[castleofargh]

Here's another question: Is there an inherent sonic difference between driver technologies?
I often see electrostatic drivers described as being more transparent than planar drivers, planar as having "faster" bass than dynamic drivers, dynamic as having "more natural," bigger bass, and better extension on both ends than balanced armature drivers, and balanced armature as being more detailed than dynamic. I've also seen 1,001 other claims of how each technology might be different and/or superior to the others. It seems like planar is the most popular request, while e-stat is the most aspirational. I think dynamic is more sought after than BA.
I don't think I have enough experience with all of the different technologies to know, and, even if I did have experience, I'm not sure I'd be able to tell the difference between individual tunings and technologies. Further more, I'm not sure what "transparent" means in regards to headphones (less distortion?), I don't know what "faster" sounds like, and I mostly dismiss "natural" as being a nonsense term, detail I can mostly wrap my head around (though, to be honest, even that seems like it's a kinda catchall term)... But that hardly matters because those terms are really only the tip of the iceberg when it comes to the differences people perceive.
My experience is that dynamic drivers are capable of being hugely detailed and seemingly very transparent (though, again, I'm not sure I know what that sounds like or if it even applies), that BA drivers are capable of massive amounts of bass that, to my ear, sounds very natural, etc.

So, is there an inherent sonic difference between technologies? If so, what are those differences and what do they sound like?

PS— I wonder if this is actually a question of semantics and jargon and what the hell people actually mean when they're talking about audio. A thread addressing those questions is one I've been wanting to start for a while, but I honestly think it'd take a linguist to parse it all out. Do we have a linguist in the house??!

same opinion as bigshot. different techs are different, sometimes drastically so, but pretty much all headphones are also different.
limited to IEMs but I went on a rant about the same basic ideas here:
https://www.head-fi.org/threads/iem-dynamic-vs-balanced-armature-bass.911504/

 

[megabigeye]

same opinion as bigshot. different techs are different, sometimes drastically so, but pretty much all headphones are also different.
limited to IEMs but I went on a rant about the same basic ideas here:
https://www.head-fi.org/threads/iem-dynamic-vs-balanced-armature-bass.911504/

Oh yeah. Oops! I wasn't going to ask this question because it was already kind of addressed in that thread. I forgot! My question is a little different, I guess, because I was asking about all driver types and all sorts of performance, not just bass (even though most of what I mentioned was bass).

So my takeaway is that so long as there's a difference in SNR, distortion, etc. there can be a difference, but that it's hard to know if it's due to the inherent nature of the driver technology or simply differences between individual drivers. I would guess that any type of driver could fall anywhere along a spectrum from really poor performance to really excellent performance, which would make comparisons between them kind of a moot point. Is this what you guys are saying, more or less?

 

[bigshot]

Anything mechanical is a tradeoff. You balance advantages with weaknesses and come out with some sort of happy compromise. But there's a million different ways to compromise, and one kind of compromise might work better for one person than another. For instance, with speakers, it's desirable to have efficiency, small size and deep bass, but it isn't easy to get all three at once. So you make the cabinet bigger to make an efficient speaker with deep bass- that's the theory behind old style box speakers. Or you make a small speaker that is inefficient with deep bass like modern designs. Some kinds of drivers tend to be better at some things and some others. But because they are mechanical, it's never absolutely perfect. It's all a trade off. Transducers are the wild card. You should build your system around that, rather than the other way around.

 

[castleofargh]

Oh yeah. Oops! I wasn't going to ask this question because it was already kind of addressed in that thread. I forgot! My question is a little different, I guess, because I was asking about all driver types and all sorts of performance, not just bass (even though most of what I mentioned was bass).

So my takeaway is that so long as there's a difference in SNR, distortion, etc. there can be a difference, but that it's hard to know if it's due to the inherent nature of the driver technology or simply differences between individual drivers. I would guess that any type of driver could fall anywhere along a spectrum from really poor performance to really excellent performance, which would make comparisons between them kind of a moot point. Is this what you guys are saying, more or less?

