MP3 encoding, how lossy is it?

[jayshah]

I hope this is technical enough.

Ok, lets say you have a WAV file, We'll say it's at 100% quality. You encode it to MP3 (192k CBR, lets say). Now it's at say 90% quality (I'm just making these numbers up). If you convert this MP3 to WAV it will still be 90% quality. Now, what happens when you convert this new WAV back to 192k CBR MP3? Will it stay at 90% quality, or will it degrade further? If it degrades, how many times can you repeat this process before it it stabalizes?

 

[blahblah99]

i dont think the wav file degrades the second time around...

if i recall correctly, mp3 compression cuts out the frequencies that aren't audible to humans and just "compresses" the rest... so if you encode a wav file the second time around, it won't have anything to cut out, and will just compress the data as in the first time you encode it.


if you really want to test it out, just get a sample wave file, encode it to mp3, decode it to a different file name, and finally do a file compare. Then encode/decode it again to third filename and do another file compare. Post your results here

 

[tomcat]

I'm pretty sure it will degrade with each succession. The wave file you started with the first time is not the same wav file you have the second time, so the mp3 algorithm will choose to throw out different data to meet the 192kbit bitrate.

 

[SsZERO]

I'll second what tomcat said. You can get a better idea of what will happen with repetitive encoding by taking a bitmap image, and saving it as a JPEG. Then take the JPEG you saved and save it as a JPEG again, using the same program with the same settings. Eventually the picture will begin to look blocky and a lot less sharper.

-= SsZERO =-

 

[AnthraX101]

I dissagree. If I recall corectly, the lossy portions of MP3 compression involve specific frq's. The rest of the compression is lossless. However, the file size will probably be different due to the conversion, adding small errors that are indetectable.

The thing about JPEG compression programs is that they usualy have a specific ammount they are allowed to cut out, usualy about 89-90% of it's orig. quality. This means that each time you recompress it, it will go down by about 10-20% in quality. You can set the program (if it is a good one) to your %'s.

Armani

 

[spidey07]

Sure you'll continue to degrade the true WAV file if you continuelly run it through a compression/decompression that deliberately throws away information. Each time you compress it with MP3 you lose information. That is why they sound so crappy on the first compression, I'd hate to hear a decompress/recompress.

hope this helps.

<edit> man I can't spell

 

[SsZERO]

You can also control the extent of MP3 encoding by raising or lowering the bitrate. This is the same as adjusting the level of JPEG compressing. If you compress a JPEG at 40%, that means that 60% of your image has been chucked away by the JPEG algorithm. When you decode something that has been encoded, the decoder reconstructs the original file based on the data that is present (which is less than the original source). If you have a 40% JPG open in Photoshop, then it is contained in the memory as a bitmap with no encoding. If you save it again as a 40% JPEG, you are throwing away another %60 of the picture's data...

The reason I mentioned JPEG was to make it easier to see the loss of quality over repetitive encodings. Of course, you can experiment for yourself and see what happens. If you repetitively decode a wave from MP3, then re-encode it, the MP3 will begin to loose its stereo separation (if it was a stereo encoding), additionally, the highs and lows will begin to sound really distorted. The highs will get that swooshing effect and the lows will have a crackling sound, like when you play too much bass through crappy speakers.

-= SsZERO =-



<< I dissagree. If I recall corectly, the lossy portions of MP3 compression involve specific frq's. The rest of the compression is lossless. However, the file size will probably be different due to the conversion, adding small errors that are indetectable.

The thing about JPEG compression programs is that they usualy have a specific ammount they are allowed to cut out, usualy about 89-90% of it's orig. quality. This means that each time you recompress it, it will go down by about 10-20% in quality. You can set the program (if it is a good one) to your %'s.

Armani >>

 

[Becks2k]

"That is why they sound so crappy on the first compression" = retarded statement

 

[spidey07]

???

Well they do sound crappy. I was just listening to a MP3 cd on my stereo last night. Sounds crappy. compressed soundstage, loss of dynamic range, just crappy.

If MP3 is such a great format then why hasn't the recording industry picked up on it? Because you LOSE information, which is a nono in a professional recording. DAT has ruled for quite a while now.

 

[CTho9305]

what was the bitrate / sample rate of the MP3??? on good speakers, up to 128kbps is easily distinguishable - but around 160-192kbps, MP3s are pretty much indistinguishable from the oringinal CD (IMHO, of course)

 

[spidey07]

bitrate was 192 CBR on one album, 192 VBR on the other. This is DVD player hooked up to a digital receiver. I think the encoder was LAME. Guess I'm kinda picky, i've also heard CDs that sound pretty bad.

<edit> I've also done a lot of tests by decoding the MP3 into a WAV and making an audio CD. Same effects.

 

[SsZERO]

The decoders built into DVD players still kinda suck, so I wouldn't be using that as a basis for comparison. If you properly encode a MP3, using a high quality encoder like LAME and VBR with the correct settings, then you will barely be able to tell the difference between the original and the MP3 version. Of course, you need to be using a good DECODER as well, and that is something not present in today's crop of DVD players. Try connecting your computer to the receiver and download MAD MP3 decoder for Winamp...MAD is better than Nitrane at decoding MP3s.

