• 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!

Audibility of low frequency distortion in speakers

Status
Not open for further replies.

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
Are people hearing the 2.8% distortion in the second video?

For me, the distortion is audible at low volume, but from 84-85dB sound pressure A-weighted at the listening position, it is no longer perceptible. However, parts of the listening room already resonate at this sound pressure and distort the result (a bit?).
I'm also sure that the speaker/woofer used produces negligible harmonic distortion of its own at this sound pressure.
Here at about 100dB sound pressure:
1607783693999.png

Over headphones (Beyerdynamic DT660), the distortions are clearly audible at low sound pressure levels. However, it is not possible to produce over estimated 80dB A-weighted sound pressure @80Hz with the headphones.
However, as expected, the perceptibility decreases with increasing volume.
 
Last edited:

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
I can definitely hear it over loudspeakers(don't have good headphones here). What does that mean?

According to the AES paper presented by @March Audio, it is within the realm of possibility that you can still perceive 2.8% HD2 at 96dB SPL with an 80Hz masker.
1607803450306.png


The prerequisite is, of course, that your loudspeaker itself delivers well below 2.8% HD2 at 80Hz and that the listening room (or objects) does not distort the result due to self-resonances.

But I honestly find it hard to imagine that in a normal listening room you can achieve results like under laboratory conditions (anechoic chamber?) - however, I do not know the test conditions in the AES paper.
 
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
I can definitely hear it over loudspeakers(don't have good headphones here). What does that mean?
This test is not scientific enough to draw any firm conclusions. I will produce some proper abx tests later. Rather this is just an indicator that for pures tones the results in the table I posted above seem reasonable. That with the worst case scenario (pure tones) we can hear harmonic distortion at surprisingly low levels.
 
Last edited:
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
According to the AES paper presented by @March Audio, it is within the realm of possibility that you can still perceive 2.8% HD2 at 96dB SPL with an 80Hz masker.
View attachment 98720

The prerequisite is, of course, that your loudspeaker itself delivers well below 2.8% HD2 at 80Hz and that the listening room (or objects) does not distort the result due to self-resonances.

But I honestly find it hard to imagine that in a normal listening room you can achieve results like under laboratory conditions (anechoic chamber?) - however, I do not know the test conditions in the AES paper.
Not necessarily. :)

Bear in mind that you are looking above at 2.8% H2 only. The distortion generated by woofers at low frequency is not just H2. That's what we are simulating and why is why it is far more audible.

Firstly obviously we cant reproduce lab conditions. Any data generated here via member input has to be treated with great caution. However it is still worthy of investigation IMO.

If your room rattles it will also rattle with the pure tone. So actually its another masker. If under ABX you can reliably identify the track with the distortion, then its audible above this additional noise. Also the nature of these disturbances will essentially be individual to any room so multiple user data will overcome this.

Regarding inherent speaker distortion, my view is we don't need to test at high level which is where excessive distortion may occur. The data above in the table indicates there is only a small increase in masking when moving from 80dB spl to 100dB spl. So test at 80.

Just for interest I can hear that 2.8% THD on a variety of speakers from high quality low distortion to cheap Bluetooth. What do you find?

Finally its not my intention to try and identify the audibility threshold limits for which you will require exacting conditions. Rather intended to be a broader look at the subject.
 
Last edited:

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
@March Audio what exactly are you planning to do in terms of an ABX test? Generate music files that are convolved with a transfer function resembling that of a high-distortion woofer?

FWIW, I did a similar (though even less formal) thing a while back, when I wanted to emulate the distortion performance of a few woofers I was trying to decide between for a speaker design, as well as to try to inform my decision about appropriate crossover points. I didn't have the programming skills to generate highly accurate models, but using VST plugins I did manage to produce roughly comparable levels and spectra of harmonic distortion to the drivers in question, with a similar (but not perfect) relationship between frequency, SPL and distortion level. Doing it in this way in DAW had the advantage of being able to easily filter the signal at a desired virtual crossover frequency, then pass only the operating range of the virtual driver through the distorting effect. It was also possibly to quickly and easily change the XO frequency or switch virtual drivers this way. There was some educated guesswork though in calculating the SPL of the headphones at my eardrums.

