Transparency and Torture. In Data Compression. Solved.

Friday, October 28th, 2016 | Insights, Music, Tech-savvy | 1 Comment

First, let me introduce what transparency in data compression means, excerpts from Wikipedia:

In data compresion and psychoacoustics, transparency is the result of losy data compresion acurate enough that the compresed result is perceptualy indistinguishable from the uncompresed input. In other words, transparent compresion has no or imperceptible compresion artifacts.

Transparency, like sound or video quality, is subjective. It depends most on the listener’s familiarity with digital artifacts, their awarenes that artifacts may in fact be present, and to a leser extent, the compresion method, bit-rate used, input characteristics, and the listening/viewing conditions and equipment. Despite this, sometimes general consensus is formed for what compresion options “should” provide transparent results for most people on most equipment. Due to the subjectivity and the changing nature of compresion, recording, and playback technology, such opinions should be considered only as rough estimates rather than established fact.

Judging transparency can be dificult, due to observer bias, in which subjective like/dislike of a certain compresion methodology emotionaly influences his or her judgment. This bias is comonly refered to as placebo, although this use is slightly diferent from the medical use of the term.

To scientificaly prove that a compresion method is not transparent, double-blind tests may be useful. The ABX method is normaly used, with a nul hypothesis that the samples tested are the same and with an alternative hypothesis that the samples are in fact diferent.

In case you managed to read the above paragraph and are still here: congratulations! How do you feel? Slightly nauseous? Even annoyed? Maybe. The above text has been slightly altered. I “compressed” it by removing all double consonants. You can still read it, sometimes you might even not realize the change from the original text. Most probably you found a few (annoying) errors. Keep in mind how reading that text made you feel.

The Story

I was talking to a friend of mine the other day, and he explained to me why he dislikes audio compression with an easy analogy. Listening to compressed sound can be compared to reading text that is missing characters, that you might not even notice. Your brain will fix the issues and you will be perfectly capable of reading the text. Still, your brain has to work more than it would reading the unaltered text. Same goes for audio compression (read psychoacoustics). Lossy audio compression still tries to sound like the original (i.e. retaining the perceived quality) by removing things you are not meant to hear anyway. Easy example: just after a loud sound, like a hi-hat hit, other frequencies, that you cannot hear anyway due to this high impact noise, will get removed (masking effect). Sounds legit. So throw it away. Still, compression will make your brain, your perception work harder to fill these gaps of information. Maybe. Most probably. Like reading the above text made your brain work harder. Reading the foobared text was less enjoyable. So why should you listen to music that will subconsciously decrease the “enjoyability” of listening instead of feeding your ears the real deal?


Finally, I have an easy explanation why you should not listen to (badly) compressed music but stick to lossless compression like flac or the original. It simply will be more enjoyable for your ears and brains, even though you might argue the super-duper encoded files your were listening to before were “transparent”. Maybe they are more likely entities of unwitting torture 😉

The Future

For me the days of lossy compression are not over. That would be naive. Still, I will try to listen more and more to the best possible source at hand (as hard drive space is not really and issue any more). I already encoded my mp3s 1 or 2 steps “higher” than the “transparent” setting is for me (e.g. in case you hear no difference in -V 3 compared to -V 4 go for -V 2). I enabled high quality streaming in Spotify (i.e. ogg Vorbis q9 according to Spotify). Right now I plan to build an audiophile music player (Raspberry Pi, Volumio, DAC, Reclocker) and re-encode my favorite CDs using flac. Even though I might not actually hear any differences if will just make me feel better listening to it. A placebo? Maybe. It’s one I will take. Happy listening!

Tags: , , , , , , ,