THD for headphones. What is it?

[AustinsGuitar]

I see people online talking about the hd650's having some legendary <0.05 THD number. What does that mean in practice terms? Sorry its hard to understand what total harmonic distortion means from a listenable stand point. I have a degree in physics, but I am very new to audio. Also what is an audible noise floor? I see that measured in a lot of the dac and amp reviews on here under FFT level(DBrA vs Frequency) as a negative number. I would like to know this.

Thanks a lot in advance <3. Just got the hd650's and i absolutely love them .

 

[RayDunzl]

I don't have a degree in physics, but here are some thoughts:

If measuring, Harmonic Distortion is the generation of tones, by a converter, an amplifier or transducer, related by integer multiples of the fundamental frequency, that are not present in the source sine wave...

Blow on a beer bottle, it nearly produces a clean sine wave. The same "note" (fundamental frequency) would sound very different on a trumpet, due to the harmonic series the trumpet produces. You would generally prefer that the audio equipment not add its own harmonics. All sounds other than a pure sine wave are simultaneous combinations of different frequencies at different amplitudes.

0.05% THD means the harmonics are at least 66.02dB quieter than the fundamental being measured.

An audible noise floor is noise you can hear, especially when no music is playing. -60dB , might be an audible level, depending on the frequency spectrum of the noise. Lows are harder to hear than highs (see Fletcher Munson)

Decibels are a logarithmic scale, and relative to some chosen level. 0dB, when talking signal, is usually defined as the full signal value, lesser values are related by negative decibels.

The relation can be positive (greater than) or negative (less than) whatever reference you have assigned as 0dB. For Sound Pressure Level, 0dB is defined.

I have HD650 too. I think of them as "the safe choice".

Here is the spectrum of my voice saying "ooooohhhhhh". Fundamental about 200Hz, harmonics at multiples of 200Hz, and some noise...

Here is blowing across the top of a bottle. The fundamental stands alone, a low level of harmonics, much closer to a sine wave.

The moral of the Harmonic Distortion story:

If you add to or change the harmonic series (Harmonic Distortion), you change the sound.

And the waveforms:

Testing uses sine waves extensively, as they can be analyzed very precisely for their level of perfection.

 

[solderdude]

I see people online talking about the hd650's having some legendary <0.05 THD number. What does that mean in practice terms? Sorry its hard to understand what total harmonic distortion means from a listenable stand point. I have a degree in physics, but I am very new to audio. Also what is an audible noise floor? I see that measured in a lot of the dac and amp reviews on here under FFT level(DBrA vs Frequency) as a negative number. I would like to know this.

Thanks a lot in advance <3. Just got the hd650's and i absolutely love them .

THD = Total Harmonic Distortion is the difference between the actual signal and all the distortion products in the audible range.
It is either displayed in percentage or dB.
The lower the number the less distortion the better the fidelity.

Measuring distortion requires A: a very quiet room to test in, especially for low frequencies so must be an isolated chamber.
Its also requires a microphone that distorts less than the driver even at high SPL.
Furthermore it requires a very low noise microphone and microphone pre-amp.
Also the amplifier driving the headphone must have low distortion and be able to power the headphone.

Measuring distortion properly thus is a difficult task.
Rtings does a great job at this. Compare headphone performance HERE
My distortion measurements suck as I don't have A: an isolated chamber (could make one, too lazy) B: have a cheap crappy mic C: have a noisy pre-amp. So I cannot measure below 0.2%. BUT what it can show is problem areas that are quite likely audible and that's what I want to know.

What you will see is some headphones will perform better below 1kHz and others better above 1kHz. Our ears are most sensitive around 3-4kHz so you want distortion lowest there. We are less sensitive to low frequencies so a rise there is less important.

You will also see that some headphones distort less at higher SPL and others distort more at lower SPL.
When you see lower distortion at higher SPL than the distortion products are lower than the noise floor at 90dB SPL measurements.
When you see distortion numbers go up at higher SPL the driver is going out of its linear range (too much excursion) or the material for the membrane starts to warp.

