Audibility of low frequency distortion in speakers

[March Audio]

That's interesting. I guess 10Hz is a special case in that the harmonics themselves need to be quite loud to pass the absolute threshold of audibility.

True, but only for a fundamental of 10Hz! At 20Hz, you can see that the thresholds stay relatively stable as SPL changes, and by 50Hz, sensitivity is already increasing as SPL decreases.

Also, at least for music, I think that the 10Hz fundamental data point can safely be ignored.

Music just doesn't have content that low.

 

[March Audio]

This time a different typical small speaker at 96dB
THD is about 8.5%

Same monitor at 86dB
THD is about 2.8%

 

[richard12511]

Also, at least for music, I think that the 10Hz fundamental data point can safely be ignored.

Music just doesn't have content that low.

Some music does. Practically, though, I agree. That music is very, very, very rare.

There was a thread on another forum documenting songs with sub 10Hz material, though I haven't been able to find it recently. I think Phoebe Bridger's song "Punisher" has sub 10Hz content in the last minute or so. There are several notes so low that you can't hear any of them, but you can tell that the frequency is getting lower by the fact that your body is vibrating slower.

 

[March Audio]

Are people hearing the 2.8% distortion in the second video?

 

[thewas]

Are people hearing the 2.8% distortion in the second video?

 

[March Audio]

View attachment 98638

What were you listening on? Easy or difficult to hear?

 

[Doodski]

Are people hearing the 2.8% distortion in the second video?

I think I can't hear a difference and then sometimes it seems to be louder when the 2.8% distortion is engaged. I think it is my imagination. Using Sennheiser HD 598 headphones off my ASUS Z370E motherboard.

 

[thewas]

What were you listening on? Easy or difficult to hear?

Funnily my small and quite SPL limited KEF LS50 desktop system (with room correction).
Its still quite easy to hear compared to other distortion tests, except at very low listening levels.

 

[Savi]

Are people hearing the 2.8% distortion in the second video?

Easy to hear using my phone, dense adapt dac and focal clear headphone (low level). Do not know if the test is relevant though.

 

[Pawelekdabek]

Hmmm, actaully with music this 2.8% would be very hard to notice if not impossible. Especially if higher frequencies maintain often much lower levels of distortion.

With 8.5 I feel more like I'm listening to distortion than to the actual sound.

 

[Tangband]

We are not especially sensitive to harmonic distortion in the bass-area, but we are sensitive to phase distortion in that area ( 100 Hz ). That means that a bad integration subwoofer/main loudspeaker can be worse sounding than the measured THD.

 

[March Audio]

Hmmm, actaully with music this 2.8% would be very hard to notice if not impossible. Especially if higher frequencies maintain often much lower levels of distortion.

With 8.5 I feel more like I'm listening to distortion than to the actual sound.

Yep with music distortion will be more difficult to detect. We will be moving on to looking at music .

 

[March Audio]

We are not especially sensitive to harmonic distortion in the bass-area, but we are sensitive to phase distortion in that area ( 100 Hz ). That means that a bad integration subwoofer/main loudspeaker can be worse sounding than the measured THD.

The research shows that we are far more sensitive than is generally stated.

Have a look at the numbers in this post.

https://www.audiosciencereview.com/...ency-distortion-in-speakers.18291/post-597347

 

[ctrl]

I replicated (as far as possible) the distortion signature of a typical small speaker playing at 96dB SPL 1m....
Its best to set the playback to HD and full screen so you can see the numbers.

Cool that you're getting to the bottom of this.

The graphic in post#1 uses narrow band noise as the typical masker at 150Hz, as far as I know, rather than a single tone.
When using single tone masker, the masking is less good.
On the other hand, we would like to simulate the extreme case here in the forum. Like your chosen example, this would correspond to a piece of music with, for example, a synthesizer passage that simply consists of a single tone at 80Hz.
But we should keep in mind that this is not comparable to a typical piece of music.


To rule out other effects such as IMD and Doppler distortion, did you choose a similar test setup as described in Axiomaudio's listening test?

...were played in stereo on a pair of wide-range loudspeakers (Axiom M80ti's) operating with an Axiom EP600 subwoofer. Pure sine-wave tones at fixed frequencies were played over a third M80 speaker and EP600 subwoofer (all speakers and subwoofers were concealed behind an acoustically transparent but visually opaque curtain), at gradually increasing loudness levels until the listener detected their presence and signaled by a raised hand that something in the music doesn't seem right.

