|
|
Post by monkumonku on Mar 11, 2013 12:32:16 GMT -5
I agree with you that a lot of the music currently available isn't mastered very well. But you must remember that this has always been true. Just like most tube fanatics conveniently fail to remember all the pathetically BAD sounding tube amps that were out there fifty years ago, record lovers conveniently forget all the crummy sounding records. If anything, the drawback to digital audio is that, by making it easier and cheaper to make a recording, it encourages people to do so who are less than qualified. (When it cost $50k to master a record, at least that limited the number of total incompetents who were able to raise the money to do a pressing.) As for vacuum tubes, I think that there is simply a rather sad misunderstanding going on there. Tubes sound different because they make large amounts of second harmonic distortion. Most tube power amps also have very low damping factors, which makes them sound different in a different way (they interact differently with various speakers to produce different inaccuracies). Please understand that I have nothing against any of this, and happily defend the rights of tube lovers everywhere to choose the sound that they like. Let's just avoid muddying the waters by pretending that the difference isn't simply a coloration introduced by the tube equipment. (We age bourbon in oak barrels to get the flavor from the wood, and nobody complains about the wood mucking up the taste; this is the same thing; and not at all a bad thing - as long as we all understand the facts.) As for vinyl, my LEAST fond memories of vinyl include lots of record hiss, a few million crackles, and flinching every time a loud POP made me wonder if the damage was permanent or would come off with a good careful cleaning. I find it funny how record lovers never seem to remember THAT side of records. To me, a few pops per minute make a record downright unlistenable - no matter how good the quality of the audio between them. (And, yes, I know that, if you're very, very, VERY careful, you can avoid MOST of the pops... but not the surface noise.) That's my whole point to begin with. On a measurement plot it seems like digital artifacts are way too small to lose any sleep over them. However, their measured size says little or nothing about their value in terms of audibility by humans, let alone in terms of musical impact. The measured distortion caused by vacuum tubes is preposterously huge compared to that of a 16-bit 44.1 kHz CD, yet there are people who prefer listening to 24-bit 192 kHz digital recordings through vacuum tubes regardless. There is not a linear relationship between the measured accuracy of an audio signal and the perceived sound quality or musical satisfaction this accuracy engenders. I have heard several thousands of audio CDs, and not a single one of them comes even remotely close to the sound quality of the best LPs. As for 24-bit 96 kHz (and 192 kHz) digital, yes of course it sounds much better than vinyl records, but then not all music is available in digital Hi Res... Worse, even when it is available anyway after all, it still very often falls short compared to vinyl mostly due to p!$$ poor mastering. The mind is a powerful thing and of course influences our perceptions of everything. Here is an example: Back in junior high, I had a bunch of 45's. I'm still very fond of these songs. One of them, "Cherish" by the Association, I would play over and over on my lousy record player because I had dedicated it to this certain girl on whom I had a huge crush. It got played so much with that crummy needle that it developed all sorts of noise and scratches on the record. All of that noise and all of those scratches became part of the song for me. When I got the "Association's Greatest Hits" album and played it, the song sounded very different to me because it was lacking in noise and scratches. In fact, it sounded dull - like something was missing (which in fact there was something missing, all the noise from the 45's). I was so familiar with the scratches that I could tell what song it was going to be just from hearing the scratches before the song started. Even to this day when I hear the song, I expect to hear the noise along with it. Hearing the 45 evokes a different feeling in me compared to hearing a "pristine" version of the recording. Maybe that's the same with vinyl - you grow up with it and get used to hearing all the surface noise, ticks, pops, hiss, etc. and to hear a CD that is lacking in those things makes you think the CD is unnatural. Cold, sterile, inorganic..  |
|
|
|
Post by mgbpuff on Mar 11, 2013 13:51:45 GMT -5
Keith said, "Tubes sound different because they make large amounts of second harmonic distortion."
