The difference in sound in iTunes & DAW

[murrayhill]

Hey guys I have a question.

Have you tried this before? Try importing an mp3 from your iTunes library into Logic or whatever DAW I suppose and listen to the difference.
Technically speaking, what is that differnece coming from? Playback engine?

[Sirocco]

since itunes has the ability to eq tracks and other things, im pretty sure it is even affecting the signal when its not turned on, by a small amount

[Drumhum]

Coo! Never noticed that before. There seems to be quite a difference.

First thing to note is that logic converts the MP3 to a non compressed format (whatever you've set in prefs - like wav or aiff etc). Logic also uses a very different playback engine to itunes too. Logic playsback/mixes in a 32 bit floating point 'world'. Itunes has to convert (uncompress) the compressed format type of MP3 as it plays back. So in perspective, from an MP3 in itunes you are hearing one type of conversion process played back in one type of audio engine while in logic you hearing a different conversion in a different audio engine.

I know I'd "trust" logic's sound more than itunes but this is an interesting observation. I too would like to know some specifics regarding what is going on here!

Sorry Sirocco, but with the greatest respect, your theory does not hold up for me. Simply adding some computer code to provide some feature such as a basic graphic eq wont affect the sound like adding an analogue graph.eq in an analogue audio path might! I accept that there may be some intended eq-ing going on in itunes: it is a non-pro, "consumer" application after all and its possible apple think the eq will please average joe. I'd be surprised if this is the case though. There are bound to be folk using itunes to playback uncompressed files in an "audiophile" set-up and if there was any eq going on I feel it would have been pointed out by now.

I'll try this when I get the time:
Import a track from a CD as a 16 bit, 44.1kHz wav file in Logic and do the same in itunes and now compare the sound.
Export the wavs to mp3 in both logic and itunes, play both back in quicktime and compare.
import the mp3s into logic again and compare
etc etc

Try to find where the changes in sound occur.

[max cooper]

Originally Posted by Sirocco

since itunes has the ability to eq tracks and other things, im pretty sure it is even affecting the signal when its not turned on, by a small amount

Yeah. It sounds like dreck no matter what you do.

You can import all your CD's into your DAW, but easier to just get a CD player.

If iPods had digital outs...

[murrayhill]

glad to hear that there actually was a difference. I thought I was trippin for a sec there.

Here's another one.

You really can't hear the difference as much when you bounce your logic sessions into mp3s and play them in iTunes. What's up with that???

[pan]

You don't have by accident activated "sound enhancement" in i-Tunes prefs???

It screws the image - some kind of stereo enhancement/compression...

[Drumhum]

Murrayhill,

Its important to understand that Mp3 conversion (from or to) can be done with lots of different techniques. There are lots of different conversion algorithms: some good some not so good. Which one is used depends on many factors. For example you would use a different algorithm/technique in an ipod which has to convert an MP3 to "sound", in real time, to some utility for converting mp3 to a wav file. The latter can spend more time (or less!) doing the process. Note the differences in VBR and CBR and the fact there is often an option for "normal" and "best" conversion regardless of bit rate etc.

Most of apple's apps will use the quicktime converter. I'm not sure what algorithm Logic uses but it could be a different one. You can get plenty of apps on the net that will use different algorithms too. Mp3 is a standardised format but how you create an mp3 is not.

Your latest observation does not involve itunes converting wav to mp3 as logic has done the conversion. Any differences between itunes and logic here is down to just the mp3 to PCM conversion. As Mp3 is a "lossy" codec, its the conversion ~to~ mp3 which makes the most difference to sound quality: the algorithm has to decide what to keep and what to discard.

IMHO Mp3 is pretty rubbish in terms of sound quality. Its good and useful in some ways (mostly that its commonly adopted) but if your prime concern is sound quality then leave MP3 alone.

[murrayhill]

drumhum, thanks for the lesson.
one more question for ya.

What would converter and format do you recomend for iTunes use?

[UnderTow]

Originally Posted by Drumhum

Coo! Never noticed that before. There seems to be quite a difference.

