What objectivists and subjectivists can learn from each other

[andy_c]

Jack, very perceptive observation about the distinction between "consumerist objectivists" and so-called "pure objectivists". I have noticed that AV people tend to be in the "consumerist objectivist" camp, rather than the so-called "pure" one. Why? I think it's because, in AV systems, especially multi-channel ones, there is a lot more stuff to buy, therefore a more urgent need to determine what's worth the extra money and what is not.

The audio-only objectivists seem less concerned with consumerism and more concerned with good measurements, even though the benefit may sometimes be audibly negligible. In a sense, it's audiophile nervosa on the objectivist side.

I once had an NAD preamp with a volume control (Alps Blue) whose channel-to-channel level tracking sucked. It drove me nuts! The preamp eventually failed. I replaced it with another one with a volume control having better tracking. I also got a Benchmark DAC, which has individual volume trimming pots in each channel, allowing precise adjustment of channel balance. I set it up so the power amp output voltages were precisely matched for a mono input signal at a typical volume control setting. The psychological factor of having done this cannot be underestimated. Good recordings sounded right. But when I encountered a recording with bad channel matching, I did not worry because I knew the channels were matched with a mono input signal. No more audiophile nervosa!

 

[JackD201]

Hi Andy,

Thanks. Much as I hate categorizing people, I find it important to look into the individual motivations. It's always easier to digest information when you "know where the person is coming from". I'm more of an analog guy but when I calibrate my system I always use both correlated and uncorrelated pink noise to get matched levels at the listening position. With all the things that can go askew with an analog set up the much fewer variability of a digital front end is a wonderful thing

Like others have posted here few folks, here at least, are really purely objectivist or subjectivist. Even the most diehard objectivist use subjective criteria in determining what songs and albums gets to enter their music libraries. I sincerely believe that a system should be an extension of these sensibilities and should play up to the qualities that got those tracks in the collection in the first place. I also sincerely believe that using measurements just makes things so much easier and have results that are so much more repeatable. Enjoyment (subjective) is the goal and Process (objective) is a great means to achieving that goal. I shudder to think what would happen if the Enjoyment took a back seat to the Process.

 

[jkeny]

Jkeny,

Wonder if you are going to answer my questions or not....if not, then just let me know OK.

1) I asked at what level distortions of any sort would be inaudible...how many db below reference 0db or any reference you want?

You're barking up the wrong tree - it's not about what dB below ref is audible - it's exactly about Ethan's statement in the video - a perfect null is not achievable & this is worth investigating, not dismissing as being inaudible, as he does.

2) I asked what is more right about measuring a device with music vs with sinewaves, especially a bandwidth limited thing like audio?
Tom

Read my post 304 !

Edit: Wow, Tom, I just saw your post here in which you have already answered the question you are asking me - what gives? http://www.whatsbestforum.com/showt...audio-gear-specs&p=27996&viewfull=1#post27996
This is you, right?

I believe that the best audio test for a component is an electronic (not listening test) null test using actual complex music with wide dynamic and frequency ranges and the fastest transients.
Steady state sinewaves do not cut it. FFT are helpful to show spectrum analysis. But nulling will reveal all errors, then go deeper from there to sort them out.

 

[Ethan Winer]

didn't want to wade through an hour of video - I'm sure you understand

I certainly do understand: "My mind is made up, don't confuse me with the facts."

How can you now say that the nulling test, even though it doesn't null in the cases above

You need to follow more closely. I showed three examples of a null test, and the first two nulled completely because they were time-locked. No listening needed. The two bit streams are identical. In the context of whether a FLAC file sounds different than the Wave file it came from, this is absolutely possible to disprove using a null test. Same for other situations where people believe the sound is different even when it's not. I mentioned that nulling isn't always possible because it isn't always possible. That doesn't mean it's not a valid test when it is possible! Nulling can also prove that two power amps are identical, or at least have differences too soft to be audible. Andy described that in his Post #356.

--Ethan

 

[microstrip]

(...) Nulling can also prove that two power amps are identical, or at least have differences too soft to be audible. Andy described that in his Post #356.

--Ethan

Ethan,

How do you correlate the reminders of a null test with audible differences? Can you quantify in a meaningful way (with figures) the result of an amplifier null test?

When Hafler suggested it, he wrote (quoted from the Hafler X600 manual)

The test is as sensitive as one’s hearing, and much more precise than available measurement technology because it is able to reveal the audible effect of all distortions - even those which have not yet been identified or quantified!

I have used the null method in experiments, but then I knew exactly what was the variance of the measured object that corresponded to the sensitivity of the measuring instrument.

 

[Phelonious Ponk]

Originally Posted by Keith_W
What an interesting quote. I suppose that in a nutshell is the difference between objectivists and subjectivists? One is interested in a facsimile reproduction, the other is interested in enchanting sound?

I'm interested in facsimile reproduction of enchanting music.

Tim

 

[Gregadd]

Many fields recognize different schools of thought. Often switching between the two depending on which one yields the best result for a particular application. Of late we seem to think to that it is sufficinet argument to simply name the school of thought.

