I am new here, so please spare me the flames.
This is a fascinating thread & I have enjoyed reading it. You folks obviously work in EE & understand the workings of very sophisticated A/D/A conversion technologies, which is not simple. I have but a few comments.
1) The self noise of some of the best mics made are still worse than 16/44.1K specs:
http://www.neumann.com/infopool/mics/produkte.php?ProdID=u87ai
I guarantee the consoles are as bad or worse. Old analog consoles are very “in” these days. Their specs suck. They do sound pretty good, though. Most recording studios are glad to have 24/96 Deltas these days. A few have gone to 24/192 due to pressure. The big trend now is 5.1.
2) I work with recording technologists & they use DSP. Almost all of them. Most of what you hear is run through a Pro Tools rig at one time or another. The exception being any acoustic recordings that go right to CD. But then, does the mastering lathe/processor change them, too?
3) Which brings me to a question I have for so many golden ears: What is your benchmark? I mean listening to music. One can train oneself to hear certain aspects of timbre, etc. Are the tweaks better or different? The only real benchmark is to go out & listen to the real thing. Acoustic music preformed in a good space. No reinforcement.
4) Acoustic space; it has been mentioned here, too. As well it should be as no other aspect of a live recording or sound is more important. Funny thing is studios try to deaden everything and add the reverb or whatever (often DSP) later. Many beloved recordings are coloured by the space they were recorded in. That is not bad. It makes Carnegie Hall what it is.
5) Any form of EQ is going to change (i.e. degrade) the sound quality. DSP, analog passive or active. Fact of life. It is best to avoid it, but again we go back to the space. Sometimes it is the only way to compensate for the acoustic space or the listeners' preferences.
6) I saw a panel discussion @ AES about DACs. As usual, peopled by the best minds in audio. The conclusion was 24/192K was a waste of data space. 24/96K was accepted as so close to the source audio that it was almost indistinguishable. These were the guys that developed the Delta Sigma DACs, and the older non 1-bits, too. They consider it an evolution. Also, the panel admitted that the same DACs sounded different in every implementation. So, analog design & components still are critical, as we know they are.
And speaking of the listener, is that not the real issue here? Every person I know “hears” different. Their listening needs are different. I have some very nice turntables gathering dust. I listen to CDs, due to their convenience. That is life.
Thanks for the informative tread. Sorry for the rant.
This is a fascinating thread & I have enjoyed reading it. You folks obviously work in EE & understand the workings of very sophisticated A/D/A conversion technologies, which is not simple. I have but a few comments.
1) The self noise of some of the best mics made are still worse than 16/44.1K specs:
http://www.neumann.com/infopool/mics/produkte.php?ProdID=u87ai
I guarantee the consoles are as bad or worse. Old analog consoles are very “in” these days. Their specs suck. They do sound pretty good, though. Most recording studios are glad to have 24/96 Deltas these days. A few have gone to 24/192 due to pressure. The big trend now is 5.1.
2) I work with recording technologists & they use DSP. Almost all of them. Most of what you hear is run through a Pro Tools rig at one time or another. The exception being any acoustic recordings that go right to CD. But then, does the mastering lathe/processor change them, too?
3) Which brings me to a question I have for so many golden ears: What is your benchmark? I mean listening to music. One can train oneself to hear certain aspects of timbre, etc. Are the tweaks better or different? The only real benchmark is to go out & listen to the real thing. Acoustic music preformed in a good space. No reinforcement.
4) Acoustic space; it has been mentioned here, too. As well it should be as no other aspect of a live recording or sound is more important. Funny thing is studios try to deaden everything and add the reverb or whatever (often DSP) later. Many beloved recordings are coloured by the space they were recorded in. That is not bad. It makes Carnegie Hall what it is.
5) Any form of EQ is going to change (i.e. degrade) the sound quality. DSP, analog passive or active. Fact of life. It is best to avoid it, but again we go back to the space. Sometimes it is the only way to compensate for the acoustic space or the listeners' preferences.
