Audibility thresholds of amp and DAC measurements

[nick_l44.1]

Anything that is 85 dB down from that peak in my house won't be hearable

If I understand correctly, you believe that sounds below 110-85 = 25 dB, i.e., 10 dB below the noise level, will not be audible. I highly doubt that such strict mathematics can be applied to the integral noise level. Here is an example. The level of added white noise in it is -22.5 dBA, and the peaks of the music tones in the peak spectrum are below -42 dB. Nevertheless, the melody is still audible, and different instruments can even be recognized. This is a very simple example and can be easily repeated with different variations. In my experience, the audibility of tones is lost at about 30 dB below the noise level.
The peak spectrum of non noised music fragment:

The attached example has a fairly high noise level, so be careful.

 

[j_j]

If I understand correctly, you believe that sounds below 110-85 = 25 dB, i.e., 10 dB below the noise level, will not be audible. I highly doubt that such strict mathematics can be applied to the integral noise level. Here is an example. The level of added white noise in it is -22.5 dBA, and the peaks of the music tones in the peak spectrum are below -42 dB. Nevertheless, the melody is still audible, and different instruments can even be recognized. This is a very simple example and can be easily repeated with different variations. In my experience, the audibility of tones is lost at about 30 dB below the noise level.
The peak spectrum of non noised music fragment:
View attachment 486267
The attached example has a fairly high noise level, so be careful.

Well, some apples, some oranges, but what's missing in all of this is the actual noise and signal spectra. Until you know the ERB by ERB noise and signal levels, you haven't got much of anything to argue with, let alone about. If you have the masking level of the noise, then you can compare THAT masking level to the signal level, again in an ERB by ERB fashion.

But you have to know the actual spectra first, then compare them on an ERB by ERB basis to even have a start at something worth comparing.

 

[nick_l44.1]

But you have to know the actual spectra first, then compare them on an ERB by ERB basis to even have a start at something worth comparing.

Of course, my example simply demonstrated that not everything is simple and that the RMS noise level is far from its masking level.
Unfortunately, I know nothing about theoretical psychoacoustics. But REW has ERB smoothing. This averaging in my example does not quite correspond to the masking level, but it is very close to it. (And all levels in the spectra strongly depend on the FFT length, which in my case is 128k)

Blue - noise (RMS -22.5 dBA)
Orange - signal
Green - mix

 

[rdenney]

Okay, I get all that. So, show me the data where real people can detect differences in distortion and noise levels that are 10 dB below ambient (in any spectrum) when a primary signal is being played 85 dB higher.

I can hear noise at -85 dB when there is nothing else being played, particularly if I put my ear to the speaker, and depending on the spectrum. I'm just not sure how that is relevant to meaningful evaluation of playback equipment.

Example: I'm using a B&K Reference 125.2 amp in my YouTube-watching room. When I had a good DAC plugged directly into it, there was no background noise that I could hear at all. The quiescent noise level of the amp is probably -90 dB, if I had to guess, and that's compared to its full output given that it has no attenuation, so I'm pretty confident that noise at that level is undetectable in a real room.

But when I inserted a (well-respected) B&K Sonata MC-101 preamp in between, noise appeared as a low-level hiss of the white noise variety, even with the line amp in bypass mode (the line amp, when engaged, increased the loudness of that noise level significantly, suggesting the problem is upstream from it). It's at a much higher level than the Johnson noise should be from the Noble volume pot. The preamp needs a trip to the bench--that was already noted and the reason I took it out of my main system, replacing it with a restored and fully measured Apt Holman--but I suspect it's putting out noise at more like -60 dB or even -50. When music is playing--even very quiet music--I don't hear the noise at all. Even the ambient noise of the venue where the recording was made (for classical recordings made in performance venues) covers it up. Playing vinyl, of course, buries it profoundly. I'll make some SPL measurements tonight to have better data, unless it is windy or raining (which will create higher noise levels).

The Apt Holman in the main system is inaudible when the amp is turned on. I'm using an intermediate differential amplifier to create a balanced input to the amp, and that adds about 6 dB, but it only hisses if the inputs are floating. Its SINAD is about 100 dB. When I turn on the preamp (or use shorting plugs on the inputs), that hiss disappears. The SINAD of the Holman, as measured by me, is consistent spectrally at about -90 dB with respect to a 2V input.

