>> twelve participants would be asked to come to Hanover.
It's a mistake to apply vanilla statistical thinking here. The 12 participants were not randomly drawn from the German population, they were extremely skewed towards enthusiasts/professionals: audio engineers, an owner of an actual Nautilus 801, someone who worked on MP3/AAC at Fraunhofer IIS, someone who works preparing masters for Deutsche Gramophon. If these are the people who have enormous difficulty distinguishing 256kbps MP3 from the CD original, I'm certainly not going to worry that I am going to miss out on anything with 256kbps MP3.
If 12 Grand Slam participants tell me they can't tell the difference between a standard $100 and a $1000 high end tennis racket, I'm not going to delude myself into thinking that it's going to make any difference for me.
It is still undetermined if having 12 highly-skilled professionals in the experiment is enough to have a conclusive experiment.
Also, this subject is so difficult to get right that the authors of the article themselves hedged by saying that experiment "does not support watertight conclusions".
The only one who was significantly able to tell if something was mp3 encoded or not, was a guy with a hearing damage who loved punk music. In fact, mp3 was developed for persons with normal hearing. So it is well possible that he was able to tell differences where other people were unable to.
It's quite easy to overprocess a digital audio file and wind up with something that is pristine as far as frequency response, but flat and 'pod people' like as far as emotive cues and intensity. Aliasing and cumulative losses to word length issues have a lot to do with it.
It's VERY easy to make digital stuff accurately represent frequencies like 2 Hz or 35kHz that our ears don't hear. It's a lot harder to make the digital stuff perform in the midrange when our perception can go, inconsistently and irregularly, waaaay beyond what we're used to thinking of as the limits.
By the way, the lossy compression algorithms don't try to produce exact frequency response but to leave out stuff that humans wouldn't hear anyway and compress the rest.
In my experience, headphones always yield the best results, and surprisingly it doesn't matter if I use the stock earbuds from my phone or a nice set of AKG over-ear headphones. It's not a matter of absolute sound quality, just the fact that you cut out room interactions and get the sound straight to your ears makes a big difference.
MP3 has some built-in flaws that no encoder can completely cover up, short sharp sounds like castanets really expose the pre-echo, harpsichord shows similar issues. It also has a tendency to make cymbals sound "washy" or "underwater", which all lossy codecs do to some degree, but MP3 is especially bad.
Still, at 192kbps I have to really focus to hear it in normal listening, but it's more or less always there even at 320kbps in problem tracks, if I really focus in on short sections. It just sounds subtly "off". But I hope no one actually listens to music like that, in short repeated sub-1 second sections to narrow in on a specific castanet snap ;-)
As for more modern codecs like Opus and AAC, it's generally completely transparent for me at 128kbps, and that's with a bit of playing it safe, I'm pretty sure I could drop Opus down to 96kbps. Modern codecs are really impressive.
I keep my music library in FLAC, both because I know it's CD quality and because it's an archive. I want to be able to convert the tracks to any new codec that may come along, if I need to.
My library is 280GB currently, and storage is cheap :-)
I understand the sentiment. But the reality is, if the re-encoding is not likely going to happen within the next 10 years, your hearing will probably have deteriorated so much that you probably won't hear the difference anymore anyway (assuming you can hear a difference today, which is a big assumption).
Thank you for confirming this! I record my analog synths that I play through headphones off an old mixing board, however when it comes through my ADC->iPad, stuff seems to get lost and I spend time adjusting the mix and ADSR for recording. Have been seriously mulling a reel to reel, but many others have had the same idea and the market prices are astronomical.
Recording should be done at 96kHz 24-bit or higher, to not have to meticulously optimize recording levels and to allow room for mixing and effects, without raising the noise floor to noticable levels.
Convert to normal CD quality as the last step before distribution.
That’s around ¾ the amount of music I carry on my iPhone :-)
[Anecdotal] One example of the difference between MP3 and lossless: the "image" [1] on 256kbps MP3s is worse compared to the the original uncompressed, lossless, versions (but the listening room must be appropriately prepared to reproduce a good image).
This is a highly subjective topic. IMO we'll never reach full agreement. Personally, I listen MP3 while on-the-go and lossless music at home.
Important to keep in mind the "size" of the experiment. Two interesting quotes from the article in c't magazine:
> twelve participants would be asked to come to Hanover.
> It's true that the data we collected does not support watertight conclusions, but they do provide interesting insights.
[1] https://en.wikipedia.org/wiki/Stereo_imaging