One can measure distortion at a single frequency (usual ones are 500Hz and 1kHz) and they may/can/do produce different values as well as at different SPL levels. One can also do a sweep which could tell more. Rtings sweeps and (not unimportant) weighs the frequency to generate a number closer to how it would be perceived.
In such a case the distortion number could still be low even when the lowest bass or frequencies above 10kHz show a rise.

Now the magic number. 0.03% = -70dB. The dynamic range when listening to music at average to loud levels is about 70dB. This means that distortion below 0.03% we simply cannot hear when playing music.
Unfortunately as we can see with multitone measurements there may be higher distortion products at that moment than that of a single tone.
So in some cases certain distortion may still just become audible in some specific cases.

 

[AustinsGuitar]

Thanks so much guys I think I am getting it! I like the science behind all this audio stuff. Acoustics class in college was my passion, but It really had just to do with creating good ported enclosures for speakers. It would shock people how hard it is to make a good closed speaker system!

Again thank you guys!

 

[aussie43]

Hi,
This is an old topic, but very topical for me at the moment.

Recently I bought the Sony WH-1000XM3 headphones. It was a good deal, and I expected them to be better than my current ATH-SR6BT.
It turned out - yes and no. Sure, they have noise cancellation, but when I play my favorite vocal pop and opera music, there are many spots where they distort.
To remove the distortions, I need to EQ -10db in the range between 2.8kHz - 6 kHz.

Searching for answers, I found rtings.com where they measure lots of things. They also measure harmonic distortion. The XM3's have a decent distortion. Much higher than what you say is the treshold of hearing it. And just in the frequencies I needed to EQ. (The range where I need to EQ is wider than the graph below would suggest, but that could be just due to differences between individual copies, or my ears?)

Could that be the answer to my problems, or have I misunderstood harmonic distortion?

 

[spacevector]

I don't have a degree in physics, but here are some thoughts:

If measuring, Harmonic Distortion is the generation of tones, by a converter, an amplifier or transducer, related by integer multiples of the fundamental frequency, that are not present in the source sine wave...

Blow on a beer bottle, it nearly produces a clean sine wave. The same "note" (fundamental frequency) would sound very different on a trumpet, due to the harmonic series the trumpet produces. You would generally prefer that the audio equipment not add its own harmonics. All sounds other than a pure sine wave are simultaneous combinations of different frequencies at different amplitudes.

0.05% THD means the harmonics are at least 66.02dB quieter than the fundamental being measured.

An audible noise floor is noise you can hear, especially when no music is playing. -60dB , might be an audible level, depending on the frequency spectrum of the noise. Lows are harder to hear than highs (see Fletcher Munson)

Decibels are a logarithmic scale, and relative to some chosen level. 0dB, when talking signal, is usually defined as the full signal value, lesser values are related by negative decibels.

The relation can be positive (greater than) or negative (less than) whatever reference you have assigned as 0dB. For Sound Pressure Level, 0dB is defined.

I have HD650 too. I think of them as "the safe choice".

Here is the spectrum of my voice saying "ooooohhhhhh". Fundamental about 200Hz, harmonics at multiples of 200Hz, and some noise...

View attachment 44294

Here is blowing across the top of a bottle. The fundamental stands alone, a low level of harmonics, much closer to a sine wave.

View attachment 44296

The moral of the Harmonic Distortion story:

If you add to or change the harmonic series (Harmonic Distortion), you change the sound.

And the waveforms:

View attachment 44298

Testing uses sine waves extensively, as they can be analyzed very precisely for their level of perfection.

Can you do a whistle one day please?

 

[RayDunzl]

Can you do a whistle one day please?

Whistle with mouth?

Very clean.

I will note that I haven't whistled for a long time and can barely get a note out.

Something to practice, or maybe my lips are too old.

 

[spacevector]

Whistle with mouth?

Very clean.

View attachment 99726

I will note that I haven't whistled for a long time and can barely get a note out.

Something to practice, or maybe my lips are too old.

Nice, similar to beer bottle. Much obliged.

