Audibility thresholds of S/N - signal to noise - test

[Sokel]

I did got a -70 but I had to really try,headphones this time (Audeze LCD-X) :

The key is really rapid changes.

 

[DVDdoug]

As Amir says, it not really about the ratio, but we need a reference of some kind unless we are measuring the acoustic noise at our ears.

A S/N of 40 dB might be enough and 60dB is almost certainly enough when the music/program is at full volume and drowning-out the noise.

It gets really complicated because if the noise is coming before the volume control the noise gets turned up with the volume control and becomes easier to hear (especially between songs or during quiet parts).

And audibility depends on the sensitivity of your speakers or headphones, how close you re to the speakers, and other acoustic noise in the room.

...I believe Amir measures power amplifiers relative to 5 Watts (and maybe he has standard references for line level too) so that a 500W amplifier doesn't get a better score than a 50W amp, simply by having a higher "signal".

 

[pablolie]

A high percentage of audiophile discussions go...

• "Wow this device has the best SINAD ever, awesome"... then someone counters...
• "But what's the point - no one can hear that!"
• Then someone asks for clarification "So how much is good enough?" (which is always YMMV) ...or...
• "How much is too much? Isn't this a case of complete over-engineering?"

In the end, everybody keeps their original opinion.

On the S/N front, like many said, to a large degree it depends on your environment. If I take the test above while riding Caltrain to San Francisco, I'll get very different results than if I do it in my listening room at home (very quiet when the cats behave). And if I moved 300ft underground into an abandoned salt mine with an anechoic room, my needs would again change, most likely. Plus at some point in time other measurements become more relevant than the quest for the highest S/N.

 

[Taiga]

This is exactly my problem with noise. Cannot stand it when nothing is playing. While sound is playing, sure, don't need ultra low levels.

That’s when everyone coughs during a symphony.

 

[Salt]

A high percentage of audiophile discussions go...

• "Wow this device has the best SINAD ever, awesome"... then someone counters...
• "But what's the point - no one can hear that!"
• Then someone asks for clarification "So how much is good enough?" (which is always YMMV) ...or...
• "How much is too much? Isn't this a case of complete over-engineering?"

In the end, everybody keeps their original opinion.

On the S/N front, like many said, toa large degree it depends on your environment. If I take the test above while riding Caltrain to San Francisco, I'll get very different results than if I do it in my listening room at home (very quiet when the cats behave). And if I moved 300ft underground in an abandoned salt mine with an anechoic room, my needs would again change, most likely. Plus at some point in time other measurements become more relevant than the quest for the highest S/N.

I would short it to: SINAD as low at possible for a comfortable price without compromise to comfort (longevity).
Not?

 

[pablolie]

I would short it to: SINAD as low at possible for a comfortable price without compromise to comfort (longevity).
Not?

I would totally sign off on that, but then again we also know the "comfortable price" is a huge sliding scale depending on the particular customer.

Personally, feature sets that cater to my needs seem to be the key differentiators once measured specs are "good enough" or even "overkill" (in many cases that comes easily).

 

[Mike123]

And if I moved 300ft underground into an abandoned salt mine with an anechoic room, my needs would again change, most likely.

I have a noise meter. Early in the morning with the windows closed and holding my breath I get 33 dBA of noise. Listening is measured as 60-70 dBA. So in reality, in terms of pressure, I get the need for 37 dBA of dynamic range. Okay, 53 dBA to get the measured 86 dBA. But it doesn't last long because it's uncomfortable. Isn't that why people have been enjoying music recordings for almost a hundred years, even though the noise level used to be much higher?

 

[pablolie]

I have a noise meter. Early in the morning with the windows closed and holding my breath I get 33 dBA of noise. Listening is measured as 60-70 dBA. So in reality, in terms of pressure, I get the need for 37 dBA of dynamic range. Okay, 53 dBA to get the measured 86 dBA. But it doesn't last long because it's uncomfortable. Isn't that why people have been enjoying music recordings for almost a hundred years, even though the noise level used to be much higher?

That is a factor I can subscribe to.
Another thing I notice basically every day is that the volume of the music I had last night sounds much louder the next morning. And it didn't seem loud (and wasn't), but once my ears and brain get some sleep, the thresholds seem to change. At least for me.

 

[Sokel]

Noise tests like that are better done while concentrated not at the main tone but at the sense of space each change gives.

At the headphone test -70 feels like it broadens the tone slightly compared to the pure tone. It's not like I hear the actual noise clearly.
I only know it's there because of the comparison.

