'm only interested in the "do" and only when it can be proven through double blind experimentation. That said I wouldn't bother engaging people who can't satisfy this criteria.
Hi there,
The claims I hear for audible differences existing for different electronics amount to frequency response, noise and distortion.
These distortion graphs are between the two I/V stages of a DAC :
(http://i982.photobucket.com/albums/ae301/gbyleveldt1/IC-IV_zpsb757b132.png)
And :
(http://i982.photobucket.com/albums/ae301/gbyleveldt1/DiscIVB_zps28f9f192.png)
The claim is that these DACs measure flat, but the measurements show the effects of certain filters at play. Looking at the distortion components, is this enough of a difference to cause gross audible differences? The claims I hear are that different filters used in DAC design can result in audible differences irrespective if that DAC measured perfectly flat.
What do you think?
This might be a trick question because nobody I know can tell much about linear distortion by looking at an impulse response. What little I can discern from it says that it is probably fee of audible artifacts.
The FFT of its output spectrum shows nothing that could be audible All spurious responses sum to about 97 dB down.
Ask about the results of a level-matched, time-synched DBT. ;-)
The author's name looked familiar.
http://www.stereophile.com/writer/185 (http://www.stereophile.com/writer/185)
Sorry, just to add : looking at those graphs, the THD was 0.0013 and 0.0026 respectively. That is ridiculously - no, extraordinarily low distortion. Are there any perceptual studies that show what our threshold of audibility is to harmonic distortion?
The audibility of nonlinear distortion varies tremendously depending on the musical context and the SPL of the sounds you are auditioning. For music that has maximum masking potential, even a percent or two of first or second order distortion can go undetected.
I've done some work contriving musical sounds that maximize the audibility of nonlinear distortion and came up with something that would enable detection with some effort of less than 0.1 % (-60 dB) nonlinear distortion. Someplace around 0.03% nonlinearity, even that wouldn't work. That is -70 dB.
The audibility of nonlinear distortion is generally maximized with SPLs that are around 75-85 dB. Increasing the SPL so that it is above 80-90 dB appears to make the ear more nonlinear and its own nonlinearity masks the distortion in the source.
The usual rule of thumb is that if all spurious responses due to nonlinear distortion are >=80 dB down, then nobody is going to detect it, no matter what kind of nonlinearity and no matter what kind of music, speech or sound effects your source is. If you are really conservative use 100 dB down, and then you are totally safe.
One of these days I'll book a flight to the States where I would love to meet some of you guys. Go through a few DBTs as well.... drinks on me.
I've done some work contriving musical sounds that maximize the audibility of nonlinear distortion and came up with something that would enable detection with some effort of less than 0.1 % (-60 dB) nonlinear distortion. Someplace around 0.03% nonlinearity, even that wouldn't work. That is -70 dB.
But with typical music content 0.1% nonlinear distortion is not going to be noticed?
They *can* definitely affect timbre, if the effects of the filter reach down into audible frequencies, but no properly designed filter should do that. On the other hand, there is a lot of "audiophile" gear where the rule book has been thrown out of the window and bad design is justified with voodoo.
I always assumed that perceptions of sound staging were related to SPL, the source, physical positioning of speakers and the acoustical profile of the room. How would a digital filter affect sound staging? Assuming the DAC had a flat response and harmonic components 100 dB below the musical signal, how could that possibly be?
Again, what I said was "they *can* definitely affect timbre, if the effects of the filter reach down into audible frequencies, but no properly designed filter should do that". As in the filter will only affect timbre if it's cutoff frequency or slope is low enough to affect audible frequencies - something easily visible on a frequency response plot (but not on an impulse plot or intermodulation/distortion plot).
Now I'm no EE, so I'm not qualified to make objective statements concerning this, but I would like to know from those experienced in the relevant fields what their take on this issue is. Thank you very much in advance!
Well, I studied EE before moving into CS, and have since then worked on both audio and embedded signal processing...
I've done some work contriving musical sounds that maximize the audibility of nonlinear distortion and came up with something that would enable detection with some effort of less than 0.1 % (-60 dB) nonlinear distortion. Someplace around 0.03% nonlinearity, even that wouldn't work. That is -70 dB.
