128 Kbps (CBR vs VBR) vs. higher audio bitrates

[herbstech]

Quote:

Originally Posted by bigBonehead

Why do your recordings sound like ass? - Page 13 - Cockos Confederated Forums

4th post on down... WMA is, IMO, the worst lossless format... And WMP the worst player. If you want compressed lossless files I recommend you try .flac

Worst lossless format in what way? How can any lossless format be worse than any other when mathematically they can all reproduce the EXACT same original bitstream (assuming they all started out with the same input and didn't have any noncorrectable read errors)? You could also just take a lossless WAV file and zip it. It seems to me the only variable is the amount of compression but after decompressing, they should all be identical quality which is 100% of the original CD quality. This is assuming the original music source doesn't exceed that of the encoding format such as too many tracks or channels.

Secondly, why is WMP the "worst player"? What do you mean by that? I can hear a difference using lossless WMA on WMP and 128 Kbps MP3 on WMP. I am using whatever version of WMP came with Vista Home Premium which is what I have on my Toshiba laptop computer. I have all tone enhancement settings bypassed such as EQ, volume leveling... The processor seems to be only using about 20% of it's maximum capability when playing back a lossless WMA but I'll check that soon to verify.

So please clarify what you mean by "worst" in both cases detailed above.

[bigBonehead]

All compression formats use different algorithms. If you spend any time on sites dedicated to Recording and Mastering, they all support the idea that if compression must be done, .flac is the best. Studios use uncompressed .wav at 24bit/44.1Khrz or 24bit/48Khz These files are as pristine as is needed to Master a CD. As far as WMP, download Wavosaur load an identical file into both and do an A/B comparison... These are not Opinions off the top of my head...

BTW.... much of this depends on what equipment you are using to listen to the files... A good pair of nearfield monitors will give faithful response.

[herbstech]

Quote:

Originally Posted by bigBonehead

All compression formats use different algorithms. If you spend any time on sites dedicated to Recording and Mastering, they all support the idea that if compression must be done, .flac is the best. Studios use uncompressed .wav at 24bit/44.1Khrz or 24bit/48Khz These files are as pristine as is needed to Master a CD. As far as WMP, download Wavosaur load an identical file into both and do an A/B comparison... These are not Opinions off the top of my head...

BTW.... much of this depends on what equipment you are using to listen to the files... A good pair of nearfield monitors will give faithful response.

Lossless compression, no matter what algorithms are used, are all mathematically equivalent because by definition, they can all reconstruct the original sequence of bits with no loss. Therefore, algorithm A vs. B vs. C doesn't matter but what does matter is they are all lossless thus equivalent as far as being able to reconstruct the original bits. Some may be faster than others which would make them different.

I did download Wavosaur so I will try it later tonight.

F.Y.I. I have a pair of Optimus LV-20 headphones I think made by Koss. They were inexpensive at maybe $40 like 15 years ago but I can definately hear flaws in the music with them rather easily.

[bigBonehead]

Edited 9/16/09


Btw.... all of this is very subjective. Each listener will hear different things... Especially because of the different listening environment, and the tools that are used to reproduce the sound. Oh yeah.... define loss mathematically

[herbstech]

I've given a short "definition" twice now but if you would like a better one, try this:

Lossless data compression - Wikipedia, the free encyclopedia

Here is another good one regarding lossless WMA:

Windows Media Audio 9 Lossless - Wikipedia, the free encyclopedia

For lossless compression, I can use zip, flac, lossless WMA, legos, popsicle sticks... as long as from any of those, I can reconstruct the original bitstring perfectly and the encoded bitstring is at least 1 bit less than the original bitstring. A typical CD can be compressed to about half of it's storage size using flac or lossless WMA which is significant since it is equal quality to the original (assuming you dont exceed limits of the format such as # of channels), but about half the storage size.

You made another assumption that each listener will hear different things. I would NOT assume that. Two people may have effectively identical hearing (twins for example). Also, statistically, it is likely that 2 people (out of billions) have hearing that is effectively identical. For example, if you gave an A/B audio test to hundreds of people using dozens of audio clips, 2 or more of those people might get the same exact test results, indicating that their hearing/perception is effectively identical (meaning any differences between their hearing is so minor, it can be ignored).

BTW, lossless WMA is a very convenient format for me because it is included on my Vista laptop. The sound is noticeably better than say 128 Kbps CBR MP3. Also, iTunes will accept lossless WMA as input and automatically convert to 256 Kbps VBR AAC which my 1GB iPod Shuffle will accept. I am evaluating if I can hear any difference between that and lossless WMA using the same headphones.

[bigBonehead]

This site is in French, but pages 1/2 show a graph of Flac and WMA

Comparatif des formats audio FLAC - APE - MP3 - MP3PRO - MPC - OGG - WMA

I think it shows what I was talking about. At any rate, to each his own...