 

[jkeny]

I certainly do understand: "My mind is made up, don't confuse me with the facts."

Randi style posing as evidenced in your video & posts here do not interest me in the slightest - it smacks of self- aggrandisement rather than a search for the truth. As you say your mind is made up as I suspected from the start but I didn't expect you to admit it - life is always full of surprises! But let's stop this train of posting as it is becoming unseemly & of no benefit to anyone

You need to follow more closely. I showed three examples of a null test, and the first two nulled completely because they were time-locked. No listening needed. The two bit streams are identical. In the context of whether a FLAC file sounds different than the Wave file it came from, this is absolutely possible to disprove using a null test. Same for other situations where people believe the sound is different even when it's not.

Yes, & they have no interest to me at this time.

I mentioned that nulling isn't always possible because it isn't always possible. That doesn't mean it's not a valid test when it is possible!

Yes, & when there are differences, you dismiss them in your video. Now I know that the difference in the third example on the video is in the time domain & therefore you can't get a complete lock between two sample tracks. They drift in & out of null. So here is where you seem to stop in your investigation, declaring everything that causes such time domain shifts as inaudible. I guess one difference between us is that I would continue in this investigation. For instance, rather than dismiss jitter, which I presume you dismiss, have you done a null test with two known devices - one high & one low jitter to confirm this ?

Nulling can also prove that two power amps are identical, or at least have differences too soft to be audible. Andy described that in his Post #356.

--Ethan

I don't believe that there is any such claim in his post #356 - he described the test set-up when testing amplifiers which is a necessary part of any scientific experiment.

 

[Ethan Winer]

How do you correlate the reminders of a null test with audible differences? Can you quantify in a meaningful way (with figures) the result of an amplifier null test?

This too is shown in my Audio Myths video, along with the full-res files linked to on my web site. That part starts at 32 minutes in. I created an extremely nasty obnoxious noise that nobody could possibly miss, then mixed that at various levels under music. The idea is you can decide for yourself how soft it has to be in order to not be audible, rather than take my word for it. In my experience, once artifacts are 80 dB below the music they will not be heard. This is equivalent to 0.01 percent distortion. But in many cases artifacts aren't audible even when only 40 dB below the music. This is an important part of my demo!

When Hafler suggested it, he wrote (quoted from the Hafler X600 manual)
The test is as sensitive as one’s hearing, and much more precise than available measurement technology because it is able to reveal the audible effect of all distortions - even those which have not yet been identified or quantified!

Yes, and that's the same point I'm made many times: The beauty of a null test is it reveals all differences, even those you might not be looking for. I disagree that listening to artifacts is "more precise" than measuring them, though of course you can do both. And listening does take into account Fletcher-Munson.

--Ethan

 

[Ethan Winer]

I don't believe that there is any such claim in his post #356 - he described the test set-up when testing amplifiers which is a necessary part of any experiment - describing the test set-up.

Sorry, it was Post #350. The Doug Self article is excellent, and explains a lot of this better than I can. Read it and learn.

--Ethan

 

[Gregadd]

I 've asked this question before:

I created an extremely nasty obnoxious noise that nobody could possibly miss, then mixed that at various levels under music. The idea is you can decide for yourself how soft it has to be in order to not be audible, rather than take my word for it. In my experience, once artifacts are 80 dB below the music they will not be heard. This is equivalent to 0.01 percent distortion. But in many cases artifacts aren't audible even when only 40 dB below the music. This is an important part of my demo!

Ethan-
In asking is the artifact audible, are yoiu trying to identify the specific artifact or a difference in the sound of the music as a whole? No doubt you have encountered the subjectivist claim that "it sounds better because the noise floor was lowered."Or that "jitter" bothers them even though are unable to describe what jitter is?

 

[Ethan Winer]

I think the goal is to just tell if you can hear that the audio is "disturbed" one way or the other. As you might imagine, I reject the notion that music is audibly damaged in subtle ways by very low levels of noise such as jitter, or truncation distortion. My rejection is based on having done many such tests over the years. Again, this is the entire point of my examples, and why I include the original Wave files to improve on the poor quality of YouTube videos. These examples let you listen and decide for yourself at what point the quality is harmed, and in what way it's harmed.

--Ethan

 

[jkeny]

I think the goal is to just tell if you can hear that the audio is "disturbed" one way or the other. As you might imagine, I reject the notion that music is audibly damaged in subtle ways by very low levels of noise such as jitter, or truncation distortion. My rejection is based on having done many such tests over the years.

I would like to see your tests that have led you to reject the notion that jitter doesn't matter. I note that you qualify jitter as "low levels of noise" which is misleading & incorrect. What do you mean by "truncation distortion", clipping?

Again, this is the entire point of my examples, and why I include the original Wave files to improve on the poor quality of YouTube videos. These examples let you listen and decide for yourself at what point the quality is harmed, and in what way it's harmed.

--Ethan

 

[jkeny]

Sorry, it was Post #350. The Doug Self article is excellent, and explains a lot of this better than I can. Read it and learn.