6) I saw a panel discussion @ AES about DACs. As usual, peopled by the best minds in audio. The conclusion was 24/192K was a waste of data space. 24/96K was accepted as so close to the source audio that it was almost indistinguishable. These were the guys that developed the Delta Sigma DACs, and the older non 1-bits, too. They consider it an evolution. Also, the panel admitted that the same DACs sounded different in every implementation. So, analog design & components still are critical, as we know they are.
And speaking of the listener, is that not the real issue here? Every person I know “hears” different. Their listening needs are different. I have some very nice turntables gathering dust. I listen to CDs, due to their convenience. That is life.
Thanks for the informative tread. Sorry for the rant.
the real thing
Warren,
Thanks to your post it’s getting even more interesting. IMO you hit the relevant nails right on their heads, without implying there‘s only one stairway to audio nirvana. Hence no need for flaming or bashing in a thread that’s all about trying to understand the emotional effects of technical/commercial choices (like 1- or multibit). As a music lover who sometimes likes to experiment in that tricky and spooky EE domain, I found KYW’s ‘little rant’ very honest (as a non–native speaker i saw this English word for the first time, but now I’ll never forget it 🙂 ).
Regarding benchmarking against the real thing: listening to (live) recordings that are made in familiar halls helped me lot a lot in grasping and valuating differences, and in the end, defining the qualtity of all the components in my set.
In line with Carlos’ recommendations: may I suggest to listen to a cd from Frans Brüggen’s Ocrhestra of the Eighteenth Century*? It contains two symphonies, both recorded live on 5/1985, the first one (Mozart’s #40) in the Concertgebouw (Amsterdam), the second (Beethoven #1) in Vredenburg (Utrecht). These belong without any doubt to the finest concert halls in my country (The Netherlands) and even world wide (as experts say -- I can only judge for London, Paris and Chicago), but rather different in many respects. The acoustic space is very well captured on both, IMHO. Believe me, with a well tuned system, space and music are projected in your own room in a frightening realistic way (if Carlos would be a classic music lover, beware......).
As this is an international forum, not everybody is able to second or reject this judgement (one more good reason to come to The Netherlands and visit these halls.....). Still this cd is very useful to find out if a system is capable of revealing these benchmarking differences. Maybe even better, as you are not hindered by memories of the ambience and atmosphere that specific night you were there, a kind of bias from one of those other tricky and spooky domains....
*Philips CD: 416 329-2; on vinyl (!): 416 329-1
PS BTW, I use a dCS Ring DAC, only for red book. Quote "This is a discrete, proprietary, 5 bit 64 times oversampling architecture that avoids the limitations inherent in the conventional one bit and multibit off-the-shelf converter ICs that are the basis of most other D/A converters."
You know, I really don't mind. In my system it just sounds very good, as others (from different breeds) have done. In the end it's all about synergy, isn't it?
Warren,
Thanks to your post it’s getting even more interesting. IMO you hit the relevant nails right on their heads, without implying there‘s only one stairway to audio nirvana. Hence no need for flaming or bashing in a thread that’s all about trying to understand the emotional effects of technical/commercial choices (like 1- or multibit). As a music lover who sometimes likes to experiment in that tricky and spooky EE domain, I found KYW’s ‘little rant’ very honest (as a non–native speaker i saw this English word for the first time, but now I’ll never forget it 🙂 ).
Regarding benchmarking against the real thing: listening to (live) recordings that are made in familiar halls helped me lot a lot in grasping and valuating differences, and in the end, defining the qualtity of all the components in my set.
In line with Carlos’ recommendations: may I suggest to listen to a cd from Frans Brüggen’s Ocrhestra of the Eighteenth Century*? It contains two symphonies, both recorded live on 5/1985, the first one (Mozart’s #40) in the Concertgebouw (Amsterdam), the second (Beethoven #1) in Vredenburg (Utrecht). These belong without any doubt to the finest concert halls in my country (The Netherlands) and even world wide (as experts say -- I can only judge for London, Paris and Chicago), but rather different in many respects. The acoustic space is very well captured on both, IMHO. Believe me, with a well tuned system, space and music are projected in your own room in a frightening realistic way (if Carlos would be a classic music lover, beware......).
