Highest useful bit depth and sample rate for audio files

[mgsylvestre]

Is there any information available as to what is the highest useful bit depth and sample rate for audio files above which there is no perceptible difference?

I assume that there is a limit above which there is no audible difference. Is that limit 24/192? Something above or below that?

Are there any research papers or tests results publicly available in that regard?

Anybody has any opinion on the subject?

 

[Adam]

I don't know of any papers on this, but here is my input based on my experience.

The bitrate at which someone will no longer be able to tell a difference is based upon:

• The person
• The equipment on which the audio is being played (e.g. computer speakers, car system, $50k reference system)
• The source materal (I'm guessing)

In addition, the bitrate at which a person cares about the difference can be different from the bitrate at which a person can tell a difference.

 

[sholling]

This should draw some interesting answers. Especially since many vinyl lovers consider CDs to be mid-fi rather than hifi.

 

[mgsylvestre]

OK let's clarify the question

Good points Adam, so let's start with this:

What are the bit rate and sample depth at which a group of people listening blind could tell the difference (not that they would necessarily care) more than 50% of the time (otherwise its flipping a coin)? Surely there must have been tests done about this already?

As to equipement, to try to eliminate variables, let's assume nice top-notch headphones and electronics. As to source, let's try a number of different sources encompassing most genres.

It seems to me that this question is crucial, as it more or less determines a lot of other things as far as audio file acquisition and equipment procurement is concerned. I don't mind spending money if necessary, but you can buy speaker "cable" for $7K a foot...

 

[BoredSysAdmin]

I think it's very much depend on source of audio files.
If you are ripping them off CD - lossless compression (Flac/Ape) is your single answer in the quest for high quality audio.
Flac/Ape I believe don't have bitrates per-se.
I assume it's somehow possible to rip DVD-Audio or SACD, but I'm not familiar with methods to do it.
If you are stuck with mp3 - Even at 320Kpbs some encoder limitations might showup in compressed result.

 

[Boerd]

Is there any information available as to what is the highest useful bit depth and sample rate for audio files above which there is no perceptible difference?

I assume that there is a limit above which there is no audible difference. Is that limit 24/192? Something above or below that?

Are there any research papers or tests results publicly available in that regard?

Anybody has any opinion on the subject?

I have quite a few 24bit/88Khz, 24/96, 24/192 recordings from labels like Naim/Linn/etc. To be honest I cannot tell the difference from 16/44.1 cleanly recorded CDs (from the same label) and I listened to these recordings on a few expensive setups.
I stopped buying the 24 bit version and I am currently buying the 16/44.1 versions from the online audiophile labels.
I always tell people a good 16 bit recording has all the details you need plus 90 db dynamics range.
The big problem with CDs is the loudness craziness which from a potential of 90 db dynamic range takes some CDs into the 20db range.

 

[Boerd]

the bitrate at which a person cares about the difference can be different from the bitrate at which a person can tell a difference.

Really really good point. Sad but very true.

 

[Duff man]

It will vary from person to person and listening environment to listening environment and their are too many variables to come up with a answer that is right for everyone. If you are listening to classic rock mp3's on an mp3 player with cheap ear bud headphones, any upgrade above 192 will likely not be noticed by most people.

Those with good ears who listen to high quality recordings on a decent home system will probably never be happy with mp3's, even at a 320 bit rate.

For me, it's a bit of both. I have a lot of FLAC I listen to through an "okay" set of speakers and enjoy the improved sound quality over mp3 recordings. But I can listen to the same songs while working out or during a deck party at a 192 bit rate and be just as happy.

 

[corey]

Is there any information available as to what is the highest useful bit depth and sample rate for audio files above which there is no perceptible difference?

I assume that there is a limit above which there is no audible difference. Is that limit 24/192? Something above or below that?

Are there any research papers or tests results publicly available in that regard?

Anybody has any opinion on the subject?

1) Yes, there is.
2) The limit is below that.
3) Yes.
4) Perhaps you should ask for facts (blind test results) rather than opinions.

It's late Christmas Eve, and I don't want to look it up, but Google is your friend. I think that the bottom line is that there has never been a blind test that shows a difference between 256 mbps mp3s and the source. There's lots of opinions that there is a difference, but no blind tests.

There's a site that will send you some samples, and you can participate in the testing.

You can also set up your own test, if you use someone else to help make the tracks random. Rip the same well recorded song at different bit rates, with a good ripper & coder (EAC and LAME), and take a 15-20 second sample of each. Record them back in a random order on a regular audio cd, and rate the tracks.

 

[sholling]

I think that the bottom line is that there has never been a blind test that shows a difference between 256 mbps mp3s and the source. There's lots of opinions that there is a difference, but no blind tests.

It would be interesting to see but I suspect that there are artificial limitations involved in many of those tests.

1. Complexity of the source material.
2. Familiarity with the originial uncompressed source marerial.
3. Quality of the headphones or speakers used for playback.
4. And the most critical - the hearing and pickiness of the listener.

It's fairly easy to skew results to show what you want them to if the tester has an agenda such as promoting MP3s. Simply use crappy headphones and a 192bt MP3 will sound as good as a CD. I'm not saying that all of those tests are deliberately skewed but I suspect some are while others are inadvertently skewed by the limitations that I pointed out above.