Audiophile Myths Part 1: MP3 VS FLAC, Cables, Sample Rates, Tube Amps, ETC.

[Thad E Ginathom]

Sometimes, when there is no difference ...there's just no difference.

Sometimes, a perceived difference on casual listening can disappear on critical listening.

This was my thing about sample-rate comparison: every time I heard a difference, I would go to the other sample and find the same thing there too.

Blind testing is not required to hear no difference, but sometimes it is required to show that the perceived differences are just perceived.

 

[sbhas2k]

Nice and informative thread

 

[mowgli80]

Haven't gone through the entire thread but looking at some of the posts, looks like this video may be relevant. The speaker is creater of ogg vorbis codec and probably knows more about digital audio than us forum members collectively.

 

[tcpip]

Sometimes, a perceived difference on casual listening can disappear on critical listening.

This was my thing about sample-rate comparison: every time I heard a difference, I would go to the other sample and find the same thing there too.

So true.

I still remember the story of a designer of very good valve amplifiers based in Delhi. He was happy with their sound (which was deserved) and wanted to demo to his friends how this sound was better than that from his earlier transistor amplifiers. But, as the critical listening session progressed, and people kept switching back and forth between the valve and solid-state amps, they began to converge that all the lovely characteristics they could hear from the valve amp were there in the SS amp too. Very enjoyable, and also quite humbling.

Nothing to beat blind testing.

(As a side-note, this indicates how high the quality of the two amplifiers was. Only very good valve amps and SS amps sound similar.)

 

[tcpip]

Some of my favourite links on this matter of listener expectancy and sighted listening tests:

Audio Musings by Sean Olive: The Dishonesty of Sighted Listening Tests

Matrix HiFi --> Blind testing high end full equipments

https://www.youtube.com/watch?v=G-lN8vWm3m0 -- this is a 3-minute video

 

[viren bakhshi]

As the designer, in Delhi, of those solid state amps and the early tube amps, lots of comparisons were done. The tube designs have evolved tremendously since then, as Lyrita has moved to making only tube amps - primarily single-ended designs.

Whether you want to listen to amps blind, or let the glow of the tubes suffuse you too, is totally up to you! Anyway, double blind listening is a bogey that doesn't deserve any more debate.

Regards,
Viren

 

[tcpip]

Whether you want to listen to amps blind, or let the glow of the tubes suffuse you too, is totally up to you! Anyway, double blind listening is a bogey that doesn't deserve any more debate.

Care to elaborate, considering how many of us here seem to be considering it worthy of debate?

 

[Nikhil]

With regard to MP3 I sometimes find a "ghosting" of voices specially in the vocal range of singers. On my casual listening setups (car, bedroom stereo, etc) these sounded like part of the singers vocal repertoire. In my main system they showed up as artifacts that became harder to listen to.

I am now exploring DSD and the sound is remarkably better than 24 bit /192 KHz PCM. Double DSD is even better. My DAC does not support either so I am doing a conversion to PCM in JRiver which though not ideal still sounds better than PCM. The sound is much softer with a huge reduction in harshness in the HF.

Enclosing two links from Positive Feedback by Andreas Koch who is a pioneer in the field.

DSD - the New Addiction

Resolving for Resolution


The DSD scheme is best described by this illustration below:

 

[alpha1]

Anyway, double blind listening is a bogey that doesn't deserve any more debate.

Yes, and perhaps that's why scientific methods have succeeded in ascertaining cause and effects which lead to the enhancements in the technology that we use today ... ?

 

[Nikhil]

Care to elaborate, considering how many of us here seem to be considering it worthy of debate?

Tcpip, considering that you are into DIY you can appreciate the theoretical vs the practical. In theory AB testing is good and I don't think you wil find an argument from Viren on that. The problem is in actually setting up an AB test that is free from bias. You may be surprised how non trivial the procedure is.

What a lot of enthusiasts call AB testing is better than the straight forward comparisons that we talk about. However conducting an AB Test that is statistically correct and without bias is not that straight forward.

 

[sidvee]

What a lot of enthusiasts call AB testing is better than the straight forward comparisons that we talk about. However conducting an AB Test that is statistically correct and without bias is not that straight forward.