Anyway, what I found was that, when a distorting woofer was crossed over reasonably low to a "virtual" mid-high driver that did not produce any distortion itself, a fairly high level of distortion was tolerable, both before becoming audible, and even more so before becoming objectionable. However, when the frequency of the virtual XO was raised to include the midrange, too, the distortion became significantly more audible/objectionable.

The main lesson from this for me was that, assuming a given level of woofer distortion, 3-way speakers offer significant advantages over two-way designs (ok, no real news there).

The other lesson was that, in a 3-way designed on a budget, the woofer may be the best place to shave off costs, i.e. it's more important that the woofer has sufficient bass extension and SPL capability than that it produces low distortion at a given SPL, so long as it's crossed over low enough (another relevant factor here is that the price differential between competent but unexceptional woofers and high-performance woofers tends to be very large; this is not so much the case for midrange drivers or tweeters, where excellent drivers are available at fairly moderate cost).

Conversely, for two-way designs, woofer distortion performance becomes much more important (which I presume if the point you are planning to demonstrate).

I would suggest though that the trials are conducted on headphones, for multiple reasons:
  • they tend to have lower bass distortion and better bass extension than similarly-priced speakers (at least the sealed ones do)
  • the room does not intervene to massively boost some frequencies and cut others, which may grossly affect what may or may not be audible at lower frequencies
OTOH, for most people at home it will be easier to determine the SPL at the listening position using speakers (to the extent that you can say there is even a single frequency-independent SPL at a given listening position in a normal room).

How are you planning to model the distortion profiles and apply them to the recordings? That was the hard part for me, and the reason the whole experiment remained very informal (and until now, private).
 
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
@March Audio what exactly are you planning to do in terms of an ABX test? Generate music files that are convolved with a transfer function resembling that of a high-distortion woofer?

FWIW, I did a similar (though even less formal) thing a while back, when I wanted to emulate the distortion performance of a few woofers I was trying to decide between for a speaker design, as well as to try to inform my decision about appropriate crossover points. I didn't have the programming skills to generate highly accurate models, but using VST plugins I did manage to produce roughly comparable levels and spectra of harmonic distortion to the drivers in question, with a similar (but not perfect) relationship between frequency, SPL and distortion level. Doing it in this way in DAW had the advantage of being able to easily filter the signal at a desired virtual crossover frequency, then pass only the operating range of the virtual driver through the distorting effect. It was also possibly to quickly and easily change the XO frequency or switch virtual drivers this way. There was some educated guesswork though in calculating the SPL of the headphones at my eardrums.

Anyway, what I found was that, when a distorting woofer was crossed over reasonably low to a "virtual" mid-high driver that did not produce any distortion itself, a fairly high level of distortion was tolerable, both before becoming audible, and even more so before becoming objectionable. However, when the frequency of the virtual XO was raised to include the midrange, too, the distortion became significantly more audible/objectionable.

The main lesson from this for me was that, assuming a given level of woofer distortion, 3-way speakers offer significant advantages over two-way designs (ok, no real news there).

The other lesson was that, in a 3-way designed on a budget, the woofer may be the best place to shave off costs, i.e. it's more important that the woofer has sufficient bass extension and SPL capability than that it produces low distortion at a given SPL, so long as it's crossed over low enough (another relevant factor here is that the price differential between competent but unexceptional woofers and high-performance woofers tends to be very large; this is not so much the case for midrange drivers or tweeters, where excellent drivers are available at fairly moderate cost).

Conversely, for two-way designs, woofer distortion performance becomes much more important (which I presume if the point you are planning to demonstrate).

