• 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!

Speaker Equivalent SINAD Discussion

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
I'm not talking about the linear regression lines applied to the measurement curves when calculating NBD here, I'm talking about the multiple regression analysis applied to all possible configurations of the independent variables of the model in order to find an algorithm with optimum correlation between measurement and preference. It's all in Olive's AES paper (scroll down for the correct paper).

Ok, I see. I am still sceptical about the method that requires better resolution than our ears as that would mean that you have a method that gives 2 scoring for 2 set of samples that our ears cannot differentiate.

Can't you simply calculate both on a few examples and compare the results?
 
Last edited:
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
As I said previously, this is not correct – you’ve misunderstood what Sean Olive meant in his description of LFX. Here it is again from the full AES paper (scroll down in that link for the paper):


The next two sentences in the paragraph are just explanations for the choices made above:


He is explaining why the -6 dB frequency is taken from the sound power curve and not from any of the other curves i.e. why he is using x_SP and not x_LW or x_ON – because it better defines bass output of all speakers, but particularly (not only) speakers with rear ports (presumably because bass and sound power are not directional measures, unlike e.g. the on-axis and listening window curves). The listening window however better defines the average level in the range 300 Hz-10 kHz (where we can localize sound), so this is used for the reference level ybar_LW for all speakers.

I think your misunderstanding comes from the language used in the patent application (which is generally less clear and easy to follow), in which he says ‘may be used’ instead of 'is used' in bold that I've highlighted in the above quote. This is merely because the patent application is describing techniques that may be used to calculate predicted speaker preference ratings, so he uses 'may be' throughout the paper (Adobe pdf reader says 47 times! :D) for all techniques he actually used when creating the model. Here's just one example from the end of the paragraph marked [0046] in the patent:


Contrast this with the same sentence from the end of the second paragraph of Section 3.2 of the AES paper:


I suspect this change is just some legal requirement to use very precise language when applying for a patent on a method that may be used for a particular purpose, nothing more.

So, it is definitely the mean level of the listening window (and not the sound power curve) between 300 Hz and 10 kHz that should be used for the reference level of the LFX calculation, for all speakers, including, (but not only) rear-ported speakers.
Ah, OK, the “is used” vs “may be used” is what got me, I hate that vagueness, he is justifying why it can/is used, not saying that it could be used as an alternative.
 
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
Can't you simply calculate both on a few examples and compare the results?
The difference is measurement points is what drove me mad in creating an automated version.
I’ll try using every other measurement for the NHT and see how it changes.

The paper talks about this but is unclear, it states at least 1/20, but then talks about at least 1/6.

What I don’t get is the talk about spatially averaging the curves except for the on-axis.

The discrepancy of the 1/2 octave bands is the main thing left unclear. Is the 100Hz band starting at 100Hz or centered?
 
Last edited:
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
I'm not talking about the linear regression lines applied to the measurement curves when calculating the individual variables here, I'm talking about the multiple regression analysis applied to all possible configurations of the independent variables of the model in order to find an algorithm with optimum correlation between measurement and preference. It's all in Olive's AES paper (scroll down for the correct paper).
OK, I am going to experiment with the stated Hz bands being the center and not the start/end; however, the center can be based on Base2 or Base10
https://en.wikipedia.org/wiki/Octave_band#Naming_2
Not sure which to use.
 
Last edited:
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
Oh it better not have that kind of dependency because if it does, by definition it is not perceptually correct. And couldn't correlated so well with listening tests.
Just re-ran the NHT with 1/2 the measurements; some aspects got better, other worse.

1/20 score : 2.72
1/10 score: 2.68

1/20
NBD: 0.5605657795
NBD_PIR: 0.4298273122
LFX: 1.842485049
SM_PIR: 0.2823500199

1/10:
NBD: 0.56974
NBD_PIR: 0.4223994286
LFX: 1.855987118
SM_PIR: 0.2900592965
 
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
100 Hz would be impacted by the resolution then. I remembered from Devantier paper that it was important to have high resolution there.
C42E4D8A-70A1-4BE3-95E8-8A807EB34428.png
9DA3493D-6DA0-4306-A4FA-698FFE7AFD81.png

Not that large of a difference (screenshots as I’m on mobile).

