• 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

Kyle / MrHeeHo

Active Member
Joined
Mar 30, 2018
Messages
163
Likes
183
Since waveguides are a part of speaker design I've been wondering if they have also have any place in headphone designs. Do the 'Fazors' Audeze use in their headphones have any scientific merit to them?
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
Idea based speaker design is about the same as marketing based speaker design. An idea is a hypothesis. If supported by insurmountable evidence it becomes a theory and if completely provable, a law of science.
I don't dispute that. But in this case, even if stated and tested, the hypothesis and evidence would only partially overlap what a speaker should do.

And the hypothesis is not even stated. In fact, it is vaguely assumed from the evidence! The ready availability of frequency response measurements is leading people to think that they self-evidently represent human hearing and speakers.

If they were forced to state what they thought were the 'laws' of frequency response to any sort of standard beyond vague (not even stated) 'rule of thumb', their claims would fall apart. Instead, we have 'target in-room responses' that suggest (without stating it) that human hearing is a sliding frequency response window, and an entire industry and philosophy of managing the resulting unpredictability and disappointment has developed around this.

The ideas person would know that the in-room frequency response is ambiguous and cannot tell you what is direct sound or room sound, but that a human can. They would know that, at least, a 'Spin-o-rama' measurement would allow you to get a handle on which is which and how to get a better match between them.

Or (much better) they could approach it from the other direction: define what the ideal speaker should do and design something to realise it (as close as practical). In this case, the design would not be based on evidence at all, except for final confirmation and (this being the real world) refinement.

How many people would be prepared to define what the ideal speaker should do before they go looking for evidence? Not many.
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
I can also say that I have followed this philosophy myself:

The very first pair of speakers I built were the full monty: three-way, sealed boxes, wide baffle for the woofer, DSP, active, linear phase, time alignment delays, baffle step compensation for each driver, etc. I remember thinking that there was simply no point in doing anything else. *And* I never measured the result e.g. in-room, or any measurements except for the individual drivers. I did, however, refine the result in-room by ear, having constrained the adjustment to a single variable - baffle step compensation depth.

I only later measured the step response out of curiosity - I was pretty sure it was working properly because of the way it was designed; and it was.

They're not the theoretical 'ideal speaker' but at least I think I understood what I was aiming for and what were the compromises before I started. I would know even better now, having had regular discussions on ASR about it over the last couple of years! :)
 
Last edited:
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
I don't dispute that. But in this case, even if stated and tested, the hypothesis and evidence would only partially overlap what a speaker should do.

And the hypothesis is not even stated. In fact, it is vaguely assumed from the evidence! The ready availability of frequency response measurements is leading people to think that they self-evidently represent human hearing and speakers.

If they were forced to state what they thought were the 'laws' of frequency response to any sort of standard beyond vague (not even stated) 'rule of thumb', their claims would fall apart. Instead, we have 'target in-room responses' that suggest (without stating it) that human hearing is a sliding frequency response window, and an entire industry and philosophy of managing the resulting unpredictability and disappointment has developed around this.

The ideas person would know that the in-room frequency response is ambiguous and cannot tell you what is direct sound or room sound, but that a human can. They would know that, at least, a 'Spin-o-rama' measurement would allow you to get a handle on which is which and how to get a better match between them.

Or (much better) they could approach it from the other direction: define what the ideal speaker should do and design something to realise it (as close as practical). In this case, the design would not be based on evidence at all, except for final confirmation and (this being the real world) refinement.

How many people would be prepared to define what the ideal speaker should do before they go looking for evidence? Not many.

You still have not fulfilled your burden of proof of falsifying the wave equation, as outlined to you.
 

Mad_Economist

Addicted to Fun and Learning
Audio Company
Joined
Nov 29, 2017
Messages
543
Likes
1,618
Since waveguides are a part of speaker design I've been wondering if they have also have any place in headphone designs. Do the 'Fazors' Audeze use in their headphones have any scientific merit to them?