-= SsZERO =-



<< bitrate was 192 CBR on one album, 192 VBR on the other. This is DVD player hooked up to a digital receiver. I think the encoder was LAME. Guess I'm kinda picky, i've also heard CDs that sound pretty bad.

<edit> I've also done a lot of tests by decoding the MP3 into a WAV and making an audio CD. Same effects. >>

 

[spidey07]

Thanks for the tips ssZero. YGPM - don't want to hijack this thread with my picky ears. I figured the decoder in the panasonic RP56 dvd player would be OK at best. But i've also used WAVs on a CD in my $1000 adcom CD player.

 

[SsZERO]

Well, suffice it to say that the decoder in most DVD players is about the same as 1.xx versions of winamp...not quite up there.

I've never understood why people would want a $1000 CD player, especially if it has digital playback capabilities...my DVD player is connected via the coaxial digital cable...so it streams the digital data from the CD to my receiver. I'd say the quality you hear is a question of the quality of your receiver and speakers, not so much the CD player itself...now if you are using a CD player with an analog connection, where the CD player would have the actual DAC built into it, then I could understand how a more expensive model would sound better.

-= SsZERO =-



<< Thanks for the tips ssZero. YGPM - don't want to hijack this thread with my picky ears. I figured the decoder in the panasonic RP56 dvd player would be OK at best. But i've also used WAVs on a CD in my $1000 adcom CD player. >>

 

[spidey07]

ssZero,

Keep in mind that DACs very greatly in sound. Just because you feed a digital signal to your receiver does not necessarily mean it will sound better. For example, I've hooked up my adcom Cd player to my denon 4802 digitally and via analog. the DAC in the adcom is FAR and ABOVE superior in terms of music. We're talking night and day here. This is mated to two krell mono blocks and Thiel speakers.

So, it is for that reason (DAC) that people spend a lot of money on CD players or actually buy a transport with a separate DAC.

digital is digital, it is when you actually try to move air (sound) that it counts.

 

[SsZERO]

I'm not disagreeing with you about the quality of DACs...but i'm just saying that for me, spending $1000 on a CD player with a quality DAC would be a waste, I'd rather put that money toward a quality AV receiver and feed the receiver a digital signal...then the DAC will be handled by the receiver and not the CD player. If you connect that $1000 CD player via the analog RCA audio connectors, then the audio is already analog and all the receiver does is amplify it. You know what I mean?

Good thing you realize that digital is digital. Some people go off on wild tangents about how cable quality effects the digital signal quality...digital is all or nothing, unlike analog. So if you hear music, then the cable is fine.

-= SsZERO =-



<< ssZero,

Keep in mind that DACs very greatly in sound. Just because you feed a digital signal to your receiver does not necessarily mean it will sound better. For example, I've hooked up my adcom Cd player to my denon 4802 digitally and via analog. the DAC in the adcom is FAR and ABOVE superior in terms of music. We're talking night and day here. This is mated to two krell mono blocks and Thiel speakers.

So, it is for that reason (DAC) that people spend a lot of money on CD players or actually buy a transport with a separate DAC.

digital is digital, it is when you actually try to move air (sound) that it counts. >>

 

[spidey07]

hehe, I told myself not to hijack this thread.

But guess it does pertain to digital music and quality.

ssZERO,
I wanted to share my experiences with comparing the DACs in the ADCOM and DENON because I was very surprised by the results. I simply could not believe what I was hearing in terms of differences. But even my girlfriend could tell. I never would think a DAC could vary so much, but there it was smacking me in the face. Maybe you could try the same with your receiver? Yeah, lets not get started on jitter and bit delivery.

 

[SsZERO]

I know what you are talking about. I have heard it myself. When you convert between digital and analog, there is a interpolation involved...the higher quality DACs have better interpolation algorithms and probably operate at higher speeds so they can process more data, resulting in better sound. My AV receiver is a Kenwood VR-309, and I have it connect to 2 Sony 4-way floor tower speakers. While it is nowhere near top of the line, it is more than sufficient for my 12x11 room. My DVD player is a Toshiba SD-3750, and when I play music CDs back it sounds really good. The sonic imaging is really spectacular...your Adcom is more than likely better for music playback, so I couldn't make a comparison.

-= SsZERO =-



<< hehe, I told myself not to hijack this thread.

But guess it does pertain to digital music and quality.

ssZERO,
I wanted to share my experiences with comparing the DACs in the ADCOM and DENON because I was very surprised by the results. I simply could not believe what I was hearing in terms of differences. But even my girlfriend could tell. I never would think a DAC could vary so much, but there it was smacking me in the face. Maybe you could try the same with your receiver? Yeah, lets not get started on jitter and bit delivery. >>

 

[mithrandir2001]

Uh, oh. Audiophiles. I fell into the subjectivist Stereophile/TAS crowd when was new into audio, but now that I am older and wiser I am very cynical about any comparisons that are not performed double-blind. I remember buying all of those Audio Alchemy DACs and jitter boxes and fancy digital cables, but came to the conclusion that the differences I thought I heard were in fact created by expectations of improvement, not necessarily tangible improvements. No more will I dump cash into fancy and expensive products that allege to improve the sound. It simply comes down to speaker choice, speaker placement and room effects (and the amp if you have speakers that present a tough load).