If there any large resonances or similar that got omitted by doing 1/10, that would more drastically effect it.
 

SDX-LV

Active Member
Joined
Jan 11, 2020
Messages
132
Likes
141
Location
Sweden
Wanted to stop by and say that this is an amazing development @MZKM !
Together with Amir's speaker measurements this looks like the bright future of speaker selection is within reach :)

I think soon there will be no excuse to buy speakers without first checking their scores based on Spinorama measurements by Amir or their own manufacturers who will realize that this is the only way to sell a quality product :)

Just one question because I am lazy: could this algorithm really work with just the 5 CEA2034 graphs? Or does it require additional measurement data that can not be pulled out by digitizing an existing Spinorama plot?
 
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
Wanted to stop by and say that this is an amazing development @MZKM !
Together with Amir's speaker measurements this looks like the bright future of speaker selection is within reach :)

I think soon there will be no excuse to buy speakers without first checking their scores based on Spinorama measurements by Amir or their own manufacturers who will realize that this is the only way to sell a quality product :)

Just one question because I am lazy: could this algorithm really work with just the 5 CEA2034 graphs? Or does it require additional measurement data that can not be pulled out by digitizing an existing Spinorama plot?
Well, the formula is stated/described in the first post.

NBD_ON: This looks for the accuracy of the on-axis (do you need +/-3dB or can you do +/-1dB for the spec).

NBD_PIR: Same but first the predicted in-room response

LFX: How deep the bass is.

Smoothness_PIR: Given a target slope (in this case, for the PIR), how close the actual frequency response is to it; this not only also checks for deviation like NBD, but it also checks for soundstage width (sound power level).

I’m not sure how imaging is derived from these measurements, but if you look at the Spinorama, the more similar the Listening Window graph looks like the On-Axis graph, the more precise the imaging will be. This can also be seen in my Horizontal Directivity Normalized graphs, good imaging will mean all the lines are smooth, they will naturally slope down for most speakers, but as long as its straight, it will have excellent imaging (and the closer in SPL it is the wider the soundstage).
 
Last edited:

SDX-LV

Active Member
Joined
Jan 11, 2020
Messages
132
Likes
141
Location
Sweden
Well, the formula is stated/described in the first post.

NBD_ON: This looks for the accuracy of the on-axis (do you need +/-3dB or can you do +/-1dB for the spec).

NBD_PIR: Same but first the predicted in-room response

LFX: How deep the bass is.

Smoothness_PIR: Given a target slope (in this case, for the PIR), how close the actual frequency response is to it; this not only also checks for deviation like NBD, but it also checks for soundstage width (sound power level).

I’m not sure how imaging is derived from these measurements, but if you look at the Spinorama, the more similar the Listening Window graph looks like the On-Axis graph, the more precise the imaging will be. This can also be seen in my Horizontal Directivity Normalized graphs, good imaging will mean all the lines are smooth, they will naturally slope down for most speakers, but as long as its straight, it will have excellent imaging (and the closer in SPL it is the wider the soundstage).

I saw the formulas and my interpretation without studying all inter-dependencies was that No it is not enough with just "On Axis", "Listening Window" , "Early Reflections", "Sound Power", "Sound Power DI" and "Early Reflections DI" data to get the Dr. Sean Olive scores. Or I was not reading carefully enough?
 
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
I saw the formulas and my interpretation without studying all inter-dependencies was that No it is not enough with just "On Axis", "Listening Window" , "Early Reflections", "Sound Power", "Sound Power DI" and "Early Reflections DI" data to get the Dr. Sean Olive scores. Or I was not reading carefully enough?
NBD_ON just uses On-Axis
NBD_PIR just uses predicted in-room response.
SM_PIR also just uses predicted in-room response.
LFX uses Listening Window and Sound Power.

CEA-2034 Spinorama doesn’t include PIR, I include it in the Spinorama I post though (gray line). However, PIR is calculated from ON/ER/SP, so technically the CEA-2034 Spinorama does include everything.
 

