• 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!

Evidence-based Speaker Designs

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,739
Likes
16,171
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
It must be said though that measurements of his loudspeakers are a bit embellished by using a not normalised to axial sound polar map and compensating directivity problems with the direct sound, the reality doesn't look as good http://www.princeton.edu/3D3A/Directivity/GedLee Nathan/index.html

There isn't a directivity problem - the DI curve shows it. It's diffraction at the mouth that appears as a single null only directly on axis due to an exceptionally coherent spherical wavefront. It makes for a negligible proportion of listening window, let alone early reflections or sound power. Normalising all off-axis curves to on-axis exaggerates the problem by exaggerating the small effect on overall sound power, which is smooth and flat over a broad area. Most other waveguides spread out diffraction such that there are more lower-Q wiggles with a lower magnitude, or do not emit a spherical wavefront with similar sound path length to mouth from the compression driver. The SEOS elliptical variant has no such dip as a result. It's a known compromise, but a benign one. The Revel waveguides are SOTA too and they, too, have several such interference artefacts, except they are lower-Q and lower magnitude.

https://www.audiosciencereview.com/...evel-performa-f228be.13001/page-2#post-399113

The direct sound is exceptionally smooth.
 
Last edited:

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,739
Likes
16,171
There isn't a directivity problem - the DI curve shows it.
...
The direct sound is exceptionally smooth.
Both don't hold in this case, the DI curve shows exactly the problem and the direct sound has been compromised for a smooth sound power:

1594196937593.png
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Both don't hold in this case, the DI curve shows exactly the problem and the direct sound has been compromised for a smooth sound power:

View attachment 72383

index.php


The raw measurements show the dip fills in within 10°, and then pattern control and FR is exemplary considering this is a large-format waveguide with compression driver. Normalising DI to 0° has the same problem as a normalised polar misrepresenting the overall smoothness of the radiated soundfield at LW, ER and sound power. Numerous Revel speakers would have bumpy DI referenced to on-axis as well, it's just that those bumps are spread over a wider area. See, for instance, how much better the M106 LW is than on-axis.
 

Absolute

Major Contributor
Forum Donor
Joined
Feb 5, 2017
Messages
1,084
Likes
2,125
It should be normalized to 22 degree angle, according to Geddes. This is the design-axis intended for best performance.

Here's my concern, though; if we design speakers for a particular axis to get smoothest response in a listening window, we also say that off-axis response outside of that axis isn't as important. I doubt that's true.
Unless those lines are smooth at multiple angles you will have a directivity issue. This speaker shows a lessening of directivity control over multiple angles from 4-6 khz meaning that area will be emphasized in the reflective sounds even if we change normalization-angle to the recommended one.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
It should be normalized to 22 degree angle, according to Geddes. This is the design-axis intended for best performance.

Here's my concern, though; if we design speakers for a particular axis to get smoothest response in a listening window, we also say that off-axis response outside of that axis isn't as important. I doubt that's true.
Unless those lines are smooth at multiple angles you will have a directivity issue. This speaker shows a lessening of directivity control over multiple angles from 4-6 khz meaning that area will be emphasized in the reflective sounds even if we change normalization-angle to the recommended one.

I believe both can be correct at the same time. Just because the on-axis is compromised for a sound technical reason/constraint doesn't mean that the overall soundfield isn't smooth and well-controlled, as in the case here.

edit: it should also be noted this was an older Geddes design.
 
Last edited:

Absolute

Major Contributor
Forum Donor
Joined
Feb 5, 2017
Messages
1,084
Likes
2,125
I believe both can be correct at the same time. Just because the on-axis is compromised for a sound technical reason/constraint doesn't mean that the overall soundfield is smooth and well-controlled, as in the case here.
Agreed. I don't pay attention to single lines because sometimes there's diffraction or something going on at specific angles, like the usual diffraction stuff in the Revel waveguide at 5-6 khz on-axis. If there's two-three angles in a row with the same bumps/dips I pay more attention to it because that means there's most likely a discrepancy between direct and indirect sound.

I just don't believe in this "optimal angle" argument because if you need to use a certain window of angles to get a smooth(est) line, the speaker doesn't have particularly good off-axis response.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,739
Likes
16,171
It should be normalized to 22 degree angle, according to Geddes. This is the design-axis intended for best performance.
Reminds me of calling a bug a feature. Good modern large compression driver horns like the one of the M2, some Limmer etc. show how such geometries have evolved.
edit: it should also be noted this was an older Geddes design.
If the newer ones are better then that's ok.
 
Last edited:

Duke

Major Contributor
Audio Company
Forum Donor
Joined
Apr 22, 2016
Messages
1,523
Likes
3,745
Location
Princeton, Texas
Reminds me of calling a bug a feature. Good modern large compression driver horns like the one of the M2, some Limmer etc. show how such geometries have evolved.

The waveguide on the Nathan is inherently compromised by its significantly sub-optimal size.

