Audibility

April 6, 2021
 by Paul McGowan

We often ask ourselves what’s audible and what’s not.

We can say with some confidence 10% distortion is audible, but is 0.1%?

And, compared to what? Can we hear the difference between 0.1% THD and 0.0001% THD?

There comes a point in measurements where one must decide what matters and what does not. It probably doesn’t matter, for example, that one wheel of your car might be 1/10th of an inch different in diameter than another. On the other hand, 1/10th of an inch parts variability in a Swiss watch might be the difference between functioning and non-functioning.

From my perspective, once we hit THD and IM levels below 0.1% the differences become more academic than audible.

And yet we hear major differences between differing topologies producing dramatically different levels of distortion.

What that suggests is simple. Beyond a certain point, the level of distortion becomes meaningless.

How those differing levels are achieved is where we find the bulk of sonic differences.

Subscribe to Paul's Posts

55 comments on “Audibility”

  1. It seems such observations are often a temporary matter and temporarily correct. Remember when everyone thought the 20kHz limitation of the CD doesn’t matter because it’s above the audible range. Later we knew it affects several sound degradations anyway.

    So imo it’s not that important if we right now hear the differences you describe, but if we’re sure they don’t have other implications we don’t connect to them yet.

    I guess for now you’re exactly right. But it’s always better to add the „for now“, as in less obvious fields of science, with black/white statements, otherwise one frequently has to switch opinions as knowledge evolves.

  2. What is your definition of audibility, Paul, if there are phenomenons as cognitive bias and selective perception? What do audiophile hear or focus when they claim sound improvement by different footers or antivibration platforms placed under a DAC with external power supply or when placing a record weight on a vinyl record placed on a sticky record mat or on a vacuum hold down platter? There are very strange audiophile illusions and imaginations. What I can say: every adjustment made on my digital equalizer is immediately audible. 🙂

  3. So said the manufacturer that doesn’t create an amp that delivers 0.00x% distortion at rated power. (that’s not a dig, just a reality) The question becomes, what level of distortion are you willing to live with and find fitting to the “color” of sound you want to hear. As for me, I don’t prefer the color, and have always found when it comes to Class A or Class AB electronic bi-polar amp designs, those that exhibit THD levels of 0.00x% always sound more appealing, and yes, you can hear the difference.

  4. The amplification design in general, has evolved in such a way that today extremely low THD and IMD values ​​are achieved compared to the beginning of technology, even the TIM in most of today’s products, is immeasurable using current techniques measurement.

    But: How to explain that an amplifier that has higher IMD or THD figures, sounds more pleasant to the ear than one that has values ​​with several zeros before the first digit?

    This happens when we compare tube amplifiers with those of transistors, there are even amplifiers based on transistors BP or Mosfets, in which the same happens, when compared with other BPs or Mosfets, all of them with low distortion figures.

    The same can be said when comparing amplifiers based on bipolar transistors that do not have as spectacular figures as other similar ones, however, they sound more natural to the average ear.

    Sure for someone, the easy answer is topology, but the numbers are there: all very low compared to each other.

    Is it that current measurement techniques are obsolete to explain the accuracy of the human eardespite the technological explosion? ?

    1. Fact is that we rather hear what we want and like to hear – see also cocktail party effect. Thus are there two audiophile who focus on identical aspects and elements of the music being reproduced?

  5. If an amplifier is somewhere between 0.05% & 0.006% THD (depending on the output)
    like mine is, do I really care, when my loudspeakers are measuring between
    1.2% to 0.7% distortion (depending on frequency & power)?
    In short; no.
    Ultimately I only care about how pleasing/good the whole audio set-up sounds,
    from CD (recording quality) to loudspeaker output.
    System synergy my friends.
    If you are more concerned about how your audio equipment measures
    rather than how good the music coming out of it sounds TO YOU,
    then I think that you’re missing the point of home audio in a really big way.
    Happy measuring folks! (eye rolling emoji)

    1. And this is why I’m spending my money on high quality recorded music these day’s. Extremely happy with what I’m hearing. Thank yo FatRat and Jim. By the way snake oil isn’t pleasant tasting at all and it’s indescribable 🙁

      Keep listening all 🙂

  6. Can the audible differences be measured? If traditional measurements don’t tell you a lot once you hit a certain level of exemplary behavior as confirmed by measurements, how do you tweak a circuit by ear, knowing that you’re judging the circuit by playing music with unknown distortion products through speakers with relatively high distortion?

    The goal is “straight wire with gain,” or no distortion. Can an amp designer get there by ear, or are we OK with choosing the colorations we like?