Wouldn't it make more sense to establish initial benchmarks for audibility of harmonic distortion at lower sound pressure and then compare them to the existing literature, since reproducing 96dB@80Hz at the listening position will not be possible for most people.

In addition, it must be ensured that the reproduction system (loudspeaker or headphones) does not itself produce additional harmonic and other distortions that distort the result.

At 80Hz@96dB sound pressure at my listening position, a part of my listening room audibly vibrates with the masker tone and distorts the audibility of the harmonic distortion.

Fielder and Benjamin study I believe

Is the study publicly available?

First thing that catches my attention are the figures for 10Hz frequency. At a 100dB volume level 11, 2.2, 0.6 and 0.3% distortion for harmonics 2 to 5 respectively are required to be audible. Yet at 80dB volume 110, 22, 6 and 2.8% distortion is required. Seems odd to be that radically different but its not really.
...
Our hearing sensitivity changes with volume. We can see this effect if we look at the Fletcher Munson curves....
So at lower volumes the distortion simply needs to be relatively much louder before we hear it.

The perceptual thresholds shown at high sound pressure and low masker frequency, in the Fielder and Benjamin study, do not make sense to me and seems to contradict the results of Zwicker et al. I would expect, the higher the sound pressure and the lower the Bark difference between the masking tone and the test tone, the more pronounced the masking effect.

A possible cause could be that the hatched area in the diagram below (Zwicker et al) shows an area in which a difference tone (caused by the ears own nolinear distortion) becomes audible.

If we assume as an example a masker at 150Hz, which corresponds to the central frequency at 2 bark, then according to Zwicker, for the perception threshold of HD3 at 450Hz, which corresponds to about 5 bark, we would expect the values plotted in blue in the diagram below (critical-band rate difference is -3).

The perception threshold should be
at 60dB sound pressure around 1% (-40dB),
at 80dB sound pressure around 3% (-30dB),
at 100dB sound pressure around 4% (-28dB)

What amazed me is that when the sound pressure increases from 80 to 100dB, the perception threshold shifts only slightly upward (3% to 4% at 150Hz, detection threshold for HD3).

There I was quite clearly wrong in my earlier statements (in other threads), because I assumed that the development continues as with the sound pressure increase from 60 to 80dB (1% to 3% at 150Hz, detection threshold for HD3).


In the diagram below, at 100Hz and 50Hz the perceptibility of harmonic distortion behaves about as one would expect according to Zwicker et al.

At 20Hz and 10Hz, the values obtained at 80dB, 100dB and 110dB no longer seem consistent even in the presence of a difference tone.

 

[KSTR]

IME we are very sensitive to any kind of LF abberations:

- Fletcher-Munson characteristic makes simple harmonics much larger in perception than they are, numerically

- we tend to judge LF by actual waveform, and on top of that the judgment is asymmetrical (the sign of the momentary pressure gradient plays a role). Two sines with the same amount of H2 can sound very different depending on the phase of the harmonic and that also explains why music sounds different when inverting polarity, even with the blurring phase contribution overlay of conventional speaker crossovers (and speakers+subs, notably).

- while masking helps a lot, it doesn't look viable to me to rely on it.

As for the ultra-LF stuff (10Hz and lower), I personally found that much of the perception of tones that low is an amplitude modulation of the background (noise floor or gentle music content) and strong distortion there shows up in a slightly different modulation pattern having the same base rate, only the duty cycle changes, so to speak.

 

[andreasmaaan]

At 20Hz and 10Hz, the values obtained at 80dB, 100dB and 110dB no longer seem consistent even in the presence of a difference tone.

See my earlier post (#39). The most plausible explanation (which is of course absolutely consistent with Zwicker and Fastl) is that these harmonics need to be very loud to surpass the absolute threshold of audibility.

See my markings on the graphs below. These thresholds track the Fletcher-Munson absolute audibility thresholds almost perfectly:

 

[March Audio]

Cool that you're getting to the bottom of this.

The graphic in post#1 uses narrow band noise as the typical masker at 150Hz, as far as I know, rather than a single tone.
When using single tone masker, the masking is less good.
On the other hand, we would like to simulate the extreme case here in the forum. Like your chosen example, this would correspond to a piece of music with, for example, a synthesizer passage that simply consists of a single tone at 80Hz.
But we should keep in mind that this is not comparable to a typical piece of music.


