• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

THD has a much bigger effect on sound than you think

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
Anyway I have done many blind tests regarding linear harmonic distortion and i never were able to tell below 1% THD in music. I think whatever you experienced was actually non-linear.

Think about this way, the second harmonic acts as the fundemental for the third harmonic (masking). So it's only possible to hear harmonic distortion when it's non-linear (third harmonic louder than second .etc) or if it was very high in output.

However with IMD things can build up to 10% rather quickly specially with bookshelf speakers. it's a much more valid concern.
 
OP
M

mike7877

Addicted to Fun and Learning
Joined
Aug 5, 2021
Messages
631
Likes
127
Is it, though? You need to do some reading about masking.

It is in my experience. I guess I have experience with genuinely high fidelity gear, that might be a requisite.
 
OP
M

mike7877

Addicted to Fun and Learning
Joined
Aug 5, 2021
Messages
631
Likes
127
Anyway I have done many blind tests regarding linear harmonic distortion and i never were able to tell below 1% THD in music. I think whatever you experienced was actually non-linear.

Think about this way, the second harmonic acts as the fundemental for the third harmonic (masking). So it's only possible to hear harmonic distortion when it's non-linear (third harmonic louder than second .etc) or if it was very high in output.

However with IMD things can build up to 10% rather quickly specially with bookshelf speakers. it's a much more valid concern.

The third harmonic isn't masked by the second. 1000hz to 2000hz is second harmonic, 3000hz is third. 3000 isn't double 2000

edit: the reason second harmonic isn't very audible is because it's an octave higher - it sounds like it could be there naturally, or could have been put there creatively. Third is a clash

edit 2: I think you're right though, with third being louder than second it would sound worse. Second is not always or even usually louder than third, unless you're talking about the loudspeaker drivers
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
The sound energy in music in the 1-4khz range is 30-40db lower than bass, 15-30db than midrange.

Here's the spectrum of music I was listening to.

spectrum.png


1-4 kHz range is about 20dB below bass and the song above had plenty of bass.

In electronics we see mostly 2nd or 3rd and at lower levels 4th and 5th.
Also keep in mind that instruments have lots of harmonics by themselves that are higher (and thus masking) than the ones produced by DACs and amps.

You can actually listen to distortion products when you null input and output of an amp.

Below a recent measured DAC
index.php


Now imagine the 1kHz were 50Hz and the 1-4kHz range was 20dB lower than the 0dB.
Then there would still be a distortion distance for the 5th harm (250Hz) of 105dB
Add to that music signals (incl harmonicsof the instrument itself) are just 10dB down and overpowering the distortion products by 95dB = 56000x times higher.
And the DAC above is not a special one.
When listening to music 70dB is about all the dynamic range your hearing can handle.

Play some music at impressive levels. Now attenuate 70dB... what do you hear ? and at -80dB or -90dB ? Why worry about anything below your hearing threshold ?
 

dfuller

Major Contributor
Joined
Apr 26, 2020
Messages
3,340
Likes
5,063
It is in my experience. I guess I have experience with genuinely high fidelity gear, that might be a requisite.
I...do? I've done the tests myself with very low distortion equipment, speakers especially (ATCs and Neumanns, both of which average at 0.1% THD or lower in the midrange) - I can hear about 0.5% THD in program material which is far better than most (that's 50 dB down, by the way...). The way distortion works it's almost always low orders (i.e. HD2, HD3) dominating and the other ones fall away from that.
 
Last edited:
OP
M

mike7877

Addicted to Fun and Learning
Joined
Aug 5, 2021
Messages
631
Likes
127
Here's the spectrum of music I was listening to.

spectrum.png


1-4 kHz range is about 20dB below bass and the song above had plenty of bass.

In electronics we see mostly 2nd or 3rd and at lower levels 4th and 5th.
Also keep in mind that instruments have lots of harmonics by themselves that are higher (and thus masking) than the ones produced by DACs and amps.

You can actually listen to distortion products when you null input and output of an amp.

Below a recent measured DAC
index.php


Now imagine the 1kHz were 50Hz and the 1-4kHz range was 20dB lower than the 0dB.
Then there would still be a distortion distance for the 5th harm (250Hz) of 105dB
Add to that music signals (incl harmonicsof the instrument itself) are just 10dB down and overpowering the distortion products by 95dB = 56000x times higher.
And the DAC above is not a special one.
When listening to music 70dB is about all the dynamic range your hearing can handle.

Play some music at impressive levels. Now attenuate 70dB... what do you hear ? and at -80dB or -90dB ? Why worry about anything below your hearing threshold ?

With very loud music on the second floor I can tell if my furnace is running. It isn't loud and it takes away detail. And the furnace is just noise, not harmonics which clash with the fundamentals in the music. Anecdotal of course
 
OP
M

mike7877

Addicted to Fun and Learning
Joined
Aug 5, 2021
Messages
631
Likes
127
I...do? I've done the tests myself with very low distortion equipment, speakers especially (ATCs and Neumanns, both of which average at 0.1% THD or lower in the midrange) - I can hear about 0.5% THD in program material which is far better than most (that's 50 dB down, by the way...).

