[bigshot]

What is the fastest attack in a musical instrument? Something like 3-5 milliseconds is my guess. We need to define the amounts of time we're talking about and compare that to the scale of time as it occurs in music.

Timbre and pitch are defined by the sustained part of the note, not the attack. I would think that would be obvious.

 

[csglinux]

1. Would you prefer if I'd used violas, cellos or basses, how about wind instruments? The fastest attacks are from struck (percussion) instruments but then percussion instruments are typically placed at the back of a music ensemble and when close mic'ed, are typically processed to significantly change the transients.
2. No I haven't. Using your 100m example, 20kHz air absorption would be about 52dB! Note to others, this is just air absorption of high frequencies and not related to loss of energy due to distance (roughly defined by the inverse square law) which would need to be added to the air absorption figures if we wanted to calculate total attenuation over distance, however, for this argument we can ignore the inverse square law because it affects all freqs equally. By way of comparison, 30kHz at 100m would be attenuated by air by about 94dB, 5kH by about 4dB and 500Hz about 0.3dB. The shape/freq content of a transient is therefore significantly changed by air absorption. Obviously, 100m is a bit extreme but on the other hand, in the real world we're not just talking about air absorption of high freqs, there are other materials (walls, floors, furniture and of course people, musicians and audience) which absorb high freqs many times more effectively than air.

The point I was making is that attack transients of acoustic instruments contain a wider freq content than the note itself (the sustain portion of the ADSR envelope) and change significantly in the real world, to the point of disappearing entirely. If we take another, completely different example, say a flute. At close range, you'll hear an attack which includes considerable high freq hiss, the sound of the air itself being blown by the musician across the flute's mouthpiece and this attack is fairly explosive as it's typically created by trapping and suddenly releasing air, similar to saying "Ta". At a normal listening distance in a performance environment this transient mostly disappears and we just hear relatively smooth notes. The same is true with all wind instruments and with the human voice itself (speaking or singing), where we have close proximity transients which we do not perceive at any normal listening distance. Specifically attacks/transients called plosives and essing, caused by the pronunciation of letters such as P, B, K/C, S, T and others, as any experienced recording engineer knows well, as close mic'ing the human voice typically requires significant processing of the transients.

I'm using the above and what I've stated previously to explain that transients are massively variable, to the point of virtually disappearing completely in some real life performance situations, to counter Watt's assertion that timbre and pitch is dependent on transients. If we listen to a violin section (or flute or whatever) playing a note from say the back of a concert hall 100m away and we've completely lost the transients, are we unable to discern the pitch of that note or even that it is a violin section? Admittedly, getting a seat at the back of the concert hall is not the best place to be but why would anyone ever buy a seat at the back of a concert hall if they couldn't discern timbre and therefore tell what instruments are playing (or even that they are musical instruments), what notes (pitches) they're playing or when the musicians start or stop playing them? In addition to the obvious nonsense of Watt's claims regarding pitch and timbre being dependent on transients, we've also got the issue of Watt's claims about transients and sound stage, which are potentially somewhat more valid but certainly not to the extent Watt's is making out. Again, we can remove all the transients from various sounds and instruments and still create a soundstage and transients are smeared in time by real life acoustics or signal processing of closely miced instruments, so it's hard to see what his 4us has to do with anything in a real life music mix, regardless of whether he's on about transient duration, rise times or delay?

G

Hi buddy, I promise this is not a gotcha question - I am genuinely curious to know where you got those numbers for your attenuation of a 20 kHz wave in air. I was talking about nothing other than the effects of viscosity on acoustic waves - no 1/r^2 decay, no carpeted floors, no anechoic panels in the ceiling, no other absorptive materials obscuring the direct line of sight from source to the observer - just the effects of molecular viscosity. Your numbers are orders of magnitude off from Stoke's law.

But ok, let's do the Devil's advocate thing and roll with your numbers and assume that delta_t1 is really, really large for every instrument I'm ever going to listen to. It still doesn't matter. I am not talking about errors in the reproduction of delta_t1. As per my earlier post, I am talking about errors in delta_t2, which doesn't necessarily require a very tiny delta_t1 (though there's obviously a correlation).