Level matters too, it has a sweet spot at my test, loud but not too loud, I would guess 75-80dB (C) ?

(not an enjoyable test by all means)

 

[amirm]

I have a noise meter. Early in the morning with the windows closed and holding my breath I get 33 dBA of noise. Listening is measured as 60-70 dBA. So in reality, in terms of pressure, I get the need for 37 dBA of dynamic range. Okay, 53 dBA to get the measured 86 dBA. But it doesn't last long because it's uncomfortable. Isn't that why people have been enjoying music recordings for almost a hundred years, even though the noise level used to be much higher?

There is no such thing as a single noise number, or single loudness level. Both have spectrum and that spectrum determines how audible/loud they are. I have covered this topic in detail in this video:

As to older equipment, it was annoyingly noisy. It was that noise that forced the migration to digital.

 

[pma]

I did got a -70 but I had to really try,headphones this time (Audeze LCD-X) :

Great! No cheating? Deltawave creates quite inaudible delta noise file when not turning up the volume from the normal listening level. And also the test should be done somewhere inside the files, not at the beginning, because noise addition makes discernable initial transients when you start the files from zero time. This may be the case during fast switching between the files as well.

 

[pma]

As to older equipment, it was annoyingly noisy.

In my direct studio experience (1977, Supraphon), Studer A80 was reasonably quiet machine, especially with Dolby A. There are hundreds of recordings proving that.

https://tile.loc.gov/storage-services/master/mbrs/recording_preservation/manuals/Studer%20A80-RC%20Professional%20Tape%20Recorders.pdf

 

[amirm]

In my direct studio experience (1977, Supraphon), Studer A80 was reasonably quiet machine, especially with Dolby A. There are hundreds of recordings proving that.

https://tile.loc.gov/storage-servic...Studer A80-RC Professional Tape Recorders.pdf

I very much doubt that given the SNR of ~72 dB with 3% distortion.

 

[danadam]

Apple US dongle + Truthear Zero Blue

Spoiler: abx result

foo_abx 2.2.1 report
foobar2000 v2.24.5
2025-07-31 11:00:40

File A: test200_pure.wav
SHA1: d7fff78fe16f3d5974690a2a4510053f08cac287
File B: test200_-70.wav
SHA1: bf8fa05f934aec054ce37c081ce6e95f4c88dd80

Output:
Default : Headphones (USB-C to 3.5mm Headphone Jack Adapter) [exclusive], 24-bit
Crossfading: NO

11:00:40 : Test started.
11:00:59 : Test restarted.
11:00:59 : 01/01
11:01:04 : Test restarted.
11:01:04 : 02/02
11:01:08 : Test restarted.
11:01:08 : 03/03
11:01:13 : Test restarted.
11:01:13 : 04/04
11:01:17 : Test restarted.
11:01:17 : 05/05
11:01:21 : Test restarted.
11:01:21 : 06/06
11:01:24 : Test restarted.
11:01:24 : 07/07
11:01:28 : Test restarted.
11:01:28 : 08/08
11:01:31 : Test restarted.
11:01:31 : 09/09
11:01:35 : Test restarted.
11:01:35 : 10/10
11:01:38 : Test restarted.
11:01:38 : 11/11
11:01:41 : Test restarted.
11:01:41 : 12/12
11:01:45 : Test restarted.
11:01:45 : 13/13
11:01:48 : Test restarted.
11:01:48 : 14/14
11:01:52 : Test restarted.
11:01:52 : 15/15
11:01:55 : Test restarted.
11:01:55 : 16/16
11:01:55 : Test finished.

----------
Total: 16/16
p-value: 0 (0%)

-- signature --
ed0c6ea5a987ecdef0c847d368d083750b8c43bf

( https://www.foobar2000.org/abx/signaturecheck )

As far as programme is concerned, here are 2 excerpts from the first track of Sibelius - The Seven Symphonies (in attachments). The quiet part is from 0:41 and the loud part from 8:16 of the full track:

The "16b" version is with 16-bit dither and "12b" is with 12-bit dither, so with about -72 dBFS RMS noise added. At the volume I can comfortably listen to it, the noise in the quiet part is easy to distinguish.

 

[Sokel]

Great! No cheating? Deltawave creates quite inaudible delta noise file when not turning up the volume from the normal listening level. And also the test should be done somewhere inside the files, not at the beginning, because noise addition makes discernable initial transients when you start the files from zero time. This may be the case during fast switching between the files as well.

No, no cheating (how could I cheat that?)
No sudden level changes too that I could get if that's what you're saying about transients.

