16 bit files almost unlistenable now...

[mansr]

18 minutes ago, adamdea said:

Noise could mask signal in principle but

1) you can hear a tone 20 db below the overall noise level because it will still be higher than the noise level in the relevant bin. so 16 bit allows your to hear tones at -110 dB below (theoretical) peak

I did a quick test. Using the iFi Nano at maximum volume and headphones (low-end Sennheiser), I can readily hear a 3 kHz tone at -105 dBFS and with difficulty a bit lower. With added 16-bit TPDF dither, the tone is still easily audible alongside the noise. Shaped dither at 48 kHz sample rate is difficult or impossible to hear. A 0 dBFS signal would be painfully loud, of course.

[adamdea]

1 hour ago, esldude said:

Here is a quick example.  

Pink noise at about - 40 dbFS and less than that from peak levels in the music.  Noise alone for the first few seconds and then with the music. Bonus points if you can tell me when the noise goes away.  It isn't impossible, you can hear when it is removed partway thru the snippet, but I think it is harder than some might expect.  This level of noise would be like bad cassette tape with no Dolby noise reduction.

 

 06 Heat Wave.mp3

11-12s?

[danadam]

1 hour ago, esldude said:

Here is a quick example.  

Pink noise at about - 40 dbFS and less than that from peak levels in the music.  Noise alone for the first few seconds and then with the music. Bonus points if you can tell me when the noise goes away.  It isn't impossible, you can hear when it is removed partway thru the snippet, but I think it is harder than some might expect.  This level of noise would be like bad cassette tape with no Dolby noise reduction.

 

 06 Heat Wave.mp3

I have no words to describe the first 20 s but after that it's almost unlistenable

 

Would by nice to have, let's say, -60 dBFS version to compare.

[danadam]

On 7/19/2018 at 11:45 AM, mansr said:

At 44.1 kHz (and 48 kHz) there's really no safe place to put noise. Dither shaping at these sample rates relies on our hearing being less sensitive at higher frequencies. Allowing the noise level to rise a bit beyond, say, 12 kHz gives a little more dynamic range in the lower region. The trouble is that the 10 dB or so difference in sensitivity between 5 kHz and 10 kHz is barely enough to allow any useful shaping before the increased noise level at high frequencies becomes audible. With higher sample rates, you can readily raise the noise level by 20 dB or more in the 40 kHz region without anything bad happening. Even there, though, if the level is too high, IMD will shift the noise back into the audible range.

What about 10-20 kHz region? Wouldn't the sensitivity there be low enough? There's this longer piece from gregorio @ head-fi.org about shaped dither and he claims that at 17 kHz the sensitivity is down by 50-60 dB. There is also a graph there showing spectrum of different shaping algorithms (but I don't know where it is taken from).

[Jud]

This is interesting, especially @mansr's explanation of how noise level and signal level are measured, which I now recognize as what @esldude was alluding to earlier.

 

I want to move from the more theoretical (I'm never going to intentionally listen to music at 0dB attenuation) to the everyday.  Can someone tell me what a ballpark difference in peak vs. average dB level would be for a track that gets a rating of, say, 12 on the DR Database ( http://dr.loudness-war.info/ )?

[mansr]

18 minutes ago, Jud said:

This is interesting, especially @mansr's explanation of how noise level and signal level are measured, which I now recognize as what @esldude was alluding to earlier.

 

I want to move from the more theoretical (I'm never going to intentionally listen to music at 0dB attenuation) to the everyday.  Can someone tell me what a ballpark difference in peak vs. average dB level would be for a track that gets a rating of, say, 12 on the DR Database ( http://dr.loudness-war.info/ )?

A random sampling of a few DR12 tracks suggests a typical peak to RMS difference of around 15 dB.

[mansr]

1 hour ago, danadam said:

What about 10-20 kHz region? Wouldn't the sensitivity there be low enough? There's this longer piece from gregorio @ head-fi.org about shaped dither and he claims that at 17 kHz the sensitivity is down by 50-60 dB. There is also a graph there showing spectrum of different shaping algorithms (but I don't know where it is taken from).