SDX-LV

Active Member
Joined
Jan 11, 2020
Messages
132
Likes
141
Location
Sweden
NBD_ON just uses On-Axis
NBD_PIR just uses predicted in-room response.
SM_PIR also just uses predicted in-room response.
LFX uses Listening Window and Sound Power.

CEA-2034 Spinorama doesn’t include PIR, I include it in the Spinorama I post though (gray line). However, PIR is calculated from ON/ER/SP, so technically the CEA-2034 Spinorama does include everything.

Great! Not only this confirms the claim about Spinorama being "all you need", but also makes it especially interesting to digitize all the existing Harman Spinoramas using for example https://automeris.io/WebPlotDigitizer/. I will probably try it out a bit later. Right now I am itching to look into making Spinoramas from the https://www.princeton.edu/3D3A/Directivity.html data using Virtuix CAD + Octave :)

Thank you for the excellent Excel work! :)
 
OP
M

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,244
Likes
11,487
Location
Land O’ Lakes, FL
A dedicated thread will likely be created, but here's the current rankings (again taking note that it's not designed with near-field in mind):
Preference Rating (1).png
Preference Rating Ignore LFX (1).png


The bars can be colored manually, but that's a pain, I can auto-color based on value, but then the lines get very narrow:
Preference Rating Ignore LFX (3).png
 
Last edited:

JustIntonation

Senior Member
Joined
Jun 20, 2018
Messages
480
Likes
293
THD is simply not a useful descriptor of sound quality. In designing speakers, I and other designers use THD as a way of hinting at the frequencies and levels where a transducer is running up to mechanical limitations, but other than that it simply does not correlate to what we hope it does except in extreme cases.

See Geddes' papers on the subject:
http://www.gedlee.com/Papers/Distortion_AES_I.pdf
http://www.gedlee.com/Papers/Distortion_AES_II.pdf

Powerpoint summary:
http://www.gedlee.com/Papers/The Perception of Distortion.pdf

"THD and IMD have no correlation to the perception of the distortion that they are intended to represent.
Correlation is possible with a metric that takes into account the way the ear actually functions
One of the most important implications is that distortion in loudspeakers could well be an insignificant factor"

Great summary article from Multimedia Manufacturer:
http://www.gedlee.com/Papers/THD_.pdf

I have no idea how easy it would be to derive this value from Klippel data but it would be the way to go in my opinion.

From experience I can tell you that harmonic distortion and IMD meaurements do indeed not correlate substantially to subjective experience in normal listening rooms except in extreme cases, BUT.. they DO correlate to subjective experience in very well treated rooms / studios / anechoic rooms!
The thing is, room reflections and reverb mask most low level detail and HD and IMD up to moderate levels are easily burried in the much great linear and time distortions caused by room reflections / modes and reverb. Once this massive distortion is removed sufficiently (which takes quite a bit of effort to do well) one can easily hear the differences in HD and IMD with the latter being most audible.
Sadly the people on this forum (unlike most professionals in the audio industry / studios / mastering rooms) take a different view based on very superficial work (regarding room reflections) by Toole, which is in my opinion a very big error.
 

TimDH

Member
Forum Donor
Joined
Jun 30, 2019
Messages
21
Likes
19
Thank you for this contribution. It seems very hard to argue that THD and IMD make no difference when the listening tests did not give the listeners the chance to hear very low THD and IMD- i.e. in rooms treated to permit them to hear it. Listeners might like contributions from the room, but that is not the same thing as having no preference for low TDH and IMD. The promise of cardioid speakers of course is to reduce the reflections a lot without fancy room treatment.
 

617

Major Contributor
Forum Donor
Joined
Mar 18, 2019
Messages
2,404
Likes
5,296
Location
Somerville, MA
I think it's critical to understand that the phenomena captured by THD measurement are varied, and some of the things which cause high THD are audible and some are not - and all are frequency dependent. Geddes' argument is not that nonlinearity is inaudible, it is that the measurement we call THD (or IMD or whatever) is not a meaningful perceptual metric. These things are also level depenedent. THD just isn't the right measuring stick, which sucks, because it leaves us without a way of measuring this critical and subtle aspect of sound quality.
 

