• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Audibility of low frequency distortion

March Audio

Major Contributor
Manufacturer
Joined
Mar 1, 2016
Messages
6,382
Likes
8,983
Location
Albany Western Australia
#1
@ctrl

To discuss your post in the other thread.

I'm travelling atvthe moment so please excuse dodgy phone typos and ability to access all the info I need to.

Fully accept the point about masking but let's look at another graph

1607340084643.png


So here we see H2 needs to be about 15dB below the fundamental to be audible. H3 needs to be about 28dB.

Thats 17% and 4% respectively.

The thing I'm motcallowed to mention is about 17% so yes audible imo, especially if higher orders kick off.

Also Amiirs plots showing 100% distortion at very low frequencies are probably due the speaker being driven below port resonance where the cone will become undamped leading to excessive excursions. He shouldn't be testing at high levels below those frequencies. There is potential for driver damage
 
Last edited:

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
11,424
Likes
11,916
Location
Riverview FL
#3
That graph looks odd as it would make me presume that typical instrumental timbre would be inaudible and everything would sound like a sine wave.

Or I'm looking at it incorrectly. <-- probably
 

andreasmaaan

Major Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,084
#7
@ctrl

To discuss your post in the other thread.

I'm travelling atvthe moment so please excuse dodgy phone typos and ability to access all the info I need to.

Fully accept the point about masking but let's look at another graph

View attachment 97775

So here we see H2 needs to be about 15dB below the fundamental to be audible. H3 needs to be about 28dB.

Thats 17% and 4% respectively.

The Adam is about 17% so yes audible imo, especially if higher orders kick off.

Also Amiirs plots showing 100% distortion at very low frequencies are probably due the speaker being driven below port resonance where the cone will become undamped leading to excessive excursions. He shouldn't be testing at high levels below those frequencies. There is potential for driver damage
Since masking varies highly nonlinearly with SPL, I'm not sure that graph tells us much. It really needs to show the SPL of the masker to be valid.

I tried to summarise the key aspects of masking in this post. This is the key image from that post:

1607342413013.png


As is visible, masking varies radically with SPL.

As SPL increases, downward masking decreases and upward masking increases.

And given that critical band rate increases with frequency, as frequency of the masker decreases, masking increases.

However, putting all that to one side, IMO the primary concern when a woofer shows 10%+ HD in the low frequencies is not harmonic distortion at all, but rather IM distortion (particularly in a two-way design). In the presence of a complex signal, IM products will appear all across the spectrum, making them far less likely to be masked than harmonics (which tend to be clustered in the frequency band immediately above the fundamental, and therefore most subject to masking).
 

ctrl

Addicted to Fun and Learning
Joined
Jan 24, 2020
Messages
869
Likes
2,721
Location
.de, DE, DEU
#8
So here we see H2 needs to be about 15dB below the fundamental to be audible. H3 needs to be about 28dB.
Thats 17% and 4% respectively.
Yep, @andreasmaaan has already mentioned the crucial argument in post#7:
SPL or the sensation level (SL) of the masker.
Without this information the masking cannot be classified, because it depends non-linearly on the sound pressure of the masker.


Here are two examples with SL (sensation level = dB above the individual threshold of pure tone) specification of the masker:
1607342413614.png

1607345058337.png

Source: Importance of low frequency masking


Think with the baseline values from the above diagrams it is clear that in our example case the Adam Audio T5V, the 16% THD around 70Hz at a sound level of 96dB should not be audible.
1607345712456.png
 
OP
March Audio

March Audio

Major Contributor
Manufacturer
Joined
Mar 1, 2016
Messages
6,382
Likes
8,983
Location
Albany Western Australia
Thread Starter #9
Since masking varies highly nonlinearly with SPL, I'm not sure that graph tells us much. It really needs to show the SPL of the masker to be valid.

I tried to summarise the key aspects of masking in this post. This is the key image from that post:

View attachment 97778

As is visible, masking varies radically with SPL.

As SPL increases, downward masking decreases and upward masking increases.

And given that critical band rate increases with frequency, as frequency of the masker decreases, masking increases.

