IPB

Welcome Guest ( Log In | Register )

3 Pages V  < 1 2 3  
Reply to this topicStart new topic
DVD-A and SACD--really worth it?
2Bdecided
post Jun 7 2002, 13:13
Post #51


ReplayGain developer


Group: Developer
Posts: 5259
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



What an interesting thread!

Rather than posting in great detail about this subject, I really should write a web page about my thoughts and experiences Re: DVD vs SACD vs CD - it would save me repeating myself. Unfortunately I wrote in detail about this at r3mix.net and that has now been wiped, so I can't refer to it.


On the other subject discussed here, I believe wholeheartedly that for someone to demonstrate that they can hear a difference between two audio samples, they must be able to differeniate them even when they do not know which one they are listening to. That is the essence of a blind test. Sighted tests are unreliable - even when the audible differences are huge.

However, sometimes conventional ABX isn't the best way. When you listen to something for the first time, it sounds different from the second time, and VERY different the 20th time! In fact, by the time you've listened to something repeatedly 20 times, I'd suggest that your ears are very tired, and responding very differently to the stimulus from "normal" listening. Sometimes this repetition can sensitise us to artefacts, but other times it can dull our senses. Whether it helps of hinders will depend on the nature of the artefact or difference.

Pio made a good point along these lines:
QUOTE
One of the only people that managed to ABX a 16 bits dithered file vs a 16 bits truncated file (while some other heard a difference but couldn't ABX it) performed one try each day ! Making a difference 16 times in a row can be very difficult.


I would suggest that sometimes repetition helps, and sometimes it hinders. There are differences that it is very difficult to catch in an ABX test, but they still exist.


Let me add something else to the discussion. Two collegues in the lab were evaluating the audibility of phase distortion. Traditional psychoacoustics says that it is inaudible, but some audiophiles claim that it is audible.

The test was blind: You were played clicks with no phase distortion, and then clicks with hideous amounts of phase distortion (the same as you would get from typical loudspeakers). The clicks were in pairs: distorted then undistorted, repeated three times; Or undistorted then distorted, repeated three times. Having trained yourself, you simply had to identify which set you were listening to. You did the test five times.

The first shocking result was that two of us could reliably detect which set of clicks we were hearing each time. The second shocking result was that one of us couldn't detect the difference the next day! And then the next week, he could again! No one could figure out what had changed. The equipment certainly hadn't changed. The listener (OK, it was me!) couldn't say what had changed either. No cold or flu, no loud sound exposure, no tiredness, no fatigue.

So, that's just another thing to confound it all: In a stringent blind test, one time I could hear a difference, and another I couldn't!

(btw, with real music signals (as opposed to clicks) no subjects could detect the phase distortion, but that's not to say that no one would ever be able to detect it in real music.)


As for SACD vs DVD-A vs CD: SACD will die a very slow death, DVD-A will very slowly replace CD, but most people will buy it just for its multi-channel capability, which will be a big "wow" factor when pop music starts using it creatively.

IMO: CD does have a "sound" to it. It's very subtle, but it's a "glassy" effect. It's got nothing to do with high frequency harshess - that's poor quality convertors and lousy mastering - though both of these plague many commercial CD releases. The huge audible differences between CD and SACD are mainly due to mastering, and the large amounst of HF noise that SACD pumps through your audio system. All the people who assume that most modern DACs and ADCs are "close to perfection" haven't measured one. Finally, to my ears, I think the audible difference between a $500 CD player, and the $20K Linn CD12 is greater than the audible difference between 44.1kHz and 96kHz on very high quality convertors.

You shouldn't assume that everyone in the "industry" who is making the switch up from CD is doing it for "commercial" reasons. Most will just follow the crowd, and most big companies want the copy protection and increased revenue per disc. But some of these people live and breath music and audio, and simply can't stand what CD is doing to the music. For the rest, who can't hear a difference in a fair test, the added resolution gives much more leeway to f*ck up without damaging the sound, which will give better results to the consumer, even if those results could theoretically have been achieved with CD.


Now, is 5.1 the best choice for multi-channel music? No! But that's another story...


