Encoding Theories
Nov 18, 2014 at 9:01 PM Thread Starter Post #1 of 34

DecentLevi

Headphoneus Supremus
Joined
Sep 13, 2014
Posts
4,284
Likes
2,006
Location
Seattle area
Hello, as a sound design artist and and audiophile, I thought I would share my theories on audio encoding. These are only theories and everybody may have different views / interpretations, so please remain considerate.
 
I have spent many hours over the course of at least the last 6 years in researching/comparing the difference between lossy and lossless audio encoding (MP3, AAC & OGG vs. FLAC / WAV) and their different bitrates. You really need to trust me on this: Lossless is the way to go! The higher the bitrate, the better. The general concept is that the more an audio file is encoded (converted) or re-encoded, especially if into a lower bitrate, the more degraded the dynamics, texture and the soundstage imaging become. So then you have an audio file that has new 'artificts', the bass loses it's puncyness and the treble loses it's texture. Along the way I've experimented with various combinations of VST mastering filters at an attempt to restore the audio, but with mixed results. So basically when it's downconverted too far then the copy is ruined for good.
 
I have trained my ears to listen to tiny nuances and have been able to tell the difference between MP3 / AAC / OGG / FLAC and WAV. Yes, WAV is actually better than FLAC by a little. FLAC is so-called lossless, yet the bitrate is reduced by an everage of 500kbps from WAV and I could hear some of the stereo-width and fine treble-texture that was missing in the FLAC. Also I have been able to tell the difference between varying bitrates within each codec, FLAC included, can sound better/worse than each other. However the exception would be if the master track was originally encoded into a FLAC file, thereby bypassing any secondary encoding; especially if the FLAC file is encoded in a format higher than 16/44.1 from an original source of equal or greater bit depth and sample rate. 

Furthermore, CD quality is NOT the best format (16bit bitdepth, 44.1khz sample rate, 1411kbps bitrate); it's just the defacto 'industry standard' that was decided upon by media executives back in the 1980's when they invented with the compact disc. The best format I've heard so far, as far as PCM goes (WAV) is 32bit / 96khz. That is to say, that the 32bit / 192khz seemed indistinguishable, and I have yet to try the DSD (SACD) format (am I missing anything?). In the recent year or two, I've dropped lossy files unless totally exhausting my resources to find a lossless format of the song. The only exception would be if an album was mastered poorly (Burial, Flying Lotus and older U-ziq for example), then in that case a lossy encoding can be OK since it's hard to distinguish between the source that was mastered with it's has poor dynamics, too much distortion used, etc.

Lastly, I've discovered that, contrary to common belief, that "upconversion" can seem to improve the sound quality. Take a source track at CD-quality 16bit / 44.1khz, and play it through a player such as AIMP3 and upconvert it to at least 24bit / 88.1khz and with a hi-end setup you should be able to hear added fidelity.
 
Nov 18, 2014 at 11:14 PM Post #2 of 34
Hi DecentLevi,
 
Please provide ABX test scores to back up any of your numerous claims, as all of them go against current test results and/or the basic principles behind digital audio. Or at the very least some explanations as to why these things are true, so we can better explain why you're incorrect.
 
Thanks, and welcome to Sound Science!
 
Nov 19, 2014 at 12:34 AM Post #3 of 34
Too much wrong in that to be able to answer it all. I wish we would limit things to one wrong theory at a time. It would make it easier to answer. Line by line is hard to read.
 
Nov 19, 2014 at 1:48 AM Post #4 of 34
Well for one thing, you are wrong about FLAC.  It doesn't have a reduced bit rate.  What gave you this idea?
 
I have taken songs, sent them over a SPDIF connection, recorded the received bits at the other end, and compared the received bits between WAV and FLAC.  They are totally bit for bit identical.  With them being bit for bit identical as they are played back you won't get a difference in the sound coming out of your speakers.
 
This also shows your idea about encoding of FLAC or secondary encoding to be wrong.  It would not matter if the master was initially encoded to FLAC, or encoded later or for that matter if a WAV master was FLAC encoded, decoded, re-encoded however many times.  You will get the same bits exactly either way. 
 
As for the higher the bit rate the better, and 32 bit being best, care to explain what each of those add that is audible?  What do you imagine 96 khz does better than 48 khz?
 
Can you explain what upconversion 44/16 to 88/24 does to make the sound improve?
 
Nov 19, 2014 at 3:22 AM Post #6 of 34
   
Do you have ANY idea how hard it is to echo-locate a vole in the pitch black of night, when limited to 22khz? 


Well, humans can do timing between ears of around 10 millionths of a second. 
 
On the other hand, 44/16 can do only about 40,000 times smaller than that.  SO..................................................maybe we didn't catch the vole until we had torches to see it??????
 
Nov 19, 2014 at 3:30 AM Post #7 of 34
   
Do you have ANY idea how hard it is to echo-locate a vole in the pitch black of night, when limited to 22khz? 


Do humans have echo location powers like bats now! Hooray! I'm going to go eat me some dragonflies tonight!
 
Nov 19, 2014 at 7:00 AM Post #10 of 34
  You really need to trust me on this

 
No ........ we don't.
 
I really don't know if your post was an attempt at satire or not - but why not run some of this stuff past a real expert like Ethan Winer.  I'm sure he'll get a giggle out of it.
 
Nov 19, 2014 at 9:27 AM Post #11 of 34
   
Do you have ANY idea how hard it is to echo-locate a vole in the pitch black of night, when limited to 22khz? 

 
If it's pitch black you are more likely to be eaten by a grue.
 