The on-axis "hole" is something that happens with an axisymmetric horn or waveguide when the mouth reflection arrives in-phase but about 1/2 wavelength behind the unreflected sound. It can be mitigated by a sufficiently large round-over at the mouth, which correspondingly increases enclosure size and vertical interdriver spacing.

Non-axisymmetric waveguides like the one in the JBL M2 and the SEOS waveguides do not have this issue because the mouth geometry smears out the mouth reflection in time.

The recommended setup for Earl's constant-directivity waveguide speakers, and for mine (most of which use a non-axisymmetric device), is to use aggressive toe-in such that the speaker axes criss-cross in front of the listening area, and I can explain the reasoning if you would like. Let me emphasize that the same setup geometry is recommended for my non-axisymmetric waveguide speakers. In other words the toe-in is not a "fix" for the on-axis response hole, but it does make it essentially irrelevant.

One setting where the on-axis hole would be a major issue is in a recording studio setting, where mixing (and probably mastering) are done on-axis. Dutch & Dutch evidently uses a waveguide geometry which mitigates this issue.

I have built loudspeakers which are virtually identical except for one using a SEOS waveguide and the other an Oblate Spheroid. Ime the Oblate Spheroid is the superior device regardless of listening axis, but especially so when used as intended.

It is easy to lose sight of the forest (the desired outcome) for the trees (the measurements). The desired outcome is that the direct sound be as correct as possible, and that the reverberant sound be as correct as possible. The measurements are highly useful but they are not the desired outcome in and of themselves. In my opinion.

Also, please note that the spin-o-rama measurements of a non-axisymmetric device inherently MISS what is happening along the diagonals [NO they don't; I was mistaken - see Ilkless's post below], and just as much sound is delivered along the diagonals as along the standard measurement axes. Eyeball the JBL M2's waveguide and you will see what I mean.
 
Last edited:
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Reminds me of calling a bug a feature. Good modern large compression driver horns like the one of the M2, some Limmer etc. show how such geometries have evolved.

If the newer ones are better then that's ok.

The M2 approach is literally the opposite - smearing diffraction out into numerous very low-magnitude artefacts. They are fundamentally different approaches.

And as for Limmer, as @Duke and I have said, the hole is a result of a very coherent wavefront emerging from the throat. That a different horn doesn't have such a null because the wavefront isn't as coherent doesn't necessarily mean the horn is superior.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
The waveguide on the Nathan is inherently compromised by its significantly sub-optimal size.

The on-axis "hole" is something that happens with an axisymmetric horn or waveguide when the mouth reflection arrives in-phase but about 1/2 wavelength behind the unreflected sound. It can be mitigated by a sufficiently large round-over at the mouth, which correspondingly increases enclosure size and vertical interdriver spacing.

Non-axisymmetric waveguides like the one in the JBL M2 and the SEOS waveguides do not have this issue because the mouth geometry smears out the mouth reflection in time.

The recommended setup for Earl's constant-directivity waveguide speakers, and for mine (most of which use a non-axisymmetric device), is to use aggressive toe-in such that the speaker axes criss-cross in front of the listening area, and I can explain the reasoning if you would like. Let me emphasize that the same setup geometry is recommended for my non-axisymmetric waveguide speakers. In other words the toe-in is not a "fix" for the on-axis response hole, but it does make it essentially irrelevant.

One setting where the on-axis hole would be a major issue is in a recording studio setting, where mixing (and probably mastering) are done on-axis. Dutch & Dutch evidently uses a waveguide geometry which mitigates this issue.

I have built loudspeakers which are virtually identical except for one using a SEOS waveguide and the other an Oblate Spheroid. The Oblate Spheroid is the audibly superior device regardless of listening axis, but especially so when used as intended.

It is easy to lose sight of the forest (the desired outcome) for the trees (the measurements). The desired outcome is that the direct sound be as correct as possible, and that the reverberant sound be as correct as possible. The measurements are highly useful but they are not the desired outcome in and of themselves. In my opinion.

Also, please note that the spin-o-rama measurements of a non-axisymmetric device inherently MISS what is happening along the diagonals, and just as much sound is delivered along the diagonals as along the standard measurement axes. Eyeball the JBL M2's waveguide and you will see what I mean. So it may well be that a truly comprehensive set of measurements would tell a significantly different story.

A spinorama makes enough measurements to have the diagonal data (supposed to approximate a sphere IIRC), though admittedly diagonal data isn't used to compute listening window and early reflections.
 

direstraitsfan98

Addicted to Fun and Learning
Forum Donor
Joined
Oct 1, 2018
Messages
826
Likes
1,225
@Ilkless

I know you’ve mentioned them a few times already but do you (or anyone for that matter) have any thoughts on Barefoot speakers? What are your thoughts about using them to replace a traditional passive and amp combo?
 

Senior NEET Engineer

Addicted to Fun and Learning
Joined
Jan 6, 2020
Messages
538
Likes
591
Location
San Diego
This diagram is the distribution of control room frequency response for just one speaker. There are probably more colors of mastering speakers than flavors of ice cream. Many albums are mastered with B&W 800 series speakers for example. I wonder if there is any point to a "state of the art" design for consumer use. Past a certain point the improvements may just get lost in the noise from circle of confusion.