One other thought on the 100 m (or even 50 m?) length concert hall. If that's the type of music you listen to exclusively, I can see why soundstage simply might not be all that important to you. You could well be experiencing the entire orchestra within an arc of only about 10 degrees. I have some classical binaural recordings and they're fairly pointless - very little difference in spatial imaging from standard non-binaural recordings. Recordings of acoustic guitar, piano, vocals, etc., spread out in a studio are way more effective for binaural and would, I imagine, be the types of recording that would most benefit from an infinite sinc filter in a DAC (assuming, of course, that humans really can detect these timing differences).

I get your final point about these timing numbers being very small. That point isn't lost on me. But at this stage, none of us on this thread can conclusively confirm or deny whether this might have an audible impact on the soundstage. Opinions are fine of course, but voicing them over and over again doesn't make them any more valid. And I won't mention any names here, but a single anecdotal example of one person not being able to hear any differences in any equipment they've owned since the 1930s isn't proof that all past, current and future DACs inevitably sound absolutely identical and transparent. There are several points that should be very obviously wrong with such a statement.

Anyway, I'm going to be taking a break for a while, but look forward to the vitriol, hate mail, death-threats, etc., when I return next week.

Happy Thanksgiving to my fellow Americans

 

[bigshot]

Spatial imaging is primarily a function of the mix. For playback it's possible to mess up the imaging by putting speakers in weird places, but if you have the proper placement, it is what it is... what the mixer intended it to be. Of course with headphones, there really is no soundstage. Just spatial imaging along a line through the center of the ears. Headstage if you will.

Some folks hypothesize that soundstage is controlled by microscopic slivers of time. Well, that is pretty convenient because I find that especially in this particular headphone forum, the word "soundstage" is thrown around so loosely, it's become attached to the placebo effect. It only makes sense that high end audio salesmen would latch onto that nice juicy vague term and come up with ways that they can do placebo better than the next guy. (I prefer bargain basement placebo to $8,000 placebo myself.)

So if we have a hypothesis saying that microscopic bits of time are vital to reproducing soundstage, it's time for the person presenting the hypothesis to test it with controlled tests and show a real world correlation. Until then, it's just an opinion or even worse, empty sales pitch. I see no reason to do more than nod my head and look vaguely interested until some proofs start piling up.

 

[danadam]

Rob's claim is just that simpler DAC filters can have errors in delta_t2 as large as 100 microseconds

On hydrogenaudio they have a set of test files with impulses, one with no delay between channels and others with delays ranging from 5 to 140 µs. Listening to them from my laptop's headphone out, I can say that:

• yes, 100 µs is not that hard to notice, comparing to no delay (at least with impulses, probably wouldn't be so easy with music)
• it consistently gets harder to notice the lower you go
• at each level the difference is consistent when you repeat the comparison (barring the listening fatigue)
• 20 µs is the limit that I could confirm for me

I guess that the conclusion you could make from this is that if there is any error there, it is less than 20 µs. Otherwise it wouldn't be so consistent.

So unless Rob's claim is about something entirely different, that number seems to be much too big.

 

[bigshot]

For music you can probably up it several whole orders of magnitude. In general perceptual thresholds are MUCH higher in music than they are in tones.

 

[gregorio]

[1] Hi buddy, I promise this is not a gotcha question - I am genuinely curious to know where you got those numbers for your attenuation of a 20 kHz wave in air. ... Your numbers are orders of magnitude off from Stoke's law.
[2] If that's the type of music you listen to exclusively, I can see why soundstage simply might not be all that important to you.
[3] Recordings of acoustic guitar, piano, vocals, etc., spread out in a studio are way more effective for binaural and would, I imagine, be the types of recording that would most benefit from an infinite sinc filter in a DAC (assuming, of course, that humans really can detect these timing differences).
[4] I get your final point about these timing numbers being very small. That point isn't lost on me. But at this stage, none of us on this thread can conclusively confirm or deny whether this might have an audible impact on the soundstage.