Probably not as much as Audeze's marketing attributes to them, but they do have some impact. I 3D printed a little fake Fazor (same height as the ones on the LCD2F I measured) with the correct spacing for Monoprice's M1060, whose driver is analogous to the LCD3 in its design and which I happen to have kicking around.
fakezor.jpg


For added Audeze fun, I used an LCD2 pad (because the stock pad doesn't fit with my little 3D printed "fakezor"). The results are below
fakezor vs. stock.png


Both measurements are the sum of five positions on my Brüel & Kjær 4128C HATS, compensated to diffuse field. Impacts below 4khz are equivalent to the variation from placements. Above there, we see more distinct changes

fakezor difference.png


The largest below 10khz is in an area where the M1060 (and the LCD3, I believe) has a quite substantial resonance, and this may limit the meaningfulness of the apparent change there. Other than that, behavior around 8-10khz is quite significantly different. This area is quite sensitive to positioning, but the use of multiple placements should decrease this effect somewhat, and I would feel comfortable attributing at least part of this change to the "fakezor". The same may be true of the 11-15khz change, but this area is particularly variable with even very small adjustments, so I'd really want to do more tests to make a statement there - and it's late and I should really be going to bed. If this is of interest to anyone, I could do some additional measurements later.

Edit: Note that the difference plot is "fakezor" minus the lack thereof - flip it over to see the FR change from adding them, I was too lazy to redo it the right way around :p
 
Last edited:

MSNWatch

Active Member
Joined
Dec 17, 2018
Messages
142
Likes
171
Very interesting thread. I have owned a pair of Technics SBC-700 with a single sealed 12 inch SVS sub (I know I need a pair) with Anthem Room Correction applied to 500Hz and below and 4th order crossovers set at 80Hz and I have to say the sound is as good as it gets - and great horizontal and vertical dispersion also. I have many other concurrent speakers set up at the moment which measure well - NHT Xds, Infinity IL-60s, Totem Mani-2 signatures (the best measuring of the brand), and Energy Veritas 1.8s and 2.2s and the Technics system is the most pleasing for me to listen to. I have also owned many others over the years from Celestions and Spicas years ago to Apogees, Flatline 175, Martin Logans, several custom pairs, B&Ws, Sony ES and many others and have kept what I currently have.

I'm intrigued enough by the discussion here and in its own thread that I am now considering the Epique CBT24s. At least in my experience (and perhaps preference) the speakers I prefer tend to have very wide sweet spots which I realize are engineering decisions made by the speaker makers and this pair looks like it would offer the widest sweet spot of them all. I would run it with a sub with Anthem Room Correction to try to get the most out of the pair.
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
You still have not fulfilled your burden of proof of falsifying the wave equation, as outlined to you.
I wasn't aware that I had tried to overturn any law of physics..?

You said this stuff:
The user went so far as to question the validity of power response, when it is a direct representation of the actual, physical acoustic radiation of a loudspeaker system - with baffle-step being a natural off-shoot of the wave equation. Cosmik has not managed to falsify the wave equation. This is his burden of proof for his proposition that our current understanding of acoustics - and speaker design based on it - is merely another sui generis idea to be reckoned equally with other abstract conjectures absent direct experimental verification.

I don't quite follow it, but within it I think you are saying that 'power response' tells you everything about a speaker..?

My dispute is with the idea that a frequency response accumulation at some place in a room tells you everything about the speaker. This is why the Harman people developed the 'Spin-o-rama' (or rather they probably worked it out logically and rationally).

It may be that 'speaker design' often *is* based on an in-room frequency response measurement or variant thereof, in some vague way, but it doesn't mean that such a 'target response' approach is valid.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
I’m really struggling to see how all this discussion of target response is relevant to the thread.

The evidence mostly suggests that simply designing to a target in-room response is not the best approach. So I don’t understand why it keeps coming up here..
 