--Ethan

The Doug Self article shows no test results for amplifiers - it shows the setup & configuration for such a test but no results as far as I can see! Again making a statement that "Nulling can also prove that two power amps are identical, or at least have differences too soft to be audible. Andy described that" & later that Doug Self shows this in the linked document is a gross distortion of the truth of both of these references. You are either deliberately trying to mislead readers or you have a bad habit of mis-interpreting what you read? It is these types of energy wasting posts of yours that is so frustrating.

 

[NorthStar]

Missynchronisation, mistiming; aren't they type of distortions?
...Creating annoying noises of some sort? ...Jittering.

* Please correct me if I made any typo(s).

 

[jkeny]

Essentially jitter comprises of two forms
- random jitter which is more benign & said to sound like noise
- deterministic jitter which correlates with the signal in some non-random way. This appears to be less benign & presents itself as audible distortion.

A single number for jitter is as meaningless as a single number for THD - the spectrum of the jitter appears to be a much more important factor, as is also the case with THD.

 

[microstrip]

This too is shown in my Audio Myths video, along with the full-res files linked to on my web site. That part starts at 32 minutes in. I created an extremely nasty obnoxious noise that nobody could possibly miss, then mixed that at various levels under music. The idea is you can decide for yourself how soft it has to be in order to not be audible, rather than take my word for it. In my experience, once artifacts are 80 dB below the music they will not be heard. This is equivalent to 0.01 percent distortion. But in many cases artifacts aren't audible even when only 40 dB below the music. This is an important part of my demo!



Yes, and that's the same point I'm made many times: The beauty of a null test is it reveals all differences, even those you might not be looking for. I disagree that listening to artifacts is "more precise" than measuring them, though of course you can do both. And listening does take into account Fletcher-Munson.

--Ethan

I see that your conclusions rely on your experience, and should be considered as such.
My experience is different from yours, we will follow different routes. Thanks for your time.

 

[NorthStar]

Essentially jitter comprises of two forms
- random jitter which is more benign & said to sound like noise
- deterministic jitter which correlates with the signal in some non-random way. This appears to be less benign & presents itself as audible distortion.

A single number for jitter is as meaningless as a single number for THD - the spectrum of the jitter appears to be a much more important factor, as is also the case with THD.

Thank you John! :b

 

[fas42]

This too is shown in my Audio Myths video, along with the full-res files linked to on my web site. That part starts at 32 minutes in. I created an extremely nasty obnoxious noise that nobody could possibly miss, then mixed that at various levels under music. The idea is you can decide for yourself how soft it has to be in order to not be audible, rather than take my word for it. In my experience, once artifacts are 80 dB below the music they will not be heard. This is equivalent to 0.01 percent distortion. But in many cases artifacts aren't audible even when only 40 dB below the music. This is an important part of my demo!

I'm pleased that you did that, Ethan, I've been meaning to do something similar for ages, but have been too "lazy"! The numbers you quote agree exactly with what I would have estimated, demonstrating that CD's 96dB range is plenty good enough to do the job of "perfectly" capturing dynamics. My own experience is that when the musical signal, what you're listening to drops to -60dB and below then you have to work pretty hard to hear things: volume turned up to max, and ear next to the speaker type of effort ...

Frank

 

[Keith_W]

I'm interested in facsimile reproduction of enchanting music.

Which is why the words "clinical" and "sterile" are not part of the objectivist lexicon. Can't be measured, but every subjectivist knows what it is.

I suspected this would happen. A whole bunch of people fail to grasp my point, and a torrent of indignant posts result. If people deny that sterile sound even exists, and instead focus on the measured performance of a system ... the resulting sound will be sterile. I have heard a dozen examples of this phenomenon.

While we are here, we may as well ask - fidelity to what, anyway? Fidelity to a recording, or fidelity to a live performance? Firstly, any recording which features anything synthesized, or anything with artificial reverb, or anything using an amplified source, or digitally corrected singing - is fidelity to what some recording engineer thinks it should sound like. In other words, it is fidelity to a subjective impression of what music should sound like. Don't forget he is making this decision on studio monitors or headphones - nothing like the system you have at home.

If we are talking about fidelity to a live performance - well, anybody who has been to a live performance knows there is no such thing. Your experience will vary greatly depending on where you are sitting. This is unless we are talking about very simple music - say, a violin solo. Recordings of live performances are subject to as much twiddling of knobs to get the balance correct as well.

If there is a type of music that drives me up the wall, it is Diana Krall type singing. Voices are heavily autocorrected, there is artificial digital reverb all over the place. I call it "karaoke machine sound". I think it is fortunate that the kind of music that I do like (classical) is free from the most egregious digital tinkering like this, and most CD's sound natural. But, if you happen to like this type of music ... and the sound of digital reverb drives you up the wall ... what are you to do?

Unless you are the type who only collects audiophile recordings, anybody who has a real collection of music knows that the sound from CD to CD varies greatly. Recording engineers are not Gods. If I don't like the sound that is coming from my CD - well it is my CD and my system. I can do what I want to make it sound better to my ears.

Earlier, someone made the point that audio equipment was designed by objectivists. Maybe, but it's the subjectivists who are making your CD's.