Cheers,
David.

http://www.David.Robinson.org/
Go to the top of the page
+Quote Post
Garf
post Jun 7 2002, 14:00
Post #52


Server Admin


Group: Admin
Posts: 4886
Joined: 24-September 01
Member No.: 13



QUOTE
Originally posted by 2Bdecided

IMO: CD does have a "sound" to it. It's very subtle, but it's a "glassy" effect. It's got nothing to do with high frequency harshess - that's poor quality convertors and lousy mastering - though both of these plague many commercial CD releases. The huge audible differences between CD and SACD are mainly due to mastering, and the large amounst of HF noise that SACD pumps through your audio system. All the people who assume that most modern DACs and ADCs are "close to perfection" haven't measured one. Finally, to my ears, I think the audible difference between a 0 CD player, and the K Linn CD12 is greater than the audible difference between 44.1kHz and 96kHz on very high quality convertors.


After reading this I remain with the question: what is the techical cause for this 'glassy effect' ?

--
GCP
Go to the top of the page
+Quote Post
2Bdecided
post Jun 7 2002, 14:52
Post #53


ReplayGain developer


Group: Developer
Posts: 5259
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



That's the $64,000 question!

It's not even possible to answer this queston:

Is the audible "signature" of CD quality sound due to (a) a fundamental limitation of the format, or (b) current implementations of the format?


I sense that you may be thinking "Since there's no technical explanation for this, then it's all imaginary". However, the problem is that there are too many possible explanations for this, and it's difficult to prove or disprove any of them at the moment.

For example, in a real room with real audio equipment, you can measure the signal at the listener's ears (e.g. use a dummy head). And you can show that including information above, say, 20kHz in the recording that you replay over the audio system will cause different audible information to be received by the listener. If you examine closely the information BELOW 20kHz that reaches the dummy header, it will be different due to the inclusion of information ABOVE 20kHz in the recording. This is due to the non-linear effect of everything from the loudspeakers to air itself.

The problem is that it's difficult to quantify the subjective effect of this. With pure tones we can predict the distortion, and predict the audibility of it with a steady-state masking model. But what about complex musical stimuli and a good stereo recording? Here, microsecond timing differences between the ears give rise to perceived location. Also, the onset of sounds gives crucial information to the auditory system - and the transient information at the onset of a sound is very difficult to analise - certainly we don't understand quite how the ear and brain carries out this analysis.


Another example: In a band limitted system, though all signals below the nyquist limit can be accurately represented, the time resolution of these signals decreases as you approach nyquist. For example, the CD limit is 22050Hz, which means that you can store a 22049Hz signal (in a perfect system). But this signal has a time resolution of 1 whole second! That means, the signal can fade up over a second, and fade down over the next second - but any faster switching would produce harmonics over 22050Hz, which could not be stored in the system. As you move down from Nyquist, this problem falls away rapidly, but even in the high-teens of kHz, you need 1 or 2 cycles of the waveform before it responds perfectly - by which time the signal onset is long gone, and your brain has done its processing.

You can look at the implicit lowpass of the ear, and the bandpass induced ringing of the auditory filters and say "it can't matter - the ear smears the signal anyway". But then you can look at the response of the inner hair cells, and realise that the ear tries to regain the time smearing - and if you're looking at a wide-band rather than tone-like stimulus, it does this very very well.


So, there's two reasons why higher sampling rates might sound better (or at least different!): equpiment distortion, and insufficient time resolution. But the real reason they sound better may have nothing to do with either of these: it may be just that they move the essential real-world compromises further away from the audio band.

Cheers,
David.
http://www.David.Robinson.org/
Go to the top of the page
+Quote Post
Garf
post Jun 7 2002, 15:11
Post #54


Server Admin


Group: Admin
Posts: 4886
Joined: 24-September 01
Member No.: 13



QUOTE
Originally posted by 2Bdecided

Another example: In a band limitted system, though all signals below the nyquist limit can be accurately represented, the time resolution of these signals decreases as you approach nyquist. For example, the CD limit is 22050Hz, which means that you can store a 22049Hz signal (in a perfect system). But this signal has a time resolution of 1 whole second! That means, the signal can fade up over a second, and fade down over the next second - but any faster switching would produce harmonics over 22050Hz, which could not be stored in the system. As you move down from Nyquist, this problem falls away rapidly, but even in the high-teens of kHz, you need 1 or 2 cycles of the waveform before it responds perfectly - by which time the signal onset is long gone, and your brain has done its processing.


I don't understand this, can you elaborate?