The thread is about amps and DACs, but I've never heard a DAC make noise that I could detect in any circumstance. And every functioning amp I've owned has needed some sort of visual power indicator for me to even know it's turned on, which tells me quiescent noise is far more likely to be a source problem rather than an amp problem, but not when the only source is a DAC. The noise floor of a preamp doesn't get louder as the preamp is turned up, depending on where the noise is being produced, nor does the noise floor of an amp (at least those without attenuators or when the attenuators are turned all the way up). Noise on the input side of a preamp or in the source devices plugged into the preamp are therefore the likeliest candidates for adding noise. I may have more to say once I've gotten that MC-101 onto the bench.

Now, to distortion. When I compare distortion levels of sine waves in the sensitive spectrum, I can detect small differences in distortion. I can, for example, easily distinguish a 1 KHz sine wave being played by my Tektronix CFG250, which has a THD of about -45 dB (as measured by me), from the -100(ish) dB oscillators in my distortion analyzers (an HP339A and an HP8903B). These were clearly distinguishable played through an old Kenwood amp that has a measured THD of about -60 dB and through old Canton loudspeakers. I also have an old Kikusui 455 function generator that measures at more like -25 dB, and I can clearly hear the overtones, not just a difference. But when I compare my ability to distinguish differences in THD levels in actual music, using proper ABX procedures into good headphones through a good DAC, I cannot distinguish differences in distortion below about -36 dB. I don't think it takes much masking to cover harmonic distortion.

There seems to me a distinction between what is possibly detectable under certain extreme circumstances and what might remotely affect the playback of actual music in a real room, and that's the difference between the high and low levels suggested at the start of the thread. That was the issue I was responding to.

Rick "informed and data-driven observations, but still anecdotal" Denney

 

[danadam]

In another thread, people were adamant that 96dB of dynamic range and 144dB of dynamic range were Audibly THE SAME and any music over 16/44 was moot, waste of time and money?? (16/44 vs 24/96)

I don't know about other people but I know that it is for me. In attachment there is 16-bit dither pulsing (pulsating?) every 1 second. On HD650 and RME Adi-2 Pro (hi-power disabled) I start hearing this at volume somewhere between -5 and 0 dB. I don't think I ever listened to anything at volume higher than -10 dB (classical music) and usually I stay below -20 dB.

And that's flat dither (and without any other signal that could mask it). With shaped dither the headroom would be even bigger.

In the second attachment there is 3.5 kHz tone at -100 dBFS, beeping every 0.5 second. At that -10 dB volume I start barely hearing it.

 

[krabapple]

There seems to me a distinction between what is possibly detectable under certain extreme circumstances and what might remotely affect the playback of actual music in a real room, and that's the difference between the high and low levels suggested at the start of the thread. That was the issue I was responding to.

There certainly is, but a lot of 'high end' depends on promoting hysterical anxiety over the latter.

 

[bmc0]

My living room has an ambient noise floor of about 35 dBA. The absolute peaks of music at the very loudest I can tolerate (and then only when my wife is at least 50 miles away) are at maybe 110 dBA. Anything that is 85 dB down from that peak in my house won't be hearable over the freezer in the basement or the radon ventilation fan (or the tinnitus, or the blood pulsing in my ears).

It's possible to hear much lower levels at high frequencies than one might think in a reasonably quiet domestic space: Dynamic-Range Issues in the Modern Digital Audio Environment. Here's a relevant extract (from section 2.2; emphasis mine):

A determination of the level of white noise that is just audible in quiet listening conditions was performed by the author [4]. 13 listeners were exposed to a monophonic acoustic noise source that produced noise closely approximating white noise in the frequency region of 1–10 kHz. The noise source cycled the noise on and off smoothly at a 2-s rate while the listener adjusted the level to reach the "just audible" level. Each listener performed this experiment in his or her preferred listening environment. The results of these experiments are shown in Fig. 3.