[herbstech]

Dude, what are you talking about? None of those graphs show lossless WMA. You stated previous that in your opinion, lossless WMA was the worst of the lossless formats. I still dont see anything that backs that up and I totally disagree. There is no "best" lossless format because for "simple" tasks such as a 2 channel (stereo) CD music using 44.1K 16 bit samples per second, they are all identical as far as reproducing the original bitstring 100% accurately.

[herbstech]

Quote:

Originally Posted by bigBonehead

BTW.... Windows Media Player is not a good comparison tool. It colours the sound. If you want a free player that is as clean as I've found in years, try this: Wavosaur

It appears that the software you recommended doesn't handle lossless WMAs so I cannot do an A/B comparison with Wavosaur and WMP.

You still haven't stated how WMP colors the sound with proof to back it up.

[JohnE.]

WMA lossless got a bad rap (firstly in the UseNet lossless groups and then it spread to other audio forums) in it's early iterations due to the fact that it wasn't truly lossless.

When WMA lossless first came out the decoded files could not pass an MD5 checksum test. Some speculated that the encoder modified bits in the wave header before passing it through the encoder, but afaik this was never substantiated. I never noticed any difference in sound quality myself, but the fact it couldn't pass a simple checksum test made many people suspicious and they shunned it and Apple lossless, which also suffered from this problem, like the plague. Perhaps this fact has gotten garbled over the years and become a general perception that WMA lossless "sounds bad" when it was really a matter of "encodes/decodes bad" back in the day?

I don't know if Microsoft or Apple have corrected this issue, as I haven't tested it since the early days. FLAC and Monkey's Audio have no problem with an MD5 test and suit my needs well enough that I haven't looked at WMA or Apple lossless in a long time. Now that I have a Mac I may look into Apple lossless again.


"The assumption part is that "most people" do not realize this. How would you or anyone else know what "most people" know regarding this? Did someone take a survey and ask them?"

You're right, that part is an assumption, but I agree with the author and believe it to be essentially correct. Many discussions about this very topic in alt.binaries.mp3 territory and on several audio and bittorrent forums over the years has made clear to me that the majority (that care about these things anyway) still believe 320 CBR to be the best quality choice for encoding mp3's, when the facts say otherwise. There have been several surveys (and subsequent discussion) on many of the forums I frequent and 320 CBR has always won the popularity contest, so far. 320 CBR was the best option at one time, before LAME optimized it's VBR algorithm but, just like I found FLAC and stuck with it, seemingly, so too have the 320 CBR folks.

[herbstech]

Well when you say best you need to define your term. Best in what sense? Any lossy compression format is a tradeoff. One thing really bad about lossy compression is that someone may then take that MP3 (let's say it is a true 320 Kbps MP3), and compress it again, maybe down to 128 Kbps. I doubt that file would be the same as ripping the same exact original recording from CD quality directly down to 128 Kbps because of the intermediate step. Kinda like lightly compressing a BMP file to JPG then more aggressively compressing that JPG to a much smaller JPG. You may have "double artifacting". With memory being so cheap these days, the only real reason to even consider 128 Kbps would be for uploading and downloading (especially on dialup).

Here is another thing to consider. Do they have a music format that is a fixed bitrate (such as 1024 Kbps CBR) but is encoded in such a way so that every other bit can be skipped for a bitrate of 512 Kbps? It could even be set up so 3 of every 4 bits can be skipped for a bitrate of 256 Kbps. That is, interleave the bits so that throwing some way as described would result in a valid music file, just less accurate. That way, only one original file encoding would be needed but several different bitrates could be extracted directly from it.

An example would be suppose someone posts a very high quality lossless 1024 Kbps music file of a song they made from their "garage" band. Someone with a slow internet connection may opt to download the lowest bitrate possible (say 64 Kbps) to sample it first. Then if they like it, they can redownload a higher bitrate. An even more "slick" thing would be to allow the 2nd download to "fill in the gaps" since part of the "1024" file is actually already downloaded.

So to be clear, I would like to see a music format that starts at say 1024 Kbps (1 Mbps), but layed out so that someone can download it in the following bitrates from that same initial file just by skipping some bits.

Kbps: 512, 256, 128, 64

Maybe there is a format out there now like this that I just dont know about. This whole idea about having to "rerip" files to different bitrates is wasteful. I rip to lossless so I always have "perfect" quality (assuming no errors while ripping). From there, I can compress down to 320 Kbps AAC VBR for my iPod. I have a lot of trouble telling the difference between the two but that doesn't mean they are equal. Someone else with better ears and/or better brain and/or better equipment might be able to tell.

With my "new" proposal, only 1 initial "rip" (to 1024 CBR) would be needed. From there, special software could just "skip" bits to make a smaller but valid music file. Just like they have progressive JPG, they should have progressive audio files too.

Maybe someone who reads this will design a format like I am describing and make people's lives better because of it.