it's not that hard to take a headphone and another headphone and measure many variables to know the differences in sound. even easier, we can record the sound itself and clearly find out how different it is from the other headphone. you can look for many measurements of many headphones and look for trends depending on the tech involved.
there are a few issues for me here and most really are about properly defining what we're looking for and when we should consider something a trend or just variability among a range of possible results. taking the most basic of all variables, the frequency response, would you say that a certain driver tech gives a specific type of FR? in a way we can easily see correlations, like how on fullsize headphones and DD drivers, we very often will find the resonance for the driver at low freq(better there than in the midrange) causing a bump in FR, and we also often see such drivers roll off fairly soon in the sub frequencies. on the other hand we can find several orthos and electrostats with almost dead flat low end until very low frequency. so it can seem fine to associate that FR pattern with the driver type. but some DD headphone have a massive boost in the low end and still a lot of amplitude at very low frequency. should we count that as exception? start to classify those headphones as part of another design or sub category of design? but then you'll soon have as many sub categories as you have variables to measure, perhaps more.
or once we've found a few DD with a lot of low freq amplitude(closed back or some that just drop by less than the threshold you decided to count as significant), and also found some electrostats that roll off strongly in the low end(whatever the actual reason why, probably crappy seal or whatever), should we just drop that correlation as not really being one? if so how many such headphones do we need to find before we reject the otherwise solid trend? sooner or later all this is going to beg for some sort of statistical analysis and very specific criteria.

there is such a wild range of sounds coming out of DD headphones that it's hard IMO to have them all fit in just one box of caricatured specs and behaviors. but I'm not trying to suggest that a given design doesn't have a set of consequences including some for the sound. there are definitely some. my personal issue comes from deciding on the relevant parameters(something that will probably end up being somewhat of an arbitrary decision), and the reliability of the data we're going to use(spoiler, the guy who talks about fast bass is out of any study I would ever trust).

 

[bigshot]

The typical range of manufacturing variance for frequency response in headphones is greater than +/-3dB, which is quite a bit, so even measurements on the internet might be different than the copy you get when you order it from Amazon. I think the best approach to it is to buy, return if you flat out don't like it, eq if it's close.

 

[gregorio]

Here's another question: Is there an inherent sonic difference between driver technologies?

The first thing to understand about drivers is that they're quite different from anything else in the recording and reproduction chain (with the exception of microphones). Instead of dealing with digital data and/or an electric current, we're dealing with converting an electric current into a physical phenomena (sound pressure waves) and this is a hugely inefficient process. It's a bit like incandescent light bulbs, where the vast amount of the power consumed ends-up as heat and only a small percentage ends-up as what we actually want, light. Same with speakers, if you have a 100w speaker it's probably only capable of outputting around 3w of acoustic energy, the other 97w is simply lost. Different technologies can be somewhat more efficient but at best never more than about 25% and being more efficient doesn't necessarily mean "better".

While the answer to your question is essentially "yes", it's really not that simple because each of the different technologies have both advantages and disadvantages and many of the different manufacturers address (try to compensate for) the disadvantages differently and in addition, we've got the issue of manufacturing tolerances that bigshot mentioned. Taking into consideration the inefficiencies involved, the laws of physical motion and physical properties (such as inertial and the compressibility of air for example) that don't affect digital data or an electric current, it's impressive that headphones/speakers manage the sonic quality they do but we've had over a century of research and more importantly, unprecedented financial incentive.

In case it's not clear, I'm basically agreeing with bigshot and castleofargh.

Another consideration, typically overlooked by audiophiles and something you touched on previously, is what about the opposite end of the chain, recording? After all, that's what we're trying to reproduce. Microphones are essentially speakers in reverse, they convert sound pressure waves into an (analogous) electric current, as opposed to converting an analogous electric current into sound pressure waves. And just like speakers, there are different mic technologies (dynamic, condenser, ribbon, back-electret and pressure zone), all of which have different advantages and disadvantages, different implementations by different manufacturers and different sonic characteristics. However, contrary to audiophile assumption (and their goals/practice with speakers/HPs), mic's are not employed on the basis of their fidelity and indeed the very term "fidelity" has an ambiguity that's not really applicable to mic use (so we tend to use "accuracy" instead). Sure, there are occasions when we're looking for a mic with maximum accuracy but most audiophiles would be relatively shocked at how rare those occasions are. Instead, we choose mics essentially according to their inaccuracies/weaknesses, or to put it another way, according to the "colouration" which results from their inaccuracies. In rock/pop genres for example, we virtually never use the most accurate mics, we often use some of the least accurate and, it's not just the mics, the most prized mic pre-amps are also very coloured, way more so than even pretty cheap consumer amps. Also, none of this has any correlation with cost, a highly accurate mic would likely cost say around $1,000 or so, while some of the most prized could cost several (up to about 50) times more but be significantly less accurate. One of my first experiences ("behind the glass") in a world class studio taught me that (in that specific circumstance) a couple of fairly battered $100 mics gave a superior/preferable result compared to a pair valued at nearly $200,000, although neither pair was anywhere near accurate (but they were inaccurate/coloured in very different ways). The resultant commercial release went on to be regarded by many as a definitive recording (of it's type) and many of the musicians with whom I discussed the recording were very surprised (and some were horrified) when I told them we'd used just about the cheapest studio mics on the market!