For example, you say 'but any faster switching would produce harmonics over 22050Hz'. What is the problem of that? If you are sampling at 44.1khz, you are not expecting to store (or hear) anything above 22050Hz anyway, so I don't understand what the problem is sad.gif

--
GCP
Go to the top of the page
+Quote Post
gdougherty
post Jun 7 2002, 18:19
Post #55





Group: Members
Posts: 195
Joined: 10-February 02
From: One Mile Up
Member No.: 1299



QUOTE
Originally posted by Garf


I don't understand this, can you elaborate? 

For example, you say 'but any faster switching would produce harmonics over 22050Hz'. What is the problem of that? If you are sampling at 44.1khz, you are not expecting to store (or hear) anything above 22050Hz anyway, so I don't understand what the problem is sad.gif

-- 
GCP


It's actually something I found while reading a paper to better understand nyquist and the theorem's effect on audio sampling. 2B's example doesn't quite work since filtering generally occurs at about 98.something percent of nyquist to eleminate these high harmonic issues. Over very short periods of time the calculations become less error prone as the calculations "lock" onto the signal. Until then you have errors that produce small amounts of distortion in the sampling process. These occur more noticeably as you approach the nyquist frequency.

Still the assertion that the higher harmonics affect the reproduction of the signal is interesting, though depending on the equipment the effect would be fairly miniscule given most equipment's inability to produce signals over 20Khz. Some speakers can produce frequencies up to 30Khz, but I'd imagine in most recordings these higher harmonics aren't significantly there since most mics roll-off at 20Khz, maybe slightly higher.
Go to the top of the page
+Quote Post
Cygnus X1
post Jun 7 2002, 19:12
Post #56





Group: Members (Donating)
Posts: 676
Joined: 5-June 02
From: New York
Member No.: 2224



Here's another related question regarding high-frequency content in music: I've oft heard area audiophiles talk about how the combination of harmonics in excess of 20Khz can create a discenable "beat" at lower frequencies. Since I am a symphony musician on the side, I can attest to hearing the effects of this so called "beat" at lower frequencies. For example, when a flute and clarinet play together in their high register, and are a third apart in pitch, a harmonic resonance occurs at a different pitch than the fundamental and third. However, the harmonic is within the range of hearing. . . . so does this hold up for ultrasonic harmonics, as I have been told? Not that I would be able to hear it anyway. . . my hearing is already seeing the effects of tinutitis (sp?)--when I was a kid I could hear the 19Khz switching signal from most TV's quite loudly. Now, my hearing sharply rolls off at @16.3Khz and I can't detect anything above 17.7Khz, period, and anything above 17Khz has to be very loud in order to detect it. Makes me wonder if having DVD-A solely for the ultra-hf content would be a waste since I don't know it's there to begin with??
Go to the top of the page
+Quote Post
bryant
post Jun 7 2002, 19:46
Post #57


WavPack Developer


Group: Developer (Donating)
Posts: 1292
Joined: 3-January 02
From: San Francisco CA
Member No.: 900



QUOTE
Originally posted by Ruse
Bryant, I'm sure you are not arguing that the scientific method is flawed, are you? (Don't read this as me implying that the double blind test = the scientic method.) There is no middle ground when it comes to the results. You can use all the intuition you like to design the experiment, but the final proof must be based on cold, hard logic.

No, I am definitely not arguing that the scientific method is flawed. In fact, my brother and I have been discussing the design of some experiments that might show this effect if it exists (he has almost completed his doctoral studies in cognitive psychology at UCSC). If we could show that there was some level of some distortion that was impossible for a subject to identify in an ABX test, but that would show an effect in some other test that did not rely on reporting by the subject (for example, a reaction time test), then that would shed considerable doubt on the value of ABX testing as the final word on audio testing.
QUOTE
Originally posted by Garf
As other people have already pointed out, if they have an effect, they can be ABX-ed. What ABX does, is measure if, in any way, a listener can notice a difference between two clips. This means that the listener doesn't have to perceive anything wrong with either individually, or perceive (in the common sense of the word), a problem. If in any way they can determine a difference, they will be successfull.

If in no way they can determine a difference, the clips _are_ 'identical' to them for any means and purpose.
GCP
The hearing system is incredibly complex and involves layer upon layer of information processing between the raw data that comes from the nerves in the ear and the summary report that gets delivered to the conscious mind. Certainly this process involves throwing away a lot of information and comparing the information from the different ears and all sorts of other analysis that we don't understand. And there is even conscious control over the actual processing parameters when, for example, you strain to hear something specific.