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,690
Likes
6,013
Location
Berlin, Germany
edit: To keep this thread on topic, I have updated the HEDD entry in OP and placed it on the main list.
In my view, Type05 and Type07 don't fully qualify as thoroughly engineered compared to Type30 and Type20 models. The reason is that 05/07 were the initial desgins to start the company and therefore several corners have been cut, you have to start with something and have limited time and resources. The main difference is that the 05/07 crossover alignments have not been designed systematically from ground zero by defining proper acoustic targets for the drivers from which the actual crossover is derived, rather the responses have been established more by an old-school approach using a dbx controller, dial in some PEQ and highpass/lowpass starting from educated guesses "until it looks and sounds good". The acoustic phase of woofer and tweeter isn't fully coherent throughout a wide range arround the XO point and the response isn't as smooth as it gets, as a result. Still very acceptable in absolute terms because Mr.Heinz is of course a very experienced and seasoned overall speaker and speaker driver designer.
On the contrary, the 30/20 XOs have been created the other way round in a much more time-consuming process, define the targets first (the "educated guesses" part moves to here), measure the raw drivers anechoic under a multitude of angles, apply weighted averaging vs. angle on these to obtain a processed "raw" response that represents a listening window wider that strict on-axis**), then finally sythesize the filter blocks needed to get these averaged raw responses on target (within the constraints of the filter hardware which have been rather severe). Therefore, the magnitude response is smoother and the phase coherency is much better which, together with being 3-way, payed off significantly in percieved sound.... but as the partial designer and person in charge of many relevant aspects of those models, I'm biased of course.... not affiliated with HEDD anymore, btw.

**) this is different from eg the Neumann approach where the design goal is perfectly flat on-axis free field (anechoic), with the baffle-step and any diffraction effects fully compensated but with the risk of increased disturbance off-axis. Whereas Type30/20 don't measure flat on-axis anechoic but average out nicely over some range of listening angles.
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
From that article:

From a stereophonic imaging standpoint, the square loudspeaker enclosure (1) also lets us down, because the acoustic impedance change at the sharp corners causes sound from the tweeter to be re-radiated. The re-radiation interferes with the direct sound to make the directivity pattern periodic. From an imaging standpoint, the sound source is the width of the speaker instead of a point. This aperture effect destroys the imaging ability as no virtual sound source can be narrower than a speaker.

Could this explain why some of us percieve wide bodied/big driver speakers produce wider/bigger/fuller sonic images?

Well, I'm not going to make any difinitive statement here but an empirical observation regarding my speaker design. They aren't commercial just something I did for me.

They are typically thin floor standers, (3 way active dsp xo) however have a 25mm radius around all front corners. The drivers were all chosen to have the widest dispersion.

The number one thing multiple people have stated when they hear them (it's the first thing that hits them) is regarding how wide, enveloping and defined the soundstage is.

So in my case, thin isn't an issue.

There are other factors, the room is treated with a mixture of diffusion and absorption, including first reflection point, and has diffusion on the ceiling. This is going to affect the sense of envelopment, but I don't know what the big factor is here, dispersion, speaker corner rounding or room acoustics.
 
Last edited:
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
The acoustic phase of woofer and tweeter isn't fully coherent throughout a wide range arround the XO point and the response isn't as smooth as it gets, as a result. Still very acceptable in absolute terms because Mr.Heinz is of course a very experienced and seasoned overall speaker and speaker driver designer.
On the contrary, the 30/20 XOs have been created the other way round in a much more time-consuming process, define the targets first (the "educated guesses" part moves to here), measure the raw drivers anechoic under a multitude of angles, apply weighted averaging vs. angle on these to obtain a processed "raw" response that represents a listening window wider that strict on-axis**), then finally sythesize the filter blocks needed to get these averaged raw responses on target (within the constraints of the filter hardware which have been rather severe). Therefore, the magnitude response is smoother and the phase coherency is much better which, together with being 3-way, payed off significantly in percieved sound.... but as the partial designer and person in charge of many relevant aspects of those models, I'm biased of course.... not affiliated with HEDD anymore, btw.

**) this is different from eg the Neumann approach where the design goal is perfectly flat on-axis free field (anechoic), with the baffle-step and any diffraction effects fully compensated but with the risk of increased disturbance off-axis. Whereas Type30/20 don't measure flat on-axis anechoic but average out nicely over some range of listening angles.

What do you think of the Lineariser? Does it mean there only needs to be a "good enough" analog active crossover for things that can't be corrected upstream (directivity, non-linear distortion, compression), and leave everything else to undergo DSP upstream without having to incur an AD/DA conversion?

Also, there is indeed such a risk if only on-axis is used because of local interference effects that bear little relevance to the direct sound perceived by the listener - its the reason Harman also uses such spatial averaging of the listening window. However, Neumann, by any measure, has avoided the pitfall of designing around these local interference effects to the detriment of off-axis. The on-axis response is flawless, as are the horizontal polars.

