WmAx

WmAx

Audioholic Samurai
LDanix said:
You could be right. However, there are no tests that I have found that have determined that these codecs are truly "lossless" (if you can find one, please pass it on).
This can be confirmed by encoding a wave file, decoding it, and them comparing the two(original source vs. decoded) wave files. The decoded data will be, byte for byte, identical, when compared to the source file.

-Chris
 
L

LDanix

Enthusiast
Chris- Since I am only planning on using that particular hardware to listen to my music, I figure a simple listening test with that equipment will suffice. If I was going to listen to the music on several different systems, I would have listened to the samples on all the systems. And if I planned on using the best equipment known to man, then I would have been anal enough to look for inaudible differences between the files created and the original PCM data. Since my usage of these files is not going to be that universal or highly detailed there would be no reason to spend the time going through all that extra effort. I came to my decision based on my listening needs and expectations.

Referring to your post about lossless codecs- What application would you recommend for comparing two files at the bit-for-bit level? Do you believe that the differences would be audible?

What are those "codec buster" samples you refered to a few posts ago?

Based on your highly detailed questions, I am assuming that you have worked with digitally recorded audio at these minute levels for quite some time along with having highly trained ears and a vast understanding of the audio. What have your experiences with MP3, WMA, AAC and/or any other codecs been like? If you were recommending a codec and bitrate for a system with limited bandwidth (wireless media server) or limited space (iPod Suffle or other Flash-based MP3/WMA player) which would be your choice and why?
 
M

MDS

Audioholic Spartan
What application would you recommend for comparing two files at the bit-for-bit level? Do you believe that the differences would be audible?

diff.exe, comp.exe, windiff.exe - any program that can compare binary files (the fact that it is audio is irrelevant).

WmAx didn't go into why all lossless codecs will sound the same (if they are truly lossless - flac, monkey's audio, shorten, wma lossless, apple lossless, etc are all truly lossless), so I'll give you the background should you wish to research further.

I believe the reason for confusion among some is that many believe that a lossless codec is similar to a lossy codec (mp3, aac, wma, etc) but does a 'better' job of encoding so that it sounds indistinguishable from the original. That is not the case and is comparing apples to oranges.

Lossy codecs use a technique known as 'perceptual coding'; in layman's terms they analayze the signal and discard parts that a model of human hearing says we would not perceive and that is how a reduction of file size is achieved. For example, a soft sound (below the 'masking threshold') that follows a loud sound will not be heard because it is masked by the loud sound - thus it can be discarded. The encoded file is vastly different than the original, but for most purposes 'sounds the same' (debate to the contrary notwithstanding).

Lossless codecs in comparison do no such thing. They achieve a reduction in file size by data compression. This is analogous to the way zip achieves a reduction in file size, although the algorithm is vastly different. The simplest way for a lossless codec to reduce the file size would be to use the same type of algorithm that zip uses (based on Huffman Encoding) but that achieves very little compression because there isn't much redundancy in an audio file that could be exploited by huffman encoding.

As huffman encoding reduces file size by replacing sequences of byes ("symbols") with a shorter sequence, lossless codecs typically do the same type of thing but use a technique known as rice coding, which is itself a form of 'predictive coding'. In principle, it is similar - replace long sequences of identical bytes with a shorter sequence. Additionally, they try to exploit properties of the audio file such as identical information in both channels (no need to store it twice). In practice it is a bit more complicated, but that is the gist of it in layman's terms. Google for rice encoding or lossless encoding if you really want to learn more.

So, if a codec is truly lossles, upon decoding you will get back the exact bytes you started with. If they are exact, there is no way they can sound different than the original.

As for streaming via wireless, using a lossless codec will only help if the device to which you stream has a built-in decoder. Thus the amount of data being streamed is smaller and all the decoding work is done in the receiving device. If you were to use something like the Squeezebox, it does the decoding on the server and streams the uncompressed wave to the receiver - thus no gain whatsoever in terms of the amount of bandwidth required to stream to the receiver. The advantage is solely in reduced storage requirements on the server.
 