As this is an international forum, not everybody is able to second or reject this judgement (one more good reason to come to The Netherlands and visit these halls.....). Still this cd is very useful to find out if a system is capable of revealing these benchmarking differences. Maybe even better, as you are not hindered by memories of the ambience and atmosphere that specific night you were there, a kind of bias from one of those other tricky and spooky domains....
*Philips CD: 416 329-2; on vinyl (!): 416 329-1
PS BTW, I use a dCS Ring DAC, only for red book. Quote "This is a discrete, proprietary, 5 bit 64 times oversampling architecture that avoids the limitations inherent in the conventional one bit and multibit off-the-shelf converter ICs that are the basis of most other D/A converters."
You know, I really don't mind. In my system it just sounds very good, as others (from different breeds) have done. In the end it's all about synergy, isn't it?
Gentlemen, thank you for sharing your thoughts in this great discussion.
Carlos, which one of Dan Steely's recording are you referring to ? I found three in all - two by Warner Brothers one recorded by MCA.
Still searching for Philips 416 329-2...
Carlos, which one of Dan Steely's recording are you referring to ? I found three in all - two by Warner Brothers one recorded by MCA.
Still searching for Philips 416 329-2...
percy said:
Sadly I don't remember, as the disc is not mine.
I can ask a friend.
I would also love to have it.
carlosfm said:
Exactly.
But I don't have any kind of animosity against SACD.
I just say that a good PCM recording is superiour.
Even then, I gave an example of a very good sounding SACD disc (although from a PCM master).
Please, if you find it, buy it.
What I do thing is that the story repeats again and again.
People are pushed to new formats by the marketing guys, while sometimes it's not a step forward.
It happened with CD (against vinyl), it happens again with SACD, the story is the same.
Reading the maketing talk is all advantages, it's a "new and very superiour, high-resolution format".
Is it?
YES !
TNT said:
Well, the evolution of the sound quality from CD since the beginning/mid 80's was huge.
The problem was not the players, it was on the recordings / CD pressings, which sounded horrible.
It is to be expected that the new formats take the same route, IF they survive.
And if the major labels stop joking with the consumer.
The typical consumer who buys these "super resolution" formats wants and demands a better sounding disc, not only just the same music on another format.
And if it sounds worse, you can fell like assaulted.
I can't help but thinking that new formats are always released too soon.
Indeed. But what I would like to know (and it's something you tech heads might be able to answer) is if the SACD is so inherently "bad" that it's a lost cause, or can it be improved on?
Unfortunately, we here will have no say about the future of the SACD, or DVD-A. That is in the hands of the masses. But I hope the SACD will survive in one shape or other, like vinyl has. I think it can be a great format for smaller record companies and music artists that perhaps want to experiment with surround sound on a more serious level.
phn said:
In my experience, and take it just as that, SACD can sound good it taken from 24bit PCM masters.😀 😀 😀
Otherwise it sounds excessively smooth and unnatural.
But some people like it this way.
Maby this situation will also improve.
phn said:
Yes we have.
We are the consumers.
And quite a demanding consumers, as we demand quality.
Realize that we are a very small minority.
Most people don't care about quality, a download from the net is fine, even paying the same price for miserable quality compressed music.
Here in Lisbon, at Fnac (one of the most complete stores to buy any kind of music here) I could only see 4 DVD-A (unknown) discs mixed with DVD-V and 15~20 SACDs.
Are these successful formats?
I remember that CD was a boom since the first years.
I know... now the discs look the same on all formats.
Also, labels have an incredible bad taste to pick the music they release on the new formats.
As I like to listen to music above all, and 99.9% of the music I love is on CD only, that's what I buy.
They don't give me any other choice.
I don't buy a disc just because it sounds good.
If I don't like the music, that's not for me.
I'm not a typical audiophile with the room full of test discs.
I concur that early CDs were poor. The dithering (from the mastering equipment) was troublesome & listening to an early CD starting from music to a very quiet passage demonstrates this issue. They are much better now. The MP3 craze amazes me. They are tolerable for certain reasons (unavailability of material being one of them) but are rite with artifacts.