KH120A vs a popular German kit that's as well-designed:

dxt_mon_kh120_deg_hor_vgl.gif


Neumann's own product page also links to a review from Sound and Recording with detailed measurements. The only reason I don't already own a Neumann is because the KH120A is too small for my needs (prefer a 7-inch + waveguide), but the KH310A out of budget.
 

Erik

Active Member
Joined
Jul 1, 2018
Messages
137
Likes
271
KH120 and KH310 horizontal directivity plots.

1551490597400.png 1551490617600.png

Intermodulation distortion of a two-way and a three-way loudspeakers.

1551490891200.png 1551490911300.png

These IMD graphs are from Neumann glossary, so I suspect they've put their own products data on them.

KH310 should have lower IMD level than KH120 but their horizontal dispersion looks worse.
Is there any point in choosing KH310 over KH120 in terms of sound quality? Except of KH310 playing lower and louder.
 

Sancus

Major Contributor
Forum Donor
Joined
Nov 30, 2018
Messages
2,923
Likes
7,616
Location
Canada
Is there any point in choosing KH310 over KH120 in terms of sound quality? Except of KH310 playing lower and louder.

Is it just me or is that directivity not really all that hot for a $2250 speaker?
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Is it just me or is that directivity not really all that hot for a $2250 speaker?

The problem is because the driver configuration is not symmetric. Neumann has already pushed that configuration to its limit, here are third-party measurements. As good as it can get for this sort of configuration - and well above what remains the norm in this market. Dynaudio makes similar speakers with nowhere as good directivity (look at that ridiculous 5.5kHz tweeter crossover) as does HEDD. A more familiar vertical TMW would yield both the advantages in IMD, distortion, compression etc without giving up horizontal dispersion, but at the expense of a larger baffle size and a far less compact speaker given the driver sizes. And indeed they do it for the flagship KH420, which has polars as good as the KH120.
 
Last edited:

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,690
Likes
6,013
Location
Berlin, Germany
What do you think of the Lineariser? Does it mean there only needs to be a "good enough" analog active crossover for things that can't be corrected upstream (directivity, non-linear distortion, compression), and leave everything else to undergo DSP upstream without having to incur an AD/DA conversion?
The Linearizer originally was my idea (actually a rather old one since I'v been doing this privately for more than a decade) but I left the company during the conceptual stage and had no control of what went in there finally.

The main point of it is phase unwrapping because that makes no sense to do in analog as it compromises the design right from the start in most any other regards. If you want to design true phase-coherent crossovers they cannot be transient-perfect at the same time, this is a mathematical impossibilty, systems theory thing (unless you resort to a huge peaking allpass cascade which is PITA to design and implement, PSI Audio from Switzerland is one of the few companies who do this).

However, a linear phase crossover with 100% phase coherency is simply established from the minimum phase analog XO by applying the time-inverse of the system allpass function with zero side effects (yes, on of the few instances of a free lunch, almost, some digital headroom is lost but that isn't an problem with 24bit).

Further, phase unwrapping can be extended to the low-frequency roll-off to "speed up the bass" which is relevant for bass-reflex system with not so super-low cutoff (> 40Hz'ish) notably when the order is high (HEDDs have the typical 2nd-order electric system highpass with gives a 6th order total). Full unwrapping to linear phase causes too many artifact and increases latency but an intermediate target is useful, for example a 2nd order phase target around the cutoff frequency. It is important for rendered signal quality that the correction kernel is derived analytically (either by simple equivalent functions or by curve fitting) rather than from measured data directly because this assures a clean kernel with no artifacts whatsoever (see Grimm LS1 whitepaper for this). I don't know how the final correction kernels for the Linearizer were created.

Since the processing capabilities are there anyway it is tempting to do minor magnitude corrections as well. HEDD seems to have opted to correct for a more flat on-axis response which is sort of reasonable industry-standard for monitors (and it definitely sells better to the spec-sheet focussed crowd) and cover some small design flaws in the analog crossovers as well, as noted.

Since this preprocessing can only affect overall global response but not the individual drivers, it cannot fix any inter-driver issues like phase incoherency. Therefore, the analog crossover must at least care to implement the transition regions between drivers as perfect as it gets but you can cut corners on the total magnitude response.