1. National Physical laboratory - "The attenuation of sound in air due to viscous, thermal and rotational loss mechanisms is simply proportional to f 2. However, losses due to vibrational relaxation of oxygen molecules are generally much greater than those due to the classical processes, and the attenuation of sound varies significantly with temperature, water-vapour content and frequency. A method for calculating the absorption at a given temperature, humidity, and pressure can be found in ISO 9613-1 (1993)." I used this online calculator, and the amount of attenuation is in agreement with my personal observations.
2. No, I don't exclusively listen to orchestral music, I listen to many/most genres.
3. In virtually all studio recorded music, timing is relatively unimportant. Left/Right positioning is achieved almost exclusively by panning (level differentiation between left and right speakers) of mono sources, rather than by timing differentiation. As most studio music is multi-tracked, the timing is usually all over the place, many milli-secs. Likewise depth/presence is achieved with a combination of artificial reverb (algorithms), EQ, volume and compression, all of which (except volume) change and smear transients in time anyway.
4. Except apparently Rob Watts! It's still rather unclear what timing errors he's referring to, jitter and timing errors between channels is way out of the picture or not even in the picture. If he's talking about delta_t2 how does that make any difference if everything has a delay of t2? OR, how does a filter or anything else know what is a transient and what is another waveform and only apply delta_t2 to a transient? He can't be referring to linear phase FIR filters because they are, well linear phase and indeed he states he's talking about NOS, IIR and minimum phase filters. So, he's talking about post-ringing, although later slides go on about pre-ringing but then he say's that everyone is wrong that pre-ringing is bad? He seems to be saying that his new filter can achieve exactly the same accuracy as a linear phase FIR filter but without any pre-ringing. Great, so he is curing something which isn't "bad" to start with and most DACs oversample anyway!

G

 

[bigshot]

By the way, a defined soundstage is probably more important in orchestral music than any other genre. There is a very specific placement that classical recordings shoot for. You want to have the orchestra spread out in front of you with violins on the left and cellos and bass on the right. The woodwinds should be lined up across the middle. Ideally you should be able to tell the exact placement in the middle where the flute is as opposed to the oboe. This kind of precision is a lot easier with a center channel than it is with 2 channel systems with a phantom center. The other night I was watching the new blu-ray of Dr Dolittle from the 60s. It was released in 6 track Westrex sound originally. It had scenes where characters would be talking and they would get up and walk across the screen. The dialogue followed their position exactly. That is harder to do with 2 channel too.

Most people around here who use the term "soundstage" don't know what it really means. It isn't L/R channel separation. It's using the distance of the speakers from the listener to create a flat plane 15 feet or so ahead of the listening position where sounds are placed across the plane. You can't do that with regular headphones, because the sound is a straight line through the middle of the head.

 

[csglinux]

1. National Physical laboratory - "The attenuation of sound in air due to viscous, thermal and rotational loss mechanisms is simply proportional to f 2. However, losses due to vibrational relaxation of oxygen molecules are generally much greater than those due to the classical processes, and the attenuation of sound varies significantly with temperature, water-vapour content and frequency. A method for calculating the absorption at a given temperature, humidity, and pressure can be found in ISO 9613-1 (1993)." I used this online calculator, and the amount of attenuation is in agreement with my personal observations.
2. No, I don't exclusively listen to orchestral music, I listen to many/most genres.
3. In virtually all studio recorded music, timing is relatively unimportant. Left/Right positioning is achieved almost exclusively by panning (level differentiation between left and right speakers) of mono sources, rather than by timing differentiation. As most studio music is multi-tracked, the timing is usually all over the place, many milli-secs. Likewise depth/presence is achieved with a combination of artificial reverb (algorithms), EQ, volume and compression, all of which (except volume) change and smear transients in time anyway.
4. Except apparently Rob Watts! It's still rather unclear what timing errors he's referring to, jitter and timing errors between channels is way out of the picture or not even in the picture. If he's talking about delta_t2 how does that make any difference if everything has a delay of t2? OR, how does a filter or anything else know what is a transient and what is another waveform and only apply delta_t2 to a transient? He can't be referring to linear phase FIR filters because they are, well linear phase and indeed he states he's talking about NOS, IIR and minimum phase filters. So, he's talking about post-ringing, although later slides go on about pre-ringing but then he say's that everyone is wrong that pre-ringing is bad? He seems to be saying that his new filter can achieve exactly the same accuracy as a linear phase FIR filter but without any pre-ringing. Great, so he is curing something which isn't "bad" to start with and most DACs oversample anyway!

