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

[pma]

It is not cheating if your result is not based on testing just at the beginning or just at the end of the files. Another kind of cheating would be to use a tweeter only, or a cell phone speaker only, which would lead to suppression of the 200 Hz tone. Web tests can always be cheated somehow, but I hope it is not the goal. The goal is to show the readers what is the audible impact of increasing / decreasing S/N. The test is supposed to be performed on the main home audio system.

 

[danadam]

There was no fade in/out applied, it is a result of mixdown of 200Hz sine with noise signal, some initial transient product.

If you say so, though I don't see how that could be possible. The "pure" signal starts almost at 0.1 (-20 dBFS) and "-70" signal starts at 0. Even -40 dBFS RMS noise can't do that much change, let alone -70 dBFS RMS.

BTW, Audacity is not the best tool to show it.

IMO good enough in this case.

and the test interval may be chosen inside the files.

Sure, that's what I did.

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.

With start position at 0 and "Keep playback position" unchecked there's a distinct difference in the initial click for me.

Similarly if I set the end position at the end and the start position shortly before it. Then the click at the end is different.

 

[rolfrander]

Opening the files in REW, there seems to be quite a bit harmonic distortion. I am wondering if it is this I can hear, and not the noise. However, the magnitude seems to be the same. I tried using the EQ in foobar to remove the main tone and just listen for the harmonics, but that made the noise quite audible.

 

[Sokel]

With start position at 0 and "Keep playback position" unchecked there's a distinct difference in the initial click for me.

Similarly if I set the end position at the end and the start position shortly before it. Then the click at the end is different.

No click for me, the sound I hear is more like a low port thumb, as if the driver (woofer for me, crossed to 250Hz at midrange, active) attacks and then settles instantly.
It's faint but it's there at all combinations.

I'll check if I hear it with headphones too.

Edit: yep, with headphones too.

 

[danadam]

Opening the files in REW, there seems to be quite a bit harmonic distortion. I am wondering if it is this I can hear, and not the noise.

It's a bit strange but they are in both files so it shouldn't affect anything. Simply subtracting one file from the other shows attenuated main tone and only noise:

The presence of the attenuated main tone is strange too, but it also shouldn't affect anything. At this level it translates to only 0.01 dB difference in the 2 files.

No click for me, the sound I hear is more like a low port thumb, as if the driver (woofer for me, crossed to 250Hz at midrange, active) attacks and then settles instantly.

In attachment there's a capture (through RME Adi-2 Pro) of me switching between A and B, starting at position 0 and with "Keep playback position" unchecked:

and starting at position 0.000001:

 

[danadam]

I must admit that doing the ABX (in the middle of the files) with Sennheiser HD650 would be more difficult than with Truthear Zero Blue or Sennheiser HD 380 Pro.

 

[pma]

Direct subtraction of test200_pure.wav minus test200_-70.wav files. First 1600 samples shown. Please note the difference in the first 50 samples, as a result of softwre mixdown of the sine and noise files. This is what is audible at the beginning and similarly at the end of the files during ABX.

 

[tmtomh]

Noise is always important only when related to useful content. That's why I keep the same 200Hz tone in all samples. Otherwise one would be tempted to turn up volume until he would hear the noise. It does not make sense.

One can also turn volume down and repeat the same test with posted files. This says how S/N audibility is related to programme volume.
Regarding 200Hz, I have chosen it because the ear is less sensitive than to 1kHz - 3kHz range of frequencies and thus the noise is more audible. Of course the speaker is to be able to reproduce 200Hz, cell phones are prohibited in this test, it would be cheating.

Yes, but @amirm 's point is that silence is useful content when it comes between two tracks of music (and for that matter, when it comes in the form of a pause or rest within a single piece of music).

To be clear, I understand your point here - for the purposes of testing what you have set out to test, we of course need a primary musical signal so that the volume of the noise remains relative to a typical, or at least sane and tolerable, signal volume.

I think your test here is very useful as far as it goes, and I thank you for it. However, it is limited in its usefulness because there is some masking effect, even of a 200Hz signal, and our music listening experience includes silence.

Again, I am not saying this is a bad test or that you should not have set this up. I am only saying that this test does not actually demonstrate the audibility threshold of noise for the purpose of music listening.

One interesting variation on this test would be to provide the same file with the same 200Hz signal, but with a few second of silence in the middle. So folks could start the clip using the 200Hz sound as their volume reference, and then see if their audibility threshold for the noise became more sensitive during the silent part.

P.S. I just had a very quick, non-blind/uncontrolled listen to the files on my iMac - so just computer speakers. Noise was easily audible down to -50dB. I could not detect -70, and I'm pretty sure I couldn't detect -60 either. I'd have to try a proper blind test on my main speakers and/or with headphones to determine if I could hear -60 or -70. (I don't claim to have exceptional hearing - I just don't know since I haven't tried it yet.)

 

[rdenney]

Listing through cheap computer headphones with office background noise, I could not reliably distinguish between -60, -70, and the clean sine wave.

Amir is right that we expect silences to be silent, but in the real world there is no such thing--the gaps in the music are filled with ambient room noise. That can be gated off but that has always been very difficult to hide except maybe for the spaces between tracks. I've tried that with live recordings, with the gate set to -60 dB and even -70 dB to eliminate obvious ambient noise when no music is playing, but those experiments have all failed. Ambient noise is, in fact, natural and can be conspicuous by its absence.

