A compelling argument for uncompressed audio and hi-def audio

[asliarun]

I came across a very interesting article in Ars Technica about research conducted on how sensitive humans are to time and frequency. While it is commonly known that we perceive loudness in a non-linear manner, this research proves that we also hear timing (between two sounds or notes) and frequency (pitch) in a non-linear manner.

Interestingly, most subjects of the study were able to perceive timing and pitch 10 times better than the conventionally accepted linear model of our hearing! Some subjects also excelled dramatically in either their perception of timing or pitch (but not both). I know little about mp3 encoding but I understand from the article that compression techniques in audio (such as mp3) are based on this linear model which may turn out to be based on totally wrong fundamentals. mp3s might actually be butchering music (or the way we hear music) to a much larger extent than we thought.

I've held a view that 320kbps mp3s are "good enough" for casual listening in most cases, but this study really makes me question that notion.

 

[ranjeetrain]

There is nothing surprising in that article. All it says is (1) Humans have better audio perception than the previously regarded uncertainty limit and (2) the previously model used for audio compression (linear model) is not the most appropriate.

And I don't find anything surprising in both the facts at all. The fact no 1 is, in fact, a very easily witness-able phenomena. How come two people listening to same system, some can hear a difference others can't? Is one of them lying? Perhaps not, perhaps one of them really hears a difference and the other doesn't. If two people can run at difference speeds, if two people need different amount of food intake per meal, if two people can have different accuracy of vision, why can't two people have different levels of hearing? Simple as that, isn't it?

I find nothing new or surprising in the first part.

For the second part: Again, we all know none of the audio compression methods available currently are perfect. If there was a perfect compression method, it would have become de-facto standard by now. But, since that's not the case, we see new compression methods appearing on the horizon every once in a while. It became popular with MPEG compression, then other contenders jumped into foray. OGG Vorbis, Advanced Audio Coding etc etc. Each subsequent method tried to eliminate the deficiencies with the previous methods. And some of them succeeded as well. After spending weeks of extensive A/B, I did establish (for myself) that AAC compression was better (more efficient, it removed less information for the same bit-rate and sounded better to me).

At some point of time, I decided to rip all my media all over again and encode them with AAC. But didn't do due to lack of time. In the meantime, the storage prices have come down so heavily that there is NO NEED of using compressed files anymore. For example, a 32 GB card costs as little as Rs 1.5k, which is good for storing 70 hours uncompressed audio. It couldn't have been so only a few years ago. I had to shell out upwards of 5k to get around 100 hours of compressed audio storage.

My point is -- that article says nothing new, nothing significant. With semi-conductor based storage devices becoming dirt cheap, even phone/portable players users NEED NOT store compressed audio (if their device is good even to tell the difference between compressed and uncompressed audio that is.)

As far as 320 kbps music being good enough -- I am with you on that. When I leave my golden ears aside, 320kbps is more than adequate for listening on anything but very good systems. I have many tracks that are so good musically they hardly draws my attention towards the fact that they are compressed at 128 kbps.

 

[asliarun]

There is nothing surprising in that article. All it says is (1) Humans have better audio perception than the previously regarded uncertainty limit and (2) the previously model used for audio compression (linear model) is not the most appropriate.

And I don't find anything surprising in both the facts at all. The fact no 1 is, in fact, a very easily witness-able phenomena. How come two people listening to same system, some can hear a difference others can't? Is one of them lying? Perhaps not, perhaps one of them really hears a difference and the other doesn't. If two people can run at difference speeds, if two people need different amount of food intake per meal, if two people can have different accuracy of vision, why can't two people have different levels of hearing? Simple as that, isn't it?

I find nothing new or surprising in the first part.

For the second part: Again, we all know none of the audio compression methods available currently are perfect. If there was a perfect compression method, it would have become de-facto standard by now. But, since that's not the case, we see new compression methods appearing on the horizon every once in a while. It became popular with MPEG compression, then other contenders jumped into foray. OGG Vorbis, Advanced Audio Coding etc etc. Each subsequent method tried to eliminate the deficiencies with the previous methods. And some of them succeeded as well. After spending weeks of extensive A/B, I did establish (for myself) that AAC compression was better (more efficient, it removed less information for the same bit-rate and sounded better to me).

At some point of time, I decided to rip all my media all over again and encode them with AAC. But didn't do due to lack of time. In the meantime, the storage prices have come down so heavily that there is NO NEED of using compressed files anymore. For example, a 32 GB card costs as little as Rs 1.5k, which is good for storing 70 hours uncompressed audio. It couldn't have been so only a few years ago. I had to shell out upwards of 5k to get around 100 hours of compressed audio storage.

My point is -- that article says nothing new, nothing significant. With semi-conductor based storage devices becoming dirt cheap, even phone/portable players users NEED NOT store compressed audio (if their device is good even to tell the difference between compressed and uncompressed audio that is.)

As far as 320 kbps music being good enough -- I am with you on that. When I leave my golden ears aside, 320kbps is more than adequate for listening on anything but very good systems. I have many tracks that are so good musically they hardly draws my attention towards the fact that they are compressed at 128 kbps.

To me, what was new and surprising was the extent to which people are sensitive to either pitch variations or timing variations. I personally haven't had that much experience doing critical listening and trying to evaluate prat, speed, musicality etc. Hence this article was quite informative for me. I am also personally very sensitive to tonality and pitch variation and my teeth get on an edge when I hear subtle off key - in fact the subtler it is, the worse it gets. I am fairly tolerant to timing though.

I can now better appreciate why people spend so much effort trying to get these aspects right and why there can never be a one size fits all for audio systems.

I do agree with you about not needing to bother with compressed audio anymore.

 

[Thad E Ginathom]

Just to be pedantic... the problem is not with data compression, which has to do with methods of storage rather than audio, but with lossy data compression.

What is most interesting in this kind of work is that it is helping to explain how we hear. That's interesting whether we compress or not --- but I guess it will also be useful stuff for those that invent the compression codecs of the future.

 

[asliarun]

Just to be pedantic... the problem is not with data compression, which has to do with methods of storage rather than audio, but with lossy data compression.

What is most interesting in this kind of work is that it is helping to explain how we hear. That's interesting whether we compress or not --- but I guess it will also be useful stuff for those that invent the compression codecs of the future.

Yes, and it also goes beyond that. This research makes a case for audio systems that are configured differently to cater to different end user sensitivities, or that are revealing in different ways.