ATC is great :D.

I'm not talking immediately obvious. Try 0.4% for a week on everything, then take it off. Then 0.3%, 0.2%, 0.1% 0.05%. I expect once the odds get to about 1/4 the level of the speakers, further decreases would be inaudible
 
OP
M

mike7877

Addicted to Fun and Learning
Joined
Aug 5, 2021
Messages
631
Likes
127
Think about it this way: you can tell that the ATCs and Neumanns are clearer than lesser speakers. Some of what makes them sound better is their flat frequency response free from peaks, and some is their wider than average dispersion from reasonable crossover points. And some is the very, very low harmonic distortion.
Of course they make very good speakers with excellent drivers, but they are not perfect, and one day they will be bested by new technology and better materials lowering distortion and crossover points further
 

voodooless

Grand Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
10,229
Likes
17,811
Location
Netherlands
I think average might not be the best way to look at that, lower notes tend to be a lot more transient than higher notes.
By definition, they are not.

With very loud music on the second floor I can tell if my furnace is running. It isn't loud and it takes away detail. And the furnace is just noise, not harmonics which clash with the fundamentals in the music. Anecdotal of course
How about you stop with the anecdotes and platitudes and actually demonstrate what you claim? If you actually want to convince people, that is the very best way.
 

Killingbeans

Major Contributor
Joined
Oct 23, 2018
Messages
4,089
Likes
7,552
Location
Bjerringbro, Denmark.
I think average might not be the best way to look at that, lower notes tend to be a lot more transient than higher notes.

Isn't the whole point of low notes that they are slow? "Fast bass" is a misnomer?

A transient is a sum of a spectrum. It can't exist without high frequency content.
 

Dro

Active Member
Joined
Apr 8, 2018
Messages
221
Likes
205
That does not stop low frequencies from peaking very high, especially compared to the rest of the spectrum at that time.

What I am trying to say is that low notes, while physically not able to be "fast" like higher notes tend to be very punchy (bass drum), whereas higher notes tend to come with more sustain in actual music, especially music where you have bass at levels that get interesting for distortion. So looking at a full song spectrum or entire decades of music seems foolish when try to determine how loud bass peaks get compared to the rest. That does not mean I think 0.1% THD is a requirement in bass. Unless maybe it is 0.1%, five octaves higher, but neither amplifiers nor speakers behave like that.
 
Last edited:

Pdxwayne

Major Contributor
Joined
Sep 15, 2020
Messages
3,219
Likes
1,172
Here's the spectrum of music I was listening to.

spectrum.png


1-4 kHz range is about 20dB below bass and the song above had plenty of bass.

In electronics we see mostly 2nd or 3rd and at lower levels 4th and 5th.
Also keep in mind that instruments have lots of harmonics by themselves that are higher (and thus masking) than the ones produced by DACs and amps.

You can actually listen to distortion products when you null input and output of an amp.

Below a recent measured DAC
index.php


Now imagine the 1kHz were 50Hz and the 1-4kHz range was 20dB lower than the 0dB.
Then there would still be a distortion distance for the 5th harm (250Hz) of 105dB
Add to that music signals (incl harmonicsof the instrument itself) are just 10dB down and overpowering the distortion products by 95dB = 56000x times higher.
And the DAC above is not a special one.
When listening to music 70dB is about all the dynamic range your hearing can handle.

Play some music at impressive levels. Now attenuate 70dB... what do you hear ? and at -80dB or -90dB ? Why worry about anything below your hearing threshold ?
Your spectrum is based on how many seconds of song? Since this just represent the peak levels of the whole time period, it can't really tell second to second bass to treble balance, correct?

Here is a 10 seconds analysis of a song I like. Once you start reducing the total time period if analysis, you can get more difference:
30_40_seconds.PNG.png
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
whole song.
Yes, with shorter pieces you see a bigger dynamic difference.

Let's analyze this plot and assume the amp used has 0.01% THD (-80 dB) where 2nd, 3rd, 4th and 5th are present.
We see bass notes 28Hz (A0) and 32.7Hz (C1)
There will be the following harmonics in the instrument (levels will differ)
56, 65.4, 84, 98.1, 112, 13.8, 140, 163.5 of course there will be more frequencies higher up as well all the way up to kHz range in increasingly small values.
-80dB is for the total harm energy so the individual harmonics will be lower in amplitude. Let's say -85dB each.
The level of the fundamental is -18dB and -20dB.
So the harmonics generated by this (not that great performing) amp will be around -103 and -105dB.
Now lets look at the plot in the range between 50Hz and 200Hz.
We see -40dB as the lowest signal. This is the musical content.
The distortion is 60dB below those signals so masked by the harmonics of the instrument itself.
The distortion signals of a 0.01% THD amp thus is 1000x lower than the music content.
Nope... you are not going to hear that.
Note that those lines will be dancing up and down.
When you would mix in other frequencies (or noise) not related and the dancing 'poles' are absent at those extra frequencies or are constant.. yes you can probably hear those quite far down.