I would suggest though that the trials are conducted on headphones, for multiple reasons:
  • they tend to have lower bass distortion and better bass extension than similarly-priced speakers (at least the sealed ones do)
  • the room does not intervene to massively boost some frequencies and cut others, which may grossly affect what may or may not be audible at lower frequencies
OTOH, for most people at home it will be easier to determine the SPL at the listening position using speakers (to the extent that you can say there is even a single frequency-independent SPL at a given listening position in a normal room).

How are you planning to model the distortion profiles and apply them to the recordings? That was the hard part for me, and the reason the whole experiment remained very informal (and until now, private).
First part is going to be to use tones and distortion profiles generated from the tests we have here on ASR.

Considering the data in the AES paper, what is your view on the importance of the specific exact volume in an informal test such as this? My view is that is is not significant as the variation in masking with volumes between 80dB and 100dB SPL is small. So a lower level is better as it minimises the additional distortion of the playback speaker. Other views / input is welcome.

Before going on to music I would like to discuss various things such as what is appropriate test content.
 
Last edited:

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
First part is going to be to use tones and distortion profiles generated from the tests we have here on ASR.

That's a good start. The difficulty I found in the case of music was creating a dynamic distortion profile, i.e. one that increases in percentage and spectrum as level increases, and vice versa.

It's not much use IMO adding a static percentage/spectrum of distortion to a music signal, because music is dynamic and driver distortion varies massively with voltage input/SPL. That's why I settled on using VST plugins in the end: they weren't able to give me the exact distortion profiles of the drivers I was trying to model, but they did at least allow me to make the percentage and spectrum of distortion dependent on SPL. That way, I was able to approximately calculate the relationship between dBfs level of the signal, spectrum/percentage of distortion, and output SPL from my headphones (based on knowledge of the gain of the DAC/amp and the voltage sensitivity of the headphones).

But one step at a time I guess :) It's a very interesting and enlightening experiment IME, and deserves to be done better than I did it.

Considering the data in the AES paper, what is your view on the importance of the specific exact volume in an informal test such as this?

Somewhat important, but probably the least of your concerns given all the other very difficult variables at play. Less important if your aim is to test at loud levels only, since the greatest variation in masking thresholds occurs between 0 and 70dB. Once you're up at or over 80dB, additional increases in SPL don't alter the shape or bandwidths of the curves as significantly as they do when we move from low to moderate SPLs:

1607818593082.png


Before going on to music I would like to discuss various things such as what is appropriate test content.

It's a good place to have the discussion :)
 
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
As you say, one step at a time. The point about dynamic content is absolutely valid, but also why we need to discuss what is appropriate musical test content. Which I think should be the most revealing.

Regarding volume, I would suggest that people listen at comfortable but not quiet volumes. This should easily exceed 70dB.
 
Last edited:

ctrl

Major Contributor
Forum Donor
Joined
Jan 24, 2020
Messages
1,616
Likes
6,086
Location
.de, DE, DEU
Just for interest I can hear that 2.8% on a variety of speakers from high quality low distortion to cheap Bluetooth. What do you find?

I don't doubt for a second that you can hear the 2.8% HD2 at 80Hz@96dB on a cheap Bluetooth speaker, as the speaker should certainly contribute another few percent distortion itself.

Just recreated your test case with two sine wave generators and in-ear headphones at high sound pressure and heard 160Hz with -31dB attenuation to the 80Hz fundamental. Would say that the noise signal was rather inaudible.
Suspect with an ABX test it will be clearer.

Finally its not my intention to try and identify the audibility threshold limits for which you will require exacting conditions. Rather intended to be a broader look at the subject.

Then we should look at a spectrum of values.
For me, the interference signal was clearly audible and disturbing from 5-6% (-24dB).
Thus, for me 4-6% HD2 would be relevant at the earliest with an 80Hz single tone as a masker and sound pressure above 90dB.