    1. But measurements will tell you that there are clear audible differences when there are different measurements – see flat frequency response versus a frequency response with bass boost (often seen with bookshelf speakers). Or a frequency response with a dip in the most sensitive frequency region of our hearing ability. Not to mention improper impulse responses and the lack of natural punch of percussion instruments or plugged string instruments or pianos.

  7. “THD and IM levels below 0.1% the differences become more academic than audible”
    Tell that to companies who have a totally different view on this matter, e.g. Bryston.
    Well, it’s a fact PSA devices are not the best when it comes to measurements. At least now we know why.
    I agree with Barsley (3:20 comment).
    There are a lot of things you can hear (like different footers, less distortion) but cannot put into words what exactly the difference is that you hear.
    But the fact that we cannot put into words what we hear/experience does not mean it’s not there.
    Can person A describe the taste of food to person B (who never ate that food) in such a way that person B exactly “knows” the taste of that food ? No, impossible.
    Person B has to taste it for himself. And then it’s possible he doesn’t taste what person A described.
    But, what person A described is real to him, notwithstanding (who invented this word ?) the fact person B does not taste the same.
    To person B it may taste like snake oil.
    In audio it’s exactly the same. A lot of people call everything they cannot hear “taste” (hear) snake oil, or imaginary.
    Well, in that case there’s always the old, faithful equalizer. Crude, but it works (for some…). 🙁

              1. FR, I like self deprecating humour but their products would have to be beyond excellent to tempt me. I think I’d always be left with the feeling that I’d been ripped off.
                As for ‘Audiovile’ I’ve already got a whole product range in my head 🙂

    1. There’s 2 things to think about—differences and improvements. Sticking you fingers in your ears makes music sound different, but not an improvement.

  8. I can not taste and hear snake oil…so it doesn’t exist.
    That’s a pity.
    But wait…there are more ways to fool yourself 🙂
    You guessed it…the equalizer

    1. An equalizer will not add the subtle harmonics that a vacuum tube will add to a signal, as they are microphonic. Spending a couple of decades at a very respected brand and one of the oldest in the industry that built and sold both tube and electronic amplifiers, from a few watts to kilowatts, spanning multiple brand lines, some from more than 1% THD to the models that have 0.00X% THD range at rated power, when you build 10’s of thousands a year and have to deal with on a daily basis this myriad of models, you come to learn fairly quickly that it’s rather easy to discern. It’s not snake oil, it’s synergy.

      1. Anything that”adds” means it was not there in the musical signal. I don’t want to add to the signal. I want to hear what the musicians sent to me.

        1. Let’s assume that there is a perfect recording and mixing process based on perfect microphones and perfect studio monitors. But fact is that there is no perfect audiophile loudspeaker. Thus you need an equalizer which compensates for the dips and peaks in the frequency response of your audiophile loudspeaker!

          1. Not really. You may want to listen with the same speakers they used to master the recording. Or if we get more uniformity with the speakers used by pros you can get pretty close.
            However, I don’t want to change the version they sent to me.

            DSP is primarily used to tame some frequencies in YOUR room, not so much the speakers. If the speakers need a lot of DSP, you should get new ones.

          2. There are certainly a number of variables that go into a listening area. The room is the other half of the speaker. My system embraces both old and new technologies, as well as 7.2.4. I don’t run any EQ or DSP in my 2 channel listening. If you start with a decent speaker in the front 2 channel config, and provide a dedicated 2 channel amp, you can perform as well as any other system out there. The 2 channel amp I use exhibits 0.003% worst case, and averages 0.001% THD at full rated power, having over 160,000µF of tank circuit with well over 3dB of headroom. (and it’s a 1983 vintage)

  9. This is all interesting and good. But as always, without concrete answers.
    Are the IM and THD measurements made using a static load? If in fact that is the case then it’s not a true representation of the real world reactive load. When the measurements are made are they using a single tone or sweep? Again neither one of them represents the complexity of music.

    So in the end it becomes both the designers and consumers choice as to what road they pursue and why.

    Personally I’m straddling the fence…. as I believe

    1. Mike,
      Fence straddling sounds painful.
      Do you have one on each side or do you keep ’em together?
      😉

      I’m guessing static load, only ‘useful’ for vague comparisons with other amps.
      Like dB/W/m or dB/2.83v/m.
      Who’s gonna sit exactly 1 metre from their floorstanders?
      (I’m not talking nearfield listening here)
      ‘dB/W/m’ is purely for a measured comparison with other loudspeaker
      so that the customer can get an idea of loudness…but you already know this.
      I’m just pointing it out as an example of why we measure audio equipment after the fact.

      1. FR,

        Especially a picket or barbed wire fence 🙁 It’s all in the technique….