This is not a true statement. Tubes in fact are more linear in their operating range than semi-conductors. The 2nd and higher even harmonics come from the clipping characteristic of tubes vs that of transistors. Tubes clip more gradually giving more even harmonic distortion components than the sharp clipping of the transistor which yields higher odd harmonic distortion components. The other reason for differences in harmonics between tube and ss amps come from the many single ended tube amps available. SS amps are not typically single ended (although they can be). Single ended amps tend to clip asymmetrically causing a lot of even ordered harmonics and there is no cancellation of those even harmonic components like occurs in push-pull design. Push-pull design is almost universal in ss amp design and causes the even harmonic distortion components to be automatically eliminated leaving only odd order distortion harmonics. This plus the high odd harmonic content clipping character of semi-conductors can give a hard sounding aspect. Unclipped push-pull tube and ss amps should sound very similar.
|
|
|
|
Post by yves on Mar 11, 2013 15:51:48 GMT -5
One problem with ABX testing is it can only be used to prove a difference can be heard. Meaning, it can NOT be used to prove NO difference can be heard. As a direct consequence, people who portray themselves as "true objectivists" by claiming differences too small to be detected in ABX testing should be categorized as way unimportant to the music listener are typically nothing but subjectivists in denial. www.avguide.com/forums/blind-listening-tests-are-flawed-editorialOn a measurement plot it seems like digital artifacts are way too small to lose any sleep over them. However, their measured size says little or nothing about their value in terms of audibility by humans . . . . Correct. To determine whether they are audible you need to do ABX testing. You might find this of interest: www.computeraudiophile.com/blogs/mitchco/16-44-vs-24-192-experiment-163/No doubt. That simply tells you that some people prefer distorted signals to clean ones. Probably because they have become accustomed to the former and accept it as "normal." True, for the reason above. |
|
|
|
Post by garym on Mar 11, 2013 16:19:51 GMT -5
One problem with ABX testing is it can only be used to prove a difference can be heard. Meaning, it can NOT be used to prove NO difference can be heard. That makes no sense. Either a difference can be heard or it cannot. The two outcomes are mutually exclusive. Well, Yves, it is difficult for me to understand how anyone could deem a difference he can't hear as important to him. www.avguide.com/forums/blind-listening-tests-are-flawed-editorialSee the comment in your link from gmgraves. Also keep in mind that the "differences" that no one can hear cannot be seen on test instruments either. |
|
|
|
Post by yves on Mar 11, 2013 17:33:47 GMT -5
I was referring to Fourier analysis limited by the computational power of current DAW workstations, as found in most professional studio applications today. When presented smallband test tones, the human ear can only discern the difference between frequencies that are farther than 3.6 Hz apart. However, likely due to a phenomenon called beating, people have been able to reliably hear differences between broadband audio signals where a frequency had been shifted by as little as 1 Hz. I was using the above example as a methaphor to further illustate how the outcome of a listening test often depends more on the specific context in which the listening experiment was conducted than it does on that which the experiment was designed to examine in the first place. Another fine example of how things can easily tend to get misleading IMO is the noise floor of vinyl records. It's not just the measured amount of noise that matters, but also, to a very large degree, the characteristics of the noise as well as how exactly said noise is perceived in the presence of other imperfections, and vice versa. Because, some types of imperfections have a psychoacoustic tendency to modify the way we perceive other types of imperfections. This can be either good or bad, but we should never underestimate the brain's ability to adapt and fill in missing information. This brings me back to my point, which boils down to the fact I am not only interested in trying to prove a difference can be heard, but also in whether what I am doing has made an improvement. Rather than listening to one thing and then another thing without knowing what it is I am listening to, I go into listening tests to determine when I stop hearing a distortion. I try to find answers to questions like "did it bring me closer to what the artist intended?" and "was I better able to tell what the instruments are?" I can't always do that if I am not somehow in control of the parameters. Further, ABX tests fail to eliminate things which I find important on alot of time scales and, because I have memory, if I listen to one thing and then another thing it's going to inform me when I go back, so therefore I can't perform the same test twice. The fact I am already familiar with the test sound before I actually take the ABX test can easily spoil things so, contrary to popular belief, listening fatigue is not the only limiting factor here. As a result, I don't tend to do alot of ABX tests. I agree that Hi Res digital has come a long way and is vastly superior to analog, but CDs aren't exactly Hi Res and