I can't say it was very loud too, E-MU cannot drive Audeze loud (same with my similar ancient at-ATH-M2 as they have similar impedance, about 27 Ohm) .
Maybe this interaction as E-MU's spec is as high as 22 Ohm?
Shouldn't be a problem for FR though.

Maybe I should try with my old Salas DCG3?

 

[danadam]

And also the test should be done somewhere inside the files, not at the beginning, because noise addition makes discernable initial transients when you start the files from zero time. This may be the case during fast switching between the files as well.

No sudden level changes too that I could get if that's what you're saying about transients.

Foobar's ABX comparator applies fade-in and fade-out when switching between the files and also when starting/ending playback but only if the start/end position is set to something different than the files' start/end.

Unfortunately it doesn't apply the fade when staring/ending playback if the start/end positions are not changed (i.e. full files is played). If the files don't have fade-in/out baked in, this will create discontinuity click when starting/ending playback. I'm guessing that pma's worry is that adding noise to one file may have audibly changed the character of this discontinuity click.

Though the actual problem with these specific files is that they are different at the start/end, regardless of noise. Looks like some kind of fade-in/out was applied to the file with the noise but not to the "pure" one:

 

[pma]

There was no fade in/out applied, it is a result of mixdown of 200Hz sine with noise signal, some initial transient product. But I believe there is no reason of cheating and the test interval may be chosen inside the files. The goal is to find signal to noise audibility, not to be cheating with several samples at the beginning. BTW, Audacity is not the best tool to show it.

 

[rolfrander]

I'm not sure I did this correctly, I could identify X without listening to A or B:

HK80 connected to FOSI SK02

foo_abx 2.2 report
foobar2000 v2.1.4
2025-07-31 13:13:19

File A: test200_-70.wav
SHA1: bf8fa05f934aec054ce37c081ce6e95f4c88dd80
File B: test200_pure.wav
SHA1: d7fff78fe16f3d5974690a2a4510053f08cac287

Output:
Default : Hodetelefoner (Fosi Audio SK02)
Crossfading: NO

13:13:19 : Test started.
13:13:25 : Test restarted.
13:13:25 : 01/01
13:13:31 : Test restarted.
13:13:31 : 02/02
13:13:36 : Test restarted.
13:13:36 : 03/03
13:13:45 : Test restarted.
13:13:45 : 04/04
13:13:50 : Test restarted.
13:13:50 : 05/05
13:13:57 : Test restarted.
13:13:57 : 06/06
13:14:01 : Test restarted.
13:14:01 : 07/07
13:14:08 : Test restarted.
13:14:08 : 08/08
13:14:12 : Test restarted.
13:14:12 : 09/09
13:14:17 : Test restarted.
13:14:17 : 10/10
13:14:21 : Test restarted.
13:14:21 : 11/11
13:14:27 : Test restarted.
13:14:27 : 12/12
13:14:31 : Test restarted.
13:14:31 : 13/13
13:14:37 : Test restarted.
13:14:37 : 14/14
13:14:48 : Test restarted.
13:14:48 : 15/15
13:14:53 : Test restarted.
13:14:53 : 16/16
13:14:53 : Test finished.

 ----------
Total: 16/16
p-value: 0 (0%)

 -- signature --
ab20f799b7bb3527bc54f01459ef67785a489ba2

 

[Sokel]

Foobar's ABX comparator applies fade-in and fade-out when switching between the files and also when starting/ending playback but only if the start/end position is set to something different than the files' start/end.

Unfortunately it doesn't apply the fade when staring/ending playback if the start/end positions are not changed (i.e. full files is played). If the files don't have fade-in/out baked in, this will create discontinuity click when starting/ending playback. I'm guessing that pma's worry is that adding noise to one file may have audibly changed the character of this discontinuity click.

Though the actual problem with these specific files is that they are different at the start/end, regardless of noise. Looks like some kind of fade-in/out was applied to the file with the noise but not to the "pure" one:
View attachment 466683
View attachment 466684

That's my settings, as always, I choose a small duration so I can fast rapid them.

Now that you say it though, there is a low "thumb"-like sound at changes but its the same at any combination (tried them all) and it's more audible with speakers.

 

[rolfrander]

I'm not sure I did this correctly, I could identify X without listening to A or B:

HK80 connected to FOSI SK02

Redoing the test with Salnotes Zero was eaven easier, I could pick identify the X of each test alone without comparing to A or B. To my ears the samples have a different timbre which I find easy to recognize. I don't know if this is "cheating" or not, as per the discussion above.