Total Harmonic Distortion
Jul 21, 2019 at 1:50 PM Thread Starter

#### SilentNote

A quick search did not reveal a clear explanation. In simple language, what is distortion, harmonic distortion, and total harmonic distortion?

How much THD does it take to be audible? Does it increase with increasing volume? Does it matter if I never listen past 90dB if or where the THD is relatively low?

Does it have an impact on tonality? What is it’s relationship with frequency response if any?

Last edited:
Jul 21, 2019 at 2:25 PM
Briefly:

In an ideal amplifier, the output is an exact replica of the input, only larger. Any deviance from that is called distortion.

If you feed a pure sine wave of frequency F into an amplifier, the output will typically show additional artifacts at integer multiples of F. These can be observed and measured by running the output into a spectrum analyzer, which shows the energy at a range of frequencies in a signal. These additional artifacts generated by the amp at integer multiples of the input signal are called harmonic distortion. The energy at frequency 2F is second harmonic distortion, at 3F is third harmonic distortion, etc. Adding up (after a fashion) all of these artifacts gives the total harmonic distortion (THD).

The audibility of harmonic distortion varies depending on the listener, of course. Higher harmonics are generally considered more audible, and odd harmonics are thought to be more euphonic (and thus less noticeable/objectionable) than even ones. Since THD is a sum of all the components, it really doesn't tell you the full story of how you'll perceive the harmonic distortion. THD generally does increase at higher volume. At the level where the amp starts to clip, it increases exponentially. At what dB level the listener starts to notice it depends on the listener, type of music, and other factors.

Generally speaking, intermodulation distortion (IMD) has a more deleterious effect on sound quality than harmonic distortion, since the frequencies generated are the difference between frequencies, and thus generally not harmonic to either. In most well-engineered modern (particularly solid state) amps, neither THD or IMD is likely to be egregious, but some listeners may still perceive them.

Harmonic distortion does have a small effect on frequency response in an amp, but other factors generally contribute more to that.

Last edited:
Jul 21, 2019 at 2:44 PM
https://www.rane.com/note145.html

to talk in a more intuitive way, 1%THD means 40dB below the original signal. such a value doesn't fully characterize the harmonic distortions, it's a little like when we discuss jitter, we just give a value for the max impact of whatever is going on. if it's low we don't care. if it's high, we'll probably care more about getting a new device than a fully detailed representation of how bad it is ^_^. so for a consumer at least, just a basic value is usually good enough to make our mind and avoid garbage products.

Jul 21, 2019 at 3:34 PM
In the first AES video in my sig file, there is a test where they take a horrible buzzing sound- the worst kind of sound possible- and they lower it progressively under music until it can’t be heard any more. You can download the test samples. The links are in the YouTube description. -40dB is the point where it becomes pretty much inaudible for normal listening. It’s hard to think of a form of distortion that would be worse than this buzzing noise.

Last edited:
Jul 21, 2019 at 5:08 PM
In the first AES video in my sig file, there is a test where they take a horrible buzzing sound- the worst kind of sound possible- and they lower it progressively under music until it can’t be heard any more. You can download the test samples. The links are in the YouTube description. -40dB is the point where it becomes pretty much inaudible for normal listening. It’s hard to think of a form of distortion that would be worse than this buzzing noise.

I’ve watched the video. Very informative indeed. I can hear the very conspicuous distortion down to -50dB on the pop mix and -70dB with the quiet passage. Though that is on cue. I’ll most definitely miss it without knowing it’s there. Also I assume most distortion don’t sound nearly as bad so that’s the absolute worst case scenario.

What affects the timbre of instruments? Is it mostly frequency response?

Last edited:
Jul 21, 2019 at 5:55 PM
What do you mean by on cue? The way to listen to music is to set the volume for a comfortable peak level and listen to it all the way through at that level. With noise at -40dB, you might faintly hear something in the dead grooves between tracks if the volume is up, but it isn't going to make a difference while music is playing, no matter how softly. Recorded music doesn't generally have a dynamic range far beyond 50dB. I don't believe you can hear -70dB with anything. That would be down with the air conditioning room tone in the recording studio. Also, doesn't the sample in the video just go down to -45dB and stop? Where did you get the -50 and -70 samples from?

I bet you listen to things with distortion levels well above 1% all the time and you don't even notice it.

Last edited:
Jul 21, 2019 at 6:12 PM
The video started with -20db -30db etc and I could still hear the distortion when the video shows that it’s at -50db. But that’s because I was paying attention to the timeline and seeing when the distortion was playing. That’s what I meant on cue.

The second portion of the video was a softer music playing and it did the same thing starting from -30db down to -75db.

Without the visual cue I can semi-reliably pick out the noise when the video shows -40db and -60db respectively.