Fig. 3 is a histogram of the results of the threshold determination for the 13 listeners and indicates that the mean threshold was 3.8 dB SPL for 20-kHz low-pass filtered white noise. The detection levels span the range from -2 to 9 dB SPL and have a statistical standard deviation of 3 dB. The low level of detected noise and the tight spread in values are surprising in that the listening rooms have wide-band noise levels in the range of 20–35 dBA SPL. In this study it is determined that the listener was detecting the presence of the noise using the noise energy within the 3–7 kHz band, and the effect of room noise was unimportant in most circumstances. This indicates that the spread in threshold values is governed by individual hearing acuity rather than environmental variations.

Of course, this means basically nothing with respect to audibility of nonlinear distortions.

 

[j_j]

Of course, my example simply demonstrated that not everything is simple and that the RMS noise level is far from its masking level.
Unfortunately, I know nothing about theoretical psychoacoustics. But REW has ERB smoothing. This averaging in my example does not quite correspond to the masking level, but it is very close to it. (And all levels in the spectra strongly depend on the FFT length, which in my case is 128k)
View attachment 486279
Blue - noise (RMS -22.5 dBA)
Orange - signal
Green - mix

So, your levels have nothing to do with actual SPL if they change with window length. You're missing a basic calibration.

 

[nick_l44.1]

You're missing a basic calibration.

No equipment that can be calibrated is used for this analysis. It is a direct fft of pre-prepared audio files, where 0 dB strictly corresponds to 0 dBFS. The levels of any noisy spectra always depend heavily on the length of the FFT, like this:

 

[solderdude]

Unfortunately, I know nothing about theoretical psychoacoustics

Psssstttt... J_J does .... a LOT and about the practical side.

 

[j_j]

No equipment that can be calibrated is used for this analysis. It is a direct fft of pre-prepared audio files, where 0 dB strictly corresponds to 0 dBFS. The levels of any noisy spectra always depend heavily on the length of the FFT, like this:
View attachment 486429

How to normalize an FFT was known somwhere in the realm of, um, oh, 1950 or so.

 

[flipflop]

is anyone still editing it?

I was able to keep the OP updated for a while because ASR had no (or very generous) restrictions on editing back when the thread was created. The 'Edit' option is no longer available to me.
If someone with extended editing privileges wants to make changes, they have my blessing.

 

[ahofer]

I was able to keep the OP updated for a while because ASR had no (or very generous) restrictions on editing back when the thread was created. The 'Edit' option is no longer available to me.
If someone with extended editing privileges wants to make changes, they have my blessing.

Thanks, BTW, this is one of the most important scientific aspects of our hobby, and much ignored in the broader audiophile world. Sometimes even in the more “objective” corners as well.

 

[donjoe0]

I was able to keep the OP updated for a while because ASR had no (or very generous) restrictions on editing back when the thread was created. The 'Edit' option is no longer available to me.
If someone with extended editing privileges wants to make changes, they have my blessing.

I was thinking I could simply use the Report function to suggest the edit, but somehow it still seems wrong because your name will still be attached to whatever the text ends up being.

 

[j_j]

How to normalize an FFT was known somwhere in the realm of, um, oh, 1950 or so.

What's more, that FFT does appear to be properly normalized. You don't seem to understand how the power spectrum adds across frequency. None the less, using an ERB-scale analysis would show you what you want. Even if you simply picked the closest power of 2 for each band, that would be revealing.

 

[cay-uwe]

I am new to this forum and basically I find this thread approach very good. Therefore I would like to thank @flipflop for the efforts made to create a more practical view to measuremnts.

Anyhow, considering that this thread is already open for over 7 years and full of different opinions, shows, how difficult it is to get a common guideline for measurements value and meaning for a consumer.

As long the HiFi or Audio industry does not adapt to a common standard, any measuremnts one sees are meaningless, unfortunately ...

 

[ahofer]

I am new to this forum and basically I find this thread approach very good. Therefore I would like to thank @flipflop for the efforts made to create a more practical view to measuremnts.

Anyhow, considering that this thread is already open for over 7 years and full of different opinions, shows, how difficult it is to get a common guideline for measurements value and meaning for a consumer.

As long the HiFi or Audio industry does not adapt to a common standard, any measuremnts one sees are meaningless, unfortunately ...

No, not meaningless. I think one could adapt the “strict threshold” here and never look back.

Furthermore, this site has given me plenty of education on how to read and interpret measurements as presented, and has encouraged several efforts to document common measurement standards (see spinorama.org). Measurements are more and more meaningful as a result.