First thing to note is that logic converts the MP3 to a non compressed format (whatever you've set in prefs - like wav or aiff etc).

This shouldn't make a difference. With today's comouters, there is plenty of power to do real-time decoding.

Logic also uses a very different playback engine to itunes too. Logic playsback/mixes in a 32 bit floating point 'world'.

This also should make absolutely no difference unless something is broken in one of the playback engines. When you play back a 16/24 bit stereo file, the mix engine shouldn't affect the sound at all. It should come unchanged out of the 32/48/64 mix engine.

Itunes has to convert (uncompress) the compressed format type of MP3 as it plays back.

Should make no difference. See above.

So in perspective, from an MP3 in itunes you are hearing one type of conversion process played back in one type of audio engine while in logic you hearing a different conversion in a different audio engine.

I think the main difference is the MP3 decoder being used in either application. iTunes might also do some other processing under the hood but I doubt it.

And of course, there is the ubiquitous placebo effect ...

Sorry Sirocco, but with the greatest respect, your theory does not hold up for me. Simply adding some computer code to provide some feature such as a basic graphic eq wont affect the sound

In theory you are correct but you never know what coders are doing under the hood unless you have tested the application for bit accuracy.

I'll try this when I get the time:
Import a track from a CD as a 16 bit, 44.1kHz wav file in Logic and do the same in itunes and now compare the sound.
Export the wavs to mp3 in both logic and itunes, play both back in quicktime and compare.
import the mp3s into logic again and compare
etc etc

I would suggest you don't listen but instead load the exported files into a wave editor and do a phase reversal null test. Our ears and brains are just not trustworthy enough.

Alistair

[Drumhum]

Murrayhill,

By "lesson" I hope I didn't come across as "preachy". I was more thinking about the subject by typing ("thinking aloud" is the wrong phrase!) but I'm glad you found some value from my waffle.

Sorry, but I can't provide some real, hard recommendations for MP3 converters. I've just not been up-to-speed regarding this for a while. When I "went into" this I kind of concluded that MP3 was not a good idea and tried to resist the move to audio data compression. But then Apple brought out the iPod:-) I know the situation is much better than the early days of mp3 when some software sounded terrible! There's been a lot of lessons learned since then. I'd be surprised if itunes was 'bad' at the job considering how many folk use it. I'm pretty happy using Itunes to convert my stuff but then I don't really care too much about achieving the perfect Mp3. In fact I tend to use ACC rather then MP3 for my own use. I'm sure there'll be lots of stuff on the web about this.

[Drumhum]

Undertow,

"This shouldn't make a difference. With today's comouters, there is plenty of power to do real-time decoding."

Not sure what you mean by this. If you mean Logic's method of converting to a wav (say) file first is un-necessary with "todays computers" I'm not sure if Logic's designers would agree. It makes sense to have all your audio in a Logic project in a unified format for many reasons. Whether its done in realtime or not the mp3 still has to be converted and we are discussing how mp3 conversion affects the sound.

"This also should make absolutely no difference unless something is broken in one of the playback engines. When you play back a 16/24 bit stereo file, the mix engine shouldn't affect the sound at all. It should come unchanged out of the 32/48/64 mix engine."

erm.. well I agree with your statement about how the audio engine/mix engine ~shouldn't~ affect the sound but the fact is different audio/mix engines ~do~ sound different. I believe Logic is pretty good and if there is a difference between itunes and logic I'd expect logic to be more "right" - which is where I was coming from. I wasn't suggesting that Logic somehow distorts/affects sound in some way. How software mixes audio streams is pretty complex - its more than just a summing of numbers. when you reduce the volume fader of a track you are not reducing the bit resolution exactly - for eg. Floating point operation is quite a subject!


"Should make no difference. See above."

Again I'm not sure what you are getting at here. I'm sure this is the sentiment that the designers of mp3 converters/audio engines are working towards


"In theory you are correct"
Thankyou :-)

"but you never know what coders are doing under the hood unless you have tested the application for bit accuracy."