Makivirta+and+Anet+2001.png
 

tuga

Major Contributor
Joined
Feb 5, 2020
Messages
3,984
Likes
4,281
Location
Oxford, England
This diagram is the distribution of control room frequency response for just one speaker. There are probably more colors of mastering speakers than flavors of ice cream. Many albums are mastered with B&W 800 series speakers for example. I wonder if there is any point to a "state of the art" design for consumer use. Past a certain point the improvements may just get lost in the noise from circle of confusion.

Makivirta+and+Anet+2001.png

From a consumer perspective I see nothing wrong with offering loudspeakers designed with somewhat different tonal balance flavours to cater for different tastes; but there's no excuse for performance in other parameters not to be as good as permited by the topology and the price point.

It is only us 0.0000001% – the audiophiles – that really worry about how accurate or how pleasing/euphonic a speaker sounds; and manufacturers who's raison-d'être or goal is to sell as many pairs as possible.
Most people just want speakers to be as small as possible, look as good as possible, and work.
 

pierre

Addicted to Fun and Learning
Forum Donor
Joined
Jul 1, 2017
Messages
962
Likes
3,046
Location
Switzerland
This diagram is the distribution of control room frequency response for just one speaker. There are probably more colors of mastering speakers than flavors of ice cream. Many albums are mastered with B&W 800 series speakers for example. I wonder if there is any point to a "state of the art" design for consumer use. Past a certain point the improvements may just get lost in the noise from circle of confusion.

Makivirta+and+Anet+2001.png

To be fair, that's 2001 data. Most of the studios I have been in the last few years are using room correction (some GLM, a lot with Dirac and Sonarworks). So I believe situation is on average better. Many studios are also paying for Dolby certifications or similar process which improves homogeneity of studios.

After that, many still do not believe in a flat accurate response and prefer to listen and adapt (after some time, you get used to your room and mix such that it "translates" well to most speakers. I think this is much easier to do so if you start with a flat FR with a RT60 within bounds, plus it is easier to adapt to a new room.

A good question, is how flat it needs to be such that mix and mastering are homogeneous and translate equally well? I hope it is within +/- 1.5dB because I have yet to see a studio which is flat 20Hz-20kHz +/-1 dB without FIR filters.

It is also not clear to me if studio are consistent using a flat on axis target or using a harmann curve (or similar, and in this case which one) for mastering.
 
Last edited:

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
It is also not clear to me if studio are consistent using a flat on axis target or using a harmann curve (or similar, and in this case which one) for mastering.

Many studios in Europe still use a flat in-room target response (although the practice is increasingly less common IIUC - and for good reason). I think this probably explains why so many setups in the pic posted by @Senior NEET Engineer measure flat(ish), or are even slightly upward-tilted.

I also understand that with the increasing popularity of automated correction products (like Sonarworks), Harman-like target responses are being employed by more and more (usually smaller) studios. I find this to be a concerning development, personally.

Take what I say with a grain of salt, though. It's based on anecdotal experience only (in Germany).
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,739
Likes
16,171
I also understand that with the increasing popularity of automated correction products (like Sonarworks), Harman-like target responses are being employed by more and more (usually smaller) studios. I find this to be a concerning development, personally.
Although just to add, Sonarworks' default for loudspeaker correction is unfortunately also a flat target.
 

pierre

Addicted to Fun and Learning
Forum Donor
Joined
Jul 1, 2017
Messages
962
Likes
3,046
Location
Switzerland
Many studios in Europe still use a flat in-room target response (although the practice is increasingly less common IIUC - and for good reason). I think this probably explains why so many setups in the pic posted by @Senior NEET Engineer measure flat(ish), or are even slightly upward-tilted.

I also understand that with the increasing popularity of automated correction products (like Sonarworks), Harman-like target responses are being employed by more and more (usually smaller) studios. I find this to be a concerning development, personally.

Take what I say with a grain of salt, though. It's based on anecdotal experience only (in Germany).

that’s my experience as well. Still if all studios are using a flat target, you can learn to mix for that and it will be consistent. if people start to use a tilt provided by Sonarworks or similar but they do not agree on the slope we will not improve.
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,739
Likes
16,171
that’s my experience as well. Still if all studios are using a flat target, you can learn to mix for that and it will be consistent.
Not really, as the FR slope at the LP depends on the directivity of the loudspeakers, room reverberation curve and listening distance, while the perception above transition frequency is more relating to the direct sound, that's why that should be flat and not the LP sum measurement of direct and reflected sounds.
https://www.audiosciencereview.com/...ut-room-curve-targets-room-eq-and-more.10950/
 

pierre

Addicted to Fun and Learning
Forum Donor
Joined
Jul 1, 2017
Messages
962
Likes
3,046
Location
Switzerland
Top Bottom