Plus there are so many other random variables that effect accurate listening. I read an article where a persons mood/state of relaxation effects the way he interprets the quality of sound, an effect I have heard myself, the days I am stressed I can hardly enjoy music or the times couple of beers make the shrill highs smoother. So how do you find test subjects with good moods on the day of the test? How do you conduct AB/blind/double blind tests without having an accurate measurement system ergo the listener? It is ok in other fields because the variables are somewhat controlled (like pharma where chemicals are reacting with the blood pressure etc. not something subjective)but I find it weird every time someone disagrees about audio they immediately throw up DBT testing as if it is a panacea. I am sure the test can be conducted to some degree of accuracy in a proper lab with proper equipment and test subjects that have been monitored over a period (controlled subjects), but all this costs big money and imo is stupid to prove that Amp a is better than Amp b. It is ok if we are looking for a life saving drug, but to say speaker cable 'A' is bad or overpriced against another is downright illogical.
Cheers,
Sid

 

[tcpip]

Tcpip, considering that you are into DIY you can appreciate the theoretical vs the practical. In theory AB testing is good and I don't think you wil find an argument from Viren on that. The problem is in actually setting up an AB test that is free from bias. You may be surprised how non trivial the procedure is.

I am fully aware that a lot of double-blind testing is inconclusive because either it wasn't really as blind as necessary, or because the comparison environment did not support careful comparison. Why, people sometimes even lack the facility to accurately set the gain of two playback systems to the same value.

But calling double blind listening a "bogey" because of difficulties in setting up the experiment will make all experiments in quantum mechanics "bogeys", won't it? The super-colliders must all be scrapped, because after all, they are all (expensive) "bogeys"? Since when did we, educated in the rationalist school of philosophy, begin to discredit the science because of the difficulties of setting up a good experiment?

Of all people, a person who actually designs and builds amps will know this, won't he?

What a lot of enthusiasts call AB testing is better than the straight forward comparisons that we talk about. However conducting an AB Test that is statistically correct and without bias is not that straight forward.

Completely agree. On similar lines, straight and simple comparisons between "MP3 and FLAC" or "DSD and PCM" are often meaningless because basic care has not been taken to bring both sources on an equal footing before comparison. I keep asking the MP3-naysayers, "Do you have any idea of the parameters used to encode the MP3 file?" I almost never get any answers. I also ask, "Are you sure that the MP3 you are listening to is encoded from the same uncompressed source which you are comparing against?" Once again, I get no answer. What is the value of such comparisons? :lol:

Some people tell me "I switched between the SACD layer and the CD layer on the same CD, on the same player, and the SACD sounded better. Case closed." I ask them, "How did you verify that both layers generate analog output at the same amplitude, to let you do a fair comparison?" No answer. I ask, "Do you have any basis to believe that the exact same digital mastered and mixed data has been encoded into both layers? Do you know that most hybrid SACDs come with different mastering for the two layers?" Once again, no answer.

Pretty tiring, this. Science, if not done right, becomes a pseudo-science.

 

[tcpip]

How do you conduct AB/blind/double blind tests without having an accurate measurement system ergo the listener?

The listener in a double-blind listening test is not a "measurement system". He/she is a comparison system. There is a massive difference between the two. In electronics, we have a building block called "comparator", which compares two values without knowing the absolute value of either input. That's the function of the listener in an ABX comparison test.

Also, your comments about a listener reacting to music differently depending on mood are true -- we have all seen it. But it is expected that an ABX test is conducted in a short span of time, without changing the setting or environment, therefore the listener will bring the same bias to both the A and B samples, thus cancelling them out. Moreover, you try to neutralise biases of individual listeners by having many listeners and analysing all their responses for statistical relevance. Statistics is not an immature science, and they have evolved many methods to minimise impact of precisely this sort of bias. There are many fields much more contentious than audio where statistical methods are used to sift through potentially inaccurate responses for statistical relevance and correlation. A lot of sociology uses such methods.

Science is not as stupid as some of us seem to make it out to be.

 

[jls001]

To even out the differences in the critical listening abilities of test subjects, the state of their emotions/moods, and various other variations that may creep in, a large number of tests can be carried out to make the sampling more closely representative of the larger population. At least that's what is followed in statistics. How practical or economical that proposition is, is an altogether different matter. Now, how do we decide that the minimum number of unique test subjects makes it statistically and scientifically significant? Is this tool practically and scientifically usable by Joe Audiophile for, say, choosing between two interconnect cables?