Also, there is indeed such a risk if only on-axis is used because of local interference effects that bear little relevance to the direct sound perceived by the listener - its the reason Harman also uses such spatial averaging of the listening window. However, Neumann, by any measure, has avoided the pitfall of designing around these local interference effects to the detriment of off-axis. The on-axis response is flawless, as are the horizontal polars.
[...]
Neumann's own product page also links to a review from Sound and Recording with detailed measurements. The only reason I don't already own a Neumann is because the KH120A is too small for my needs (prefer a 7-inch + waveguide), but the KH310A out of budget.
Neumann's are for sure one of the best speakers in engineering / spec-sheet terms and they managed to achieve world-class directivity for the choosen design concepts. The KH120 is a small nearfield monitor that I wouldn't use for anything else than modest levels in a near-field desktop environment. They quickly sound stressed when forced to fill a room. The 310 doesn't go that much louder and lower but the sound doesn't fall apart as quickly because it's a 3-way. IMHO, still needs a sub to fill a room, though. Leaves us with the KH420...

In the end you must always listen for yourself as specs and standard measurements alone don't tell the whole truth.
 
OP
Ilkless

Ilkless

Major Contributor
Forum Donor
Joined
Jan 26, 2019
Messages
1,757
Likes
3,437
Location
Singapore
Neumann's are for sure one of the best speakers in engineering / spec-sheet terms and they managed to achieve world-class directivity for the choosen design concepts. The KH120 is a small nearfield monitor that I wouldn't use for anything else than modest levels in a near-field desktop environment. They quickly sound stressed when forced to fill a room. The 310 doesn't go that much louder and lower but the sound doesn't fall apart as quickly because it's a 3-way. IMHO, still needs a sub to fill a room, though. Leaves us with the KH420...

In the end you must always listen for yourself as specs and standard measurements alone don't tell the whole truth.

Yes, but the choice of "design concept" is ultimately selected based on application (bass extension, listening distance, SPL), and it is possible to evaluate absolute performance within the same class of "design concept".

The abstract and arbitrary notion of "filling a room" ultimately boils down to empirical, quantifiable concepts like bass extension, listening distance and SPL, in relation to the spectral content of the input signal.
 
Last edited:

MattHooper

Master Contributor
Forum Donor
Joined
Jan 27, 2019
Messages
7,195
Likes
11,808
Well, I'm not going to make any difinitive statement here but an empirical observation regarding my speaker design. They aren't commercial just something I did for me.

They are typically thin floor standers, (3 way active dsp xo) however have a 25mm radius around all front corners. The drivers were all chosen to have the widest dispersion.

The number one thing multiple people have stated when they hear them (it's the first thing that hits them) is regarding how wide, enveloping and defined the soundstage is.

So in my case, thin isn't an issue.

There are other factors, the room is treated with a mixture of diffusion and absorption, including first reflection point, and has diffusion on the ceiling. This is going to affect the sense of envelopment, but I don't know what the big factor is here, dispersion, speaker corner rounding or room acoustics.

Thanks for your reply!

The description you gave isn't the sound characteristic I've been trying to describe.

It's almost typical for thin floor standing speakers to do a disappearing act with great soundstaging. One of my favorite speaker brands - Audio Physic - is known for that type of presentation, and pretty much every speaker I've owned has produced wide immersive soundstaging.

What I'm talking about is the impression of size, body and weight of the actual instruments and voices within a soundstage. The sonic richness, the sense that you are hearing something closer to a full sized, fully-weighted acoustic guitar, or sax, or piano, or drums, vs a smaller or less subsantial instrumental reproduction (even if placed in an immersive soundstage).
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
Thanks for your reply!

The description you gave isn't the sound characteristic I've been trying to describe.

It's almost typical for thin floor standing speakers to do a disappearing act with great soundstaging. One of my favorite speaker brands - Audio Physic - is known for that type of presentation, and pretty much every speaker I've owned has produced wide immersive soundstaging.

What I'm talking about is the impression of size, body and weight of the actual instruments and voices within a soundstage. The sonic richness, the sense that you are hearing something closer to a full sized, fully-weighted acoustic guitar, or sax, or piano, or drums, vs a smaller or less subsantial instrumental reproduction (even if placed in an immersive soundstage).

If I understand what you are getting at, and I'm not sure I do :) I would suggest this has more to do with the microphone set up on a recording.

If you are talking tonal richness then that's more frequency response. A wide baffle will have different baffle step characteristics.
 
Top Bottom