I agree that the technology is relevant, but not the most important aspect of the experience. The art of music is. Also, many of the specifications that I have seen bandied about are the product of testing devices that use A/D conversion to do their job. I use several of these devices in my work and all use conversion technologies to attain aural information (SMART, RTA, FFT, TEF, etc.). These are the very devices that we use to indicate what works well or not. Somewhat cyclic reasoning there. Personal auditioning of the music is the true test.
Steely Dan is a great source for testing for me; I know their music very well, I have heard it on several good systems & headphones & even live! They were/are one of the best produced rock bands ever recorded.
Lourens:
Thank you for your kind words.
There are several very fine halls (and great museums) in your country, which I intend to visit someday. I have an interest both informally and professionally in architectural acoustics, so this has immediate bearing for me. I will look up the recordings that you have mentioned. As they are from the early 1980’s, I gather they were early digital or analog.
If you come to New York City, New York, USA in the near future, let me show you a fine hall we have here. I have some professional association with Carnegie Hall (The Stern Hall and the new Zankel Hall) and I would be happy to assist in that endeavor. Please feel free to drop me a PM.
Warren
I agree that the technology is relevant, but not the most important aspect of the experience. The art of music is. Also, many of the specifications that I have seen bandied about are the product of testing devices that use A/D conversion to do their job. I use several of these devices in my work and all use conversion technologies to attain aural information (SMART, RTA, FFT, TEF, etc.). These are the very devices that we use to indicate what works well or not. Somewhat cyclic reasoning there. Personal auditioning of the music is the true test.
Steely Dan is a great source for testing for me; I know their music very well, I have heard it on several good systems & headphones & even live! They were/are one of the best produced rock bands ever recorded.
Lourens:
Thank you for your kind words.
There are several very fine halls (and great museums) in your country, which I intend to visit someday. I have an interest both informally and professionally in architectural acoustics, so this has immediate bearing for me. I will look up the recordings that you have mentioned. As they are from the early 1980’s, I gather they were early digital or analog.
If you come to New York City, New York, USA in the near future, let me show you a fine hall we have here. I have some professional association with Carnegie Hall (The Stern Hall and the new Zankel Hall) and I would be happy to assist in that endeavor. Please feel free to drop me a PM.
Warren
In the same rack as the Audio Precision gear I use at work, I've got a CD player with Aja loaded into it. Even if the AP gear says that the equipment I design and the DSP code that I write is "good", I invariably listen to the output with real music.warren o said:
And you'd be surprised how many different ways you can screw something up in DSP software that can make music sound *awful*, and still have the AP gear put out a perfect frequency response, SNR and THD plots. 😀
carlosfm said:
Besides that, unfortunately, the supposedly "big" names do not even learn from their mistakes.
CD was a quantum jump. From walkman to discman. No tangled tapes, no degradation over time...etc.. This "convenience" was promised and delivered hence the success - imo. I guess the new formats really need to deliver on their promise of "better sound" otherwise they are doomed. Not many (me for one) care much about surround sound for music so that leaves only one thing - "better" sound.
If you consider you as single individual as a too small sample size, then you miss the point of ABX. You can never prove universal applicability, but you can make sure that the placebo effect is eliminated from _your_ results.
If the difference is so obvious, it should be no problem to show .05 significance.
Where do you get that from?
Type-2 error is huge only if your number of trials is too small (<20), *and* is considered only if you fail to show hearing difference confidently. But, being intelligent person, you must get the gut feeling that something is wrong when you predict reliable hearing of difference when you know what you listen, but in blind ABX do no better than random chance. Its the unbelievable fact that this *does* happen why both: snake oil placebo and ABX test exist.
ABX test was NOT created to "prove" that golden-ears are wrong. It was created to *distinguish* genuine audibility from brain playing tricks. Its main goal is to know that audible difference is real. Number of trials is *increasing* sensitivity to subtle differences. Its much worse to claim you hear subtle differences by simple few trials, and be done with it. In that case *both* your type-2 error is huge, *and* you "reliably eliminate subtle but percieved differences from detection."
