Ever since I've been collecting music on my computer (years now) I've aimed to keep it at 128kbps or better, and never been too fazed about higher qualities like 192kbps, 320kbps, etc. I'm also never noticed any difference because I listen to it mostly through crappy computer speakers, or headphones. I'm wondering when I get a car and a sound system, if I then play my 128kbps mp3's through the system and turn it up loud will I then start to notice a loss of quality?

I've never thought about this before as it only recently occurred to me, so I'm curious to what quality I should have my music as. Like if I had it as 320kbps, would there be any noticeable difference from having it as 192kbps? Anyone else noticed a difference and had to upgrade their music quality? Please discuss anything else relevant to this issue that I may be overlooking. Thanks.

this has been discussed again and again (try searching). yes there is is a diffrent between 128 and 320. there is alot more quailty in original cd's compared to mp3 aswell

yes, it will sound worse.

that being said, sometimes the cds sound terrible too.

REally? That's so bad.

it also depends on how good ur car stereo is. it may not actually be good enough for u to discern the difference in quality.
but generally on an average/decent stereo you will notice quality loss with 128kbps mp3 tracks.
most of my music is at 192kbps or higher, and i still notice a slight loss of quality to the original recording (cd)
going up to 320kbps, in my opinion, u need to have two things - 1. a very good stereo, and 2. very good ears to be able to tell the difference in quality. the difference is definitely there, but whether you will notice it or not, i probably doubt

Ahh, the good old days when 128Kbps was great and ripping took hours!

Funny how we used to be happy with 12-15 tracks on a regular CD and now 700 mb of mp3 just isn't enough!

What i want to know - is there an additional loss of data when a song is compressed to mp3 then uncompressed to WAV then recompressed to mp3 (assuming the same bit rate, encoder etc)?

yes there is....the data that is lost during the initial compression to mp3 cannot be recovered
there is absolutely no quality GAIN when converting an mp3 to a wav file

Might want to read the question again champ

The answer is yes.. kinda...

Theoretically, if you get an MP3 file and convert it back into a wav file, it will only retain the original encryption of the mp3 data, and kinda "fill in" the blanks so to speak, enabling it to be played on a conventional CD player.

If you then use this wav file and re-convert it with a different protocol, even the same bitrate MP3 (although most programs use lame now days), then there is a good chance that the file will be slightly different to the original MP3... Converting it into a WMA or a different bitrate protocol makes it even more prevelant.

Re-converting the MP3 using the exact protocol to the original MP3 will still create a different file, as the file being converted wasn't from the original CD, but from the WAV file re-converted from the MP3 of the original CD.

So therefore you are gradually losing more and more of the original data within the file by re-encoding it over and over. This will create more noise, and in turn be of less quality than the original MP3 file.

Eventually the MP3 file will reach a state which will have a minimal difference to the previous data, due to the algorithims being in an increasing simplex state (for lack of a better word).

This is why you can sometimes download a 128 bit file and it sounds reasonable, and then download another to find it sounds horrid... The reason the former sounds so bad is because it has probably been converted more than once already (maybe as a .zip, a wav, wma, MP3)....

Even lossless files can suffer this fate if they are converted to a different format too many times...

thanks ben

austin-towers, perhaps u should understand what goes on when compressing files

I didn't see the problem with your answer , so I basically elaborated on what you said . In the end I realized why you kept your answer short n simple... It can get pretty complicated to explain


Well I thought so anyways.......

To make it even simpler compare this first photo of a woman which is uncompressed.



The second photo shows her as she would look if she was compressed into a 128 bit MP3 file...

Thanks Ben, but i'm still not certain, unless you're absolutely sure.

My understanding is that an mp3 encoder discards data it thinks can't be heard by the human ear (as dictated by the algorithm and user configuration). It then compresses the remaining information to make a smaller file.

I guess what i'm thinking is that in the first encode the bulk of the 'unheard' data is lost. Then the wav decompression fills in the gaps as you say. But if it's then re-encoded, won't it just discard the same data that was lose in the first place?

BTW, it's not that i like to do this very much but occassionally i get a long (eg 1 hour) mp3 that i want to split and always wondered if there was an issue as the program i've used does it this way (maybe they all have to?).

Spyne - my question was about the encode-decode-encode cycle, i understand that there's no gain when going back to wav format.

bahahahhahahahahahahahahahahahahhhahahahhaha

thats so good, im gonna steal that and use it in another thread

bahahahaahhahaahhahhahhahahahhhahaha

I may have missed/forgotten a section, and therefore misinterpreting what you are asking about.

But on the encode-decode-encode/compress-uncompress-compress cycle; as you acknowledged, compressing/encoding to MP3, even at 320k still gets rid of some portions of the music. By decoding or uncompressing, how can you get the information that isn't there. The MP3 does NOT have the data, so then decoding/uncompressing can't bring the MP3 back to sound the same as the original file from the CD.

