A challenge for those who want one

[Vincent Kars]

they don't SOUND any different.

Most of the time they don't, assuming sufficient bitrate.
But some times they do, the famous killer samples.
Observe that even the makers (Fraunhofer et. al.) never claimed 100% transparency.
https://hydrogenaud.io/index.php/topic,77128.0.html

 

[Koeitje]

You can identify MP3 if you know what artifacts to listen for. I don't know, so I don't hear it but a friend can pass an ABX test on certain tracks.

 

[JSmith]

You can just use this?

There's a few similar listening tests online.

The one you posted I got 4/6 correct using laptop earbuds. One incorrect choice I picked 128kbps, but rushed that one a bit... the other I selected 320kbps, which was basically indiscernible from the WAV. I listened to each sample, comparing, at least 10 times each and there are differences, but very subtle... need to listen for certain things. Correct and sounding "better" doesn't always equate either, depends on the music and what sound you're trying to isolate. It's easier to pick 128kbps out as more different than 320kbps or the WAV, for me.

You or others may way to try some of these which are more ABX than pick which is labelled what;

http://abx.digitalfeed.net/list.html

I just tried 96kbps to start and got all correct over 20 tests as it's easy to hear the compression at such a low bitrate, sounds horrible and empty;

Going to 128kbps using the laptop and earbuds still was more difficult, but able to pick differences, however the music had some life back in it;

Onto 160kbps and believe it or not I got exactly the same score as 128kbps, same differences to listen for but less prominent;

However... once I got to 192kbps it was a different story;

I still thought I could hear some of the changes I did earlier at a much reduced level... but the results show otherwise, more flip of the coin. Either that or listening fatigue (especially the same small sample over and over), maybe concentration lapsing. 160kbps seems like a compromise between 128kbps and 192kbps.

I've spent a few hours doing this with short breaks here and there. As I progress through the test the differences I hear become harder to pick, seems around after 4 or 5 attempts it starts for me... if take a break for 5 or so mins it becomes easier again. I put this down to auditory memory, where my brain knows I've listened to this over and over. It knows the "easier" sound, so based on the memory of it, fills in the perception loss... but after the 5 or so mins break that fades, which IMO is why the difference becomes more apparent again. So yeah stopping for now and will try the higher bitrate tests another time using proper headphones.

What is funny though (and expected) is it's easier to hear the distortion in these earbuds than the differences in A/B when critically listening.

The Killers sample is quite apt... starts with "simple", middle "what are you made of" and ends with "running out of time".

Whilst this is fun from a personal testing perspective, it's not doing or finding anything new in general whatsoever...

Exactly what needs to be retained to preserve perceptuality is retained though for almost all listening situations, most equipment and music, even with a "lossy" codec like MP3 encoded at 320kbps (assuming the encoder is doing it's job properly).

Many people are just listening to music streams or files using cheapo earbuds on a phone... so those that say there is no case for audio compression algorithms clearly are out on a limb as most phone accounts have data limits/caps. Not all countries have high speed data connections available either and not everyone can afford large data plans. Even though FLAC compresses, it's not efficient (I know it's not meant to be) and thus still takes up unnecessary data bandwidth or storage space, encoding unnecessary or non-existent inaudible higher frequency components.



JSmith

 

[abhijitnath]

I found that the classical ones were easier to pick out, compressed rock, less so. So source material (of course!) matters.

 

[Mikko Dee]

https://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality. You can just use this?

I got 3 out of 6….

 

[Mikko Dee]

Imo, storage is now a relatively cheap commodity. I store all my music on 2 1tb ssd drives in lossless formats. Why? Because I can and I don’t need/want/care about the actual or perceived loss in quality using a compressed or lossy format. This ain’t the 80’s….

 

[ferrellms]

I got 3 out of 6….

Not really valid since they use 128Kbs in their comparisons. You need to go to 192Kbs for complete transparency. This is what Fraunhofer has found in their blind codec testing. One would expect some recognition of a difference with 128K included. Of course, it is really hard to find a streaming service that uses 128k these days - 192 seems to be the minimum.

 

[ferrellms]

Most of the time they don't, assuming sufficient bitrate.
But some times they do, the famous killer samples.
Observe that even the makers (Fraunhofer et. al.) never claimed 100% transparency.
https://hydrogenaud.io/index.php/topic,77128.0.html

Whatever Fraunhofer claimed, in their blind tests none of the expert listeners could hear a difference with 192kbs up. Have you tried the test?

 

[ferrellms]

There's a few similar listening tests online.

The one you posted I got 4/6 correct using laptop earbuds. One incorrect choice I picked 128kbps, but rushed that one a bit... the other I selected 320kbps, which was basically indiscernible from the WAV. I listened to each sample, comparing, at least 10 times each and there are differences, but very subtle... need to listen for certain things. Correct and sounding "better" doesn't always equate either, depends on the music and what sound you're trying to isolate. It's easier to pick 128kbps out as more different than 320kbps or the WAV, for me.

You or others may way to try some of these which are more ABX than pick which is labelled what;

http://abx.digitalfeed.net/list.html

I just tried 96kbps to start and got all correct over 20 tests as it's easy to hear the compression at such a low bitrate, sounds horrible and empty;

View attachment 133553

Going to 128kbps using the laptop and earbuds still was more difficult, but able to pick differences, however the music had some life back in it;

View attachment 133554

Onto 160kbps and believe it or not I got exactly the same score as 128kbps, same differences to listen for but less prominent;

View attachment 133564

However... once I got to 192kbps it was a different story;

View attachment 133565

I still thought I could hear some of the changes I did earlier at a much reduced level... but the results show otherwise, more flip of the coin. Either that or listening fatigue (especially the same small sample over and over), maybe concentration lapsing. 160kbps seems like a compromise between 128kbps and 192kbps.

I've spent a few hours doing this with short breaks here and there. As I progress through the test the differences I hear become harder to pick, seems around after 4 or 5 attempts it starts for me... if take a break for 5 or so mins it becomes easier again. I put this down to auditory memory, where my brain knows I've listened to this over and over. It knows the "easier" sound, so based on the memory of it, fills in the perception loss... but after the 5 or so mins break that fades, which IMO is why the difference becomes more apparent again. So yeah stopping for now and will try the higher bitrate tests another time using proper headphones.

What is funny though (and expected) is it's easier to hear the distortion in these earbuds than the differences in A/B when critically listening.

The Killers sample is quite apt... starts with "simple", middle "what are you made of" and ends with "running out of time".

Whilst this is fun from a personal testing perspective, it's not doing or finding anything new in general whatsoever...

Exactly what needs to be retained to preserve perceptuality is retained though for almost all listening situations, most equipment and music, even with a "lossy" codec like MP3 encoded at 320kbps (assuming the encoder is doing it's job properly).

Many people are just listening to music streams or files using cheapo earbuds on a phone... so those that say there is no case for audio compression algorithms clearly are out on a limb as most phone accounts have data limits/caps. Not all countries have high speed data connections available either and not everyone can afford large data plans. Even though FLAC compresses, it's not efficient (I know it's not meant to be) and thus still takes up unnecessary data bandwidth or storage space, encoding unnecessary or non-existent inaudible higher frequency components.



JSmith

You evidently have perception ability beyond that of the many expert listeners Fraunhofer used in their blind tests! Good for you.