probably that's the important reason why, most of the time, LPs still sound much better to me than CDs. I am thoroughly confused with your comment about Fourier analysis. Fourier analysis is a mathematical process performed on data. It is as accurate as the original data and the computational accuracy at which it is performed. As such, it can be as accurate as you like - with no limit whatsoever. In contrast, our ability to recognize differences in the frequency domain does indeed have some measurable limit. The ability of we humans to SEE distortion on an oscilloscope trace (by actually looking at the shape of the wave) is limited to distortion over about 5% in most cases. ALL statements about noise masking and such are heavily dependent on the conditions under which they are measured. Compression schemes like MP3 and Vorbis were worked out with the intent that MOST PEOPLE WOULDN'T NOTICE THE DIFFERENCE WITH MOST MUSIC. Nobody, including in most cases the inventors, would actually suggest that there was no difference audible to anyone with any source material. However, if you want to evaluate the "viability" of digital music reproduction, you must also consider the limitations of ALL the alternatives. Is it possible to distinguish a digital artifact that is at -100 dB below full scale? Personally I suspect that it probably is not... but maybe it's possible. (Are we turning the gain up so loud that a 0 dB signal would kill us? If so, is that a "reasonable circumstance" under which to compare things?) What I can say for certain is that, if our sample were recorded on a vinyl record instead, that signal at -100 dB would be BURIED 30 dB BELOW THE RECORD SURFACE NOISE... And a typical CD has a THD in the 0.00x% range, whereas a record is probably in the single digit percentages at higher recording levels (with luck). So, are you suggesting that a digital system, with a really good S/N ratio, and some artifacts at -100 dB that might or might not be audible to some people, isn't superior to an analog recording, where we KNOW that there is a noise floor at only -70 dB, which is audible to everyone? Or are you suggesting that the 0.003% THD on a DAC is MORE audible than the much higher level of THD that you get from vinyl. I would agree that characterizing anything as "perfect" is going a bit too far, but it seems obvious to me that digital has gotten way closer than analog.... and that it also has the potential to get even closer (while analog technology is pretty well "maxxed out"). As for feedback from the brain adjusting the sensitivity of the receptors in your ear... it's been pretty well known for a long time that your brain/ear implement a form of dynamic range compression to improve sensitivity at low sound levels... when things get really quiet your ears get a lot more sensitive... and this sounds like they've finally worked out some details about how it occurs. I agree this video is MUCH more explanatory than that article that claimed 192 kHz music downloads make no sense. However, it still fails to address the undeniable fact substantial evidence found in modern auditory neuroscience indicates severe flaws in our age-old understanding of what's audible under what circumstances and what isn't, as well as the fact inaudible sounds affect brain activity in such way they become part of our listening experience as they alter our perceptions. Our ability to discern differences in the frequency domain is TEN TIMES more accurate than Fourier analysis and likely about a hundred times more sensitive than a thirty-year-old oscilloscope. Hair cells PHYSICALLY react to nerve impulses coming from the brain. Think about that for a moment when you hear someone say oh it's more than 100 dB below the music so it SHOULD be inaudible. |
|
|
|
Post by yves on Mar 11, 2013 17:42:38 GMT -5
As a matter of true fact, I prefer listening to vinyl after it's been carefully transferred to digital Hi Res and click and pop sounds have been (manually) removed in software such as ClickRepair and iZotope RX Advanced 2. The mind is a powerful thing and of course influences our perceptions of everything. Here is an example: Back in junior high, I had a bunch of 45's. I'm still very fond of these songs. One of them, "Cherish" by the Association, I would play over and over on my lousy record player because I had dedicated it to this certain girl on whom I had a huge crush. It got played so much with that crummy needle that it developed all sorts of noise and scratches on the record. All of that noise and all of those scratches became part of the song for me. When I got the "Association's Greatest Hits" album and played it, the song sounded very different to me because it was lacking in noise and scratches. In fact, it sounded dull - like something was missing (which in fact there was something missing, all the noise from the 45's). I was so familiar with the scratches that I could tell what song it was going to be just from hearing the scratches before the song started. Even to this day when I hear the song, I expect to hear the noise along with it. Hearing the 45 evokes a different feeling in me compared to hearing a "pristine" version of the recording. Maybe that's the same with vinyl - you grow up with it and get used to hearing all the surface noise, ticks, pops, hiss, etc. and to hear a CD that is lacking in those things makes you think the CD is unnatural. Cold, sterile, inorganic.. |
|
|
|
Post by yves on Mar 11, 2013 18:30:27 GMT -5
