What is the "just noticeable difference" in audio loudness?

[esldude]

I found an article that addresses this question specifically here

In it a chart of research results over 60 years is given & the lowest JND is 0.25dB F.E. Toole and S. Olive, "The Modification of Timbre by Resonances: Perception and Measurements", JAES vol 36, # 3, March 1988, pp 122-142

The author states "Toole and Olive, on the other hand, in their 1988 study used pink noise for their acoustic signal source and determined that a 5 kHz resonance, with Q = 1 was just detectable at .25 dB." But later "The .25 dB figure quoted from the Toole & Olive research seems to contradict this (his JND of 0.75dB or 1dB), but consider the filter Q = 1. That's a pretty broad chunk of the audible spectrum over which that resonance exists. With the ear-brain combo performing an integration across that broad a portion of the audible spectrum than its easy to see how a large amount of acoustical energy is captured, leaving a change that small noticeable. However, pink noise is not real world and one thing my research has shown is that the hearing process reacts very differently to different types of sound; a .25 dB detectable difference using real music just isn't plausible and the research supports that."

He goes on to state "In this particular article, I settled on a minimum discernable difference dB value of .75 - 1.0. My experience has shown that this is what the average listener, under average listening conditions, listening to music played back through typical consumer-grade audio gear will be able to clearly identify - and do so repeatedly."

So, I'm still looking for any research which backs up the claim that 0.1dB amplitude difference is noticeable as a quality difference when playing music? Any links, ESL?

I am not at home and can't provide any now. It should be pretty easy to find some references to that.

If you match to .1 db your tests won't be corrupted by loudness issues. Is it possible to be more at some frequencies than others? Maybe, the JND (which btw is not .1db again the confusion) results indicate that is possible. Just match to .1 db and you need not worry.

The broad Q .25 db response differences are discernable with pink noise and some other signals. However, I don't know that they are heard as louder. They are heard as different. JNDs for hearing loudness are larger than for hearing two signals as different.

Why not make some pink noise and put some broad gentle peaks in the response and listen blind. Should be interesting and educational for you. Ditto with some music files. Make one .5 db louder and do an ABX in Foobar on it. See what happens and how it sounds/feels for you. Then try a .25 db.

 

[rbbert]

...
The broad Q .25 db response differences are discernable with pink noise and some other signals. However, I don't know that they are heard as louder. They are heard as different. JNDs for hearing loudness are larger than for hearing two signals as different...

I think this is a key point perhaps first proposed when ABX testing came into being. Regardless of the origin, it also ties in with the article by JJ referenced earlier.

 

[jkeny]

I am not at home and can't provide any now. It should be pretty easy to find some references to that.

Yes, I was hoping that it would be easy to find but so far I haven't turned up any references other than what I posted.

If you match to .1 db your tests won't be corrupted by loudness issues.

I know that's the claim & I'm wondering what supporting evidence there is to back up this claim.

Is it possible to be more at some frequencies than others? Maybe, the JND (which btw is not .1db again the confusion) results indicate that is possible. Just match to .1 db and you need not worry.

I know it's not the JND that will be noticed as a loudness difference, the claim is that it will be perceived as a quality difference.

The broad Q .25 db response differences are discernable with pink noise and some other signals. However, I don't know that they are heard as louder. They are heard as different. JNDs for hearing loudness are larger than for hearing two signals as different.

Sure, that's understood

Why not make some pink noise and put some broad gentle peaks in the response and listen blind. Should be interesting and educational for you. Ditto with some music files. Make one .5 db louder and do an ABX in Foobar on it. See what happens and how it sounds/feels for you. Then try a .25 db.

As I said already broadband signals which span a number of ERB filters are additive so the 0.1dB actually could be perceived as louder due to this addition of signal power in each ERB that the signal spans. So, yes noise as the test signal, may well be perceived as of a noticeably different quality if different by 0.1dB but I can't see this applying to music signals. I will do some testing but before I do - have you done these music tests & proved to yourself that 0.1dB amplitude difference is discernible?

 

[esldude]

Yes, I was hoping that it would be easy to find but so far I haven't turned up any references other than what I posted.

I know that's the claim & I'm wondering what supporting evidence there is to back up this claim. I know it's not the JND that will be noticed as a loudness difference, the claim is that it will be perceived as a quality difference.

Sure, that's understood

As I said already broadband signals which span a number of ERB filters are additive so the 0.1dB actually could be perceived as louder due to this addition of signal power in each ERB that the signal spans. So, yes noise as the test signal, may well be perceived as of a noticeably different quality if different by 0.1dB but I can't see this applying to music signals. I will do some testing but before I do - have you done these music tests & proved to yourself that 0.1dB amplitude difference is discernible?

First off, .1 db is not discernible with music or other signals. That is the point to match so close it has no effect on the outcome of testing.

