Banned
Joined 2019
Thanks Pavel, sorry I missed this when you posted it.
The wav file in your zip seems to contain nothing at all, except right at the end. Is that what you got?
Bvvrati. The files should align perfectly once the offset is taken into account. There is no AD/DA conversion, no reclocking. Look for the 4410 Hz maker at the beginning of the files.
The wav file in your zip seems to contain nothing at all, except right at the end. Is that what you got?
Bvvrati. The files should align perfectly once the offset is taken into account. There is no AD/DA conversion, no reclocking. Look for the 4410 Hz maker at the beginning of the files.
No, you got it wrong. Turn up the volume, there is a silent music. I have already described it two times in this thread, without having your attention.
Low Level Detail: An experimental search and test.
So my conclusion is you were not interested.
For perception tests of things that should be entirely down in the noise floor don't use uncorrelated signals (like mixing a different track under the given one), this is masked out in perception like any neighbor's kitchen radio playing, our own breathing noise etc. Rather, we must use at least losely correlated but scrambled signal.
As for the original intent of this thread, quantifying low level detail (or loss of it) is really a tough measuring task. A way it might be possible is to measure FR and distortion per full logsweep at contiuously decreasing levels, which calls for precise time-domain averaging multiple takes (thousands, literally) once we decrease level so much that noise is getting dominant in a single take. Then look for changes in the FR/distortion profile vs level, at some point we might see the effect of some non-linear / non-continuous friction or similar. Different test signals might be used as well, including real music. Additionally we could overlay the test signal with uncorrelated noise (which is cancelled out by the averaging to an exactly known amount) and see if this "noise bias" does change anything...
As for the original intent of this thread, quantifying low level detail (or loss of it) is really a tough measuring task. A way it might be possible is to measure FR and distortion per full logsweep at contiuously decreasing levels, which calls for precise time-domain averaging multiple takes (thousands, literally) once we decrease level so much that noise is getting dominant in a single take. Then look for changes in the FR/distortion profile vs level, at some point we might see the effect of some non-linear / non-continuous friction or similar. Different test signals might be used as well, including real music. Additionally we could overlay the test signal with uncorrelated noise (which is cancelled out by the averaging to an exactly known amount) and see if this "noise bias" does change anything...
LOL. OK then. I don't think I've ever heard Aaron Copeland's Rodeo described as "only brutal drumming digital absurdity" before.
Did the Sinatra song with the hidden wind quintet strike you equally?
The myth of low level detail revealed
I read this excellent thread from beginning to end, and when I got to this post it was just perfect.
The point of testing with "low level detail" which you absolutely cannot hear is to demonstrate the system including speakers may be better than our hearing. If this is the case, it might be interesting to see if these speakers renowned for having superior low level detail might just have some frequency response variation rather than actual lower distortion or noise or other resolving ability.
I read this excellent thread from beginning to end, and when I got to this post it was just perfect.
The point of testing with "low level detail" which you absolutely cannot hear is to demonstrate the system including speakers may be better than our hearing. If this is the case, it might be interesting to see if these speakers renowned for having superior low level detail might just have some frequency response variation rather than actual lower distortion or noise or other resolving ability.
The whole purpose of MP3 compression is to cut data rate by discarding low level "noise" that is masked by a strong signal in the same sub band, so there really should not be anything
Well, hopefully not completely nothing. One would hope that the higher bit rate schemes leave something there. I wasn't able to pull anything out of even a high bitrate MP3, but it seems that Pavel can.
The embedded signal test may not be the best, but at least it's an easy start. The flaws that I hear in medium to good compression are a loss of ambient details and a warming and pushing forward of the vocals. I suppose those are related. Non listening based tests for that could be tricky, but it would be nice to find such tests.
The embedded signal test may not be the best, but at least it's an easy start. The flaws that I hear in medium to good compression are a loss of ambient details and a warming and pushing forward of the vocals. I suppose those are related. Non listening based tests for that could be tricky, but it would be nice to find such tests.
It is very important. It makes voices with skin on them and allows for the telling of subtle clues about an instrument, but also the tiny details that allow for a good 3D soundstage/image (if on the recording).
The difference between good hifi and great hifi. Some trining to hear these things does not hurt.
dave
@KSTR & @planet10
I posted along the same lines as you guys earlier in the thread.
I was looking to investigate the problem from the other end of the issue - finding what type & level of low level distortion is perceptually audible & in what way?
My idea is to try to find a way of generating a noise track correlated to the signal & mixing it with the signal at various levels to establish what conditions of noise/level/mix are audible.
Any ideas?
I posted along the same lines as you guys earlier in the thread.
I was looking to investigate the problem from the other end of the issue - finding what type & level of low level distortion is perceptually audible & in what way?
My idea is to try to find a way of generating a noise track correlated to the signal & mixing it with the signal at various levels to establish what conditions of noise/level/mix are audible.
Any ideas?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.