But with typical music content 0.1% nonlinear distortion is not going to be noticed?
True.
This may be a silly question, but in terms of harmonic distortion (the "nice" kind) what would be a typical figure that one would expect to cause audible differences? 1%, 5% etc? I realise that one can't get a complete picture from a single THD measurement, but I would like to try and put things to perspective by comparing the miniscule harmonic distortion in the graphs to real world tube amps that people claim make the sound so lovely to listen to.
I realise the answer is probably going to be "it depends on ..." but any input on this would be appreciated. Thanks!
This may be a silly question, but in terms of harmonic distortion (the "nice" kind) what would be a typical figure that one would expect to cause audible differences? 1%, 5% etc?
Its not a silly question, and that applies to all honest questions no matter how basic.
The audibility of nonlinear distortion as you guessed is very dependent on the order of the nonlinear distortion and the sound being distorted.
The myth that even order distortion is more euphonic odd order distortion probably due to a number of facts:
(1) The first even order is second and the first odd order is third and as we all know 3>2. This relates to masking. Any frequency component that is closer to the fundamental is more likely to be masked by it. The second harmonic is always closer to the fundamental than the third harmonic and therefore it is more likely to be effectively masked. Note that push-pull amplifiers are inherently resistant to generating even-order distortion.
(2) Musical instruments are themselves generators of harmonics. The motion of the sound generators in musical instruments such as reeds, lips and strings produce varying amounts of nonlinearity. Plucking strings produces a sharp-edged stimulus that is something like an impulse, and a true impulse contains all frequencies. The commonly-used resonantors usually respond to harmonics as well. The natural harmonics in the music thus mask any harmonics that are created by failings of the equipment. It is possibly ironic that some musical instruments deliver much more energy as harmonics than as the fundamental component of the musical note.
The counterpoint is that any nonlinearity that produces harmonics also produces intermodulation. Intermodulation tends to produce component frequencies at frequencies that are not harmonically related to the fundamental. Note that most musical instruments have individual mechanisms for producing (different strings, different pipes, different resonators) each note. That tends to minimize intermodulation products and makes them more noticeable when produced by electronic or acoustic equipment.
I realize that one can't get a complete picture from a single THD measurement, but I would like to try and put things to perspective by comparing the miniscule harmonic distortion in the graphs to real world tube amps that people claim make the sound so lovely to listen to.
I realise the answer is probably going to be "it depends on ..." but any input on this would be appreciated. Thanks!
Under ideal conditions for masking, up to 10% or more THD can be masked by the music.
For example 10% THD is commonly used as a performance limit for subwoofers. This probably has a lot to do with the fact that bass notes then to be played one at a time and that most bass-producing musical instruments produce more energy as harmonics than fundamentals.
So there are your common limits of audibility for nonlinear distortion - from 0.1% to 10%. Current accepted practice seems to be to get under 0.1% by a factor of 2 or 3 wherever possible. Getting under 0.001% is always done for the purpose of inventorying and displaying leading zeroes! ;-)
So if I told the DAC designer that his distortion graphs show harmonic products a 100 dB below the fundamental that basically invalidates his graphs as evidence of his user preference? You reckon 70 dB down is enough for distortion to be largely inaudible given the masking effects of music.
Sorry Arnold, how do you work out the harmonic percentages in terms of how far below the signal is from the fundamental? I noticed 0.1 would be -60 below the fundamental and 0.03 would be -70. Please explain how that works. Thanks!
The dB scale is logarithmic. 20 dB means a difference by a factor 10. So 60 dB is a thousandfold difference and 70 dB is approximately three times that.
Sorry Arnold, how do you work out the harmonic percentages in terms of how far below the signal is from the fundamental? I noticed 0.1 would be -60 below the fundamental and 0.03 would be -70. Please explain how that works. Thanks!
Wikipedia: Decibel (http://en.wikipedia.org/wiki/Decibel)
The myth that even order distortion is more euphonic odd order distortion probably due to a number of facts:
One factor for it not being a myth is that the 2nd and 4th harmonics are the same note as the fundamental, just 1 and 2 octaves up. The 3rd isn't. Neither is the 6th, but by the time you get to the 6th harmonic the level is usually pretty far down.