        I suspect static loads also. A standard yes, but real world no. So those specs are somewhat useful if I need distortion free heater. And to some degree, some representation of an amps performance.

        So back to the fence, ouch damn it. 😉

  10. Where in the frequency band that distortion is occurring probably affects us more than the average level. Our individual hearing varies. We probably have different tolerance levels for distortions at different frequencies. For example, small amounts of, say, high frequency distortion add to our biological hearing frequency distortion, putting the sum of distortion over the threshold of irritability. A tiny amount of distortion could tip the scale. What bothers you might not bother me, or the other way around.

      1. Thanks for your corroborative response, even though you meant it as a put down. Since I’m correct I don’t need to read all that literature that you know so much about.

  11. I’ve spent a lifetime conducting wine tastings, and it has become very evident that clean wines, defined by their “impeccable” numbers, almost always taste boring. They lack character. They fail to ground themselves in the grapes, soil, climate and other characteristics of the region. So it is with stere systems.

  12. The profile of the harmonic products has a lot to do with how audible these non linear products will be in an amplifier and once the threshold of audibility has been surpassed, how innocuous the coloration will be.

    The harmonic products also can vary in amplitude in respect to one another as the signal level is increased. It is more desirable to have an amplifier that exhibits a similar distortion profile throughout a reasonable output power range.

    Now why not just remove most of the distortion so that its all way below the audibility of distortion? The answer is that you can, but most designs that have accomplished this have done it at the expense of other critical performance characteristics. In other words, compensation is audible even when keeping phase and gain margins to acceptable levels. 😉

  13. The distortion numbers don’t carry a lot of weight for me. What is more important is how they achieved those low numbers. If the designer is using a lot of feedback I don’t care how low those numbers are. Chances are the amp with zero feedback will sound better than the lower distortion amp with significant feedback.

  14. Paul,
    I am sorry, but I think you are being purposely misleading. 0.1% distortion is just 60dB below the signal. That is not a lot and most likely audible in certain frequencies. You want to have your electronic equipment to get to what is measurably audible. And this is likely 100 to 120 db below the signal. So, you need to “target” a 0.001% to be 100dB below the signal and feel comfortable that it will be “transparent” to the ear. And this is for the combination of noise and distortion.

    What is fascinating is speakers at different levels have distortion that is much higher in level. You re lucky if you can get distortion 50dB below the signal at most frequencies and much higher in the bass when listening at 100dB loudness. But, even with those figures, you can likely “hear” a distortion that is 60dB below the signal in the electronics. Our brain ca do that. We are also much more sensitive to frequency variations as well. Just 1dB at mid frequencies is likely “hearable”.

    You just continue to fuel mythology with statements like this. They also contradict in part what your speaker designer Chris states.

    Many people like their electronics to be part of the “production” of music. I want them to be part of the reproduction, so a target to be below 100dB of the signal is good enough for me. Chasing a number beyond that is not really necessary. Even is my Class D amp and DAC do measure significantly better than that. It was no by my target to get that.

    1. I am certainly not intentionally misleading in my words. That’s just not me and it never has been. You may not agree with me but I’d appreciate a little courtesy of not suggesting I am making misleading statements.

      Thanks. I know it’s hard not to take jabs at what folks think is happening “behind the scenes” but just because we do not see the other person’s point does not logically follow that there’s malice in one’s words.

      1. Sometimes these posts and replies are difficult to read. Differences and beliefs in approaches to 2 channel audio are abundant. Questioning someone’s integrity or honesty in this type of forum just seems to be uncalled for. Whether it’s about other posters and what some see as redundant and disagree with, or whether it’s about Paul and PSA. A criticism should be welcome, but a question of integrity or motivation crosses a line. It shouldn’t be hard to express an argument of why one feels one approach is better than another. It also shouldn’t be hard to realize that differences exist and should be easy to keep people to people bias out of the equation.

        I know that emotions can run high at times depending on the subject matter.

        The fun part of reading and responding here is the occasional chuckle, the thought process of approaches, and most importantly the ideas of improvements that can be made to existing systems.

        1. Thanks, Mike, I couldn’t agree more. I love differing points of view and the controversies that arise from them. Its what keeps it interesting and helps us grow. Sometimes even a little bit of change seeps in and that’s great!

          1. Paul,
            If I believed that you did not know the difference between distortion (and noise) measurements in dBs or percentage, then I would gladly give you the benefit of the doubt.
            However, I completely assume that you do know the difference. That you are an engineer well versed in the issues. That, as you mentioned a few days ago, the engineering talk gets you excited.
            Thereby, I cannot believe that you were completely honest in your statement or got confused. Unless, you also believe that you cannot hear distortion figures around 60dB below the signal. So, which one is it? Can you hear distortion at 60dBs below or you used an inappropriate figure of percentage? That may be the third scenario too.

