• 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!

Spinorama! (also known as CTA/CEA 2034 but that sounds dull, apparently)

Dave Zan

Active Member
Joined
Nov 19, 2019
Messages
168
Likes
486
Location
Canberra, Australia
Amir has posted his first speaker review based on Klippel measurements and it includes a lot of excellent information about the CTA (formerly CEA) 2034 standard.
I have a few queries and ideas about that standard and probably other people have some interest in it too.
So I have started this thread as a place for CEA 2034 discussions rather than tie it to a specific speaker review.

For context, the standard is heavily based on work by Floyd Toole and other researchers associated with JBL. so if you have his book it should be familiar.
A free copy was obtainable from the web if you were prepared to fill out a few details, probably still is :)
There is a new version in preparation by the CTA, if anyone knows of a review copy I would like to check it.

The copy I have is CTA 2034-A and it seems to have some poorly drafted and edited bits.
So some of my queries are about possible typos, other are more conceptual.
Just to start, on p.12
"Early Reflections" are defined as
Floor Bounce 20, 30, 40 down
Ceiling Bounce 40, 50, 60 up
Front Wall Bounce 0, +-10, +-20, +-30 horizontal
Side Wall Bounces +-40, +-50, +-60, +-70, +-80 horizontal
Rear Wall Bounces 180, +-90 horizontal

This is as shown by Amir on his charts for the JBL 305.
The standard then defines a new list of terms
"Vertical Reflections"
Floor Reflection -20, -30, -40 vertical

(in other words identical to the "Floor bounce" defined above)
Similarly with
"Horizontal Reflections"
Front
the same
Side the same
but
Rear +-90, +-100, +-110, +-120, +-130, +-140, +-150, +-160, +-170, 180
Not the same, looks like the first is a mistake and the second is a correction, can any one confirm this?

*UPDATE
It was confirmed by T Welti that the first is a mistake.
A new revision of the standard is underway and should correct this.
The Klippel software is based on the standard and hence is also inconsistent with the research.
The intention is to correct this too.


As I checked the above numbers I also noticed one quirk.
The direct sound is 0, +-10 vertical,
The early reflections include +40, +50, +60 v.
So +20 and +30 just kind of vanish! (OK, they do appear in the calculation for sound power.)
This may be deliberate but it seems odd.
Whereas the sidewall bounces start where the direct sound ends. and similarly for the floor bounce.

On a conceptual note.
CTA 2034 only measures 2 arcs, a vertical one and a horizontal.
A speaker could put substantial power into diagonal anomalies and sound poor but would measure fine.
Not a problem with circular horns but maybe an issue with JBL M2 style horns.
The Klippel measures full spherical data, would spot this but not show it in the Spinorama.
Does Klippel, or anyone else, address this?

The standard has a few more issues with data presentation but I may hold that until I see if any one is actually interested.

David
 
Last edited:

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Amir has posted his first speaker review based on Klippel measurements and it includes a lot of excellent information about the CTA (formerly CEA) 2034 standard.
I have a few queries and ideas about that standard and probably other people have some interest in it too.
So I have started this thread as a place for CEA 2034 discussions rather than tie it to a specific speaker review.

For context, the standard is heavily based on work by Floyd Toole and other researchers associated with JBL. so if you have his book it should be familiar.
A free copy was obtainable from the web if you were prepared to fill out a few details, probably still is :)
There is a new version in preparation by the CTA, if anyone knows of a review copy I would like to check it.

The copy I have is CTA 2034-B and it seems to have some poorly drafted and edited bits.
So some of my queries are about possible typos, other are more conceptual.
Just to start, on p.12
"Early Reflections" are defined as
Floor Bounce 20, 30, 40 down
Ceiling Bounce 40, 50, 60 up
Front Wall Bounce 0, +-10, +-20, +-30 horizontal
Side Wall Bounces +-40, +-50, +-60, +-70, +-80 horizontal
Rear Wall Bounces 180, +-90 horizontal

This is as shown by Amir on his charts for the JBL 305.
The standard then defines a new list of terms
"Vertical Reflections"
Floor Reflection -20, -30, -40 vertical

(in other words identical to the "Floor bounce" defined above)
Similarly with
"Horizontal Reflections"
Front
the same
Side the same
but
Rear +-90, +-100, +-110, +-120, +-130, +-140, +-150, +-160, +-170, 180
Not the same, looks like the first is a mistake and the second is a correction, can any one confirm this?

As I checked the above numbers I also noticed one quirk.
The direct sound is 0, +-10 vertical,
The early reflections include +40, +50, +60 v.
So +20 and +30 just kind of vanish! (OK, they do appear in the calculation for sound power.)
This may be deliberate but it seems odd.
Whereas the sidewall bounces start where the direct sound ends. and similarly for the floor bounce.

On a conceptual note.
CTA 2034 only measures 2 arcs, a vertical one and a horizontal.
A speaker could put substantial power into diagonal anomalies and sound poor but would measure fine.
Not a problem with circular horns but maybe an issue with JBL M2 style horns.
The Klippel measures full spherical data, would spot this but not show it in the Spinorama.
Does Klippel, or anyone else, address this?