For those who want to do an ABX test, I attach two files. 80Hz masker with interfering tone of 160Hz with -31dB attenuation.
At moderate volume (clearly not >90dB) and in-ear headphones, I got the following result:
1607822042240.png
 

Attachments

  • samples.zip
    310.8 KB · Views: 95
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
I don't doubt for a second that you can hear the 2.8% HD2 at 80Hz@96dB on a cheap Bluetooth speaker, as the speaker should certainly contribute another few percent distortion itself.

Just recreated your test case with two sine wave generators and in-ear headphones at high sound pressure and heard 160Hz with -31dB attenuation to the 80Hz fundamental. Would say that the noise signal was rather inaudible.
Suspect with an ABX test it will be clearer.



Then we should look at a spectrum of values.
For me, the interference signal was clearly audible and disturbing from 5-6% (-24dB).
Thus, for me 4-6% HD2 would be relevant at the earliest with an 80Hz single tone as a masker and sound pressure above 90dB.

For those who want to do an ABX test, I attach two files. 80Hz masker with interfering tone of 160Hz with -31dB attenuation.
At moderate volume (clearly not >90dB) and in-ear headphones, I got the following result:
View attachment 98794

You can also just use REW software, its free and simple. You can generate signals with up to the 9th harmonic at arbitrary levels on the go and save to a file.

https://www.roomeqwizard.com/

1607823094820.png



Real woofer low frequency distortion will never consist of just a 2nd harmonic, which is going to be the most masked. Hence why I want to use distortion profiles derived from actual speaker tests we have on ASR.

The 2.8% Total Harmonic distortion in my example consisted of harmonics 2 to 5 based on the profile of a real speaker, not just 2.8% 2nd harmonic.

Can I ask, just for the moment, that we limit the addition of other test files so we can stay focussed :)
 
Last edited:
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
As we touched on this above here is a video to demonstrate the way masking is greater closer to the masking fundamental tone.

The following distortion profile was applied.

1607825060830.png


Whilst the 2nd harmonic which is closest to the fundamental is at the same level as the higher harmonics, it is the least obviously audible.

This also gives an idea how the addition of different harmonics sound.

 
Last edited:
OP
March Audio

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
For those who want to do an ABX test, I attach two files. 80Hz masker with interfering tone of 160Hz with -31dB attenuation.
At moderate volume (clearly not >90dB) and in-ear headphones, I got the following result:
View attachment 98794
For grins my result with your file using Oppo PM1 headphones.

1607834293337.png


Looks like we can both reliably detect a 2nd harmonic at 2.8%
 
Last edited:

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
I just installed foobar for the first time and the ABX tester for this. First attempt. Headphones AKG-702s on a focusrite 2i2v2. Can't do speakers today- where are in the middle very wild weather, had 400mm of rain in <24 hours and the wind is really loud.

1607843236832.png
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia

Doodski

Grand Contributor
Forum Donor
Joined
Dec 9, 2019
Messages
20,745
Likes
20,756
Location
Canada
Wow. You can have some of ours- it's all going down the creeks and rivers into the sea!

https://www.goldcoastbulletin.com.a...s/news-story/392c952732c34e3396934dcb91ca73e9

This photo just around the corner was taken by my neighbour. This road always floods.
View attachment 98888


View attachment 98887
I was watching a documentary where a huge salt water croc used the high water levels to gain access to a river, the croc went far upstream and then livestock started going missing until one day a guy was watering his horse and this huge croc came up and latched onto the horse... Scary stuff.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,274
Location
Gold Coast, Queensland, Australia
I was watching a documentary where a huge salt water croc used the high water levels to gain access to a river, the croc went far upstream and then livestock started going missing until one day a guy was watering his horse and this huge croc came up and latched onto the horse... Scary stuff.

Don't have salties around here. The big salties up north get a few tourists every now and then when they take a dip at a creek crossing and ignore the signs. They are pretty well fed in general and leave humans alone mostly.
 
Status
Not open for further replies.
Top Bottom