            By the way, I am sure you know of these people. They stated something similar to my point years ago:
            https://benchmarkmedia.com/blogs/application_notes/interpreting-thd-measurements-think-db-not-percent

            1. Thanks. In my experience the threshold for audibility of simple THD is about 0.1%, maybe 0.05%. That may run counter to your experience but I’ve spent a fair amount of time working with people (and listening myself) to what matters and where that cutoff lies. So, basically, yes, I still stand by that statement.

              However, that was never the point of the post. The post was to point out, as Darren confirmed, that the biggest contributor to sound quality is not how much (to a point) but how it got there. If one builds, for example, a circuit with zero global feedback but plenty of degeneration and balancing techniques to lower distortion, that will almost always sound far better than a more classic circuit that relies upon global feedback to attain the identical distortion number.

              That’s the point of the post. Wheteher that point is 0.05% or 0.005% isn’t relevant in the bigger picture. What’s relevant is HOW one gets there.

    2. Great points. I think Paul was careless, but to say it was intentional seems unknowable, unlikely and unnecessary. It was a short blog post, so let’s cut him some slack.

      The line that jumped out at me in Paul’s post was this: “From my perspective, once we hit THD and IM levels below 0.1% the differences become more academic than audible.” That’s certainly not true — or at least it requires a lengthy explanation. As Darren says above, the trick is to lower overall distortion without introducing more audible kinds of distortion.

  15. The “zero distortion” amps — how do we like them? QUAD feed forward 405 was never a favorite. The Halcro zero point…. was well received. Now we have the THX Aaa789 module from Laurie Funchal lab with unmeasurable distortion. Benchmark power amp used it. Opinions? I got the headphone amp and love it, don’t miss those euphoria second harmonics.

    1. Hi Peter,
      For what it’s worth, I have a friend in England with a pair of Harbeth 40.1 loudspeakers
      & a Yamaha AS-1100 who borrowed a Benchmark AHB-2 for the weekend.
      Even though the AHB-2 had ridiculously low distortion figures he reported to me that
      the difference in sound quality was negligible & therefore he was not willing to ‘upgrade’.

      1. Indeed. You have to optimize the room acoustics and the loudspeakers before marginal differences in sound quality of today’s power amp designs are resolved and become audible. My favorite audio dealer and me we couldn’t even hear significant differences in sound quality between my old (25 years) TacT Millenium integrated amp and a state of the art Devialet pizza box both driving a megabuck loudspeaker costing 180 k$. And isn’t it strange that most recording studios are operated using active loudspeakers?

  16. So the devil makes me ask…
    Is he a bit deaf
    Is the rest of his gear not up to it?

    Someone has to ask.
    Or as Darren points out, how you handle the distortion matters.

    Do catch his weekly “thehigipodcadt”

  17. Ears hear very differently than machines. THD is an EASY machine measurement, but fairly meaningless to ears, or at least poorly correlated with audibility.

    I participated in a great anecdotal experiment in audibility at a meeting of the New York Audio Engineering Society. The presentation was a distortion box that generated variable amounts of different distortion modalities.

    Even order harmonic distortion was hard to detect on simple signals up to 5%, even for a room full of veteran audio engineers – unless the fundamental and harmonics are widely spaced on the loudness curve (which is why you need low distortion subwoofer drivers).

    Odd order harmonic distortion was more audible, with our hands going up between .5% and 1%.

    Inter-modulation, as in standard tests, was audible at .1%. Crossover distortion provoked negative responses at .002-.01%.

    The rule is ANHARMONIC DISTORTION can be heard by ears at barely measurable levels; and amplifier distortions in general are more audible than speaker distortions, in part because speaker distortions are nearly universal.

    There are also distortions that increase with nearly every technique of bringing THD down: temporal and transient distortions; and then there are distortions that are implicit like the spatial distortions of speakers and rooms – not to mention bass modes, comb filtering, etc.

    If you never hear music without these distortions, they become like water to fish – transparent.

    So measure TIM, Cepstral response, phase response (ears are good to 3 microseconds differential), Doppler Inter-Modulation, transient envelope distortion, response to MNoise (18dB crest factor impulses in top octave), waterfalls over 4PI steradians and inter-modulation between the anharmonic overtone series of skins and metallophones, and then we can talk about who is getting it right.

    Marketing types want to reduce audio quality to a handful of scalars – but that will never be correct. Hearing is too complex for the consuming public to grasp and respond properly to audio data, and their ears are so conditioned by bad content, acoustics and audio gear they can’t recognize audio quality beyond mid-fi either.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301
1-800-PSAUDIO

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram