Recently, in a non- hydrogenaud.io forum I stated: "<0.1% of people would be able to distinguish competently-encoded 256kbps .mp3 from uncompressed, and only about 1% would be able to distinguish 192kbps. 128kbps is garbage, however."
This was an educated guess. I expected it to be fairly straightforward to find actual studies about this. However, without exhaustive efforts (and without membership in certain organizations), I've only managed to find scattered data (some of which is ABX) and some non-peer reviewed "studies" that don't satisfactorily answer the question.
Are there any published studies about this, peer reviewed or even just "reasonably competent"?
Thanks.
or even just "reasonably competent"?
Have you seen http://wiki.hydrogenaud.io/index.php?title=Hydrogenaudio_Listening_Tests ?
Are there any published studies about this, peer reviewed or even just "reasonably competent"?
Not many studies looking at 25 year old audio formats. You might find a few papers on early MP3 encoders, but later ones are much better. Instead, take a look at the HA and other community tests.
Have you seen http://wiki.hydrogenaud.io/index.php?title=Hydrogenaudio_Listening_Tests ?
Yes. Unfortunately, those appear to be comparisons of perceived quality vs. MP3 encoder.
What I'm looking for is:
For a given MP3 encoder (assuming CBR), at what bitrate does a statistically significant % of listeners become able to distinguish between uncompressed sources and the MP3 compressed version?
For a given MP3 encoder (assuming CBR), at what bitrate does a statistically significant % of listeners become able to distinguish between uncompressed sources and the MP3 compressed version?
A rating of 5 in those tests means that the evaluator could not distinguish from the lossless file. To answer your question, most people had trouble distinguishing between lossy and lossless around 128k for the better MP3 encoders and for the subset of tracks included in those tests, but most could notice minor differences by 96k.
Usually not a good idea to present statistical figures without evidence that supports them. The subject at hand is definitely no exception. Pulling figures out of the air in a way that polarizes the topic is even more foolhardy. It certainly won’t sway those who already hold an opinion.
Anecdotally, I’ve come across many people who can regularly differentiate 256kbit mp3 made with a competent encoder from the original source. I have no reason to doubt any of them.
Usually not a good idea to present statistical figures without evidence that supports them. ...
Yeah, I think I've learned that now ;-)
Anecdotally, I’ve come across many people who can regularly differentiate 256kbit mp3 made with a competent encoder from the original source. I have no reason to doubt any of them.
But if they claimed that here, you would demand that they post ABX results, right?
Are there any published studies about this, peer reviewed or even just "reasonably competent"?
Not many studies looking at 25 year old audio formats. You might find a few papers on early MP3 encoders, but later ones are much better. Instead, take a look at the HA and other community tests.
It might be a 25 year old format, however is it still the most popular lossy format in use (accounting for around 50% a few years ago), so is still very relevant.
or even just "reasonably competent"?
Have you seen http://wiki.hydrogenaud.io/index.php?title=Hydrogenaudio_Listening_Tests ?
Is there enough interest for another listening test?
I suspect that audio enthusiasts don't care about MP3, and non-enthusiasts don't care about listening tests.
Anecdotally, I’ve come across many people who can regularly differentiate 256kbit mp3 made with a competent encoder from the original source. I have no reason to doubt any of them.
But if they claimed that here, you would demand that they post ABX results, right?
Right, though more often than not by members who know what they are and how they're used, ABX tests support their claims. Many know to present them ahead of time.