G

Many thanks for that link! I tried to find the derivation of the equations used in this applet, but when you click on the link for the "ISO 9613-1 Formulae for the attenuation(absorption) of sound during propagation outdoors", it takes you to a site for some English travel agency(?). There's maybe(?) a link to the formulas used, but with no explanations or derivations: http://www.sengpielaudio.com/LuftdaempfungFormel.htm There's some suggestion that their formula is valid only for 50 Hz to 10 kHz, but even for a 10 kHz wave, their attenuation coefficient differs by two-orders of magnitude from that of Stokes': https://en.wikipedia.org/wiki/Stokes'_law_of_sound_attenuation In this sengpiel formula, we have: P_t = P_i.exp(-x.a.s), with s = distance, a = attenuation coefficient and x = 1/(10.log_10(e^2)) ~ 0.115, so their effective attenuation per meter is actually x.a and this is the source of the discrepancy. I'm struggling to believe that Stokes has been wrong for the last 150 years or that any humidity or temperature variations would explain differences this large. Can anybody explain the discrepancy? I did notice one issue in the definition for 'a' in the sengpiel link. They have dimensions of ~(t^{-2}) and ~(t^{-1}) for the first and second terms of 'a', respectively. These are both inconsistent with the dimensions of 'a', which should be nepers per meter, or ~(1/m) in SI.

For the sake of the Rob/Chord DAC debate, let's assume that Stokes was wrong. It still doesn't matter. Somewhere in the depths of recorded music there's going to be a starting and a stopping of a note or a sound that takes place over a relatively short period of time, given that we're able to perceive that at one moment we don't hear the sound and the next moment we do. If a twig snaps in the forest - there's no way we could reliably locate it only from the teeny tiny difference in amplitude reaching our left and right ears. If that were true, our brains would have to be constantly adjusting for external factors we couldn't possibly know about, i.e., wind velocity and the fact that part of our hat had just slipped slightly to one side. Statements like "virtually", or "most of the time" (or the terrible errors we currently make in recorded music) are no doubt quite correct, but if we're making errors AT ALL in the D->A process, then there's room for improvement and Rob may have a valid argument. I'm not disputing that a large part of the perceived directivity comes simply from the amplitude delta reaching the left and right ear, but - again - this is a "virtually/most of the time" argument that doesn't cover all our bases. The link that @danadam posted to the time-delayed impulse responses makes for a good demo of the twig-snapping experiment. You aren't hearing any difference in amplitude in any of these cases - only a difference in timing. You should be able to clearly hear the shift in the position of the click - at least for the larger times. Even down to 20 microseconds, @danadam - you should be patting yourself on the back; if our cilia really worked like a sum of Fourier modes, you effectively just heard up to 50 kHz, which, considering we can't actually hear past 22 kHz, is impressive, no?

The issue is that the errors in delta_t2 from most DAC filters aren't constant over playback of a complex signal. If they were, it wouldn't be a problem, because all you'd have done is permanently shifted the violin section a little to the left. However, if the timing error itself varies as a function of time, there's no way to triangulate a single time-invariant location of that sound.

I couldn't find the part where Rob talks about zero pre-ringing. I don't think that's what he's saying, because a perfect reconstruction (up to Nyquist) must contain both pre- and post-ringing around a sufficiently-sharp transient. But maybe I missed something here. (If I did, please point me to the relevant post.)

 

[ironmine]

DACs all sound the same, the differences are due to expectation bias. They are not differentiated at all in an ABX/DBT.
Sonic differences between DACs is the myth.

You gotta be kidding.

My current W4S DAC-2 ($1500) sounds very different (better) than my former April Music Stello DA100 ($900).

When my friend offered to me his W4S DAC-2, I took it for auditioning and comparing against April Music, but I was highly skeptical. I thought that I already had the dac that was "expensive enough and good enough".

But the moment I replaced one dac for another, I instantly heard a difference (more presence, better separation, superior soundstage). That difference mattered to me, I bought W4S DAC-2 and sold Stello DA100. With the new DAC, escaping from my room into a new audio reality is just easier. It's more convincing, W4S fools my brain faster and more masterfully.

With April Music, achieving the same "trance-like" state was more difficult: everything had to be just perfect (my physiological mood and physical condition, the recording had to be superb, etc.). Even then it took my mind longer to "adjust".

But with W4S, I am able to get involved into music almost every time I sit down for a listen. It just opens a door to it and invites me to step into it.

 

[Killcomic]

You gotta be kidding.

My current W4S DAC-2 ($1500) sounds very different (better) than my former April Music Stello DA100 ($900).

When my friend offered to me his W4S DAC-2, I took it for auditioning and comparing against April Music, but I was highly skeptical. I thought that I already had the dac that was "expensive enough and good enough".

