I wonder when I should play FLAC instead of 320 Kbps: size of room, type of music, etc. I saw plenty of discussions about 320 Kbps MP3 vs. FLAC on Internet but none with reliable source. Are there any scientific studies (e.g. blind experiments) or at least expert opinions (e.g. by some famous DJ) on that matter?
Well, the thing I found requiring surprisingly large MP3 bit rates was encoding my accordion's tremolo registers. Imagine three reeds sounding with lots of harmonics, almost equal volume, and slightly different pitches, hand-tuned so that all of the respective slow beatings combine into a pleasant sound. That's actually a somewhat unusual combination since high-quality accordions tend to have one reed in a tone chamber, giving it a different volume and/or using only two beating reeds, while folksier accordions use rather heavy beatings in order to get a piercing rather than pleasant sound.
At any rate, a high quality recording of something played slowly with that sound reencoded at 128kbs does not sound "slightly lower quality, discernible by experts" but totally awfully crappy. Like coal grain microphone crappy. With 192kbs, you start getting tolerable. Tolerable!
Ogg/Vorbis faired decidedly better regarding the quality/bitrate ratio. So this kind of thing caught me on the wrong foot.
As a rule, I'd stay away from any lossy encoding until you are finished mastering and listening. You never really know what kind of actually-not-all-that-relevant audio component will end up hogging the bitrate.
Some music styles might be less touchy here. I think it is mostly the high frequency content: in my case it was getting too unmusical, but it is similarly distracting when fine-grained noise (like a hihat) is getting too musical, having discernible and wobbling tones in the noise (actually, "musical noise" is a well-known compression artifact).
"lossless" means you get your waveforms reconstructed exactly, bit for bit, identical to the original.
The biggest problem of MP3, and people keep ignoring this as if it really doesn't matter, is that it smears transients. internally, MP3 is split into frames of fixed size. The same math is used to reconstruct the entire frame. Even though the frames are small (1152 samples = 1/40th of a second), the duration of one frame is sufficient for the ear to detect that some sounds have been "altered".
320kbit still means you keep 1/4th of data only. Come transient time (snares, percussion, hihats, etc) that loss of data shows and what may be a beautifully crunchy attack is sacrificed because the encoder math optimised the data to cover the entire frame's needs, not necessarily those few important milliseconds.
I've never found audible difference between 320kbit and 256kbit mp3. As long as mpeg audio is a bunch of matrices through which data is passed back and forth, adjusting the bitrate primarily affects how much data you're going to discard at the entry point.
If your ears can't detect this, stick with MP3. If they can, go for FLAC. My favourite is ALAC - less open, but more supported on devices I actually use.
First off, your question isn't really about FLAC, but about MP3 vs. any lossless audio. FLAC, ALAC, etc. all end up as raw PCM in the end, and the same PCM that went into the encoder, bit for bit accurate. To answer your question:
When does playing 320 Kbps MP3 instead of FLAC matter?
- When you care about having good quality sound for those who hear the difference.
(Everyone else covered the differences... read those replies for details.)
- When you plan to time-stretch the audio or otherwise manipulate it.
Starting out with an original source will leave you with the best quality. You don't want to be DJ-ing with MP3s, stretching out artifacts.
- When you want to re-encode the audio with another lossy codec.
I do a lot with internet radio, which is often a lower bitrate MP3 (~128kbit). There are some very good sounding (comparatively) stations at even this bitrate. They sound better than other stations at that bitrate because their source material is lossless to begin with. If you start out with lossy material and re-encode it again, you will lose even more fidelity. Even worse, the algorithms may decide that the artifact sound is something to highlight, and will take up more bandwidth to reproduce the artifacts. (This isn't a real serious problem until the third or fourth encoding, but that's not uncommon when a station plays MP3s through an MP3 encoder sent off to a server to be transcoded to different bitrates and codecs.)
type of music, etc.
Some sounds compress better than others. Sine waves are great. Accordions are often used in reference material as they get pretty bad with MP3. For me, the worst of the compression artifacts comes from smearing in the cymbals and other percussion. Timing gets off a bit because the frame size is so large. I also find brass in MP3 particularly bad.
Are there any scientific studies
Yes, many. Check out the references section on the Wikipedia page... there are lot of good cited sources. http://en.wikipedia.org/wiki/MP3#References
FLAC as its clear is : Free Lossless Audio Codec , that means the compression applied on the sound file does not affect on the sound quality... but we have compressions in 320 Kbps that we will lose some frequencies over 16 KHz to have a lower file size , but is that make a real huge difference? as a music producer and a sound engineer I can say yes! it will make a real difference for me and its because I'm very accurate on sounds because of my work , experience and also my equipments (I mean accurate monitoring speakers and more...) , but for peoples there is no difference even on a high-end hifi system they can't sense it, I also asked this time over time to my listeners and costumers but all of them say no! there is no difference! but I also know some frequencies are lost , but if they can't recognize that, so its ok , you have less size files on your disk. but in professional music production ,all DAWs are using lossless audio to make a good output quality and when you are going to export as mp3 you need to apply a dithering on it if you don't want to have mp3 compression artifacts (sometimes dithering applies automatically embedded in DAWs).
(thanks to Alex Basson)
While there are a lot of good answers here, I got the impression that you are talking about whether there's really a difference between 320kbps Mp3s and .flacs when one is spinning them as a DJ. No one's really said anything about that situation and I happen to have quite a bit of personal experience with that as I've been "spinning" mp3s since Final Scratch first came out. I also know quite a bit about sound, recording and the like, so I'm gonna put in my two cents. Remember, this is my PERSONAL EXPERIENCE;
In a nutshell, my opinion on the matter is that, for DJing, there is no difference in playing a flac or an mp3. In FACT, I find that there is really not much difference between a 192kbps mp3 and a 320kbps one either. Little system, big system, well-tuned system, it doesn't matter. Hell, 192kbps is considered "CD-transparent". People swear to me that they only use 320kbps mp3s because they used 192kbps before and it sounded like crap, but I think that must have been due to other reasons. Like the mp3 encoding algorithm used to encode the mp3 in the first place, or the quality of the original recording that is being encoded. Also, most tracks that are for DJs are made and mastered as such that the mp3 encoding process doesn't affect their sound that much, if at all.
It can depend on the quality of the original recording, the playback equipment and even room acoustics. In the car I hear no difference between MP3 and cd. But in my home theater system, which is mid-range quality, there is a clear improvement when playing Flac files in 5.1 channel. MP3 discards some of the ambience that comes thru the rear speakers. With some music it is not so noticeable, but with an excellent recording it is.
Listen to music with good bass speakers (sub-woofer)...MP3 totally cuts the lower frequencies. It's true that we can't hear those, but we can FEEL the boom vibration. Good -professional or really hifi- sub-woofers (e.g. Bose 320) make rooms tremble from a CD or FLAC (or live music), but input MP3 and that vibration is gone.
I'm a volunteer sound technician at a 300 member church, with Bose 820 and 320, like 3000W sound total, with good mixing tables and professional amplifiers (Bose too) and crossovers cutting the high frequencies for the 320s, and cutting the low frequencies for the 820s. Every time somebody brings MP3s or we play Spotify, the booming is gone.