That makes no sense. Either a difference can be heard or it cannot. The two outcomes are mutually exclusive. They are mutually exclusive, but "hearing no difference" is an outcome that's effectively biased by the circumstances in which an ABX test is conducted. Listening tests, double-blind or not, cause people to listen specifically for differences rather than to just enjoy music. This not only causes stress, but it also alters our perception as we focus on different aspects of the sound than what we'd normally focus on (i.e., what we'd focus on during a normal music listening session). Just because we aren't always consciously aware of information that reaches our brain via our ears, doesn't also mean our brain cannot respond to it. jn.physiology.org/content/83/6/3548.longIf you read other comments in that same discussion thread, perhaps you will see why others disagree. Further, there is no reason to assume test instruments are sufficiently sensitive to measure every important aspect of sound (well... not unless you can prove it, which, let me guess... you can't, so this is clearly yet another perfect reason for me to question your socalled "objectivism" even more, if you catch my drift). |
|
|
|
Post by garym on Mar 11, 2013 19:03:30 GMT -5
They are mutually exclusive, but "hearing no difference" is an outcome that's effectively biased by the circumstances in which an ABX test is conducted. Listening tests, double-blind or not, cause people to listen specifically for differences rather than to just enjoy music. This not only causes stress, but it also alters our perception as we focus on different aspects of the sound than what we'd normally focus on (i.e., what we'd focus on during a normal music listening session). I.e., there are differences we can hear only if we are not listening for them. If we listen for them, we can't hear them. Sorry, but that argument fails the Popper criterion. It is unfalsifiable. Oh, a great deal of information reaches our ears, and other senses, of which we're not consciously aware. We have selective attention --- a good thing, since otherwise we'd be overwhelmed with (mostly inconsequential) sensory input. Our brains may even respond in some electrochemical way to these data. But remaining unaware of sensory signals to which we're actively attending is another matter. Hearing music is a conscious experience. Subconscious events occurring in the brain that do not effect that conscious experience are no more relevant to it than the many other events that may simultaneously be going on in the brain, such as those controlling digestion or respiration. Or if there is an effect on the conscious experience, the burden of proof for that hypothesis --- identifying and demonstrating that effect ---would fall upon its proponents. |
|
|
|
Post by arthurz on Mar 11, 2013 20:07:51 GMT -5
One problem with ABX testing is it can only be used to prove a difference can be heard. Meaning, it can NOT be used to prove NO difference can be heard. That makes no sense. Either a difference can be heard or it cannot. The two outcomes are mutually exclusive. Gary is right. You can't prove that no one in the whole world can't hear the difference, but you can prove it for your test subjects (and then, by induction, the whole world if your sample is large enough). The way you prove it is you look at the distribution of answers. If there's no difference, people will either say explicitly they heard no difference, or guess, in which case they'll be approx. 50% correct. There's a good probability that _some_ people will be 80% correct, but if the proportion of those people matches the binomial distribution, you still assert that the result is pure chance, even though the audiophile subject would surely claim to have golden ears (easily falsifiable by continuing the experiment). |
|
|
|
Post by arthurz on Mar 11, 2013 20:19:21 GMT -5
Further, there is no reason to assume test instruments are sufficiently sensitive to measure every important aspect of sound (well... not unless you can prove it, which, let me guess... you can't, so this is clearly yet another perfect reason for me to question your socalled "objectivism" even more, if you catch my drift). I think there's EVERY reason to assume a sufficiently sophisticated instrument will outperform human senses. There's not a single sense I'm aware of where this wouldn't be the case. Is there even a shred of evidence to the contrary. As Ethan Winer notes ( www.skeptic.com/eskeptic/10-01-06/ ), we know precisely what we need to measure to understand everything about audio fidelity: noise, frequency response, distortion, and time-based errors). On which of these parameters are you suggesting humans can outperform machines? How exactly? So far, you suggested the ability to hear noise below exceeding 100 dB SNR, but (a) machines CAN definitely do it, and (b) you had no evidence whatsoever that humans could. |
|
harri009
Emo VIPs 
ReferenceAnalog.com
Posts: 1,425
|
Post by harri009 on Mar 11, 2013 20:24:49 GMT -5
Some also say compression in the ears cause a/b testing to be invalid due to the relative swaping. They say it works with other things as well, like colors with our eyes. If you take 2 colors of close shade and A/B them you cannot see a difference. Only when you compare side by side can you see the difference :/
|
|
|
|
Post by arthurz on Mar 11, 2013 20:33:05 GMT -5