G

 

[megabigeye]

Many months later: @gregorio, thanks for that answer. That helps my understanding quite a bit. Sometimes it's easier to understand one thing by learning about an analogous thing (pun kind of intended).

I have a few other questions:
1) What are attack, decay and transient? I think I understand the concepts in theory, but I don't think I have a practical understanding and I don't know what they sound like when I'm listening. Are there examples (like YouTube?) of what each sounds like? People often drop these words in reviews and impressions and general audiophile speak, but I don't really know what they're talking about and I often wonder if they know either.
If necessary, you can treat this as three separate questions.

2) What is (generally speaking) the greatest dynamic range of music? And how is that range measured? Is it measured equally above and below an average loudness (0dB?), so, say, -20dB and +20dB for a total of 40dB? Is dynamic range a measure of all sound that can be considered part of the music, or does it include the noise floor?...

...And for that matter...
3) What is noise floor exactly? I imagine that an easily audible one might be a hiss or a hum, but how else does it manifest itself? Where does it come from? Is there a difference between a recording's noise floor and that of a piece of equipment?

4) And finally (for now), what are the different measurements and thresholds that must be passed in order for a piece of equipment/media to be considered audibly transparent? Jitter, signal-to-noise ratio, distotrion... What else, and at what point can we no longer distinguish a difference for each?

Thanks in advance to all that answer! I've really enjoyed having this resource. I often read the Sound Science forum, sometimes for information, sometimes for a laugh, sometimes to figure out if gregorio has saint-like patience, or if his house is filled with strangely head shaped holes in the wall.

 

[aukhawk]

:
1) What are attack, decay and transient?

:
I would suggest 'attack' is the speed with which a sound-producing thing (such as a loudspeaker or other transducer, or an instrument such as a piano or a guitar or a violin, or even an entire space such as a cathedral) can respond to a stimulus (such as a very fast rise in voltage hitting a transducer, or a finger hitting a piano key). Among instruments a rim-shot on a snare drum has a very fast attack.
Similarly 'decay' is the speed with which such an object can respond to an absence of stimulus. Something like a bassoon has a very short decay. A cathedral has a very long one. Hendrix' guitar, even longer.
A 'transient' is sort-of those two things in combination - a fast attack immediately followed by a fast decay.

:2) What is (generally speaking) the greatest dynamic range of music?

Depends a lot on how close you are to the source. A trombone player in an orchestra sitting just in front of the percussion section probably experiences a huge (maybe damaging) dynamic range. Someone in a non-ideal distant seat in the same auditorium probably experiences a slightly unsatisfactory 40db or so (since the 'noise floor' is quite high - in the midst of an audience).

 

[megabigeye]

:
I would suggest 'attack' is the speed with which a sound-producing thing (such as a loudspeaker or other transducer, or an instrument such as a piano or a guitar or a violin, or even an entire space such as a cathedral) can respond to a stimulus (such as a very fast rise in voltage hitting a transducer, or a finger hitting a piano key). Among instruments a rim-shot on a snare drum has a very fast attack.
Similarly 'decay' is the speed with which such an object can respond to an absence of stimulus. Something like a bassoon has a very short decay. A cathedral has a very long one. Hendrix' guitar, even longer.
A 'transient' is sort-of those two things in combination - a fast attack immediately followed by a fast decay.



Depends a lot on how close you are to the source. A trombone player in an orchestra sitting just in front of the percussion section probably experiences a huge (maybe damaging) dynamic range. Someone in a non-ideal distant seat in the same auditorium probably experiences a slightly unsatisfactory 40db or so (since the 'noise floor' is quite high - in the midst of an audience).

Thanks for the answers!
For 1), I mostly understand this, but how does it translate to listening to headphones? For instance, I often see the claim that the slower decay of a dynamic driver makes the bass sound more "natural" than a BA driver. To my mind this doesn't really make sense. Wouldn't a faster speaker decay be more accurate, since it'd be able to represent both slow and fast decaying sounds more accurately? Maybe "accurate" and "natural" aren't synonymous.