AES Audio Myths Workshop @ 32:50

Last edited:
Jul 22, 2019 at 2:16 AM
If you need your eyes to hear, I would suspect that bias might be affecting your discernment. I doubt you'd be able to hear -50dB in the quiet music at a normal volume.

Jul 22, 2019 at 2:32 AM
If you need your eyes to hear, I would suspect that bias might be affecting your discernment. I doubt you'd be able to hear -50dB in the quiet music at a normal volume.

Yes, that’s what I’m saying.

Without my eyes and not knowing what to look for I will not be able to pick out -50db on the pop mix and -70db on the concerto.

Thank you all for sharing this information freely!

Last edited:
Jul 22, 2019 at 2:42 AM
You should try doing it blind. Use the downloadable FLAC files on the page linked from the description.

Jul 22, 2019 at 2:52 AM
You should try doing it blind. Use the downloadable FLAC files on the page linked from the description.

I don’t have to do a abx blinded test to realize that in real world scenario where I’m focusing on the music + the distortion noise wouldn’t be as obvious + I usually listen at ~70db, -40db is as good as inaudible.

Jul 22, 2019 at 7:31 AM
I don’t have to do a abx blinded test to realize that in real world scenario where I’m focusing on the music + the distortion noise wouldn’t be as obvious + I usually listen at ~70db, -40db is as good as inaudible.

I am trying not to post anything in this thread because I am so non-expert. So I will only ask questions. A few things are puzzling me. Aside from the general idea that harmonic distortion is generally not audible on decent modern solid state amps, are we not leaving some unsettled or unverified matters on the table?

Doesn’t the cited video state that you should download the samples because YouTube will compress the files? Would you believe me if I told you that I get absurd results with the video through YouTube even casually listening through an iPad speaker, extremely similar to what @silvernote was reporting? Do you think this might be due to the YouTube compression? If so would you think that listening to the test on my iPad remotely achieves the intent of the test, as opposed to downloading the files and making a blind comparison?

Could someone please provide a reference for the idea that lower-order even harmonic distortion is less euphonic than higher order harmonic distortion? Doesn’t a lot of music use parallel octaves because of their pleasing effect? So why would 2nd order harmonic distortion (wouldn’t that be distortion one octave up?) not be euphonic?

What level of harmonic distortion is typical for a pair of headphones?

What level of harmonic distortion is typical for a pair of loudspeakers?

How does this compare to the amount of harmonic distortion being created by a competent modern solid state amp?

Can someone give me an example of a properly operating modern solid state speaker amplifier being used as intended and not clipping and not being audibly transparent over loudspeakers when listening to music?

Don’t tube amps sometimes introduce impedance anomalies so severe as to introduce audible changes in frequency response when driving a pair of loudspeakers that is not a good impedance match for that particular amp? Could this change in frequency response be mistaken for harmonic distortion in those circumstances?

Jul 22, 2019 at 7:46 AM
FYI I don't think my results are that crazy.

With visual cue (listening with the eyes):
- Pop song with -1dB: I can notice down to -50dB
- Concerto at -30dB: I can notice down to -70dB
In both cases the noise is -40~50 dB under the signal

Without visual cue (eyes closed):
- Pop song with -1dB: I can notice down to -40dB
- Concerto at -30dB: I can notice down to -60dB
In these case the noise is -30~40dB under the signal

Real world (estimated, harmonic distortion instead of noise, focus on music, masking effect etc.):
- Pop Song with -1dB: Probably notice down to -30dB
- Conceto at -30dB: Probably notice down to -50 dB
In this estimate it'd be only -20db~30dB under the signal

Which means i probably can't reliably pick up 5-10% harmonic distortion in real life.

Jul 22, 2019 at 7:53 AM
Not sure why folks would signal a harmonica out for distortion

Jul 22, 2019 at 9:16 AM
FYI I don't think my results are that crazy.

With visual cue (listening with the eyes):
- Pop song with -1dB: I can notice down to -50dB
- Concerto at -30dB: I can notice down to -70dB
In both cases the noise is -40~50 dB under the signal

Without visual cue (eyes closed):
- Pop song with -1dB: I can notice down to -40dB
- Concerto at -30dB: I can notice down to -60dB
In these case the noise is -30~40dB under the signal

Real world (estimated, harmonic distortion instead of noise, focus on music, masking effect etc.):
- Pop Song with -1dB: Probably notice down to -30dB
- Conceto at -30dB: Probably notice down to -50 dB
In this estimate it'd be only -20db~30dB under the signal

Which means i probably can't reliably pick up 5-10% harmonic distortion in real life.

Is the distortion in that video harmonic distortion?

Did you test in a way the video said you should not test?

Were your results sighted or blind?

Is there any reason to think that your calculations might not be valid?