The myth that even order distortion is more euphonic odd order distortion probably due to a number of facts:
One factor for it not being a myth is that the 2nd and 4th harmonics are the same note as the fundamental, just 1 and 2 octaves up. The 3rd isn't. Neither is the 6th, but by the time you get to the 6th harmonic the level is usually pretty far down.
While that is a fact, it might not keep the claim from being a myth.
While the case for the ugliness of aharmonic spurious responses seems pretty clear, the fact that the transposition doesn't end up at the same note seems to have a far weaker case.
Sorry Arnold, how do you work out the harmonic percentages in terms of how far below the signal is from the fundamental? I noticed 0.1 would be -60 below the fundamental and 0.03 would be -70. Please explain how that works. Thanks!
Wikipedia: Decibel (http://en.wikipedia.org/wiki/Decibel)
Yes That table in the upper right hand corner of the cited page makes it all pretty clear. Neat reference!
What do you guys think of this :
"Designers at ESS and other DAC manufacturers consult with these audiophiles to ascertain which dacs required work amongst their Sigma Delta designs to make it a class leader today.
They literally had to invent new methods of diagnostics looking outside the traditional tools of audio analysis. They have moved way past measurements of frequencu response, distortion and signal to noise!
Longish video and somewhat technical but your jaw will drop as all our differences, disputes and concerns have and are being addressed compehensively in an ongoing fashion. The short version: some people can hear differences that others can't and the designers have come to accept it and working with the "golden ears" now on their staff or actively consulting they are engineering improvements to their dacs that are superior to previous ones."
http://www.youtube.com/watch?v=1CkyrDIGzOE (http://www.youtube.com/watch?v=1CkyrDIGzOE)
Let me know what you think of the video.
It's true some people can hear things others can not. They are called "delusional". [I didn't watch the video since they are either kooks or crooks. Either way, a waste of my time, and yours.]
But mzil, just take your time to watch it. It might be delusional crap, but just tell me what you think. When you have time. That DAC designer is now asking me to provide evidence that distortion 100 dB down is inaudible, to contradict his findings.
Tell him you'd like him to disprove the existence of the Flying Spaghetti Monster.
Seriously, you are wasting your time. I would really consider what you're hoping to achieve arguing with a religious nut who will likely always have a gaggle of clueless disciples.
But mzil, just take your time to watch it. It might be delusional crap, but just tell me what you think. When you have time. That DAC designer is now asking me to provide evidence that distortion 100 dB down is inaudible, to contradict his findings.
Worrying about things that are 100 dB down is like thinking your car will have discernibly better acceleration if you shake the dirt out of the floor mats to reduce the overall mass.
edit to add: Here's an on line demo which sort of shows you how quiet -100 dB is compared to normal levels, at least in a rough way. Be careful not to click on the top part of the chart untl you get a feel for how it works. [I take no responsibility for damage to hearing or equipment. Read the warnings there too.]
Hearing Test (http://www.phys.unsw.edu.au/jw/hearing.html)
Everything else being equal, the car will have better acceleration. I don't have a better analogy at the moment, but if I did it would also account for the topic title.
Just out of curiosity, did any of you watch the whole presentation? What did you think? Any good points there .. or is it basically snake oil?
I watched all of it, sometimes pausing to read the slides, and I found it very interesting. Not because he claims that a small minority of people in some extremely rare corner cases is able to hear surprisingly small differences, but because of all the other interesting stuff he explains along the way.
Apart from the obvious goal to make people choose ESS's products over the competitor's, as far as I'm concerned there's no snake oil in the presentation. I would recommend it to anyone with a general interest in audio technology, perception or chaos theory.
Worrying about things that are 100 dB down is like thinking your car will have discernibly better acceleration if you shake the dirt out of the floor mats to reduce the overall mass.
edit to add: Here's an on line demo which sort of shows you how quiet -100 dB is compared to normal levels, at least in a rough way. Be careful not to click on the top part of the chart untl you get a feel for how it works. [I take no responsibility for damage to hearing or equipment. Read the warnings there too.]