For 2), I meant recorded music, but I wasn't very clear.

 

[bigshot]

The fastest attack in live music is at least an order of magnitude slower than what headphones can reproduce. When you hear audiophiles talk about "speed" and "time error" it's usually hooey because they haven't taken the time to figure out how big the slivers of time they are actually talking about are relative to music.

If a transducer can reproduce 20kHz, by definition it is able to resolve 20 cycles per millisecond. Since the fastest that humans can even perceive attack is about 30ms, you can see that it's like an ant and an elephant.

 

[megabigeye]

The fastest attack in live music is at least an order of magnitude slower than what headphones can reproduce. When you hear audiophiles talk about "speed" and "time error" it's usually hooey because they haven't taken the time to figure out how big the slivers of time they are actually talking about are relative to music.

If a transducer can reproduce 20kHz, by definition it is able to resolve 20 cycles per millisecond. Since the fastest that humans can even perceive attack is about 30ms, you can see that it's like an ant and an elephant.

Ah. Interesting (and surprisingly easy) way to think about it. Thanks. Does that mean that attack (and decay, I guess?) is a function of frequency response? Or is there something else that might affect how a driver reproduces attack or decay?
When I think of decay, I imagine something like a guitar string resonating for a moment after it's plucked; unintended reverberations.

 

[castleofargh]

It's probably best to stick to the subjective side of a feedback and disregard the made up objective justification that comes with it about speed, transient and what not. Because a subjective impression is usually all they have and all they should be talking about anyway.

2) What is (generally speaking) the greatest dynamic range of music? And how is that range measured? Is it measured equally above and below an average loudness (0dB?), so, say, -20dB and +20dB for a total of 40dB? Is dynamic range a measure of all sound that can be considered part of the music, or does it include the noise floor?...

The greatest dynamic range is the greatest that your file can support. There will always be some trickster who will go from no signal to full scale just because he can. How it's measured depends on what you're looking for, but by definition the dynamic range is the ratio between a highest and lowest values. Usually the lowest value is going to be a noise of sort.

On a record, turning down the volume by 30dB makes it subjectively a lot quieter already, so to give the sense of loud and quiet there is no need for huge dynamic range. Imagine having typical albums with 120dB of dynamic. Even if it was possible, it would result in a lot of the music being inaudible once you have set the level for the loud parts. And you'd have to play it uncomfortably loud to try and hear as much as you can handle. Most people would be unhappy so most albums aren't aiming for huge dynamic. Classical albums will often be where you'll find the most dynamic. I find anything beyond 60dB of dynamic to be annoying to listen to.

Also, there are practical limitations, like how even a quiet studio probably has some 20 to 30dB SPL of ambient noises, so the dynamic range will start from there even in the best high dynamic scenarios.

About noises. Noise kind of defines anything that isn't music or the distortions forced onto it. So room noises, thermal noises, quantization noises, the sounds from your own body. All of those are noises in this context.

 

[bigshot]

Does that mean that attack (and decay, I guess?) is a function of frequency response?

Sound is a combination of frequency, amplitude and time. They're all related and they all affect each other. Any deviation from the proper presentation of each can be called distortion. Over the past century, distortion has been reduced greatly in home audio. Has it been reduced enough? That's a subjective judgement call. There's always going to be some degree of error. The question is, is it enough for you to perceive it? In order for timing error to reach the level of audibility, it would probably have to affect frequency response as well.

The easiest way to look at it is to compare reproduced sound fidelity to live music. The frequency response, amplitude and time accuracy of live music all fit easily into normal CD specs. The error in headphones is higher, but less so than the distortion created by the effect of a room on the sound fidelity of a live band. But the error created by the room is desirable, and error in transducers isn't.

Human ears are quite forgiving to certain kinds of error, and they aren't discerning enough to perceive any distortion below a certain point. Headphones are capable of reproducing sound with a remarkably low amount of distortion... better than speakers in that regard. Yet speakers sound better overall because the distortion created by the room is euphonic. The problem with transducers isn't so much fidelity as it is the lack of euphonic envelope wrapped around the sound. Headphones don't add any envelope, and speakers in a room have too many variables to the envelope to precisely control it in most living rooms. That envelope is harder to create and control than just eliminating audible distortion.