Some also say compression in the ears cause a/b testing to be invalid due to the relative swaping. They say it works with other things as well, like colors with our eyes. If you take 2 colors of close shade and A/B them you cannot see a difference. Only when you compare side by side can you see the difference :/ This is invalid as a blanket statement against A/B testing. For example, to test the difference between 16/48 and 24/192 audio, you could easily play a song, and seamlessly, but randomly switch between the two types of sources. If people can't detect when the switch takes place, there's no audible difference. Surely you'd agree it's similar to your side-by-side color comparison example. |
|
|
|
Post by yves on Mar 11, 2013 20:36:29 GMT -5
I.e., there are differences we can hear only if we are not listening for them. If we listen for them, we can't hear them. Sorry, but that argument fails the Popper criterion. It is unfalsifiable. Ethan Winer has published an interesting video at the start of which James "JJ" Johnson talks about expectation bias. However, the very first thing he points out is if you listen for high frequencies, you might not hear low frequencies and, if you listen for low frequencies, you might not hear high frequencies. Meaning, our perception can be altered, or "steered", depending on what aspect of sound we try to listen for, or "focus on". If we do NOT try to compare sounds, we focus on aspects other than the ones we focus on if we DO try to compare sounds. This introduces bias, and it's the kind of bias ABX typically fails to eliminate. In that same video, Poppy Crum explains how the brain can fill in missing information based on cognitive response. The brain remembers what things SHOULD sound like, and it alters our perception accordingly so that we actually hear sounds that don't even exist. It just fills in the blanks though we're not consciously aware that it does. Same thing happens when we listen to music. Most of us are more or less equally familiar with music than with spoken language, so the principles of psycholinguistics are closely related to the way we perceive music. Here is the video. So, how exactly were you going to eliminate expectation bias using ABX tests without fundamentally changing the effect the music itself has on our perception of this music? Use test tones rather than music as the test sound for ABX? No can do, since we're trying to examine the effects the listening equipment has, or doesn't have, on our music listening experience aka not test tones. ...Which is exactly why I am referring to subconscious events that DO affect our conscious experience. |
|
harri009
Emo VIPs
ReferenceAnalog.com
Posts: 1,425
|
Post by harri009 on Mar 11, 2013 20:37:08 GMT -5
Side by side is at the same time.
|
|
harri009
Emo VIPs
ReferenceAnalog.com
Posts: 1,425
|
Post by harri009 on Mar 11, 2013 20:37:35 GMT -5
Which would of course be A and B not A then B. But unpractacle in audio. Heck I have own the equipment I own because I can clearly hear a difference. I do usually A/B things, but can see where if someone is listening at reference levels A/B would be worthless
|
|
|
|
Post by GreenKiwi on Mar 11, 2013 20:43:39 GMT -5
|
|
harri009
Emo VIPs
ReferenceAnalog.com
Posts: 1,425
|
Post by harri009 on Mar 11, 2013 20:50:45 GMT -5
Data knows where it's at. I wonder if he is listening to SACD, 24/192, or MP3. Probably not Vinyl. |
|
|
|
Post by yves on Mar 11, 2013 20:52:11 GMT -5
If ABX tests are fundamentally flawed, which is a valid hypothesis as it is backed up by top experts in auditory neuroscience, your claim is nothing more than subjectivism in denial. That makes no sense. Either a difference can be heard or it cannot. The two outcomes are mutually exclusive. Gary is right. You can't prove that no one in the whole world can't hear the difference, but you can prove it for your test subjects (and then, by induction, the whole world if your sample is large enough). The way you prove it is you look at the distribution of answers. If there's no difference, people will either say explicitly they heard no difference, or guess, in which case they'll be approx. 50% correct. There's a good probability that _some_ people will be 80% correct, but if the proportion of those people matches the binomial distribution, you still assert that the result is pure chance, even though the audiophile subject would surely claim to have golden ears (easily falsifiable by continuing the experiment). |
|
|
|
Post by yves on Mar 11, 2013 21:02:41 GMT -5
The brain is part of the hearing system. No machine known to mankind can perfectly measure everything that goes on inside the brain, re perception and the conscious present. I think there's EVERY reason to assume a sufficiently sophisticated instrument will outperform human senses. There's not a single sense I'm aware of where this wouldn't be the case. Is there even a shred of evidence to the contrary. As Ethan Winer notes ( www.skeptic.com/eskeptic/10-01-06/ ), we know precisely what we need to measure to understand everything about audio fidelity: noise, frequency response, distortion, and time-based errors). On which of these parameters are you suggesting humans can outperform machines? How exactly? So far, you suggested the ability to hear noise below exceeding 100 dB SNR, but (a) machines CAN definitely do it, and (b) you had no evidence whatsoever that humans could. |
|
|
|
Post by monkumonku on Mar 11, 2013 21:58:42 GMT -5
I guess my question for this discussion is, how much of it is really just academic eggheadism?  |
|
