Audibility thresholds of amp and DAC measurements

[j_j]

Hi @j_j
Apparently the quietest anechoic chamber is at Microsoft and it was measured at -20 dB(A) SPL. Do you know how to correctly interpret and compare those numbers?
(I guess maybe A-weighting can account for some of the difference)

They estimate the EXCESS NOISE by subtracting the "atmospheric noise". Yeah, that's kind of hinky. The atmospheric noise will be there at 1 atmosphere pressure. Always.

I'm quite aware of how the room is spec'ed and built. There's no magic, it's a way to make impressive numbers. Weighting would have some effects, but they can calculate the actual noise floor, and then they measure how much above the noise floor the room is, and subtract the thermal noise floor to find the excess noise.

Also, of course, if you measure in any one ERB, you'll get a smaller number. That would also be, well, true but misleading.

 

[DonH56]

<elided>
Again, tonal vs. narrow-band-noise is not very different, but yes, you can, at your ear canal resonance, get into small negative SPL's, but only if you're young and your organ of corti is un-blasted-to-bits.

Well, crud, I lose on both counts...

 

[j_j]

Well, crud, I lose on both counts...

Yeah, welcome to the modern world, where the noise level on the street exceeds the normal indoor talking level. Yeah, I measure that s**t.

 

[NTK]

This may be a little off topic. I had been looking for this Auditory Demonstration CD by the Institute for Perception Research (IPO) at Eindhoven University of Technology (TU/e), Northern Illinois University, supported by the Acoustical Society of America.

Auditory demonstrations on compact disc

In 1984, the Committee on Education in Acoustics of the ASA discussed the possibility of reissuing the “Harvard Tapes,” a series of audio demonstrations on tape

pubs.aip.org

I finally found it at the TU/e site. Free download!

Auditory demonstrations

research.tue.nl

 

[Hayabusa]

So is narrow-frequency noise. You do understand that the 6dB is white noise, across 20-20K, yes? So the energy in any 1 ERB is much smaller.

I would suggest you go watch through my "hearing tutorial" at www.aes.org/sections/pnw in the "meeting recaps" area. This may help you to realize you just compared an apple to a grapefruit.

nothing wrong with that, in this case we ARE comparing apples and grapefruits. Wideband noise and tonal information. A 6dB SPL white noise level suggest for others maybe that that's the absolute lower limit, but its not. That's why I made my statement. Maybe someone else found it useful.

 

[abm0]

at your ear canal resonance, get into small negative SPL's, but only if you're young and your organ of corti is un-blasted-to-bits.

Hmm, Wikipedia gives me "The threshold of hearing is frequency-dependent and it has been shown that the ear's sensitivity is best at frequencies between 2 kHz and 5 kHz, where the threshold reaches as low as −9 dB SPL." after I tried one of those geepeetees and it answered something completely stupid that was positive numbers in dB.

But anyway this should all be made irrelevant by environmental noise anyway. The lowest I've measured with some old-school indicator needle second-hand sound level meter (of admittedly unknown calibration) in my residential-area apartment in the dead of night was 20 dBA, so I should probably demand no better than about 10 dBA of unwanted content out of my reproduction chain since that adds up to no more than 20.4 or so dBA and should be just below the threshold of being able to detect a difference vs. pure environmental noise.

 

[Blumlein 88]

Hmm, Wikipedia gives me "The threshold of hearing is frequency-dependent and it has been shown that the ear's sensitivity is best at frequencies between 2 kHz and 5 kHz, where the threshold reaches as low as −9 dB SPL." after I tried one of those geepeetees and it answered something completely stupid that was positive numbers in dB.

But anyway this should all be made irrelevant by environmental noise anyway. The lowest I've measured with some old-school indicator needle second-hand sound level meter (of admittedly unknown calibration) in my residential-area apartment in the dead of night was 20 dBA, so I should probably demand no better than about 10 dBA of unwanted content out of my reproduction chain since that adds up to no more than 20.4 or so dBA and should be just below the threshold of being able to detect a difference vs. pure environmental noise.