So it decoding and then encoding again, I would imagine the music will lose more sound, as it is compressing the file from an already compressed file. Sorta like a tightly wound spring, if you cut some off it, it can still fill the same amount of space, but it just has less of the actual spring there. You probably wouldn't notice if too much was gone, but if you cut a lot out, whilst it still has the same space taken then you will hit some problems.

Or perhaps like the fire sprinklers. Even with less water flowing through it looks the same, but it doesn't have the same effectiveness (thanks numb3rs )

Edit - Ben, those comparisons are bitchin

Thanks for the fast reply. Want to reiterate - i understand that there's no going back to the original quality! That's not what i'm asking about.

I just want to know definitely if there is a continual data loss (from the original mp3) following subsequent compression/decompression cycles.

At first glance, i agree with you Louie, but i'm till thinking that the algorithm will just keep trying to take out what it thinks can't be heard, and if there isn't any data that fits the bill then surely it won't take more out?

Ah yea, I get what you mean now. On that one, I have no clue. I don't know if it's just a simple cut at certain freq's or whatever for the compression. I guess if someone knows that, they should be able to answer that one. Would be interesting to know how it decides what to discard and how it compresses. But I do think that there is a continual loss, at least to a certain extent. I don't have solid proof, but it just seems to make sense that the more you encode, even after decoding, you will still be losing something, what you lose I don't know, but something still will be lost IMO

The conversion process of an MP3 involves filtering and also an elaborate elimination process to cut out redundant and less important information which is within a sample....

MP3's compression method makes use of the phenomenon called auditory masking together with some clever algorithims to help keep the information needed to re-create the file to a minimum. For more information on how the MP3 process achieves this read Psychoacoustics. .

For a simple example of how data can be lost or distorted check out the pic I made up below, illustrating the conversion process, showing the resultant wave created by the process of coding, and re-coding a "sample" of music:

* In the first waveform you can see that the MP3 file has certain points of data (green dots) that give the "road map" for the recreation of the sound sample. It's important to understand that the MP3 (player/coder/de-coder) program doesn't see the brown line joining the dots together... only the green dots, which help it re-form the sample.

* the second waveform shows the re-intergration of redundant data, so it can again become a WAV file for use in a CD player. Notice that the MP3 program simply provides the same sample numerous times (depicted by the green dots) to fill in the gaps, as it cannot see the slope of the waveform (depicted by the brown line) as it re-samples the data into the larger WAV file.

* the third waveform shows the re-sampling of the WAV file into the same format as the original MP3 file was.

* the last waveform shows the resultant sample which, even though it was sampled using the same bitrate as the original file, it has changed ever so slightly and therefore cannot be considered a perfect replica of the original MP3. This will continue until the MP3 process breaks the waveform down to a point which is simple enough to intergrate perfectly.... This waveform would look more like a bar graph than a music sample .

Of course... the MP3 process is actually much more involved and complex than this, and the MP3 process actually attempts to simulate the interlinking line to some extent, but it's not perfect by any means...

:edit:

Your right in thinking that there is only so much data that the MP3 process can discard before there is none left that is irrellevant, for eg: cutting out under 20Hz, and over 16KHz would only have to be performed once.
It's the sampling process that creates the most flaws, as well as the process of removing masked sounds... What may have been considered an important peice of data in the first sampling process, amy be considered less important on the next re-sampling process simply because when the waveform was resampled the data produced was slightly different.

No worries, that works for me!

So essentially, when i break up a long mp3 into shorter ones by a one-time decode/re-encode, it won't be a perfect copy but if the original bitrate/sampling rate is high enough then there shouldn't be too much damage done?

Nice one, thanks

Correct, it's exactly like compressing jpeg images over and over again, you get an accumulation of compression artifacts that eventually will be noticeable and ruin the picture (or sound in this case).

Ben said it well and as correct as I have read. Copies are copies and are never as good as the original!

There is however a limitation that has not been discussed in this thread that I found. Perhaps in others but to be honest I'm not up for an hour of searching so here goes.

The limitation I am talking about is the human ear. There is a threshold where the digital information surpasses our ability to hear. Not to mention the ability of the equipment we use to reproduce it.

The algorithims used to encode or remove redundant data are based on the abilities of the human ear. Some "masks" are better than others.
It has been my experience that useing old school software to encode my Cd's results in a better quality sample. Yes it takes longer, I have to apply all my own tags but in the end the SQ enjoyment is worth the effort.

Cheers

Wierd... I guess that the programs are trying to subtract more and more out of the original data, which wouldn't be noticed on an average setup, or say a n iPod through earphones (for example)... On a good setup you can probably notice the difference I suppose...

Hmmmmm, I'm gonna crack out an old mp3 convertor and compare it, both in how much it can compress as opposed to a new type, and how it sounds... Stay tuned

How the music is returned to WAV depends on the decoder used also. Some will use funky algorithms to try and 'put back' information which is not there - so corrupting the data further. A bit like oversampling in a DAC.
my motto on this is "If it's not there, it's not there. No point go over it 128x - it's still not going to be there."

btw, I stole those pics for another thread too

What if you convert it using lossless format. Is it really lossless or will there still be a loss of quality?