MP3 has come a good way. Most 128kbps CBR files still sound horrid from an archival perspective, but invoke some LAME VBR action and you'll realize they are getting pretty darn close to transparency. If an MP3 I encode has a subtle artifact or two, I don't fret because the whole purpose of MP3 is portability and ease of transfer. If CD is a 10 and cassette is a 1, I can make MP3s that sound like a 9.5, if not a 9.9. If you can get that close at a 10:1 compression ratio, I'll take it! The closeness to the original has become quite uncanny if you use a great encoder like LAME and a fine decoder like WinAMP 2.7x. I have been pleasantly surprised by how LAME not only takes care of the frequency/distortion issues, but also important attributes like imaging, soundstaging, transients and hall ambiance. MP3 is a flawed format by its design (wastes too much space for high-frequency encoding), so it can only get better with newer lossy formats.

But to the original poster, reencoding and redecoding over and over again will degrade the sound file because the MP3 format is lossy.

 

[Special K]

I took a 45MB wav file that I ripped from a CD and encoded it into a CBR 128kbps mp3 file, whose size was about 4MB. I then took that MP3 file and converted it back into wav. The size of the wav was about the same as the original. When I convert a MP3 back to wav, the quality of the wav file is the same as that of the MP3, correct? I mean, the MP3->WAV convertor has no way of knowing what data was removed from the original wav file, right? If that is true, then why is the file size of the new wav identical to the old one, even though it contains the same data as the compressed MP3? What is the point of all the "wasted" space?

 

[mithrandir2001]

<< If that is true, then why is the file size of the new wav identical to the old one, even though it contains the same data as the compressed MP3? What is the point of all the "wasted" space? >>


Because that WAV file uses 176KB for each second of 16-bit 44.1KHz audio, regardless of quality. Two WAVs files that each contain 5 minutes of music - even if they are different songs - will be the same length. That's just the format. If you create a wallpaper bitmap that is solid black, it will still take up a ton of room. WAV is not compressed.

 

[Roguetech]

<< The reason I mentioned JPEG was to make it easier to see the loss of quality over repetitive encodings. >>

I thought JPG and MP3's were the same technology? Hence JPEG and MPEG?

 

[Colt45]

<< << The reason I mentioned JPEG was to make it easier to see the loss of quality over repetitive encodings. >>

I thought JPG and MP3's were the same technology? Hence JPEG and MPEG? >>



yeah.. i thought so too.

besides we all know that vinyl sounds better than mp3's or cd's

 

[SsZERO]

MPEG = Motion Picture Experts Group
JPEG = Joint Photographic Experts Group

Basically, MPEG and JPEG are acronyms for groups of people who develop encoding algorithms and establish spec standards...like specialized versions of the ISO (International Standards Organization). The algorithms used in JPEG and MPEG Audio Layer is similar, but not the same...the JPEG algorithm creates a specialized palette and removes pixel data for pixels that are either not used at all, or are not easily perceivable by human vision...for example, if you have blue 132 with blue 140 in some small areas, it would remove some or all of the blue 140. If you have a 24-bit bitmap image in 800x600, that means you have 480,000 pixels, each requiring 3 bytes of memory (1.5 megabytes total). Most pictures do not have 16.7 million individual colors...so the JPEG algorithm can safely remove a large portion of pixel data without reducing the image quality, it then proceeds to remove the lesser amount of similar colors.

MPEG encoding of audio uses quantization, meaning it breaks the wave file apart into "frames", then each frame is passed through the algorithm. This algorithm first removes data for sound that is not audible to human ears at all, it then removes data for sounds that are not audible due to another sound being louder (example, if a song has a string section for ambience, you will not hear the string section when a bass drum is struck, so the algorithm would remove the data for the string section for each occurence of the bass drum). The extent to which the algorithm removes data is determined by the bitrate...so if you are encoding at 128kbps, then the algorithm will remove data from the frame until it is in spec with 128 kilobits.

Bonus: This is why CBR encoding sucks...you are applying a "once size fits all" kind of algorithm to the sound. Say a song starts out with just cymbals and a bass drum...you don't need 128 kbps to store data for those sounds...with VBR encoding, the bitrate will adjust to accomodate the requirements of the song. After the song gets going, you might have bass and vocals...along with other ambient sounds. This will require more than 128 kbps to sound good. Remember, the algorithm can adjust bitrate for each frame it processes, so you make the most efficient AND best performing use of MPEG Layer 3 by using a quality VBR encoder, such as LAME.


-= SsZERO =-



<< I thought JPG and MP3's were the same technology? Hence JPEG and MPEG? >>