However, putting all that to one side, IMO the primary concern when a woofer shows 10%+ HD in the low frequencies is not harmonic distortion at all, but rather IM distortion (particularly in a two-way design). In the presence of a complex signal, IM products will appear all across the spectrum, making them far less likely to be masked than harmonics (which tend to be clustered in the frequency band immediately above the fundamental, and therefore most subject to masking).
Well this is why I initially said its 'debateable" ;)
 
OP
March Audio

March Audio

Major Contributor
Manufacturer
Joined
Mar 1, 2016
Messages
6,382
Likes
8,983
Location
Albany Western Australia
Thread Starter #10
Yep, @andreasmaaan has already mentioned the crucial argument in post#7:
SPL or the sensation level (SL) of the masker.
Without this information the masking cannot be classified, because it depends non-linearly on the sound pressure of the masker.


Here are two examples with SL (sensation level = dB above the individual threshold of pure tone) specification of the masker:
View attachment 97777
View attachment 97779
Source: Importance of low frequency masking


Think with the baseline values from the above diagrams it is clear that in our example case the Adam Audio T5V, the 16% THD around 70Hz at a sound level of 96dB should not be audible.
View attachment 97780
Withbthat data maybe.

However as Andreasman said its not as simple as simpl thd.
 

Frank Dernie

Major Contributor
Forum Donor
Joined
Mar 24, 2016
Messages
5,664
Likes
12,759
Location
Oxfordshire
#11
IM products will appear all across the spectrum, making them far less likely to be masked than harmonics (which tend to be clustered in the frequency band immediately above the fundamental, and therefore most subject to masking)
Surely harmonics, by definition, are multiples of the fundamental frequency, not clustered near it?
Also, since the sound signature of individual instruments is due to the relative levels of harmonics it produces, and timbre the more subtle variations in this pattern of harmonics surely all musical instruments would sound the same if this were true?
For me personally the most important parameter when choosing a speaker has always been its reproduction of instrumental timbre, so insensitivity to high levels of harmonic distortion has always been something of which I have been somewhat sceptical.
Maybe the harmonic variations causing timbral differences are much bigger than I imagined.
 

andreasmaaan

Major Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,084
#12
Surely harmonics, by definition, are multiples of the fundamental frequency, not clustered near it?
Yes of course, "clustered" was a poor choice of word. What I meant was that, in transducers, lower order harmonics tend to be higher in level than higher order harmonics, and these are close(r) in frequency to the fundamental than IM products, which occur up and down the entire frequency spectrum.

Also, since the sound signature of individual instruments is due to the relative levels of harmonics it produces, and timbre the more subtle variations in this pattern of harmonics surely all musical instruments would sound the same if this were true?
Not in my opinion. If you look again at the graph from Zwicker/Fastl that I posted:

1607347281979.png


You'll see that it's only at higher SPLs that significant masking extends much beyond the second or third harmonic. Even at high SPLs, higher-order harmonics and IM products that are distant in frequency from the signal will remain unmasked.

Compare this to the harmonic spectra of various instruments and you'll notice that most instrument harmonics are unlikely to be masked (at any SPL, but especially at low-moderate SPLs).

Here are some examples I pilfered from a research paper:

1607347611779.png


1607347635905.png


Maybe the harmonic variations causing timbral differences are much bigger than I imagined.
I think this is essentially correct. It's quite normal in the case of musical instruments and human voice for the fundamental to not even be the frequency with the highest SPL for a given note.
 

ctrl

Addicted to Fun and Learning
Joined
Jan 24, 2020
Messages
869
Likes
2,721
Location
.de, DE, DEU
#13
However as Andreasman said its not as simple as simpl thd.
The whole debate in the T5V thread, referred to the low frequency harmonic distortion measured by @amirm and its comments.
Amir said:
Notice how the bass distortion stays well below its frequency response (otherwise THD would be over 100%!). So not reference quality but better than what one expects as noted.
and you then said:
Im a little lost on the bass distortion comments. Have I missed something? Just seems high to me. As expected from a small equalised speaker.
...
I think the 10% figure is debateable, but yes we certainly are less sensitive to distortion at low frequencies
...
It has high LF distortion and limited extension.
...
The distortion is clearly extremely high. You could say it might be "typical" or maybe better than typical for a cheap small speaker of this nature, but nonetheless the distortion is very high.
...
You yourself have always argued with harmonic distortions and claimed their audibility or masking was controversial - which, as shown, is actually not the case.