 

[solderdude]

Searching for answers, I found rtings.com where they measure lots of things. They also measure harmonic distortion. The XM3's have a decent distortion. Much higher than what you say is the treshold of hearing it. And just in the frequencies I needed to EQ. (The range where I need to EQ is wider than the graph below would suggest, but that could be just due to differences between individual copies, or my ears?)

The distortion plots of Rtings are 'weighed'. This means that distortion in the lows is higher than it appears to be in the plot. The intention of this weighing is to show that higher distortion in the lows isn't that audible (and indeed it isn't). What the plot shows is 'perceived' distortion.

Here's the distortion of the WM1000XM3 in its 3 different modes (passive, so powered off, on but no N.C. and with N.C. on (which lowers bass distortion by design)

 

[hctrivedi3]

I don't have a degree in physics, but here are some thoughts:

If measuring, Harmonic Distortion is the generation of tones, by a converter, an amplifier or transducer, related by integer multiples of the fundamental frequency, that are not present in the source sine wave...

Blow on a beer bottle, it nearly produces a clean sine wave. The same "note" (fundamental frequency) would sound very different on a trumpet, due to the harmonic series the trumpet produces. You would generally prefer that the audio equipment not add its own harmonics. All sounds other than a pure sine wave are simultaneous combinations of different frequencies at different amplitudes.

0.05% THD means the harmonics are at least 66.02dB quieter than the fundamental being measured.

An audible noise floor is noise you can hear, especially when no music is playing. -60dB , might be an audible level, depending on the frequency spectrum of the noise. Lows are harder to hear than highs (see Fletcher Munson)

Decibels are a logarithmic scale, and relative to some chosen level. 0dB, when talking signal, is usually defined as the full signal value, lesser values are related by negative decibels.

The relation can be positive (greater than) or negative (less than) whatever reference you have assigned as 0dB. For Sound Pressure Level, 0dB is defined.

I have HD650 too. I think of them as "the safe choice".

Here is the spectrum of my voice saying "ooooohhhhhh". Fundamental about 200Hz, harmonics at multiples of 200Hz, and some noise...

View attachment 44294

Here is blowing across the top of a bottle. The fundamental stands alone, a low level of harmonics, much closer to a sine wave.

View attachment 44296

The moral of the Harmonic Distortion story:

If you add to or change the harmonic series (Harmonic Distortion), you change the sound.

And the waveforms:

View attachment 44298

Testing uses sine waves extensively, as they can be analyzed very precisely for their level of perfection.

Hello @RayDunzl !

With reference to your answers, I'd a couple of questions. Can you please help?

1. How did you manage to plot this graphs? Is there any such tool available? Can you please share the details?
2. Is there any standard limit / reference value of THD available for "Voice Communication". i.e. How would I determine that for my voice communication system (200 - 6000 Hz), ___ value of THD is good enough?

Thanks in anticipation!

 

[RayDunzl]

1. How did you manage to plot this graphs? Is there any such tool available? Can you please share the details?

REW's Real Time Analysis - RTA - window and a UMIK-1 to feed it.

2. Is there any standard limit / reference value of THD available for "Voice Communication". i.e. How would I determine that for my voice communication system (200 - 6000 Hz), ___ value of THD is good enough?

I suspect it would be rather high.

I've truncated 16bit to 1bit and it is still intelligible.

Bellcore or ITU-T must have specified it sometime in the past for network operations.


Examples:

https://www.icc.illinois.gov/docket/P2000-0393/documents/35640/files/63067.pdf

https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-G.113-199602-S!!PDF-E&type=items

 

[hctrivedi3]

REW's Real Time Analysis - RTA - window and a UMIK-1 to feed it.




I suspect it would be rather high.

I've truncated 16bit to 1bit and it is still intelligible.

Bellcore or ITU-T must have specified it sometime in the past for network operations.


Examples:

https://www.icc.illinois.gov/docket/P2000-0393/documents/35640/files/63067.pdf

https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-G.113-199602-S!!PDF-E&type=items

@RayDunzl ,

Thanks a bunch for the prompt response. I'll go through the tool as well as documents.