Hearing Test (http://www.phys.unsw.edu.au/jw/hearing.html)
Sorry slightly off topic, but could you use the results from that Hearing Test to help determine the best setup for a parametric EQ?
I watched all of it, sometimes pausing to read the slides, and I found it very interesting. Not because he claims that a small minority of people in some extremely rare corner cases is able to hear surprisingly small differences, but because of all the other interesting stuff he explains along the way.
Apart from the obvious goal to make people choose ESS's products over the competitor's, as far as I'm concerned there's no snake oil in the presentation. I would recommend it to anyone with a general interest in audio technology, perception or chaos theory.
A lot of it is messed up. Explained earlier in the thread mentioned below. I would hope that nobody who lacks a good BS manager spends any time with it.
Start reading critique here: http://www.hydrogenaudio.org/forums/index....st&p=824801 (http://www.hydrogenaudio.org/forums/index.php?showtopic=99572&view=findpost&p=824801)
It's true some people can hear things others can not.
It is very true. For example most audio professionals who depend on listening, as they get older they acquire one or more younger engineers that they train how to listen, and then rely on their ears.
Others get really good at interpreting measurements.
I know people with well known names who can't hear stuff that is very clear to me, and I'm 66 and have had one bad ear since I was 8. I have done well in ABX tests regardless.
They are called "delusional". [I didn't watch the video since they are either kooks or crooks. Either way, a waste of my time, and yours.]
Depends if the hearing of things that others cannot passes some kind of sniff test, of which ABX is a very good one.
I never call it delusional though, I call it Illusional which is a very carefully chosen word.
It is true that delusions may set the stage for the credibility that they give their illusional experiences.
I still remember Dr Larry Greenhill (Stereophile and Audio Magazine reviewer) whose day job was being a shrink at a state mental facility comment on which well known names in audio he believed were uimmm (my word) nuts.
It is very true. For example most audio professionals who depend on listening, as they get older they acquire one or more younger engineers that they train how to listen, and then rely on their ears.
Sure, but then one can't really dismiss the claim that they can hear distortion products 100 dB below the fundamental, right? I mean the claim sounds bonkers, but if people can perceive things that others can't, then ...
Unless there any studies of human audibility thresholds that are statistically relevant that show humans can't perceive distortion below a certain point?
You may be mischaracterizing whet Arnold said. It's true that hearing varies, and that training and experience have a significant impact on the ability to hear more subtle aspects of sound, but decades of research have pretty accurately set a lower limit on what is possible in human hearing. We also have a pretty good understanding of the mechanism of hearing, and this is also consistent with what we have found in practice.
Sure, but then one can't really dismiss the claim that they can hear distortion products 100 dB below the fundamental, right?
But you can turn the burden of evidence around - if they can hear distortion products 100 dB below the fundamental, it should be easy for them to prove it under controlled lab conditions. Have they? If not, why not?
The burden to provide support falls on the people making far-fetched claims. Claiming to hear a mosquito buzzing around the business end of an active jackhammer needs supporting evidence.
Is it up to you to prove there isn't a pink elephant orbiting the planet Neptune?
Just out of curiosity, did any of you watch the whole presentation? What did you think? Any good points there .. or is it basically snake oil?
If you are curious enough you might have searched the forum in order to avoid a double-post...
http://www.hydrogenaudio.org/forums/index....showtopic=99039 (http://www.hydrogenaudio.org/forums/index.php?showtopic=99039)
I was having a debate with a DAC designer who cited these graphs as reasons for his sonic preferences, especially the first graph. He claimed that the harmonic distortion was the reason why he preferred the sound. But you are saying that the harmonic components are so far down, 100 dB's down and therefore inaudible.
Here are the impulse response graphs, respectively :
In the beginning of this thread i suggested you may look at the frequencies this happens. Did you ask the "DAC designer" you are debating with meanwhile for graphs showing this or even the frequency response of these filters? Did you try on your own?
