Hi Guys,
I am new here and I like to do a comparison between uncompressed music and MP3 192kbps.
There is no doubt in my mind that you experts have some suitable samples for me, I mean suitable so that one actually can hear the difference. I do have a decent stereo set and I can convert uncompressed wav files to MP3.
So are there any such files floating around here?
Thanks
Edmund
Are you looking for specific samples that are very hard to encode in MP3? For pretty much all music, 192Kbps is completely transparent.
If you're curious about how MP3 and the encoders behave, try taking some of your own music, and converting with a variety of medium-to-low quality settings. The results will undoubtedly enlightening.
There's a big test here (http://www.hydrogenaudio.org/forums/index.php?showtopic=70598). You might find more by searching this site for "killer sample"
Please note that we will only accept discussion of tests performed in a way that is objective, scientifically valid, and compliant with our rules (http://www.hydrogenaudio.org/forums/index.php?showtopic=3974), i.e. double-blind tests. See the Knowledgebase, pinned thread in General Audio, Wikipedia, etc. for plenty of information about ABX testing, the double-blind testing methodology that is customarily used here.
As was stated before music is fine usually @192 kbps when using a good mp3 encoder.
As for specific problems you can try these:
eig - the most critical part of it (http://dl.dropbox.com/u/2681777/problems/eig_essence.flac) as a sample of problematic impulses in electronic music, or castanets (http://dl.getdropbox.com/u/2681777/problems/castanets.wav) as a problematic sample for 'impulses' in natural music. Pre-echo problems like these are the most frequently reported issues for mp3 and other codecs even when using high quality settings, but other when listening to electronic music or castanets, even this isn't too much of concern as long as you aren't very sensitive to this kind of problem.
Temporal smearing can also be heard in harpsichord music as in harp40_1 (http://dl.getdropbox.com/u/2681777/problems/harp40_1.flac).
With rare samples an artificial tremelo can be heard. It can be annoying as in the electronic sample lead-voice (http://dl.getdropbox.com/u/2681777/problems/lead-voice.flac) (again, electronic music can be real bad to audio codecs), or - not very bad - in natural music like trumpet_MyPrince (http://dl.dropbox.com/u/2681777/problems/trumpet_MyPrince.flac).
Not bad but also not perfect can be samples like herding_calls (http://dl.getdropbox.com/u/2681777/problems/herding_calls.flac). It's voice, though a special one. As for more usual voices, female women singers (especially of the french tongue) can be a slight problem too as in Là Ou Je Suis Née (http://dl.getdropbox.com/u/2681777/problems/LaOuJeSuisNee.flac).
Trumpet can be a slight problem as in trumpet (http://dl.getdropbox.com/u/2681777/problems/trumpet.flac).
Lame development has improved upon samples like these a lot, so when using high quality settings the problematic character of most of these problems is next to nothing now to most people in a real life listening environment.
As for specific problems you can try these: <snip>
Those look interesting however they all lead to an "error 404" screen on my system/browser. Does one have to be a member of some site/forum for access? accept cookies? be in the same country/continent? or is there some other thing that is blocking me? Thanks.
As for specific problems you can try these: <snip>
Those look interesting however they all lead to an "error 404" screen on my system/browser. Does one have to be a member of some site/forum for access? accept cookies? be in the same country/continent? or is there some other thing that is blocking me? Thanks.
My bad, the links were not correct, sorry. Should work now.
Speaking shortly there are different groups of samples. The most important are two first.
1. Transients - pre-echo issues (castanets, electronic, rock)
2. Tonal (classic instruments: harpsichord, trumpets etc).
3. Speech, stereo separation, quite/loud etc.
And the combinations of these groups.
Speaking shortly there are different groups of samples. The most important are two first.
1. Transients - pre-echo issues (castanets, electronic, rock)
2. Tonal (classic instruments: harpsichord, trumpets etc).
3. Speech, stereo separation, quite/loud etc.
And the combinations of these groups.
What's the technical term for the distorion I hear I can only describe as "swirly", everyone? Listening to audience applause it would sound like the phase was being modulated slightly, once or twice per second, I'd say.
I hear it on sat. radio stations, at times, and was first exposed to it on the opening of Pearl Jam's "Daughter" (when encoded at too low a rate), a "codec killer" track recommended by David Ranada of then "Stereo Review/ Sound and Vision" magazine.
What's the technical term for the distorion I hear I can only describe as "swirly", everyone? Listening to audience applause it would sound like the phase was being modulated slightly, once or twice per second, I'd say.
I'm sure there's a more appropriate musical term, but "undercoding" is probably the best technical term for it. There were not enough bits available (given constraints like CBR.. XM/Sirius is something like 12-24kbps HE-AACv2 CBR, at least a number of years ago when I knew more about it technically). There just weren't enough bits to satisfy the demands of the psychoacoustic analyzer, and it did the best it could. This sort of artifact goes away almost entirely at either higher bitrates or when using VBR, unless the psychoacoustic model is poor.
Hey friends can you tell me what factors are changing while conversion of uncompressed audio to mp3 files.
Conversion of uncompressed audio to mp3 means going from music representation in the time domain (music considered as a sequence of musical samples collected in extremely short time intervals as defined by the sampling frequency) to a representation in the frequency domain (music considered as the frequency distribution within relatively short windows of time). The frequency distribution is simplified for getting the compression effect - frequencies which are considered to be inaudible due to masking effects of the human hearing system are ignored.
There's a lot of decision making in this process for the encoder software, for instance the encoder has to decide whether or not there are impulses in the music of the current time window.
In case there are the current time window is divided into three subordinate windows for a better temporal resolution of the music. Decisions like these are not always easy to do for the encoder. There's also some limitation in the mp3 specs so that the frequency distribution in a time window cannot always be encoded as precise as the encoder thinks it should be done.
As errors in the decision making processes cannot totally be avoided and due to the mp3 limitations mp3 music can contain moments of audible inaccuracies. This applies only to specific musical content though, and with modern encoders this is rare event when using high quality settings, and even then the inaccuracies are hardly audible usually.
Oops, where's the post I replied to?
See my edit to your first post. Your reply is very informative and may well be useful for someone else, so there’d be no point in deleting that alongside the spam; whoever deleted the latter before may have thought the same thing.