My point was that it has been shown again and again that the bold confidence in "hearing" difference mysteriously disappears as soon as being subjected to controlled ABX testing. Its that mysterious effect that raises caution. Noone blames anyone in lies. The placebo effect is absolutely provably part of the reality - we deal with it.
There is a misconception here. Null hyphotesis can NOT be proven. It can only be disproven. It is the *failure* to disprove it reliably that opens your eyes. If you start with firm confidence that you can *easily* distinguish between A,B and when ABX test shows you that you actually play dice, that proves only *one* thing - that your selfconfidence is poisoned by placebo. Because if not, you *must* have been passed the test with 95% confidence without sweat.
Yes, convenience was a big factor. It's not a surprise that the CD was a runaway success. But the SACD lacks the portability of the CD. It may not matter to the people here, but the copy protection is another problem for the average consumer. Most people like to be able to make a copy for the car. You can't do that with a SACD. But the biggest obstacle for the "hi-rez" formats may simply be that most people think mp3 is good enough. Or to quote the Stereophile page I linked to:
<<This is all reminiscent of something one of the record label executives said to me by phone after a Dolby DVD-Audio press event a couple of years back: "We should never have marketed DVD-Audio as a better-sounding format to audiophiles. You can't please them, and nobody else cares about sound quality.">>
Konnichiwa,
Is it? Have you ever ABX tested the ABX test itself? Using some items reliably audible, such as one channel with inverted polarity?
What would you say if you found that you (or others) cannot reliably detect that with .05 significance? Do you conclude that this was inaudible or do you conclude that the test is flawed? Your call. I note that NON of the widely published ABX tests used in support of that idiotic "everything sounds the same" position had first tested the audibility limit of their test/subjects.
How many trials does your ACX Test include?
And 2, I would say that the neccesary number of trials to give statististical significance is closer to 50-100 range, to be certain that insensitive test subjects etc. do not cause statistical problems I would suggest that something along the lines of 5 Trials of at least 25 - 50 Subjects with "outlier" data discarded needs to be used for statistical significance. In order to not fall into the attention span deficit trap I suggest that if more than 5 Trials are used per subject they are carried out in several seperate sittings, if you wish to take due care to find out of there is an audible difference, rather than to find support for the Null Hyphothesis.
Given that ABX Tests in general fail pretty reliably to show modest audbible differences reliably one must consider their "error budget". The usual answer that not only was no audible difference observed, but also (disregarding any issues one may take about test steups etc, which are equally many) the type B error risk was very high, usually well above 50%. An exceptionally good example are the various "challenges" Tom Nousiane published in Audio, where the charatan gloats endlessly how he tricked some "goldenears".
No, being an intelligent person I am actually aware of a variety issues with all sorts of forms of testing, including ABX. And I know enough about statististics and ABX testing to largely discard the majority of results as being both severely flawed in the execution of the test and in the statistical analysis.
I am also aware of the propensity of humans to hear a difference when in fact non is observable or equally the propensity of humans to fail to observe any audible difference despite one being present, depending upon whatever beliefs are held.
Yes.
However it from the beginniong dogged by poor statistical analysis and was immidiatly "hijacked" by people with a significant specific agenda.
ABX is a specific implementation of Double Blind testing for Audio. It raises more questions than it answers and is fundamentally flawed on several levels, as a methode to determine the presence of absence of small audible differences. It is sufficiently reliable for large audible differences only.
Correct. However the results of such tests can support or recject the null hyphotesis nevertheless. And more than enough people, including the original ABX mafia have IN WRITING, REPEATEDLY and based on VERY POOR STATISTICS claimed exactly that, namely proof for the null hypthesis.
Sayonara
wimms said:
Is it? Have you ever ABX tested the ABX test itself? Using some items reliably audible, such as one channel with inverted polarity?