Devil in the details. That single number is presumably 20hz to 20 khz. Most environmental noise is pinkish to brownish. Most of the noise energy is in the lower frequencies. So even with a measured 20 dbA, in that 2-5 khz range it might be 10 dbA or even close to 0 dbA. One way to model this is ERBs or effective rectangular bandwidth. Your ear kind of divides up parts of the bandwidth which allows it to hear a bit below the overall noise level. This is simplified as your ear can shift these bands up and down around what it hears. Roughly into maybe 30 bands. So the noise level meter might be 20 dbA, but in the few bands around that 2-5 khz range it might only be a few db or close to zero db SPL and your ear even in that noise could hear things in that range at that level.

This might lead one to hear lower level noise than you anticipate, but on the other hand with music playing it probably masks it pretty completely. If you've ever heard an LP, you can hear some of the surface noise of the vinyl LP fairly easy in the lead in grooves or silent grooves between tracks, but when the music starts you cannot hear it.

 

[Hayabusa]

Devil in the details. That single number is presumably 20hz to 20 khz. Most environmental noise is pinkish to brownish. Most of the noise energy is in the lower frequencies. So even with a measured 20 dbA, in that 2-5 khz range it might be 10 dbA or even close to 0 dbA. One way to model this is ERBs or effective rectangular bandwidth. Your ear kind of divides up parts of the bandwidth which allows it to hear a bit below the overall noise level. This is simplified as your ear can shift these bands up and down around what it hears. Roughly into maybe 30 bands. So the noise level meter might be 20 dbA, but in the few bands around that 2-5 khz range it might only be a few db or close to zero db SPL and your ear even in that noise could hear things in that range at that level.

This might lead one to hear lower level noise than you anticipate, but on the other hand with music playing it probably masks it pretty completely. If you've ever heard an LP, you can hear some of the surface noise of the vinyl LP fairly easy in the lead in grooves or silent grooves between tracks, but when the music starts you cannot hear it.

Don't forget the ears ability to hear constant tones well below the noise level also.

 

[abm0]

you can hear some of the surface noise of the vinyl LP fairly easy in the lead in grooves or silent grooves between tracks, but when the music starts you cannot hear it.

Eehhh let's just say for the purpose of buying an "endgame" DAC-amp I will choose to not rely on masking. A lot of discussion on this topic treats music as always always busy stuff with heaps of active frequencies but I listen to pretty chill stuff, a lot of it electronic, and encounter far more minimalistic sections than, say, a classical or jazz aficionado would. Pure-tone analysis is more relevant to me and masking less relevant.

 

[j_j]

Hmm, Wikipedia gives me "The threshold of hearing is frequency-dependent and it has been shown that the ear's sensitivity is best at frequencies between 2 kHz and 5 kHz, where the threshold reaches as low as −9 dB SPL." after I tried one of those geepeetees and it answered something completely stupid that was positive numbers in dB.

But anyway this should all be made irrelevant by environmental noise anyway. The lowest I've measured with some old-school indicator needle second-hand sound level meter (of admittedly unknown calibration) in my residential-area apartment in the dead of night was 20 dBA, so I should probably demand no better than about 10 dBA of unwanted content out of my reproduction chain since that adds up to no more than 20.4 or so dBA and should be just below the threshold of being able to detect a difference vs. pure environmental noise.

Well, that's why I mentioned the ear canal resonance, which is usually the ERB with the highest sensitivity. Of course it's frequency dependent.

There are basic physics issues, in particular having standard air on both sides of the eardrum at STP means you have something in the 6 to 8.5 dB SPL white noise present due to the existence of the actual atmosphere.

Since the ear is very frequency sensitive, you can't hear that. Barely. At the ear's most sensitive point, a person in a quiet place can ALMOST hear that. Those of us in their 70's, not so much.

 

[bachatero]

Why bother with DACs with THD+N 85+dB if nobody can hear the difference?