So, it seems plausible to me that there might be cases where two different stimuli could generate the same conscious result, but involve different processing to get there. It could be that the auditory system detects the stimuli differently, but simply has no mechanism to report it (perhaps because it is not a sound characteristic that occurs in nature). However, that difference in processing could still have some other effect in the mind that, even while not directly reportable, could influence our response to the music.

For a simplified example, let's assume there's a type if distortion that even when below the threshold of audibility causes some extra processing to occur. Now, let's say that after 15 minutes of exposure to this a person becomes fatigued. While one could argue that this could be detected in an ABX test if it were run slowly enough and the person could takes notes and recover between runs, I would claim that the day to day distractions of life would make this impractical. And, besides, if it could be shown that it is having some effect on the subject then it might not really matter if it could ever be directly reported.

I don't claim that any of this is actually going on, and I certainly don't believe that everything that people think they hear (but can't ABX) is real. But I am saying that with so many people complaining about the way CDs sound for so long, there might just be something to it. And I am talking about recording engineers and musicians who know what a live feed sounds like, and know what their 30 ips master tapes sound like, and bring their 30 ips master tapes home because (they claim) the final CD doesn't come close. Well, if they can't ABX it then that means that either they're hopelessly deluded or the mind is just a little more complex than we previously thought.
Go to the top of the page
+Quote Post
Pio2001
post Jun 8 2002, 00:01
Post #58


Moderator


Group: Super Moderator
Posts: 3936
Joined: 29-September 01
Member No.: 73



This is getting thick. Here are three topics :

1-Time inaccuracy of high frequencies
2-Effects of ultrasounds
3-data transmitted by the ear but not percieved.


1-Time inaccuracy of high frequencies

The time inaccuracy of high frequencies is just a mathematic trick.
A pure frequency doesn't exist in reality. In maths, a given frequency is represented by an infinite sine wave. It never begins and it never ends. When you start and stops the sine wave at given times, limiting its lenght, the accuracy of it's frequency necessarily decreases.

A 22049 Hz frequency is defined with 1/22049 acuracy.
It's higher than 22048 and lower than 22050. Therefore to represent it with such accuracy, you must be able to draw three graphs for 22048, 22049 and 22050 Hz that are different from each other.
But a 22049 sine recorded at 44100 Hz produces a beat effect of 2 Hz (try in SoundForge or CoolEdit) :



I guess that the one second lenght (or 0.5 ?) is in fact the amount of data needed for the DAC to reconstruct a pure 22049 Hz tone with no beat. In real life, I don't think oversampling algorithms used in DACs use so much samples rolleyes.gif


2-Effects of ultrasounds

There is no beat effect between audible and inaudible sounds. You can check it also in a wave editor. When you lowpass a beat effect between the two frequencies, the beat disappears. Only the lowest frequency remains with no beat at all.

The question is in fact not beat effect but intermodulation distortion.
When you play two frequencies A and B, the intermodulation are two frequencies at |A-B| and A+B. Therefore a 6 kHz square wave, which consists in 6 kHz sine + 18 kHz sine + 36 kHz sine etc, should give audible intermodulation at 12 kHz if it is distorded, therefore sounding different than a sinewave of the same frequency.
Nika and I tested this. All we could get was distortion in the playback system (in fact I actually got some, and I suppose Mike Richter god some too). We couldn't get such distortion in our ears, that would have proven the effect of unaudible frequencies on sound.
There were also tests from Mike Richter and Studioman adam too, but they didn't lead to conclusive results.
For more : http://www.musicplayer.com/ubb/ultimatebb....3;t=000822;p=33
Scroll down until Nika's third post, beginning with "Pio, Thank you for responding."


3-data transmitted by the ear but not percieved.

That's true. Learning music, we get trained to hear for example each separate tone from which a chord is made. Without having learned music, the ear gives the same data to the brain, but we perceive just one rich sound.
Being more trained we can recognize fundamental and harmonics in one note played by one instrument, while it appears as just one note for untrained people.
Another example could be the recogniton of two very close notes. Learning the rudiments of music, we try to recognize two notes separated with 1/18th of a tone. People learing more end up trying to sort out notes separated with 1/100th of a tone ! In this case, however, I wonder if the ability grows with learning. I was always able to hear 1/18th of a tone from the beginning since I learned music when I was 15 years old.
Go to the top of the page
+Quote Post

3 Pages V  < 1 2 3
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 23rd November 2014 - 09:45