I just did a quick ABX of pink noise over headphones with a laptop which differed by .25 db. I got 11 of 12. In the past somewhere around .25 db is where I can reliably do pink noise. Music is seems to vary from time to time more and type of music. When I have done it in the past .25-.40 db is enough. That will usually be done with a very short segment of 3 or 4 seconds.

Also worth noting my experience with ABX as you near the limits of differences you hear. Typically, I will be thinking I don't hear a difference anymore. Yet will do well on the blind test. Make the difference smaller and I can't discern it. Make it larger and I can discern it and also think I hear it differently. Even the pink noise I felt it was barely something. Bump it up to .75 db and it is clearly different. So in my experience you can detect a difference blind after it has gotten small enough you subjectively and consciously don't hear any difference. If I make the pink noise 1 db different I can ace it without even playing both A and B against X usually.

 

[jkeny]

First off, .1 db is not discernible with music or other signals. That is the point to match so close it has no effect on the outcome of testing.

I just did a quick ABX of pink noise over headphones with a laptop which differed by .25 db. I got 11 of 12. In the past somewhere around .25 db is where I can reliably do pink noise. Music is seems to vary from time to time more and type of music. When I have done it in the past .25-.40 db is enough. That will usually be done with a very short segment of 3 or 4 seconds.

Also worth noting my experience with ABX as you near the limits of differences you hear. Typically, I will be thinking I don't hear a difference anymore. Yet will do well on the blind test. Make the difference smaller and I can't discern it. Make it larger and I can discern it and also think I hear it differently. Even the pink noise I felt it was barely something. Bump it up to .75 db and it is clearly different. So in my experience you can detect a difference blind after it has gotten small enough you subjectively and consciously don't hear any difference. If I make the pink noise 1 db different I can ace it without even playing both A and B against X usually.

OK, thanks, I'll try that. So you are saying that 0.25dB is your JND with Pink noise & 0.25 to 0.4dB is your JND with music?
I'm interested in whether the music sounds like a quality difference & that the higher volume piece is always considered the better quality reproduction - this is the claim that is always used & I wondered where it came from or was it just another audio myth?

 

[Speedskater]

If you are interested in reading more about this subject, the AES has many papers and Power Points. Start with J.J. Johnston.

http://www.aes-media.org/sections/pnw/ppt/

 

[esldude]

OK, thanks, I'll try that. So you are saying that 0.25dB is your JND with Pink noise & 0.25 to 0.4dB is your JND with music?
I'm interested in whether the music sounds like a quality difference & that the higher volume piece is always considered the better quality reproduction - this is the claim that is always used & I wondered where it came from or was it just another audio myth?

I've seen several informal tests of which is better. In every instance louder wins. So I don't know that original origin, but it seems to be true.

 

[jkeny]

I've seen several informal tests of which is better. In every instance louder wins. So I don't know that original origin, but it seems to be true.

How informal? Tests with 0.25dB difference between the playback amplitude of music files & the louder one was considered better?
I believe that at this low a level of amplitude difference we are on threshold (if not below it) of perceptually differentiating between different playback of music & the only way this could be known to be a difference would be through the statistical proof from a blind test. That's why I'm puzzled why there is no blind test that shows this as it's the only way to back up the claim?

In other words the only way to know if this is correct is to do a blind test with enough trials - it's so close to being an inaudible difference that any rational person would query it.

So I have to conclude that in the absence of such blind test results, the claim is bogus & just another audio myth (but this time coming from the objectivists)

 

[jkeny]

If you are interested in reading more about this subject, the AES has many papers and Power Points. Start with J.J. Johnston.

http://www.aes-media.org/sections/pnw/ppt/

There's a lot there & I have read JJ's slides & text before. I think he mentions something about 0.3dB as being possible to differentiate on a clear day with the wind in your back but I'm not sure he meant with music as the signal. This figure of 0.3dB must have come from somewhere but I can't find the blind tests that show it's validity

 

[esldude]

How informal? Tests with 0.25dB difference between the playback amplitude of music files & the louder one was considered better?
I believe that at this low a level of amplitude difference we are on threshold (if not below it) of perceptually differentiating between different playback of music & the only way this could be known to be a difference would be through the statistical proof from a blind test. That's why I'm puzzled why there is no blind test that shows this as it's the only way to back up the claim?

In other words the only way to know if this is correct is to do a blind test with enough trials - it's so close to being an inaudible difference that any rational person would query it.

So I have to conclude that in the absence of such blind test results, the claim is bogus & just another audio myth (but this time coming from the objectivists)

One was testing for other effects but a file one dB louder was used as a control. Instructions were to pick best sounding. Results were random except the louder file which was chosen every time except once.

Another was to choose best sounding file in each pair. Some were the same and some pairs differed in loudness by .5 dB. Positive results on the pairings with a level difference.

No one commented they heard a loudness difference.