But the moment I replaced one dac for another, I instantly heard a difference (more presence, better separation, superior soundstage). That difference mattered to me, I bought W4S DAC-2 and sold Stello DA100. With the new DAC, escaping from my room into a new audio reality, is much easier. It's just more convincing, W4S fools my brain faster and more masterfully.

With April Music, achieving the same "trance-like" state was more difficult: everything had to be just perfect (my physiological mood and physical condition, the recording had to be superb, etc.). Even then it took my mind longer to "adjust".

But with W4S, I am able to get involved into music almost every time I sit down for a listen. It just opens a door to it and invites me to step into it.

Have you done a blind test to compare both units and see if you are able to test the difference?

 

[ironmine]

Have you done a blind test to compare both units and see if you are able to test the difference?

No, I did not. It was technically difficult to do.

But the difference is obvious. The change of the dac changed my relation to the music. Previously, I almost forced myself to listen until the end of an album (I am a disciplined guy), the sound was boring. 30% of the listening sessions were a joy, but 70% were just "so-so". I was many times unable to be carried away by music.

Now, I look forward to each evening. I know that unless something extraordinary happens, I will reach this trancelike state when the room disappears and music starts flowing in 3D images.

 

[bigshot]

No, I did not. It was technically difficult to do.

It's not hard to do at all. You just need a switch box and a way to match the line level. I've done it dozens and dozens of times with everything from a $40 Walmart DVD player to an Oppo HA-1. Everything sounds exactly the same. If it didn't sound the same, I would suspect something was wrong with one of the units. Digital audio should be audibly transparent. The specs should exceed your ears' ability to hear. If you hear a difference, something isn't transparent, so it isn't performing up to specs. I would suspect a high end DAC for performing out of spec before I would suspect a midrange player. High end stuff is sometimes deliberately colored to sound different. They call it a "house sound". I call it out of spec.

 

[ironmine]

It's not hard to do at all. You just need a switch box and a way to match the line level. I've done it dozens and dozens of times with everything from a $40 Walmart DVD player to an Oppo HA-1. Everything sounds exactly the same. If it didn't sound the same, I would suspect something was wrong with one of the units. Digital audio should be audibly transparent. The specs should exceed your ears' ability to hear. If you hear a difference, something isn't transparent, so it isn't performing up to specs. I would suspect a high end DAC for performing out of spec before I would suspect a midrange player. High end stuff is sometimes deliberately colored to sound different. They call it a "house sound". I call it out of spec.

You are so wrong and what you say runs so contrary to my experience that I don't even want to argue with you. I heard many DACs (and CD-players and sound cards), compared them in the same setup and all of them sounded different.

I feel sorry for you, truly, that you don't hear a difference. Work on your speakers, amps and room (room acoustics). If it does not help go to a doctor and let him check your ears.

 

[pinnahertz]

No, I did not. It was technically difficult to do.

But the difference is obvious. The change of the dac changed my relation to the music. Previously, I almost forced myself to listen until the end of an album (I am a disciplined guy), the sound was boring. 30% of the listening sessions were a joy, but 70% were just "so-so". I was many times unable to be carried away by music.

Now, I look forward to each evening. I know that unless something extraordinary happens, I will reach this trancelike state when the room disappears and music starts flowing in 3D images.

Yes, it IS technically difficult, but also technically necessary for proof. This has been debated to absolute death. Expectation bias is incredibly powerful. It's why placebo tablets "cure" disease. However, I will also say that placebo is completely valid for the same reason. If you think your DAC has made a positive improvement, and you enjoy it, then it actually has. I'm not going to force you to do the actual scientific controlled test to prove or disprove it. Frankly, I don't care. But your argument, as valid as it seems, lacks proof, other than your own belief system. And frankly, that trumps an ABX test because an ABX test is so difficult to do that it can be ignored in favor of a strong belief, which is not only very easy but already is in place.

 

[pinnahertz]

It's not hard to do at all. You just need a switch box and a way to match the line level.....

No, that's not correct. First, that "switch box" has to be a bit more than a switch in a box, it has to include a true blind X position, equivalent to A or B, but unknown to the tester. And nothing about how the box works can provide a "tell". It's not easy at all, and actually fairly expensive. Level matching...that will elude every casual and most serious hobbyists because it requires knowledge of electronics and test and measurement.

ABX/DBT of a pair of DACS is one of the most difficult things to pull off correctly, and that's why it's almost NEVER done. That doesn't mean it can't be, and certainly those that have been done support my assertions.