I spent some time listening to the sine wave produced by my HP 339A distortion analyzer, which is clean to at least 100 dB SINAD and probably a bit better than that. I compared it to another function generator (a Tektronix) with a measured THD of -45 dB (voltage was measured and matched, of course) and the difference even when played through an ancient cheap stereo glowed in the dark. (The stereo integrated amp itself tested to about 60 dB SINAD as I recall.) Yet I cannot hear harmonic distortion in actual music to levels below about -40 dB.

I can clearly hear the buzz being produced by something in my system to a level that is about -80 dB, but that buzz has harmonic content right in the most sensitive part of hearing. I can hear it from 6 feet away in my speakers, but if I play music at the level the system is set for when the buzz is that audible, the music is going to be very loud and the buzz will become utterly undetectable. (I think the buzz is coming from my old B&K preamp, soon to be traded out for the recently tested Holman preamp and then put on the bench, so the truth will out.)

I think the only conclusion I can draw is that a clean sine wave does a really poor job of masking either noise or distortion, and no signal at all is quite good at exposing residual noise even at very low levels. But to the extent that residual noise sounds like room noise (i.e., broadbanded hiss such as from an air handler), I doubt it's the same issue as noise resulting from a stack of inharmonic sine waves produced by some leaky electronic component.

Rick "fun with measurements" Denney

 

[telecesne]

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.

Blasphemy - everyone knows migrating to digital you lose fidelity -
*sarcasm*

 

[pma]

A lot of misunderstanding here - as always.

 

[pma]

Opening the files in REW

Then please compare them to SNR that REW is also showing. The components you speak about are far below noise level. Noise level is not the floor you are seeing in spectrum, noise level is integrated through 20Hz - 20kHz band. You cannot hear THD 1kHz of 0.0014%, that is a nonsense. There is a big difference between narrow band noise spectral density a noise level in defined frequency range.

 

[rolfrander]

Then please compare them to SNR that REW is also showing. The components you speak about are far below noise level. Noise level is not the floor you are seeing in spectrum, noise level is integrated through 20Hz - 20kHz band. You cannot hear THD 1kHz of 0.0014%, that is a nonsense. There is a big difference between narrow band noise spectral density a noise level in defined frequency range.

Ok, thank you. There is a lot I don’t quite understand here. For example: why is there harmonic distortion in a synthetically generated digital signal? (I think the answer to this might change my understanding of harmonic distortion)

And also: I obviously didn’t pay attention to the scale here: I guess the distortion in speakers is 20-40 dB above what I saw in the signal, so clearly not relevant.

 

[pma]

File analysis in REW (files are 16-bit):

20dB SN

60dB SN

70dB SN

no noise added, 200 Hz sine

direct subtraction (difference) of 200Hz sine and SN70dB file

Sine 200Hz file and 200Hz/-70dBnoise files differ only in added noise level and residual difference of 200Hz sine, which makes about -60dB (0.1%) of the sine wave amplitude.

This is the difference wav file:

 

[voodooless]

An SNR test without the S, is not an SNR test, it would just ne an N-test... they can both be valid tests, depending on what exactly you want to show.

 

[amirm]

The components you speak about are far below noise level.

Regardless of its impact, why on earth would you use a distorted tone to begin with? Where did you get it from?

 

[pma]

An SNR test without the S, is not an SNR test, it would just ne an N-test... they can both be valid tests, depending on what exactly you want to show.

I agree, but I do not understand the point. The S/N (SNR) test only makes sense in the presence of the main signal, should it be a test tone or a music sample. I have chosen the 200Hz tone for the reason of its poor masking abilities and also that it would not be suppressed in level in a normal hifi speaker or headphone system. So, if the S/N of 60dB seems to be a limit in such test, I dare to say that S/N of 60dBr (relative to useful instantaneous signal level, which varies with music flow - not to full scale level!!!!) is enough to get unaffected listening.

 

[pma]

Regardless of its impact, why on earth would you use a distorted tone to begin with? Where did you get it from?

Any tone is distorted, depending how good is your analysis SW and measuring system. That is my point. Now we debate about -140dB "distortion" is better than -120dB distortion and we make charts based on this. Our speakers and headphones cannot do better than 0.01% order. Some are even debating about -300dB distortions (no kidding). It is absurd. We are here debating about H3 0.0014% , in 16-bit files.

The files are analog generated and 16-bit digitized. There is the origin of "distorted file". Anyway, always the same file with or without added noise. The S/N rule keeps holding.

 

[voodooless]

The S/N (SNR) test only makes sense in the presence of the main signal, should it be a test tone or a music sample.

That is the point, isn’t it the comment was more towards the “but the test doesn’t have a silence crowd”.

 

[JSmith]

I think we need to separate the audible from good engineering hygiene.

Just because we can't hear something doesn't mean it isn't there.

Audio devices should be designed to ensure minimal distortion and noise within the design limits.

When there are 2 devices to choose from, 1 with more noise and distortion than the other... why choose the 1 with more noise and distortion, all other aspects being equal?

People like Rob Watts claiming he can hear distortions @ -300dB is utter madness... yet it's often that "crowd" that say measurements don't matter.


JSmith