What would you say if you found that you (or others) cannot reliably detect that with .05 significance? Do you conclude that this was inaudible or do you conclude that the test is flawed? Your call. I note that NON of the widely published ABX tests used in support of that idiotic "everything sounds the same" position had first tested the audibility limit of their test/subjects.
wimms said:
How many trials does your ACX Test include?
And 2, I would say that the neccesary number of trials to give statististical significance is closer to 50-100 range, to be certain that insensitive test subjects etc. do not cause statistical problems I would suggest that something along the lines of 5 Trials of at least 25 - 50 Subjects with "outlier" data discarded needs to be used for statistical significance. In order to not fall into the attention span deficit trap I suggest that if more than 5 Trials are used per subject they are carried out in several seperate sittings, if you wish to take due care to find out of there is an audible difference, rather than to find support for the Null Hyphothesis.
wimms said:
Given that ABX Tests in general fail pretty reliably to show modest audbible differences reliably one must consider their "error budget". The usual answer that not only was no audible difference observed, but also (disregarding any issues one may take about test steups etc, which are equally many) the type B error risk was very high, usually well above 50%. An exceptionally good example are the various "challenges" Tom Nousiane published in Audio, where the charatan gloats endlessly how he tricked some "goldenears".
wimms said:
No, being an intelligent person I am actually aware of a variety issues with all sorts of forms of testing, including ABX. And I know enough about statististics and ABX testing to largely discard the majority of results as being both severely flawed in the execution of the test and in the statistical analysis.
I am also aware of the propensity of humans to hear a difference when in fact non is observable or equally the propensity of humans to fail to observe any audible difference despite one being present, depending upon whatever beliefs are held.
wimms said:
Yes.
However it from the beginniong dogged by poor statistical analysis and was immidiatly "hijacked" by people with a significant specific agenda.
ABX is a specific implementation of Double Blind testing for Audio. It raises more questions than it answers and is fundamentally flawed on several levels, as a methode to determine the presence of absence of small audible differences. It is sufficiently reliable for large audible differences only.
wimms said:
Correct. However the results of such tests can support or recject the null hyphotesis nevertheless. And more than enough people, including the original ABX mafia have IN WRITING, REPEATEDLY and based on VERY POOR STATISTICS claimed exactly that, namely proof for the null hypthesis.
Sayonara
Its not bad, its evil.phn said:
Marketing bs that DSD is same as masters is just it, bs.
SACD is closed, its much less flexible than PCM. Its total pita to edit. Its no.1 goal is to keep consumers away from it.
It works best and reaches its full potential only when you swap all your equipment, so that it tolerates ultrasonic noise.
Its a means to seize *control* over the music industry, again. If it ever gets beyond some threshold, they will artificially destroy CD market and force everyone to switch to it. Any consumer with even slightest sanity would resist such perspective with full force. Yet even now they could do it, excusing by fact that CD-layer exists on SACDs, if not the war with competing DVDA and forces behind it.
Many people like it that way and argue that its *more* natural. Analog is smooth.carlosfm said:
But there is not much you can improve with DSD. At the bitrate they picked, there is very little you can change about it. Thus only SACD-2 will improve anything.
percy said:
Googling for that, I remembered the cover of the album.
"Everything must go".
http://www.highfidelityreview.com/reviews/review.asp?reviewnumber=13143569
wimms said:
Precisely.
Spot on.
wimms said:
It may sound more "natural" to the vast majority of people who never heard real, unamplified instruments.
The kind of sound I hear from most SACD discs is not neutral.
It's like covering the guy who's playing the trumpet with a blanket.
It's also a common misconception that analog sounds smooth.
I understand the analogy but it's not correct.
It just sounds right.
What sounds smooth is cheap/poor analog.
Surely format won't matter in the future?
Download it from the internet or buy it on cheap removable hard disk or whatever, and a plugin on you computer can do whatever decoding and conversion is necessary.
If they provided the right plugin, a company could put music out in any kind of digital format they liked.
Download it from the internet or buy it on cheap removable hard disk or whatever, and a plugin on you computer can do whatever decoding and conversion is necessary.
If they provided the right plugin, a company could put music out in any kind of digital format they liked.