Nov 19, 2014 at 1:11 PM Post #12 of 34
  Well for one thing, you are wrong about FLAC.  It doesn't have a reduced bit rate.  What gave you this idea?

Well, it kind of does have a reduced bit rate, just like a .zip file has a reduced file size. Of course, that doesn't have any impact on the contained information, and when decompressed, both a .flac and a .zip are identical (bit for bit) to the original.
 
Nov 19, 2014 at 11:39 PM Post #13 of 34
In response to the first 2 people who have replied to this post: Perhaps you are putting too much weight on purely science & math. This is a discussion on encoding theories with audio. In the realm of audio, perception is highly subjective in that different people can hear the same thing differently; hence why we all tend to prefer a different headphone and/or amp. As these are theories, this is not based on black & white data I am speaking of, but rather based upon that large grey area between black and white - a realm filled with infinite shades of grey. The realm of theories deals with the unknown - just as how many paranormal and supernatural things can't be proven by mere science alone, yet many of them are quite substantiated by actual witnesses, victims, etc. Just as how the ancient scientist claimed the earth is round, you should test some of these theories on your own before dismissing them due to so-called "popular belief" (case & point: there are many current day examples of where everybody in a society are quite literally are wrong).

And on a side note I don't know what ABX score is, but these are just my personal observations. Furthermore, instead of just doing a broad rejection of an article containing many points, it's more effective to specify which points you disagree with, and why. BTW another head-fi friend of mine, who I must admit is at a scientific knowledge level on audiophilism in general compared to my theoretical knowledge level, had read this same article and had agreed with my points.

Moving on to my concept about FLAC files, I may have misspoke. What I was trying to say is that, if you take a source WAV file at 1411Kbps  and downconvert it to FLAC at it's average of about 900Kbps, the audio loses some of it's fidelity. Upon testing many songs over a period of time between these 2 formats, (playing a section of a song on a loop, listening for tiny nuances through an AMP with good headphone brands such as Beyerdynamic, Shure, etc.), I was able to hear tiny nuance differences, mostly like the fine oscillating crackle of a synth, the punchyness of a snare within a complex mix, and an occasional stereo image of a bassline that would sound inferior on the FLAC conversion. While I am not invalidating the FLAC format, and I do use it myself - I am only saying that generally, this is a good format but if you scrutinize it closely enough, you can hear nuance differences.

In hindsight, this may not have been an error with the FLAC format itself, but because of the conversion; whereas the 500-ish Kbps were lost from the 1411Kbps. So to justify the exception to what I had initially claimed to be an error with FLAC files, my theory goes something like this: While downsampling the bitrate will cause reduction in fidelity (whether or not the conversion is to the same or another format), FLAC does have the potential to sound just as good as WAV: under the condition that the original master track has been encoded directly to FLAC, or if the FLAC has been encoded from a source track that had a greater bit depth / sample rate.

As for the questions about sample rate / bit depth: I can't scientifically explain why 96Khz would sound better than 48Khz, but with my many tests between the two, my ears always tell me that with the 96Khz (88.1 as well), everything sounded a touch more vivid and lifelike, as if more fluidic with a broader depth of the sound image. I would note however, that the above mentioned were all done with at least 88.1khz setting on the output on my soundcard, otherwise I could not hear the difference. After all, there is a reason blu-ray uses 96khz instead of the 44.1khz of CDs (as an example).

Next, for the question "Can you explain what upconversion 44/16 to 88/24 does to make the sound improve?", I would base the answer upon the above mentioned paragraph as well. But an odd thing was when using certain audio players I was able unable to hear the difference in upconversion, while I was able to hear it on others. Perhaps this is just a rare phenomenon while upconverting when playing tracks through the "AIMP3 player" specifically, as that was the audio software player I am basing that on. The program is free if you want to try it yourself, and contrary to the name it's not just for MP3's. I do think the El_Doug was agreeing with me on upconversion with his analogy; greater resolution would increase the depth perception or soundstage so that you can 'find the vole'.

As for the other points I had mentioned:
* Lossless sounds better than lossy formats (FLAC / WAV versus MP3 / AAC)
* Downconverting to a lower bitrate will reduce the dynamics, texture and soundstage imaging
* CD quality is NOT the best format
* If an album was mastered poorly, a lossy encoding can be OK

 
... is it the case that nobody agrees with these either?
 
On a closing note, I do not profess to be at a scientific knowledge level with audio engineering, yet these are only my working-theories.
 
Nov 19, 2014 at 11:43 PM Post #14 of 34
I have a question for you DecentLevi. I hope you can answer this, because you aren't the first one to come in here and do this, and I am very curious why someone would...
 
What made you come into a group called Sound Science and post a whole bunch of stuff that isn't at all scientific?
 
Are you just trying to get a reaction? Do you not understand what Sound Science means? Are you in love with your own words and aren't interested in learning anything from anyone else?
 
This is a sincere question. I hope you can explain it. For the life of me, I can't figure out what the point is for your post.
 
Nov 20, 2014 at 12:24 AM Post #15 of 34
Hello, to the previous poster
 
My post actually ended up in the Sound Science thread because it was originally posted on a thread for an upcoming head-fi meet, in response to a related comment from a fellow member who does mastering, but that thread starter had moved my posting to this new thread for me. I'm not trying to gain recognition or anything, but to just put a theory or two out there to see if anybody can back me on it, and to learn from some new views; like how I already had done about FLAC files. And I'm sure that's why we're all here - to learn from each other.
 

Users who are viewing this thread

Back
Top