I agree we don't know what is going on - which is what I said. Not sure what you mean about "bit accuracy" when you take into account we are talking about MP3 and lossy codecs.

"I would suggest you don't listen but instead load the exported files into a wave editor and do a phase reversal null test. Our ears and brains are just not trustworthy enough."

I doubt the phase test thing would be helpful here. All you will establish is that either the two apps are using the same conversion process, or they are not. It will not establish any measure of quality, or which is best. You certainly wont find an Mp3 will compare exactly with its original PCM version, no matter how the conversion was done. thats the whole point of mp3: it removes "sound" in the original that we (apparently) don't hear to reduce the data. Mp3 is by design, exploiting psycho- acoustics and the way we hear. I suggest that the ~only~ way to test if an MP3 encode is good or not is by listening to it.

[UnderTow]

Originally Posted by Drumhum

Undertow,

"This shouldn't make a difference. With today's comouters, there is plenty of power to do real-time decoding."

Not sure what you mean by this. If you mean Logic's method of converting to a wav (say) file first is un-necessary with "todays computers" I'm not sure if Logic's designers would agree. It makes sense to have all your audio in a Logic project in a unified format for many reasons. Whether its done in realtime or not the mp3 still has to be converted and we are discussing how mp3 conversion affects the sound.

I mean there should be no difference to the sound wether it is converted real-time or not. If there are differences it has to do with other aspects of the conversion and not the real-time aspect of the conversion.

"This also should make absolutely no difference unless something is broken in one of the playback engines. When you play back a 16/24 bit stereo file, the mix engine shouldn't affect the sound at all. It should come unchanged out of the 32/48/64 mix engine."

erm.. well I agree with your statement about how the audio engine/mix engine ~shouldn't~ affect the sound but the fact is different audio/mix engines ~do~ sound different. I believe Logic is pretty good and if there is a difference between itunes and logic I'd expect logic to be more "right" - which is where I was coming from. I wasn't suggesting that Logic somehow distorts/affects sound in some way. How software mixes audio streams is pretty complex - its more than just a summing of numbers. when you reduce the volume fader of a track you are not reducing the bit resolution exactly - for eg. Floating point operation is quite a subject!

I am talking about loading a single stereo file (reguardless of format) into a DAW and exporting it with no plugins, no dither and all faders at unity gain. This should result in a bit for bit identical export to the source. If not, the playback engine or mixing engine is broken.

Obviously if you are loading an MP3, it won't be bit for bit identical but the export should be identical to the decoded imported file.

"Should make no difference. See above."

Again I'm not sure what you are getting at here. I'm sure this is the sentiment that the designers of mp3 converters/audio engines are working towards

The fact that it is real-time should make no difference.

"In theory you are correct"
Thankyou :-)

Heh.

"but you never know what coders are doing under the hood unless you have tested the application for bit accuracy."

I agree we don't know what is going on - which is what I said.

I am refering to the quoted sentence: "Simply adding some computer code to provide some feature such as a basic graphic eq wont affect the sound".

In theory it shouldn't affect the sound but there are (many) hardware FX (I mean digital with digital inputs and outputs) and plugins that do not pass out a bit for bit identical signal to the input even in bypass mode. In other words: Some developers f*ck-up.

Not sure what you mean about "bit accuracy" when you take into account we are talking about MP3 and lossy codecs.

No, I am talking about the signal after the decoding process. I am talking about PCM audio going through a "bypassed" FX.

"I would suggest you don't listen but instead load the exported files into a wave editor and do a phase reversal null test. Our ears and brains are just not trustworthy enough."

I doubt the phase test thing would be helpful here. All you will establish is that either the two apps are using the same conversion process, or they are not.

And that is the first step. Just look arround these forums for people that are hearing subjective differences in files that are 100% identical. That is our brain and hearing fooling us. So step one is to determine if there really is a difference.

Then comes step two where we make a subjective choice on which app gives the better subjective sound.

Sorry for not clarifying that it was jus the first step in testing.

