Audibility thresholds of amp and DAC measurements

[bachatero]

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.

What is your calculation?

 

[j_j]

Simply compare noise floor of a quiet listening room with peak level available from a standard good loudspeaker. Not rocket science.

 

[Blumlein 88]

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

Bachatero heard a low level tone. Very doable. Then it seemed to me jimster480 was conflating that with the ability to hear -85 db distortion. Maybe I misread it. One reason I don't like SINAD or its flipside THD+N. Keep the N and THD separate. At a minimum I was very clear in my reply where I said -80 db distortion.

 

[abm0]

When pure sine waves are reproduced under ideal conditions and at very high volume, a human listener will be able to identify tones as low as 12 Hz

Also, tones as high as 28 kHz. Highest number I've seen reported so far.

Hearing thresholds for pure tones above 16kHz

Hearing thresholds for pure tones between 16 and 30kHz were measured by an adaptive method. The maximum presentation level at the entrance of the outer ear was

pubs.aip.org

 

[j_j]

Also, tones as high as 28 kHz. Highest number I've seen reported so far.

Hearing thresholds for pure tones above 16kHz

Hearing thresholds for pure tones between 16 and 30kHz were measured by an adaptive method. The maximum presentation level at the entrance of the outer ear was

pubs.aip.org

The question I have is "what level"? Skin sensation for ultrasonics at high levels is old news, but it's not "hearing" per se.

Also, at 99dB SPL the air isn't linear any more. Doesn't matter what frequency, there will be subharmonics, harmonics, and all sorts of other nonlinearities.

 

[Hayabusa]

The question I have is "what level"? Skin sensation for ultrasonics at high levels is old news, but it's not "hearing" per se.

Also, at 99dB SPL the air isn't linear any more. Doesn't matter what frequency, there will be subharmonics, harmonics, and all sorts of other nonlinearities.

ChatGPT answered:

Nonlinearity Threshold for 20 kHz and 30 kHz​

• The threshold for 1% distortion scales inversely with frequency because waveform steepening occurs faster for shorter wavelengths.
• At 20 kHz, the SPL threshold for 1% distortion is typically around 145–150 dB SPL.
• At 30 kHz, it is slightly lower, approximately 140–145 dB SPL.

Is it making an error?

 

[NTK]

Also, don't forget distance attenuation. The listening distance was 0.5 m. The sound pressure at the transducer membrane surface could easily be 30+ dB higher than at the listening distance and went nonlinear there.

 

[Hayabusa]

Also, don't forget distance attenuation. The listening distance was 0.5 m. The sound pressure at the transducer membrane surface could easily be 30+ dB higher than at the listening distance and went nonlinear there.

View attachment 420203

 

[j_j]

ChatGPT answered:

Nonlinearity Threshold for 20 kHz and 30 kHz​

• The threshold for 1% distortion scales inversely with frequency because waveform steepening occurs faster for shorter wavelengths.
• At 20 kHz, the SPL threshold for 1% distortion is typically around 145–150 dB SPL.
• At 30 kHz, it is slightly lower, approximately 140–145 dB SPL.

Is it making an error?

Not an error, a misunderstanding. 1% distortion is -40dB A subharmonic resulting from a 25kHz tone at 90 dB that's in the ear canal resonance range can be audible at 10dB SPL. That's 80dB, not 40dB. At 140dB SPL that's 100dB SPL for the subharmonic. That's earsplitting itself.

NOW, most of the problem will be in harmonics, not subharmonics, so it's not as bad as that, *but* it is important to be careful with the spectrum of the added distortion. Subharmonics are weird in that it's hard as(*&( to figure out exactly where they might land.

 

[Hayabusa]

Also, don't forget distance attenuation. The listening distance was 0.5 m. The sound pressure at the transducer membrane surface could easily be 30+ dB higher than at the listening distance and went nonlinear there.

View attachment 420203

indeed you are right, chatGPT estimates this difference to be 35dB @ 27KHz for 50cm.
So it is getting close to the non linear threshold of 140dB.
27KHz 105dB SPL @ 50cm should still be a valid level for such a test then.

 

[j_j]

They weren't using a compression driver, were then? I'm not handy to read the paper right now, but distortion in compression driver throats is often epic.

Also, remember that distortion does not START at the level you gave above, it reaches 1% there. Much less than 1% can still lead to surprising artifacts.

 

[Hayabusa]

They weren't using a compression driver, were then? I'm not handy to read the paper right now, but distortion in compression driver throats is often epic.

Also, remember that distortion does not START at the level you gave above, it reaches 1% there. Much less than 1% can still lead to surprising artifacts.

it was a PIONEER PT-R100 ribbon tweeter.
But as you can see in posts before the team that did the test made sure there were so (sub)harmonics above 20dB SPL at the listening position.

 

[j_j]

it was a PIONEER PT-R100 ribbon tweeter.
But as you can see in posts before the team that did the test made sure there were so (sub)harmonics above 20dB SPL at the listening position.

Hmmm. I have questions, but not directed to you, sorry.

 

[Snafu]

Amplifiers have this power vs distortion graph for different frequencies. When dac has volume control, is similar graph measurable ?
Or when measuring max out can we measure distortion for different frequencies and is it in audible level? Thanks

 

[j_j]

Amplifiers have this power vs distortion graph for different frequencies. When dac has volume control, is similar graph measurable ?
Or when measuring max out can we measure distortion for different frequencies and is it in audible level? Thanks

Step one. Is the DAC volume control digital or analog? If it's modifying level before the DAC, say like windows level settings for the most part, you will see the noise floor of the dac seem to rise as level goes down.

If the gain is analog, it's not entirely clear. There will be a noise floor due to the DAC, and an overload point due to the DAC. Then there will be the other noise due to the analog outputs. This can interact in a number of ways, usually not completely strange, but if the analog section is really good for a 16 bit DAC, you may scale the DAC noise level up and down without the analog system interferring. The question is "where does the gain control change the sytem gain" and "what is the analog noise floor" and "how much distortion is there".

 

[Rja4000]

For IMD, the same distortion thresholds as for FFT apply: -120 dBFS for the strict and -66 dBFS for the lenient.

Without discussing the founding of all your figures this one is obviously wrong.

IMD SMPTE is measured in relation to the -12dB high frequency component.
Therefore, it makes sense to lower your thresholds accordingly.