Sensitivity above 10 kHz is highly dependent on age and history of sonic abuse.

[Jud]

1 hour ago, mansr said:

A random sampling of a few DR12 tracks suggests a typical peak to RMS difference of around 15 dB.

 

So allowing for a greater difference between the very softest passages and the average than between the very loudest and the average, perhaps 40 or even 50dB range might not be a terribly bad guess for a track that's on the better end of popular music released currently.  Let's say 30dB for a very quiet listening room.  80dB peaks with a 65dB average doesn't seem that loud to be able to hear the softest passages.

 

I don't know what average noise/distortion level to allow for a nice quiet system, taking into account what @mansr and @esldude have said, and also allowing for the fact that noise levels would I think increase with increased overall loudness.  And while we can "hear into the noise," I wonder whether that has any perceptual/emotional effect vs. being able to hear something above the level of the noise.  I'm thinking about how effortful it can feel to pay attention to conversations at your table in a crowded restaurant, for example.

[esldude]

3 hours ago, danadam said:

I have no words to describe the first 20 s but after that it's almost unlistenable

 

Would by nice to have, let's say, -60 dBFS version to compare.

I could add that later today.

[mansr]

5 hours ago, esldude said:

Here is a quick example.  

Pink noise at about - 40 dbFS and less than that from peak levels in the music.  Noise alone for the first few seconds and then with the music. Bonus points if you can tell me when the noise goes away.  It isn't impossible, you can hear when it is removed partway thru the snippet, but I think it is harder than some might expect.  This level of noise would be like bad cassette tape with no Dolby noise reduction.

 

 06 Heat Wave.mp3

It would be better if you didn't use mp3 for such tests. The mp3 encoder will decide that parts of the noise are inaudible and discard them, and then you certainly won't be able to hear them. Here's a spectrogram of a few seconds around where the music starts:

 

 

Notice the black patches at the top.

[semente]

2 hours ago, Jud said:

This is interesting, especially @mansr's explanation of how noise level and signal level are measured, which I now recognize as what @esldude was alluding to earlier.

 

I want to move from the more theoretical (I'm never going to intentionally listen to music at 0dB attenuation) to the everyday.  Can someone tell me what a ballpark difference in peak vs. average dB level would be for a track that gets a rating of, say, 12 on the DR Database ( http://dr.loudness-war.info/ )?

 

This is from Björk's "It's Oh So Quiet" EP CD:

 

 

The dynamic swings are actually the most impressive I've seen from a rock/pop recording (-43.3dB to -0.1dB) but this is probably also due to the musical programme/melody.

 

https://www.discogs.com/Björk-Its-Oh-So-Quiet/release/28720

[semente]

Cowboy Junkies "Southern Rain" from "Black Eyed Man" CD:

 

 

These DR numbers are only realy useful to compare different mastering versions.

[semente]

Chabrier "Espana", Gardiner DG CD:

 

[Jud]

So from the examples @semente has come up with, maybe 40dB to just slightly more for really dynamic material (short of some symphonic music with modern instruments - the performances with period instruments that I've heard tend to be a little quieter).

[Jud]

A couple of thoughts about figuring out what the typical level of noise may be in your system:

 

- Measure ambient noise in front of a speaker with everything off, then turn on whatever would be in the chain if you were playing files, turn the volume to wherever it typically is when you listen, and measure again.

 

- Just to see if actually having the DAC in operation makes a difference, you could try playing a file of "digital silence" and measure again.

[adamdea]

6 minutes ago, Jud said:

So from the examples @semente has come up with, maybe 40dB to just slightly more for really dynamic material (short of some symphonic music with modern instruments - the performances with period instruments that I've heard tend to be a little quieter).

I’m not sure where you are going with this.