 

[dshreter]

Why bother with DACs with THD+N 85+dB if nobody can hear the difference?

Cynically, because people are very willing to pay for it

Practically, because it can be achieved with good engineering at a nominal cost

Philosophically, there is no reason

 

[Jimster480]

Why bother with DACs with THD+N 85+dB if nobody can hear the difference?

What are you referring to? Because 85 is pretty low. So I would say that there's definitely audibility differences that can be heard, especially with a trained ear. Up to around 105, I would say. Outside of that, I personally haven't been able to hear any difference no matter what.

 

[bachatero]

What are you referring to? Because 85 is pretty low. So I would say that there's definitely audibility differences that can be heard, especially with a trained ear. Up to around 105, I would say. Outside of that, I personally haven't been able to hear any difference no matter what.

I don't know about you, but a few hours ago I did a test to see how low distortion I could hear. So, I generated some very quiet 2kHz, 3kHz, 4kHz, and 5kHz tones in Audacity and adjusted the overall levels until I couldn't hear them anymore in my quiet room with Kali LP-6v2 monitors. That ended up being around -80dB at max volume, and the 1kHz tone at 0dB was really loud. That would mean 85dB performance would actually be 5dB better than needed, so I have no idea how people are somehow hearing differences past that.

 

[Jimster480]

I don't know about you, but a few hours ago I did a test to see how low distortion I could hear. So, I generated some very quiet 2kHz, 3kHz, 4kHz, and 5kHz tones in Audacity and adjusted the overall levels until I couldn't hear them anymore in my quiet room with Kali LP-6v2 monitors. That ended up being around -80dB at max volume, and the 1kHz tone at 0dB was really loud. That would mean 85dB performance would actually be 5dB better than needed, so I have no idea how people are somehow hearing differences past that.

Well, the thing is that you're using speakers, and speakers are trash, so that's probably the reason why you can't hear anything above 80.
If you use high-end headphones, then you can get far past 80. That's why I don't even waste time to listen to music on speakers unless it's out in the world. And that's also why smart speakers and sound bars are so popular, because speakers are for convenience, not fidelity.

 

[Blumlein 88]

Well, the thing is that you're using speakers, and speakers are trash, so that's probably the reason why you can't hear anything above 80.
If you use high-end headphones, then you can get far past 80. That's why I don't even waste time to listen to music on speakers unless it's out in the world. And that's also why smart speakers and sound bars are so popular, because speakers are for convenience, not fidelity.

Prove it! I don't believe you.

Not trying to be mean, but there are few if any people who can hear -80 db distortion, and you won't hear that or anywhere near it with music.

 

[Jimster480]

Prove it! I don't believe you.

Not trying to be mean, but there are few if any people who can hear -80 db distortion, and you won't hear that or anywhere near it with music.

How can I prove it? I mean part of how this site became popular is due to me "hearing it" and causing a whole big drama in the audio industry with Schiit and such in 2018.
Amir was the guy who believed me in what I could hear and he did the testing to prove it. This caused a whole shake up in the industry and helped to push the visibility of ASR around the net, along with me and others sharing this data on sites like Massdrop to prove that these other products actually had high performance.
So I would like to know how I can "prove it".

 

[danadam]

Prove it! I don't believe you.

Not trying to be mean, but there are few if any people who can hear -80 db distortion, and you won't hear that or anywhere near it with music.

Did @bachatero describe hearing -80 dB distortion or -80 dBFS tone? I'm not sure.

 

[bachatero]

Did @bachatero describe hearing -80 dB distortion or -80 dBFS tone? I'm not sure.

I listened to just the -80 "dB" attenuation tones so just the distortion components, and then I compared those to the 1kHz tone by enabling that at 0 attenuation.

 

[j_j]

Why bother with DACs with THD+N 85+dB if nobody can hear the difference?

In a quiet living room, one can make an argument for about 100 dB range or so. But your point still holds. I calculated an accurate number a while ago, but it's more like a bit less than 18 bits.