Head-Fi.org › Forums › Equipment Forums › Sound Science › 24 vs.16 bit, test results!
New Posts  All Forums:Forum Nav:

24 vs.16 bit, test results! - Page 3

post #31 of 41
Quote:
Originally Posted by thune View Post
 

How confident are you that the 16-bit recordings you buy have been properly dithered?  At a minimum, 24-bits means one less thing to get screwed up.

 

There's a lot to admire about the survey, but I'd like to see participants train on 10-bit, and 12 through 16-bit files compared the 24-bit source ; with training complete, weed out those who can't tell 12/13-bits from 24-bits; THEN take the 16-bit vs 24-bit test. I'm unconvinced the survey respondents know what to listen for, and it can vary by recording anyway. The results from trained listeners might still be null, but I'd find them more meaningful.


that is a good idea, so people can start looking at the real bit depth they hear. from my own experience, I can be bothered by noises around -70db, not on all songs and I'm known to be overly sensitive and obsessed with any kind of noise or hiss. but -80db I really can't hear it with music at the same time. I need to crank up the volume too much for my own good and pay attention when the music is on a very calm part. so 13/14bit would probably be enough for me for the rest of my life.

 

about the dithering, the test was done with a not great dithering setting on purpose it seems. but even without dithering at all, would people notice? to me dithering is more of an insurance than a need.

post #32 of 41

I think your figures are a little high. In order to hear -80dB, you would have to be playing your music at over 100dB. In his audio myths workshop, Ethan Winer created a music track with the most obnoxious sort of noise at progressively lower and lower levels. It disappears pretty fast, and that is with noise that sounds like a buzz saw. Not all noise cuts through like that.

 

Dithering 24 bit to 16 bit is very simple with pro equipment. I seriously doubt that is an issue at all, except with inexperienced people using shareware programs. Anyone with any experience in digital audio at all knows how to dither properly.

post #33 of 41

well that's what I somehow came up with for any kind of situation, listening quietly, or loud, with very dynamic music and long calm passages of classical music where I might want to push up the volume level because they last a few minutes...

it's a worst case scenario for my very picky self so I can say that over this level doesn't matter in complete confindence. in real practical situation, I listen pretty quietly so 20db under the music is already awfully quiet ^_^.

post #34 of 41

Well here is a really good test for it. How many bursts of noise buzz can you hear in this track?

 

http://ethanwiner.com/aes/artifacts2.wav

 

Hint: There are 16 at various levels. from -30 to -75dB. Using your figure of -80dB as your threshold, you should be able to hear all 16 (good luck!) Obviously, this is a worse case scenario with very quiet music and very obnoxious noise. In practice, with continual steady hiss, the audibility would be much less.

 

I cranked my iMac up full blast and in headphones, I can only hear the last tone in the silence at the end of the track. I hear about 8 under the music. That is -60dB.


Edited by bigshot - 7/5/14 at 1:44pm
post #35 of 41


 I used the quietest(but not silent) long parts of symphonies I own, stuck some other musics in it at different volume levels(one at a time ofc), and set the volume to "real loud for me" on those parts as it is the only moment when I sometimes, find myself turning the volume level up in real life listening.
in theory the way I tried that, I could have gone and heard -110db noise for the fun of it, as long as I was willing to turn up the volume and find a track silent enough. that wasn't the point at all. I wished to test the real limits of myself with my music in worst case yet still somehow real scenario. so well beyond what's really needed.

 

 

 your test track is on average music at average volume, so obviously we can't expect to get even remotly close to my extreme stuff. but I just thought that if my extreme stuff is covered and some more by 16bit, it would make a point of not needing more.

post #36 of 41
Thread Starter 
Castle-I agree with you but would also point out that -80 noise is still quieter than ambient noise at a live venue and for ME, even if I heard noise at that low level it wouldn't disturb me in the slightest
post #37 of 41

sure I myself usually listen to music quietly in fact and drown most of the low level sounds into ambient noise. really it was a one time test to see how far I could really go. not saying that I actually need that. on Mr Winer's track with my usual loudness I can hardly get the 7th noise and would most probably not be disturbed by it if I didn't look for it.

it doesn't take a genius to understand that 100db is really loud and that a quiet room is often above 25db in noises. so more than a -75db noise floor for quantization noises doesn't seem important just from a logical point of view.

post #38 of 41

Another thing Ethan points out in his video is that the natural noise floor of PCM is plenty, even without dithering. He plays a track bumped down dithered and undithered, and I can't really tell the difference.

post #39 of 41

Reducing this problem to absolute SPL levels doesn't take a genius, it is also too simplistic to be useful. The spectrum of noise is essential, with playback distortion and psychoacoustic masking also playing in.

 

Were people listening for noise floor in the 16vs24-bit test? cause that's probably a fool's errand. I find the differences in bit depth present mainly in the percieved character of the medium and loud sounds; their temporal, spacial, spectral, and emotional components.

 

God bless SOX: Makes this stuff easy to play with.

post #40 of 41

I've done gobs of bias controlled listlening tests.  16/44. red biook is all that is required.  When it comes to data loss, I determined that 256KBPS MP3 is indistinguishable from a CD.  192 has an audible difference but it is minor.  128 has a fairly obvious audible difference and below that things get fairly unlistenable.

 

The purpose of 24 but audio is to provide overhead for mixing and mastering.  There is no point whatsoever in cutting a recording to anything more than red book.  I haven't encountered anyone who can reliably distinguish them in a bias controlled test.  I did these tests with my personal recordings so that the only difference was the downsampling.

 

I suspect that if audible differences exist in commercial recordings, it is due to the mastering, not the excess data.


Edited by blades - 7/6/14 at 8:04am
post #41 of 41
Quote:
Originally Posted by thune View Post
 

How confident are you that the 16-bit recordings you buy have been properly dithered?  At a minimum, 24-bits means one less thing to get screwed up.

How confident are you that dithering actually matters for 16 bit recordings? A completely undithered 16 bit recording should still have inaudibly small amounts of noise. Dithering is something that is almost definitely completely inaudible, but since there aren't any real downsides, you might as well do it anyways (since it does make the quality measurably better, even if it isn't audibly better).

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › 24 vs.16 bit, test results!