Also, in audio DBT are subjects told to listen to specific traits in the audio?

 

[sidvee]

Statistics is not an immature science, and they have evolved many methods to minimise impact of precisely this sort of bias. There are many fields much more contentious than audio where statistical methods are used to sift through potentially inaccurate responses for statistical relevance and correlation. A lot of sociology uses such methods.

Science is not as stupid as some of us seem to make it out to be.

Absolutely, science/statistics is not stupid or immature, then why pray apply it to a casual hobby like 2 Channel Stereo listening. Is it because of the money involved? Then it should be applied to things like Aero Modeling, or sports Golf/Tennis cricket etc where there is huge money. Not only it is applied to audio, people get into huge arguments. Even relatively low cost cables costing perhaps $200 are demanded to be introduced into this test? To what end? Is the world going to be a better place if a $10 cable beats the $200 cable. On the same token, labs based in India, like GVK etc, falsify critical information for life saving drugs. And we are trying to gather a bunch of listeners, nomalize their bias etc, to do what?
IMO it is a useless pursuit - just a theoretical exercise, and to my knowledge, regardless of what test procedure was used/listener bias/ etc. there has been no agreement on any Blind or Double blind test relating to any audio equipment anywhere so far. They have all been inconclusive, with huge arguments for and against, and regardless people keep on buying cheap and expensive audio equipment.
If there is any test that has been accepted by all kindly point me to it.
Cheers,
Sid

 

[tcpip]

Absolutely, science/statistics is not stupid or immature, then why pray apply it to a casual hobby like 2 Channel Stereo listening.

This is not a question I am willing to try to answer. Why someone puts in money, time and energy to finding answers to some questions or build something, is a question outside the scope of what I'm trying to understand. I was trying to understand the problem of listener expectancy and the effectiveness of double-blind listening tests to tackle the problem.

 

[Nikhil]

Pretty tiring, this. Science, if not done right, becomes a pseudo-science.

Q.E.D.


tcpip, I appreciate where you are coming from. However I know that Viren is sincere.
Whether we may appreciate his work or not it was unfair to make an oblique reference to his work.
Hence my comments on testing. I also noticed your clarification on the quality of the ss amp etc.

Let's move on ...

 

[sidvee]

This is not a question I am willing to try to answer. Why someone puts in money, time and energy to finding answers to some questions or build something, is a question outside the scope of what I'm trying to understand. I was trying to understand the problem of listener expectancy and the effectiveness of double-blind listening tests to tackle the problem.

And what I am saying is that DBT applied to this particular issue has never been shown to resolve anything, regardless of what scientific rigors have been applied to it, maybe perhaps IMO because to begin with it is too subjective. So what is the point of flogging this in our humble forum? Is it just fashionable to use the word DBT? It has been done in so many venues/forums/magazines/audiophile meetings etc. and all it does is generate acrimony/inconclusive posts, that at best die and at worst are flagged OT when participants get emotional or hostile, and are closed or deleted.
Cheers,
Sid

 

[tcpip]

tcpip, I appreciate where you are coming from. However I know that Viren is sincere.
Whether we may appreciate his work or not it was unfair to make an oblique reference to his work.

My posts were about listener expectancy and the need for blind testing -- they were not about a person. I am sorry that you seem to believe they were about a person. And everything I have heard about Viren's work has been positive. Therefore, while I continue to ask what's wrong with blind testing, I really fail to see why a reference to an incident, without naming any person, entirely in passing, should be treated as "unfair oblique reference".

Let's move on ...

Yes, I guess the moving-on will happen whether we like it or not, because there doesn't seem to be any substance around which we can discuss the why-not-blind-testing question. I have seen enough case studies and other empirical evidence to indicate that blind testing is necessary. I am yet to see any such case studies or material to indicate why blind testing is worse than sighted testing. So, for want of material to chew on, I guess we'll just have to let it pass.

 

[tcpip]

And what I am saying is that DBT applied to this particular issue has never been shown to resolve anything, regardless of what scientific rigors have been applied to it,

I guess the answer to whether it has been shown to resolve anything or not depends on who you ask. A lot of brilliant engineers and designers seem to have used blind testing to put a lot of questions at rest for their own purposes. It seems to have resolved a lot of questions for them.