Worrying about things that are 100 dB down is like thinking your car will have discernibly better acceleration if you shake the dirt out of the floor mats to reduce the overall mass.
edit to add: Here's an on line demo which sort of shows you how quiet -100 dB is compared to normal levels, at least in a rough way. Be careful not to click on the top part of the chart untl you get a feel for how it works. [I take no responsibility for damage to hearing or equipment. Read the warnings there too.]
Hearing Test (http://www.phys.unsw.edu.au/jw/hearing.html)
Sorry slightly off topic, but could you use the results from that Hearing Test to help determine the best setup for a parametric EQ?
It is a crude test just meant for fun. In real audiometry they use a single blind protocol (you don't test yourself and see a score card showing your level selection from of the previous trial frequency, which may bias your next selection) and calibrated headphones playing at a pre-established, measured level [Not "Here, to start, set this knob to, er, comfortable level, whatever that means to you"] .
Unless there any studies of human audibility thresholds that are statistically relevant that show humans can't perceive distortion below a certain point?
There are shelves full of journals full of masking data probing what kind of signal makes what other kind of signal inaudible, for different listeners, at whatever level. The data has been collected for a century. Google auditory masking. Try Google Scholar to see the depth of the research (it's well mapped for many animals on this planet, not just humans!).
Cheers,
David.
It is a crude test just meant for fun. In real audiometry they use a single blind protocol (you don't test yourself and see a score card showing your level selection from of the previous trial frequency, which may bias your next selection) and calibrated headphones playing at a pre-established, measured level [Not "Here, to start, set this knob to, er, comfortable level, whatever that means to you"] .
Thanks for the info
It is very true. For example most audio professionals who depend on listening, as they get older they acquire one or more younger engineers that they train how to listen, and then rely on their ears.
Sure, but then one can't really dismiss the claim that they can hear distortion products 100 dB below the fundamental, right?
Which they?
You are missing the point. The 100 dB number is better than what the best young listeners can achieve.
I mean the claim sounds bonkers, but if people can perceive things that others can't, then ...
A key part of listener selection and training is monitoring people's sensitivity as compared to other listeners. One of the tests we applied was to look at the individual scores and see if any of the listeners did consistently better. Some people who did consistently poorly were taken aside and encouraged to not participate in the group tests. Some tests were by invitiation and we were careful who we invited, and I mean this in a good way.
When a few people did exceptionally well, we did additional trials with them to see if their results were real or a statistical anomaly. You know if you flip coins, you may get long runs of heads or tails. If you do enough flips, additional flips even things out. Same thing can happen with listeners.
We also had situations where people would do worse than guessing. We watched the tests carefully and noticed that this could happen when the listeners were unconsciously communicating, such as groaning when they were having a tough time hearing differences. One of the things about ABX is that if you are having problems hearing a difference, you generally know it long before you score the test.
Later on I did quite a bit of work with tests involving a known difference that was varied in size. I would start listeners out with large differences, and make the difference smaller in subsequent tests. Eventually, the listeners with working with the natural difference. This made it possible to see who was capable and who was trying versus those who were not trying all that hard or were simply not getting it.
Unless there any studies of human audibility thresholds that are statistically relevant that show humans can't perceive distortion below a certain point?
There are usually a number of different ways to estimate the audibility of a given kind of artifact. You can do different kinds of tests involving the same artifact. Sometimes you can estimate the audibility of an artifact from things like the Fletcher Munson curves or masking curves. In those cases where the same artifact is tested for a number of independent ways, they often give very consistent results.
For example you can estimate the audibility of an artifact using both ABX and ABC/hr tests, which are actually very dissimilar listening tests. The key is that no matter how you present the alternatives or how you ask people to compare them, when they are hearing something their responses correlate with the presence or absence of the stimulus, and when they are not hearing something their responses become random.
There are also other kinds of testing besides presenting samples. You can give people an unmarked knob that varies the level of the artifact, and let them turn it up and down until they feel that they have homed in on the level where it is just audible. This same scheme is used in audiometry - ear tests. It is arguably not exactly double blind, and a creative listener can spoof it, but it can also give very good results.
When people say that a lot of this stuff has been tested to death, they are not kidding!