 

[cay-uwe]

@ahofer I also find the measurements posted here very good and I am not complaining about that.

I see it from consumer point of view and when one looks at specifications from manufactures one will notice that a fair comparison is not really possible, because nearly in every case the measuring conditions or any standard if available is not posted. Without it, measurements are practically meaningless. I may have been a bit more specific in my statements.

 

[antcollinet]

@ahofer I also find the measurements posted here very good and I am not complaining about that.

I see it from consumer point of view and when one looks at specifications from manufactures one will notice that a fair comparison is not really possible, because nearly in every case the measuring conditions or any standard if available is not posted. Without it, measurements are practically meaningless. I may have been a bit more specific in my statements.

Possibly.


But this is one reason why the measurements done here are so valuable.

 

[billou5767]

hello :
Frequency range of human hearing
Humans cannot hear sounds of every frequency. The range of hearing for a healthy young person is 20 to 20,000 hertz.
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 (Olson, Harry F. (1967). Music, Physics and Engineering. p. 249). Below 10 Hz it is possible to perceive the single cycles of the sound, along with a sensation of pressure at the eardrums.
For these reasons, the audibility thresholds will deal with frequencies from 0 to 20 kHz.


the science :
There is actually a peer-reviewed study showing that human hearing can extend beyond the commonly cited 20 kHz limit.

In a 2006 study published in Acoustical Science and Technology, researchers Ashihara, Kurakata, Mizunami and Matsushita measured hearing thresholds for pure tones from 2 kHz up to 28 kHz.

They found that some participants were able to detect a 24 kHz tone, although the required sound level was very high (88 dB SPL or more). No reliable hearing thresholds could be obtained from all participants at 26 kHz and above.

So it would be inaccurate to say that humans generally hear up to 28 kHz. However, the study clearly demonstrates that some humans can detect sounds above 20 kHz, with measurable detection reaching at least 24 kHz in some individuals.

Reference:
Ashihara, K., Kurakata, K., Mizunami, T., & Matsushita, K. (2006). Hearing threshold for pure tones above 20 kHz. Acoustical Science and Technology, 27(1), 12–19. DOI: 10.1250/ast.27.12

Official full article:
https://www.jstage.jst.go.jp/article/ast/27/1/27_1_12/_article/-char/en

2/
The commonly cited 20 Hz lower limit of human hearing is not an absolute cutoff.

Under controlled laboratory conditions, humans can detect sounds below 20 Hz, especially when the sound pressure level is sufficiently high. However, sensitivity decreases rapidly as frequency drops, and very low-frequency sounds may be perceived partly as vibration, pressure, or pulsation rather than as a normal audible tone.

So, just like the 20 kHz upper limit, 20 Hz should be considered a practical reference point, not a strict biological boundary.

In other words: some humans can detect frequencies below 20 Hz, but they require much higher sound levels than frequencies within the normal hearing range.


i stop reading the next 47 pages your start is false

a little one :

Interestingly, recent research has shown that human ears still retain a vestigial ability to react to sounds.

Researchers recorded the electrical activity of the small muscles around the outer ear (auricular muscles) and found that they respond to sounds, particularly when people are trying to focus on a sound or when the sound comes from the side or behind.

The movements are extremely small in most people, but the underlying neural system appears to still be active. Researchers describe it as a “vestigial pinna-orienting system” — essentially a remnant of the ear-orienting reflex seen in animals.

Reference:
Schroeer et al. (2025), Electromyographic correlates of effortful listening in the vestigial auriculomotor system, Frontiers in Neuroscience.

https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2024.1462507/full

Another study:
https://journals.physiology.org/doi/full/10.1152/jn.00024.2024





One particularly interesting point for Hi-Fi discussions is that “inaudible” does not necessarily mean “biologically irrelevant.” A physiological response to a sound does not automatically mean that we consciously hear it, or that it produces an audible difference in music.

This is where research into frequencies above 20 kHz, infrasound, HRTF, the outer ear (pinna), and the human auriculomotor system becomes particularly interesting.

There is still a lot we don't fully understand about how the entire auditory system interacts with sound — especially when we move beyond the traditional 20 Hz–20 kHz model.