You didn't mention IMD once in the whole discussion in the T5V thread, and after @andreasmaaan brought it into play, you suddenly say "yes that's what I meant" o_O


But even if we take IMD into account, @amirm's statement is not wrong.
The loudspeaker examples shown by Amir in this post show audible THD values (>100%) in the range 40-50Hz, caused by the extremely large excursion below the BR tuning frequency (which the T5V does not show) and just because of the large excursion the IMD values for these speakers will literally explode under the discussed conditions (sound pressure level at 96dB) and with the T5V the chances are not bad that this will not happen to the same extent.
 

RayDunzl

Major Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
11,424
Likes
11,916
Location
Riverview FL
#15
Here's my single Bass Guitar note, for grins.

It is twice the frequency of the masking diagram.


With a sloppy curve "like" the masking curve below adjusted an octave up.

Or maybe it should be drawn differently.


1607349685235.png



1607349371422.png


So, a "real" tone has little trouble escaping the mask, and I can go to sleep now.
 
Last edited:
OP
March Audio

March Audio

Major Contributor
Manufacturer
Joined
Mar 1, 2016
Messages
6,382
Likes
8,983
Location
Albany Western Australia
Thread Starter #16
The whole debate in the T5V thread, referred to the low frequency harmonic distortion measured by @amirm and its comments.
Amir said:


and you then said:


You yourself have always argued with harmonic distortions and claimed their audibility or masking was controversial - which, as shown, is actually not the case.

You didn't mention IMD once in the whole discussion in the T5V thread, and after @andreasmaaan brought it into play, you suddenly say "yes that's what I meant" o_O


But even if we take IMD into account, @amirm's statement is not wrong.
The loudspeaker examples shown by Amir in this post show audible THD values (>100%) in the range 40-50Hz, caused by the extremely large excursion below the BR tuning frequency (which the T5V does not show) and just because of the large excursion the IMD values for these speakers will literally explode under the discussed conditions (sound pressure level at 96dB) and with the T5V the chances are not bad that this will not happen to the same extent.
No I didnt say "thats what I meant". I simply agreed with Andreasmans point :)

My whole issue in the (thing I am not allowed to mention) thread was Amirs assertions that the distortion was well controlled. It isnt. Also that his examples of 100% distortion was an indicator that it was, were erroneous. He doesnt understand the mechanism, ie the fundamental behaviour of a ported system below resonance, hence he thought it was a figure of significance.

Looking at you graphs above it isnt that clear cut. Also note the error mentioned on the 3rd data.

In fact looks to me like its very likely to be audible on (the thing I am not allowed to mention)

1607351032536.png
 
Last edited:
Joined
Nov 18, 2020
Messages
305
Likes
381
Location
USA
#17
Saying someone said something without a direct quote including the full context of that quote can be grossly misleading. It's no more possible to have a reasonable discussion using imprecise attributions than it is to achieve good engineering results using imprecise data.
 

CDMC

Major Contributor
Forum Donor
Joined
Sep 4, 2019
Messages
1,028
Likes
1,943
#18

dougi

Senior Member
Joined
May 31, 2020
Messages
419
Likes
320
Location
ACT, Australia
#19
This one of test by Axiom is interesting in that is describes and audibility curve vs frequency. link

Not sure how rigourous it really was but method seems reasonable at first glance. One figure is below:

Capture.JPG
 

andreasmaaan

Major Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,084
#20
This one of test by Axiom is interesting in that is describes and audibility curve vs frequency. link

Not sure how rigourous it really was but method seems reasonable at first glance. One figure is below:

View attachment 97859
I've pointed this out a few times when this graph has been posted, so sorry to repeat myself again, but it should be clarified (since Axiom kind of glosses over this point) that the average spectrum of music is far from flat.

So when Axiom says that the music was playing at, say, 86dB, that certainly doesn't mean that the level of the music was 86dB in the treble.

In fact, typical spectra of music look something like this:

1607373880962.png


Which interestingly is a very similar shape to the threshold curve that Axiom generated from their tests.
 
Top Bottom