To rule out other effects such as IMD and Doppler distortion, did you choose a similar test setup as described in Axiomaudio's listening test?


Wouldn't it make more sense to establish initial benchmarks for audibility of harmonic distortion at lower sound pressure and then compare them to the existing literature, since reproducing 96dB@80Hz at the listening position will not be possible for most people.

In addition, it must be ensured that the reproduction system (loudspeaker or headphones) does not itself produce additional harmonic and other distortions that distort the result.

At 80Hz@96dB sound pressure at my listening position, a part of my listening room audibly vibrates with the masker tone and distorts the audibility of the harmonic distortion.



Is the study publicly available?




The perceptual thresholds shown at high sound pressure and low masker frequency, in the Fielder and Benjamin study, do not make sense to me and seems to contradict the results of Zwicker et al. I would expect, the higher the sound pressure and the lower the Bark difference between the masking tone and the test tone, the more pronounced the masking effect.

A possible cause could be that the hatched area in the diagram below (Zwicker et al) shows an area in which a difference tone (caused by the ears own nolinear distortion) becomes audible.

If we assume as an example a masker at 150Hz, which corresponds to the central frequency at 2 bark, then according to Zwicker, for the perception threshold of HD3 at 450Hz, which corresponds to about 5 bark, we would expect the values plotted in blue in the diagram below (critical-band rate difference is -3).

The perception threshold should be
at 60dB sound pressure around 1% (-40dB),
at 80dB sound pressure around 3% (-30dB),
at 100dB sound pressure around 4% (-28dB)
View attachment 98651

What amazed me is that when the sound pressure increases from 80 to 100dB, the perception threshold shifts only slightly upward (3% to 4% at 150Hz, detection threshold for HD3).

There I was quite clearly wrong in my earlier statements (in other threads), because I assumed that the development continues as with the sound pressure increase from 60 to 80dB (1% to 3% at 150Hz, detection threshold for HD3).


In the diagram below, at 100Hz and 50Hz the perceptibility of harmonic distortion behaves about as one would expect according to Zwicker et al.

At 20Hz and 10Hz, the values obtained at 80dB, 100dB and 110dB no longer seem consistent even in the presence of a difference tone.

View attachment 98646

As I have been alluding to the videos posted to so far are of course not a serious attempt at establishing any scientific data points on the subject. Rather a demo to many that may have no idea what this simple harmonic distortion may sound like. Just scene setting really and easy for many people to access.

Pure tones are the worst case scenario of course. There is an interesting discussion to have as to what music should be used for further testing. I would advocate that it should be something revealing, not necessarily a "typical" piece of music. So long as its not completely contrived or unusual. I know we joked about it earlier in the thread but perhaps Rebacca Pigeon Spanish Harlem might well be appropriate. Simple base line.

The paper is available on AES I believe.

 

[March Audio]

IME we are very sensitive to any kind of LF abberations:

- Fletcher-Munson characteristic makes simple harmonics much larger in perception than they are, numerically

- we tend to judge LF by actual waveform, and on top of that the judgment is asymmetrical (the sign of the momentary pressure gradient plays a role). Two sines with the same amount of H2 can sound very different depending on the phase of the harmonic and that also explains why music sounds different when inverting polarity, even with the blurring phase contribution overlay of conventional speaker crossovers (and speakers+subs, notably).

- while masking helps a lot, it doesn't look viable to me to rely on it.

As for the ultra-LF stuff (10Hz and lower), I personally found that much of the perception of tones that low is an amplitude modulation of the background (noise floor or gentle music content) and strong distortion there shows up in a slightly different modulation pattern having the same base rate, only the duty cycle changes, so to speak.

Interesting. We can also play with the harmonics phase using using REW.

 

[March Audio]

Tomorrow I will make up a few Foobar ABX tests for pure tones.

 

[KSTR]

Try running two instances of REW, one generating, say, 70Hz and the other 140Hz... but not quite, use an offset of about 0.5Hz (140.5Hz). This will rotate the phase of the H2 slowly (2 seconds period) and the perceived sound will also changes slightly. I've described it before, there will be a set of phase offsets 180° apart (but not neccessary at 0° and 180°, this depends on the system/speaker/headphone's natural phase response) where the contrast is maximized and there is another set of 180° spaced offsets, now orthogonal (rotated by 90°) to the first set where the difference is mininal, even undedectable.