I guess what I am trying to say is that there are bigger fish to fry than the tiny insignificant specs that many audiophiles worry about. Does that make sense?

 

[pinnahertz]

Does that mean that attack (and decay, I guess?) is a function of frequency response?

Attack speed is a measure of how fast the envelope of the sound rises once it starts. I rimshot rises very quickly, a bowed string quite slowly by comparison. Attack speed does impact the frequency spectrum of a sound (frequency response is something else), with faster producing momentary higher frequencies. A glockenspiel hit or triangle hit have pretty fast attacks and lots of high frequency spectrum content. A kick drum wack is much slower, and lower. But here's where it gets confusing.

Attack is also perceived as the impact a sound has, which relates to the energy of the sound. Bass sounds contain much higher energy that treble sounds, and I think that's where the rather confusing term "fast bass" comes from. Bass isn't actually fast, it can't be, but to reproduce the full energy of a kick drum whack takes a bit of power, and with out it, the attack will seem weaker.

Or is there something else that might affect how a driver reproduces attack or decay?

Higher mass drivers may have some trouble with damping. Ideally, you want the driver to produce a wave front identical to the original, but since they have mass they might get sloppy and not mechanically track the electrical signal as well, over shooting a bit, or self-resonating. Shouldn't typically be much of an issue in headphones, but it sure is in speakers.

When I think of decay, I imagine something like a guitar string resonating for a moment after it's plucked; unintended reverberations.

For the clinical definition, the attached picture is of a very short guitar pluck. I've drawn in a tracing of the peak envelope. The attack portion of the envelope is yellow, the decay is red. Attack is always rising, decay generally falling.

For comparison, here's a snare hit:

 

[gregorio]

I have a few other questions:
1) What are attack, decay and transient? I think I understand the concepts in theory, but I don't think I have a practical understanding and I don't know what they sound like when I'm listening. Are there examples (like YouTube?) of what each sounds like?
2) What is (generally speaking) the greatest dynamic range of music? And how is that range measured? Is it measured equally above and below an average loudness (0dB?), so, say, -20dB and +20dB for a total of 40dB? Is dynamic range a measure of all sound that can be considered part of the music, or does it include the noise floor?...
3) What is noise floor exactly? I imagine that an easily audible one might be a hiss or a hum, but how else does it manifest itself? Where does it come from? Is there a difference between a recording's noise floor and that of a piece of equipment?
4) And finally (for now), what are the different measurements and thresholds that must be passed in order for a piece of equipment/media to be considered audibly transparent? Jitter, signal-to-noise ratio, distotrion... What else, and at what point can we no longer distinguish a difference for each?

It could quite easily take a book to answer your questions because many of the terms you've asked about are either vague to start with or have somewhat different meanings in different contexts. For this reason, some of my answers are somewhat different to the answers you've already received:

1. Attack is the initial phase of the production of a sound. For example: A finger plucking a string, a bow scraping across a string, the impact of stick/beater on a drum head, the start of the vibrations in the lips for a brass instrument or vocal chords when singing. Typically, this attack phase will last until the transient peak is reached, which can occur just a few tens of micro-secs after the beginning of the attack phase (in the case of some percussion instruments) or take several tens of milli-secs. Once that transient peak is reached, the sound/note will Decay (the amplitude will reduce) to the "Hold" (or "Sustain") phase, after which we enter the Release phase: This is the phase where the note stops being played and fades to silence. In some cases, the sustain phase is extremely short or non-existant, the decay phase and the release phase are essentially the same thing. This is typical of percussion instruments though not all, the marimba and the snare drum can be exceptions (when "rolled"). All these phases together are called the "Envelope" (and often abbreviated ADSR) and can vary enormously in duration, even in the case of a single strike on a perc instrument. The claves for example have very little resonance, no sustain phase and the entire envelope lasts just a couple of milli-secs, while the Tam Tam for example can last over 2 mins. This diagram might help you to visualise:

In some cases the release phase can start with a transient, typically when the note is "damped", using the tongue in the case of wind/brass instruments or the fingers/hand in the case of perc instruments like the timpani or triangle. Please also note that this ADSR envelope came into being with the invention of synthesizers, as all these phases could be created (synthesized) independently, however, many acoustic musicians don't really use it or use it differently. For example, "Attack" will often refer to both the Attack and Decay phases together, the Sustain phase is the same but the Release phase can often be referred to as the Decay.
If you have a simple audio editor, it's pretty easy to zoom in, isolate these phases and listen to them but particularly in the case of the A and D phases, listening to them in isolation often won't tell you much (even what instrument it is).

