hearing the difference b/w mp3 and WAV files

[dayvel]

I would recommend that everyone do some ABX tests between WAV files and 128kbps mp3s done with a good codec (like the Fraunhoefer in iTunes). On some program material it's pretty easy to tell, but on a lot of stuff you may surprised how poorly you do.

[fuu drakka]

I think a lot of it too is hearing the actual artifact for the first time. if you don't know what to listen for, it can be hard to hear, especially at the higher bitrates.
try making a really poor quality mp3, and really listen to the awful sound, then try and go up in quality, and try to hear the pixelization of the sound, even in the higher ones.
a good analogy for this for me is dither. I have always used dither, but never really knew what it sounded like. Once i got a dither that I could solo, and change in volume, I could focus in on it's sound, and can now hear what it's doing in my mixes. word.

[rimbaud234]

[quote="dayvel"]I would recommend that everyone do some ABX tests between WAV files and 128kbps mp3s done with a good codec (like the Fraunhoefer in iTunes). On some program material it's pretty easy to tell, but on a lot of stuff you may surprised how poorly you do.[/quote]

Actually the default iTunes coder is a 128 kbps version of AAC that Apple developed with Dolby especially for iTunes.

For MP3s it's important to listen to the high end - the bandwidth is usually cut as the data rate is reduced.

Also as the bit rate goes lower and lower spectral holes will appear because the encoder doesn't have enough data to code all of the spectrum. Those are the "swirrelies" you hear on really low data rate coded material.

Also listen to the transient material. The block transform encoding of MP3 causes transient smearing at lower data rates (pre-noise, etc.) To see this take a very transient signal like a repetitive hi-hat or castanet signal with a sharp front edge and encode it at around 96kbps or lower. You'll see pre-noise on the transients which is the smearing I'm talking about. It makes transients sound smooshy

As a last resort you can do a subtraction between coded and uncoded audio (as long as you've taken into account delays that can be added by the audio coding) and see the "error" signal. You won't hear all of it as the coder introduces the noise intentionally by quantizing the spectral coefficients and trys to put the noise (errors) in places that are psychoacoustically masked.

[zwolf]

Ok. I'm home now and my MP3 encoding is set at 160kbps. Reducing to 96, I hear a major difference. At 128, I'm not so sure yet. But I'm going to keep listening until my ears are trained to recognize the artifacts.

For those that have itunes, is the difference between a 16 bit WAV and a 128kbps (the "high quality" setting) MP3 really ETREMELY noticible? How does itunes encoding compare to others?

I'm still humbled.

[CropLogic]

try burning a copy of the mp3 and a copy of the original cda/wav/aiff version to an audio cd and listen to it in your car or favorite hifi system, you should hear it.

[djslayerissick]

As a last resort you can do a subtraction between coded and uncoded audio (as long as you've taken into account delays that can be added by the audio coding) and see the "error" signal. You won't hear all of it as the coder introduces the noise intentionally by quantizing the spectral coefficients and trys to put the noise (errors) in places that are psychoacoustically masked.

i just tried this with some success. i used a wav and 64bit mp3, inverted the mp3, and cancelled as much as i could.

as far as i can tell, after the main part of the real sound cancelled out, i was left with a white noise looking spectrum analysis and WAY to much high frequencies from 15k to 22k (the ones that were severely lacking in the mp3). almost no bottom end. and it sounded very distorted and white noise-y.

[TheStevens]

yeah, i think that 128 MP3s are noticibly crappier than WAV. Meaning you don't need to have monitor speakers to hear it, I can hear it well on stereos, in cars, etc. Some sounds get affected by it a lot more than others though, especially drums and cymbals. Going up to 160 makes a huge difference in my opinion.

[zwolf]

got it now. did a series of blind A/B tests and can hear the difference at 128 kbps when it comes to the wash of cymbals. It sure isn't a RADICAL difference though.

Thanks for the help.

[jajjguy]

On some of the MP3s I d/l from Limewire (o, am I not allowed to say that?) I can hear a digital crinkling sound in the super high end. Like tape hiss, but it's sounds all matrix-y. I always assumed this was due to poor encoding at low bit rates, but there may be other forces at work.

I don't think that's necessarily the mp3 encoding process, but perhaps something got digitally clipped somewhere... perhaps the CD went through a overloaded D/A/D stages somewhere, or there was some overly strong digital EQ'ing or amplifying, before the track was ripped.

No way it's clipping. It's there when the song is starting out and there's no music, and transients have no effect on it (except making it harder to hear the crinkling sound). Am I the only one who gets this?

yeah, i've heard that. low bitrate mp3s, and other aggressive data compression schemes like MINIDISC. yuk.

notice that certain bar/restaurant/supermarket jukebox setups are actually streaming ridiculously low bitrate files? you can hear some of this matrixy high end crap. not what i want to hear when i'm puttin down the brews... i can get shouty about it.

[dynomike]

notice that certain bar/restaurant/supermarket jukebox setups are actually streaming ridiculously low bitrate files? you can hear some of this matrixy high end crap. not what i want to hear when i'm puttin down the brews... i can get shouty about it.

That bugs me too. The digital music stations we have on digital cable/satellitle here are full of that stuff. Do they download their files from kazaa?

Mike

[twitchmonitor]

yeah, i've heard that. low bitrate mp3s, and other aggressive data compression schemes like MINIDISC. yuk.

notice that certain bar/restaurant/supermarket jukebox setups are actually streaming ridiculously low bitrate files? you can hear some of this matrixy high end crap. not what i want to hear when i'm puttin down the brews... i can get shouty about it.

Funny thing is I kind of like the noise...just not on tunes I'm trying to listen to. I want isolate it and stick it in some songs for fun. Maybe I'll post a file of sound and let it circulate in Limewire with horrible data compression and then download it...

[rimbaud234]

Regarding iTunes the don't use MP3. They use AAC (which you can think of as MP4). AAC is arguably much better than MP3 at similar data rates.
Apple worked with Dolby to take the standard AAC code and add improvements to make it as good as they could (which is actually quite good). The default iTunes rate (from the store) is 128 kbps AAC.

It is very good for an audio coder, but still not as good as linear PCM.

Low bit rate audio coders transform the audio into the frequency domain and then code the audio that way. The decoder takes it from the frequency domain back to the PCM domain for playback. They do this because your ears basically work in the frequency domain (with your cochlea acting as a sort of filterbank). One common trick to save bits is to not correctly code the upper frequencies as grunge and distortion up in those frequencies are less noticable than the lower frequencies. However, when comparing the coded material to source you'll hear the differences in the high end. In general it can lack definition and also they tend to not code the higher frequencies. Depending on the data rate you can have the top end cut off anywhere from 12kHz (really low data rates) to 18kHz (at the higher data rates).
Code up some MP3 files at different data rates and check out the frequency spectrums and you'll see what I'm talking about. It also helps to run a real time spectrum analysis and you can see that these changes can be dynamic (i.e. bandwidth and other things can change over time).

[Marc Alan Goodman]

Does anyone here know anything about .shn files? The have to be expanded back into Wavs before they can be played, but I wonder what kind of quality loss there is.

Thanks
-marc

[Auxillary]

Ive never had a problem. Simply turn up an mp3 of a song you like to turn up and you will immediately get annoyed, frustrated and want to kill things.

[dynomike]

Does anyone here know anything about .shn files? The have to be expanded back into Wavs before they can be played, but I wonder what kind of quality loss there is.

Thanks
-marc

Its lossless.