It will not establish any measure of quality, or which is best.

Indeed but we first have to determine wether there really is a difference. Our ears and brains are not trustworthy enough for this first test. Of course this is assuming the differences are not so blatant that we can't go wrong. Then again, I have heard people rave about being 100% sure they hear a difference in identical files and accusing everyone that doesn't of being deaf so ... just do the phase test and move on from there.

You certainly wont find an Mp3 will compare exactly with its original PCM version, no matter how the conversion was done.

No of course. I am not suggesting one does a phase test between the original wave file and a decoded MP3. I am suggesting the output of two decoders is phase tested and then move on to the listening tests.

I suggest that the ~only~ way to test if an MP3 encode is good or not is by listening to it.

After one has determined that two codecs are giving scientificaly verifiable different outputs.

Alistair

[theblue1]

Just a point of fact, here... with perceptual encoding almost all the "work" is done on the front end, removing "superfluous" (mostly HF) information following transients (where, perceptual encoding theory tells us most of the "meaningful" information is).

Little or nothing is done on playback to a conventional mp3; it's simply treated as any other PCM encoded audio. For conversion to a conventional wav or AIFF format files, the data is reformatted -- but it is not "reconstructed" in the sense that "missing" information is somehow reconstructed. (Aside from interpolation needed to fill the WAV or AIFF file format, crudely speaking.)

You can demonstrate this to yourself by renaming an mp3 suffix to *.wav (or presumably AIFF, I suppose) and dropping it into a "non-mp3" player.


Now... that said... SOME consumer players such as WMP and MusicMatch (which I've tested against presumably flat playback in SoundForge, Sonar, and classic WinAmp) have an added bass boost. (I don't know if iTunes player does or doesn't -- but it certainly wouldn't surprise me. WMP added it around version 6 or 7 and MM added a similar bass boost within the year to their next version. It is NOT to be confused with the regular EQ functions and, as far as I can tell, is not documented.)

[Drumhum]

The Blue 1,

Good point. The "sound" of the Mp3 will come from the encoding rather than the decoding - unless there is some "tweaking" going on in the playback (like eq boosts etc)

And this has been demonstrated by OP MurrayHill.

I can't go along with the idea that an MP3 file is essentially the same as a Wav file though, and that you can playback an mp3 in a wav player so long as you can "fool it" into thinking its a wav. the data structure is completely different.

for example...

In Mp3 as well as the clever psycho-acoustic/perceptual encoding stuff going on there is also some data compression (called Huffman) that is lossless. Its like the way we zip files. This part needs to be "unzipped" for the data to make sense. No wav player will do this.

Another example is how stereo files are handled (this is where "stereo", "Joint Stereo", "dual mono" options in better encoders come in). One technique to reduce data is to not duplicate identical sounds from each stereo channel - in other words the Mp3 encoder does not encode each channel separately but takes it as a whole. if the same sound exists in both channels (eg a sound panned centre) it will just encode one of them and throw away the other. the decoder re-constructs things on playback.

Note also how you need "special" software to edit mp3's: the format wont allow just a cut in the middle of a file like PCM files (like wav) do. Software that allows editing is really doing a fair amount of work - it essentially notes where you want your cut and then generates two new mp3 files from the original. when you edit a wav, you really are just moving chunks of binary around.

And then there's MP3's encoded with Variable Bit Rate which would be wholly incompatible with the CBR nature of WAV and AIFF.

There is so much more to this - I'm working off pretty vague memories from when I looked into it so don't quote me to the letter! There's loads of articles and papers to google if one's drive for geek-dom is strong enough!
It is darn clever stuff though.

Note though that just changing the dot-suffix to wav wont necessarily make software play a file like a wav. In the mp3 there is extra data that has all sorts of info to tell the player what to do - indeed another reason why I can't see how a wav player wouldn't have a chance playing back an mp3. If you found it worked for you I'm sure its because the software just 'realised' it was really an mp3 - try changing a *.doc word file to .wav and see if it plays in your software.