 

I do however feel confident you will struggle to hear 16 bit quantisation noise except by turning up the volume in a silent passage (of a test tone “recording”. For most real recordings the noise in the recording is 10 dB higher or more.  If you are interested playing at 100 dB peak level at home then the room noise level is around 70 dB down in a quietish room. 

 

What is the level below peak of self noise in most mics -80dB? Genuine question- I’m not sure but I have that figure in the back of my mind. 

[adamdea]

In relation to background noise and intelligibility-

Is it significantly more difficult to hold a conversation in a quiet room in the day than it is to hold one in a quiet room at night? 

[Jud]

12 minutes ago, adamdea said:

I’m not sure where you are going with this.

 

I do however feel confident you will struggle to hear 16 bit quantisation noise except by turning up the volume in a silent passage (of a test tone “recording”. For most real recordings the noise in the recording is 10 dB higher or more.  If you are interested playing at 100 dB peak level at home then the room noise level is around 70 dB down in a quietish room. 

 

What is the level below peak of self noise in most mics -80dB? Genuine question- I’m not sure but I have that figure in the back of my mind. 

 

Yes, I was thinking about noise levels in recordings too, though I suppose for electric instruments you could run them into the console (or just create the music as files in the computer directly, as some artists do).

 

I'm trying to get an idea of two things, what are the respective contributions of the various sources of noise coming to our ears, and whether it is possible to hear the softest passages of dynamic (though not necessarily demonstration level) 16 bit recordings comfortably at typical listening levels.

[Jud]

5 minutes ago, adamdea said:

In relation to background noise and intelligibility-

Is it significantly more difficult to hold a conversation in a quiet room in the day than it is to hold one in a quiet room at night? 

 

Interesting thought. You mean not being able to see the speaker's face in the dark - and when replaying over an audio system?  Or were you thinking of something else?

[audiventory]

19 minutes ago, adamdea said:

I do however feel confident you will struggle to hear 16 bit quantisation noise

 

In my opinion, audible or inaudible noise is not matter here: signal/noise ratio make sound quality.

 

Example:

 

[We have white paper. White font is invisible. Black one is visible.]

 

We can replace "white" to "yellow" or "blue" or other. It don't change change of 3 sentences above.

[Ralf11]

IIRC, mp3 development relied on the 'ability' of the human ear to mask sounds in various situations.

 

I don't advocate people using mp3 but a look at that R&D might help answer these questions.

[esldude]

Okay these are FLAC files.  - 40 dbFS pink noise in one and -60 dbFS pink noise in the other. 

 

 

06 Heat Wave 1.flac

 

06 Heat Wave 2.flac

[esldude]

2 hours ago, adamdea said:

snip...

 

What is the level below peak of self noise in most mics -80dB? Genuine question- I’m not sure but I have that figure in the back of my mind. 

Depends upon the microphone, recording, mixing etc. 

 

Plenty of condenser microphones have a self noise equivalent to 10 dbSPL, and can stay at less than 1% THD to 140 db.  Of course any preamp gain will amplify that noise.  The more limiting factor is where are you going to record that has a noise level down at the same low leveL?  Also the distribution of noise can be such that in the midrange 3-5 khz it might be that low. 

 

Let's assume a good condition.  I think you'll need 30 db gain to record say a good orchestra which raise that 10 dbSPL equivalent to 40 dbSPL.  You'll probably have that much noise in the room anyway.  So you're lucky to get 70-80 db dynamic range above the noise floor.  Again at midrange frequency that region can be fairly low in level.  

[danadam]

2 hours ago, esldude said:

Okay these are FLAC files.  - 40 dbFS pink noise in one and -60 dbFS pink noise in the other. 

 

 

06 Heat Wave 1.flac

 

06 Heat Wave 2.flac

Ha! So the noise in that mp3 wasn't at -40 dBFS, more like -30 dBFS?

[mansr]

21 minutes ago, danadam said:

Ha! So the noise in that mp3 wasn't at -40 dBFS, more like -30 dBFS?

Huh?