WmAx

WmAx

Audioholic Samurai
LDanix said:
Chris- Since I am only planning on using that particular hardware to listen to my music, I figure a simple listening test with that equipment will suffice.
Unfortunately, a true double blind test is not practical on such a device without a complicated routine. Actually, redundant devices would be needed, along with an ABX switching system, in order to accomplish a meaningful test in this manner. If you are more interested in possible differences caused by the hardware(not the file formats), then it makes more sense to play each file and record with a reference quality ADC into your computer. When you have recordings of two or more file formats from the hardware source device, you can proceed to match levels and synch the files in an audio editor. Next you can use a program such as WinABX, PCABX or Foobar(with optional ABX plug-in), and proceed to perform a fair double-blinded listening test.

Referring to your post about lossless codecs- What application would you recommend for comparing two files at the bit-for-bit level? Do you believe that the differences would be audible?
The prior user explained this issue(thanks Anonymous).

What are those "codec buster" samples you refered to a few posts ago?
Codec busters are sound samples that fool the perceptual coder used by a lossy codec. As such, the codec does not properly encode the sample and it has an audible difference. These codec buster samples are not represenative of normal music, though such phenomena may occur in real music(very rare).

Based on your highly detailed questions, I am assuming that you have worked with digitally recorded audio at these minute levels for quite some time along with having highly trained ears and a vast understanding of the audio. What have your experiences with MP3, WMA, AAC and/or any other codecs been like?
For saving space(portable audio applications or online data transmission), I find MP3 encoded with Lame 3.96 at 256kbps(CBR for maximum compatibility with hardware), j-stereo, default settings, is almost always transparent to my ears, per ABX tests. I generally use industry-standardized monitoring headphones(Sony MDR-7506) to carry out such tests. I do not generally use other codecs(AAC, Ogg Vorbis, etc.) due to compatibility issues.

If you were recommending a codec and bitrate for a system with limited bandwidth (wireless media server) or limited space (iPod Suffle or other Flash-based MP3/WMA player) which would be your choice and why?
MP3. Because of reasons stated above. If universal compatibility is not an issue, then you should investigate Wavepack, Ogg-Vorbis, etc.. Please refer to www.hyrdorgenaudio.org in order to research/investigate these formats.

-Chris
 

espresso

Audiophyte
Hi. I have recently downloaded wma9.1 codec and decided to try it against mp3 which I had been using.
I use easy cd da extractor to convert to mp3 and now I tried converting to wma. Bit rate was 192k (constant).
To my suprise I found that wma sounds better then mp3 at high freqencies. High tones are somehow more transparent and detailed unlike mp3.
I only have pc speakers for testing so I could not use them to test mid or low spectrum.
Then I used AudioGrabber to convert to mp3 (Lame 3.96) and same thing again when I did a compare.
The bottom line is that wma @192k sounds more like original then mp3 @192k when compared for high frequencies. The original in this case was 16-bit WAV.
 
Last edited by a moderator:
S

sploo

Full Audioholic
LDanix said:
What are those "codec buster" samples you refered to a few posts ago?
Casting my mind back many... many... years... I seem to remember reading an article comparing various lossy audio formats.

The part that I recall was that the start of Pearl Jam's "Daughter" on their Vs album was particularly difficult to encode.

Apparently this was to do with the two channels having very different contents (the intro has mainly acoustic guitar in one channel and mainly percussion in the other). I suppose this would probably upset joint stereo encoding.

How relevant that is today, well, I don't know.

Something that might be of interest (though I'd take quite a bit of it with a fair pinch of sodium): http://www.geocities.com/cfr707/mp3.htm

PS I listen to 192kbps MP3s all day at work. CDs at home. Horses for courses.
 
sleepysurf

sleepysurf

Junior Audioholic
MDS said:
If you were to use something like the Squeezebox, it does the decoding on the server and streams the uncompressed wave to the receiver - thus no gain whatsoever in terms of the amount of bandwidth required to stream to the receiver. The advantage is solely in reduced storage requirements on the server.
Actually, the Squeezebox2 does decode on the fly (at least for FLAC), thereby saving network bandwidth.
 