2. The "greatest dynamic range of music" is "pp" (pianissimo) to "ff" (fortissimo) but it isn't measured, it's judged. Pianissimo can have a wide range of different actual amplitude levels, even with same instrument and musician, depending on the acoustics of the performance venue. Dynamic range in the context of recording is the peak level relative to the noise floor (see #3). It's measured in dB but is usually somewhat subjective, as it's typically difficult to judge exactly at what point a note/sound disappears into the noise floor and it's further complicated by the fact that we can sometimes discern notes/sounds even when they drop several dB below the noise floor. Extremely few music recordings have a dynamic range greater than about 60dB and most have a dynamic range less (or significantly less) than 50dB. In the case of audio equipment, dynamic range is typically the range between it's highest undistorted output level and it's self noise. However, this can be somewhat vague/misleading, as the self noise of analogue audio equipment is higher when it's reproducing a higher level output signal than when it's idle (not reproducing a signal). The are other contexts as well though, the dynamic range of digital audio formats for example.

3. Noise floor can be a confusing term because it can BOTH refer to a single specific thing AND an accumulation of different things. For example, the noise floor on a recording is a combination of a number of different noise floors: The noise floor (self/thermal noise) of the microphones and mic pre-amps, the noise floor the digital processing (accumulated dither or quantisation error), plus of course the noise floor of the recording venue, which itself is a combination of different noise sources: External noise entering the recording venue, the noise made by the audience and/or musicians moving and breathing, the noise of air conditioning etc., and even potentially the noise that air itself makes (due to Brownian Motion). Almost always, there's a big difference between these various noise floors which comprise the recording noise floor. For example, the noise floor from the digital processing will almost always be two or more orders of magnitude less than the noise floor of the recording venue but there can be exceptions and in the case of analogue tape, the generational noise can easily become the dominant contributor to the recording noise floor, same with mics/mic pre-amps under certain conditions. The noise floor when reproducing recordings includes even more noise floor contributions: To start with, we obviously have the noise floor of the recording itself but in addition, we've also got the cumulative noise floor of all the bits of kit in the reproduction system (DAC, amp, speakers/HPs) and of course the noise floor of the listening environment, which like the recording venue noise floor is itself a combination of noise sources.

4. In practice, it's really only noise and distortion because all the potential artefacts/flaws produce either one, the other or both. For example, random jitter causes noise, while non-random jitter can create individual frequency spikes (which is a distortion of the original frequency content). For a piece of equipment to be considered "transparent", it's noise/distortion must be inaudible (at reasonable playback levels). Unfortunately, the audibility of noise and distortion varies according to the situation. For example, -40dB of noise will be inaudible in a loud, dense section of music but easily audible in a quiet, sparse section of music (if the noise floor of the listening environment is not higher) or, an audible amount of distortion at 3kHz might be inaudible at 60Hz. The only way to be sure of what is audible (and therefore not "transparent") is a DBT of that particular type of noise or distortion. In published scientific threshold studies, test conditions and signals are typically employed to highlight/exacerbate a particular artefact, which provides a good level of confidence that the observed threshold will not be exceeded under normal listening conditions. For example, some published studies have shown that jitter of just 2-3 nano-secs can be audible with specifically designed test signals but under laboratory conditions the lowest jitter demonstrated to be audible with actual music recordings was about 200 nano-secs.

Despite my apparently completely vague answer, we can make some generalisations, just be aware it's not a hard and fast rule, there can be (and are) exceptions. Assuming reasonable playback levels:
Noise/Distortion higher than -40dB will probably be audible in some/many conditions (though certainly not all!).
Noise/Distortion between -40dB and -70dB is unlikely to be audible in many/most conditions.
Noise/Distortion between -70dB and about -90dB will generally be inaudible to virtually all consumers but might be audible under some very uncommon conditions, such as: Very loud (but still "reasonable") playback levels AND particularly sensitive or trained listening skills AND extremely good listening conditions (laboratory, high quality studio or accurate and very well sealed IEMs for example) AND certain specific music recordings. It's very unlikely that noise/distortion lower than about -80dB would ever be audible to any consumer.
Noise/Distortion lower than about -90dB we can pretty safely say will never be audible. But, "pretty safely say" does not necessarily mean "say with absolute scientific certainty", although it comes pretty close

G