JustIntonation

Senior Member
Joined
Jun 20, 2018
Messages
480
Likes
293
I think it's critical to understand that the phenomena captured by THD measurement are varied, and some of the things which cause high THD are audible and some are not - and all are frequency dependent. Geddes' argument is not that nonlinearity is inaudible, it is that the measurement we call THD (or IMD or whatever) is not a meaningful perceptual metric. These things are also level depenedent. THD just isn't the right measuring stick, which sucks, because it leaves us without a way of measuring this critical and subtle aspect of sound quality.
In loudspeaker design many if not most experienced designers feel that one needs to look at the levels of the individual harmonics per frequency. Here the 2nd harmonic is viewed as least harmful, the 3rd harmonic slightly more so and the 5th harmonic and higher odd harmonics are bad and particularly important is the relation between the levels of the harmonics. Most measurements show HD up to the 5th harmonic and one wants to see a much lower 5th harmonics in relation to the 2nd and 3rd and the thought is that in absence of IMD measurements there is a correlation between HD measurements and IMD where if the 5th harmonic is close to the level of the 2nd and 3rd that one can expect high IMD.
My personal view on the audibility of HD is further that I look at where the HD products actually fall, for instance the 3rd harmonic of 1kHz falls at 3kHz and our ear is more sensitive at 3kHz than at 1kHz so I personally add the difference based on the equal loudness curve at roughly the SPL of the HD.
The above story gives a much more informed and complete picture. THD alone is indeed much too rough a metric.
As far as IMD goes without measurements of this one can make an educated guess based on what I wrote above in combination with knowledge of how badly a driver is pushed into excursion which will allways be bad for IMD (half exception to this is the new Purifi driver) but this is also usually visible in HD measurements one just needs to know that once a driver is pushed into high HD at the bottom of its range then frequencies produced by that driver higher up will suffer as well as they'll also experience an uneven force factor (the main reason for high HD/IMD at high excursion). Here 3-ways or 4-ways or 2-ways with sub ofcourse have a huge advantage.
 

617

Major Contributor
Forum Donor
Joined
Mar 18, 2019
Messages
2,404
Likes
5,296
Location
Somerville, MA
In loudspeaker design many if not most experienced designers feel that one needs to look at the levels of the individual harmonics per frequency. Here the 2nd harmonic is viewed as least harmful, the 3rd harmonic slightly more so and the 5th harmonic and higher odd harmonics are bad and particularly important is the relation between the levels of the harmonics. Most measurements show HD up to the 5th harmonic and one wants to see a much lower 5th harmonics in relation to the 2nd and 3rd and the thought is that in absence of IMD measurements there is a correlation between HD measurements and IMD where if the 5th harmonic is close to the level of the 2nd and 3rd that one can expect high IMD.
My personal view on the audibility of HD is further that I look at where the HD products actually fall, for instance the 3rd harmonic of 1kHz falls at 3kHz and our ear is more sensitive at 3kHz than at 1kHz so I personally add the difference based on the equal loudness curve at roughly the SPL of the HD.
The above story gives a much more informed and complete picture. THD alone is indeed much too rough a metric.
As far as IMD goes without measurements of this one can make an educated guess based on what I wrote above in combination with knowledge of how badly a driver is pushed into excursion which will allways be bad for IMD (half exception to this is the new Purifi driver) but this is also usually visible in HD measurements one just needs to know that once a driver is pushed into high HD at the bottom of its range then frequencies produced by that driver higher up will suffer as well as they'll also experience an uneven force factor (the main reason for high HD/IMD at high excursion). Here 3-ways or 4-ways or 2-ways with sub ofcourse have a huge advantage.

I share your intuition - as speaker designers we have to use whatever measurements we have available and just shoot for the lowest possible THD. The idea that higher orders are less desirable than lower is somewhat sensible as well but the fact remains that THD does not correlate to sound quality! We use it as a guide to speaker design excellence, but I wouldn't use it to rank speakers in isolation. Like impedance response, it's mostly a graph which allows designers to see what's going wrong rather than a graph which allows us to evaluate a design.
 
Top Bottom