I

ilostmydrink

Audiophyte
espresso said:
Hi. I have recently downloaded wma9.1 codec and decided to try it against mp3 which I had been using.
I use easy cd da extractor to convert to mp3 and now I tried converting to wma. Bit rate was 192k (constant).
To my suprise I found that wma sounds better then mp3 at high freqencies. High tones are somehow more transparent and detailed unlike mp3.
I only have pc speakers for testing so I could not use them to test mid or low spectrum.
Then I used AudioGrabber to convert to mp3 (Lame 3.96) and same thing again when I did a compare.
The bottom line is that wma @192k sounds more like original then mp3 @192k when compared for high frequencies. The original in this case was 16-bit WAV.
This is indeed the case. I also use cd-da extractor 8.2 and have done various tests myself using different audio sources. For many years (I say many, I turned 21 today) I have been a self proclaimed audiophile, researching and testing out many ideas. I begin work at the acoustics center at my college this fall.

Using Pink Floyd's "Dark Side of the Moon Album", the Red Hot Chili Peppers "Californication" album, Jay-Z's, "The Black Album", Radiohead's "Kid A" album, and my favorite Gustov work, "The Planets", I ripped each track into 3 different audio formats: mp3, aac, and wma at 320 kbps at 44.1 khz. I tested them in my 3 different listening areas: my Sennheiser HD 201 (cheap, but good cans), my Bose Audio System (which BTW, Bose is NOT high end, contrary to what they want you to believe), and my Toyota Avalon (has a JBL 5.1 system in it with an Alpine DXZ-8210 [i can't remember if that's the exact #, but it's a year old] as the head unit).

MP3 (using the LAME 3.9 encoder) produces a very positive result, with warm mid's and accurate highs up to around 16khz. Frequencies below 50 Hz were non-existant. It is nice to have them though because my car's head unit will play the MP3 format.

Window's wma format was around the same quality as that of MP3. At lower bitrates, I know for a fact that WMA is more lifelike, but differences in quality are negligable at higher bitrates. The high frequencies sounded more tinny to me, but again, my car's head unit will play these.

On to AAC. I have read in the past where many people found this codec unacceptable, so I was blown away when I heard how well it replicated the origional sound. The almost all of the frequency spectrum seemed to be there, even the low frequencies that MP3 and WMA missed in the intro song to the "Dark Side of the Moon" album, which is supposed to be a driving heartbeat. Although the bass was there, the power of the origional cd was not present. All I can guess is that the lowest part of the spectrum is still trunked to save space.

IMHO, AAC is currently the best "lossy" format of the three considered in this thread. OGG is also pretty good, but most players will not accept the format. But it all depends on your listening preferences and how your ears interpret the signal produced. You've got to judge it yourself instead of what others tell you. As for the decision between file formats on the portable player, unless you're using a pair of SHURE E3C buds or higher, you're probably not going to be able to tell the difference between formats at higher bitrates. Like stated above, the only real difference I find is that MP3 and WMA do not reproduce the lower spectrum as well as the AAC format.

BTW, the only way to achieve true "lossless" when converting from an analog signal (the origional audio data) to a digital signal (not just 0's and 1's like most people assume) is to sample the origional data at twice the speed of the higest frequency. Humans are only capable of hearing 20 Hz - 20,000 Hz, but most cannot hear that. You can feel the vibration down at 20 Hz, but most can only hear from 25 Hz ~ 16/17 KHz. That is why you see the sampling rate at 44.1 KHz.

Lee
 
Last edited by a moderator:
S

Steve1000

Audioholic
Just another data point: http://www.hydrogenaudio.org/forums/index.php?showtopic=35741 :)


espresso said:
Hi. I have recently downloaded wma9.1 codec and decided to try it against mp3 which I had been using.
I use easy cd da extractor to convert to mp3 and now I tried converting to wma. Bit rate was 192k (constant).
To my suprise I found that wma sounds better then mp3 at high freqencies. High tones are somehow more transparent and detailed unlike mp3.
I only have pc speakers for testing so I could not use them to test mid or low spectrum.
Then I used AudioGrabber to convert to mp3 (Lame 3.96) and same thing again when I did a compare.
The bottom line is that wma @192k sounds more like original then mp3 @192k when compared for high frequencies. The original in this case was 16-bit WAV.
 
Last edited by a moderator:

Latest posts

newsletter

  • RBHsound.com
  • BlueJeansCable.com
  • SVS Sound Subwoofers
  • Experience the Martin Logan Montis
Top