Besides what OP claims is nonsense for electronics unless you move into the tube amp or amp territory where intentionally huge amounts of distortion is produced.
In that case it is not the HD but the IM distortion that becomes an issue.
The comment OP made about the furnace is not relevant to the amplifier. When you can hear the furnace when no music is playing it has already reached audible levels which are much higher than any generated (and masked) distortion even a half decent amp can reach.
 

Balle Clorin

Major Contributor
Joined
Dec 26, 2017
Messages
1,287
Likes
1,185
Interesting Mike, but have you avtually listened to sound from an source that is distorted by 0.1%? I have and is not detectable by ear by me, and genrally below the hearing threshold.
Can you hear the distortion in Vinyl/ cartridge that is in the order of 1-2% in the low range an 5-10% above 10k? Very few are bothered by that when listening .Your speakers may have 10-20% distortion in the bass, you just do not notice it much
Even if anything above 0.2 is audible for single tones.
There are tools to add distortion to music and tones at different levels to find the hearable threshold , it is surprising how much is tolerable.

In fact distortion is MUCH LESS detectable than people just looking on graphs seem to believe .
Your speakers may even have 10-20% distortion in the bass, you just do not know it
 
Last edited:

NTK

Major Contributor
Forum Donor
Joined
Aug 11, 2019
Messages
2,660
Likes
5,820
Location
US East
0.1% in the same octave? No Problemo!

0.1% edit: a few octaves up? Turns into 10%. Problem!!!!!!
OP's premise was that distortions several octaves higher were more audible due to our higher hearing sensitivity and lack of masking. However, OP has yet to provide an explanation on the mystical distortion mechanism that only generates harmonics that are "a few" octaves up and nothing else.

To illustrate how quickly amplitudes of harmonics decay relative to the fundamental, let's look at how discontinuities in signals generate harmonics.
  • An impulse δ(t) has a singularity at t=0, and the amplitudes from fundamental to H∞ are all equal in amplitude.
  • Integrating the impulse gives a step function which has a discontinuity at t=0, and the amplitudes of the harmonics decrease with 1/n, where n=2 is the second harmonic, n=3 is the third harmonic, etc.
  • Integrating again gives a ramp function which has a discontinuity in the slope (first derivative) at t=0. The amplitudes of the harmonics decrease with 1/n^2.
It can be easily seen and reason that the likelihood of finding a naturally occurring distortion mechanism that will create only higher order harmonics with high amplitudes is between none to zilch.

So what it means is that if, for example, there is a high H32 (32nd harmonic, 5 octaves up from fundamental), there will be much much higher H2 to H31, and the total harmonic distortions will therefore be much much higher than when only counting H32. The scenario OP postulated is a physical impossibility.
 

peng

Master Contributor
Forum Donor
Joined
May 12, 2019
Messages
5,615
Likes
5,168
It's in the charts of many amps, I'd say at least a third of them have higher third than second harmonic distortion

Even if true, If the total level (THD+N) is loud enough it shouldn't be an audible issue for most. As to even vs odd low order harmonics, take a look of Marantz vs Denon AVRs on the ASR chart and one may be surprised to see that Marantz's pretty much always have more 3rd than Denon's relatively speaking, yet we keep seeing posts about Marantz sound is warm and musical, relative to Denon's, good figure! I bet in a controlled DBT, using analog inputs and in direct mode even the die hard Marantz warm/musical sound believers wouldn't be able to tell that little difference in the 2nd and 3rd but then who knows, it not proven until it is, I can bet all I want..

I am not suggesting Nelson Pass was right when he said the following, but I bet it may surprise many who admire him:


Many audiophiles believe that 2nd harmonic is to be preferred over 3rd harmonic. Certainly it is simpler in character, and it is well agreed that orders higher than third are more audible and less musical. However when given a choice between the sound of an amplifier whose characteristic is dominantly 2nd harmonic versus 3rd harmonic, a good percentage of listeners choose the 3rd.

I have built many examples of simple 2nd and 3rd harmonic “types” of amplifiers over the last 35 years. When I say “types” I mean that they used simple Class A circuits described as “single-ended” versus “push-pull” and so tended to have a 2nd harmonic versus 3rd harmonic in the character of their distortion, but were not made to deliberately distort.

Anecdotally, it appears that preferences break out roughly into a third of customers liking 2nd harmonic types, a third liking 3rd harmonic, and the remainder liking neither or both. Customers have also been known to change their mind over a period of time.

So much for "facts".., perhaps the real "fact" is, it's more like opinions, preference..;)
 
Top Bottom