The standard has a few more issues with data presentation but I may hold that until I see if any one is actually interested.

David

First, question, why not stick to 2034-A until 2034-B is finalized? Although that's scheduled to happen soon. CTA-2034-A is free to download from the CTA here.

I'm not clear what mistakes you're pointing out though. These all look kosher to me.

The Rear Horizontal Reflections Curve is not meant to be part of the normal Early reflections curve processing, I don't think. It's an extra curve for those that want to include it, probably intended mainly for irregular speaker designs like dipole and omnidirectional speakers.

The early reflections curve does not include the +20 and +30 (as in 20/30 degrees above the reference axis) because they are not angles for typical ceiling reflections. This makes sense- speakers are usually closer to the floor than the ceiling.

The angles chosen were selected by analyzing the angles of the first, single-bounce reflections in typical rooms. Remember we are more sensitive to horizontal bounces, and we usually have two speakers set apart horizontally, so it makes sense we'd include a lot more horizontal data in the early reflections calculation than vertical data.

Your point about diagonal anomalies is something I've wondered before, but I can only conclude researches didn't find them to be a big enough issue to be concerned about. It is something a system like the Klippel could probabl help with. But I do think that the sound power measurement weighting helps mitigate such issues.
 
Last edited:
OP
Dave Zan

Dave Zan

Active Member
Joined
Nov 19, 2019
Messages
168
Likes
486
Location
Canberra, Australia
First, question, why not stick to 2034-A until 2034-B is finalized? Although that's scheduled to happen soon. CTA-2034-A is free to download from the CTA here.

My typo, now corrected, thank you.

I'm not clear what mistakes...These all look kosher to me.

The Rear Horizontal Reflections Curve is not meant to be part of the normal Early reflections... I don't think.

I suspect it is meant to be part.
What I think has happened is that there was an error in the so-called "bounce" specifications and that the section was rewritten and renamed "reflections" but the old text not removed.
Old text seems to have been left behind elsewhere in the document too, maybe a slip up in the editor procedures.
I think what was written was probably meant to be "180 +-90" i.e 180 plus or minus 90, so 90, 100, 110,....to 270 (or -90),
but it was typed as "180, +-90" i.e. just three points.

The early reflections curve does not include the +20 and +30... This makes sense- speakers are usually closer to the floor...

Yes, I understand the reason.
It just seems a little odd that the first 10 is crucial, then a dead zone, then early reflections that are important.
It looks similar to dead zones in radio reception for waves bounced off the ionosphere.
So we can push anomalies in the vertical polar response into this window and hide them!
Seriously I wonder if the JBL M2 does this, it's verticals are a bit wobbly but the spinorama looks almost perfect.

Your point about diagonal anomalies...but I can only conclude researches...

I think the researchers would admit that it a potential issue but the problem is small for typical domestic speakers and the spinorama is already a lot of measurements to make.
I notice the professional systems do a full spherical scan.

Best wishes
David
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
I think what was written was probably meant to be "180 +-90" i.e 180 plus or minus 90, so 90, 100, 110,....to 270 (or -90),
but it was typed as "180, +-90" i.e. just three points.

No, this is not the case. The three points only is correct.The early reflections curve does not include 100,110,120, degrees etc.

It's meant to be a reflection of the most acoustically significant front hemisphere reflections because for standard speakers the rear reflections don't affect our perception all that much.

If you start to include the rear sound, the curve ends up being a lot closer to the sound power curve(I.e. tilted much more downward). Have seen this from my own measurements in practice, because I once thought the same thing. Furthermore, from Toole's book:

"Different forward-firing loudspeakers in the same room generate steady -state room curves that are well predicted at middle and high frequencies by data from only the front hemisphere. Of that data the sounds radiating at angles appropriate to early reflections contribute the bulk of the information. Full spherical data, sound power, is not necessary for loudspeakers of this configuration."

.

Yes, I understand the reason.
It just seems a little odd that the first 10 is crucial, then a dead zone, then early reflections that are important.
It looks similar to dead zones in radio reception for waves bounced off the ionosphere.
So we can push anomalies in the vertical polar response into this window and hide them!
Seriously I wonder if the JBL M2 does this, it's verticals are a bit wobbly but the spinorama looks almost perfect.

It's not that other angles aren't important, but yes, the ones included in the early reflections curve are deemed to be the most important of the vertical reflections.

Remember we are more sensitive to horizontal reflections in the first place - including more vertical angles would probably unduly weigh their significance in perceived sound. So instead we just include the most perceptually important ones.

In any case, that's what the sound power curve comes in - if manufacturers were "hiding" anomalies in the ER curve, they would show up in the sound power curve. Very rarely is the sound power curve significantly different in shape from the ER curve, other than being further tilted downward.

It *is* true that in some rooms and setups, the early reflections angles may be different. The ideal would be figuring the reflection angles in your individual room. But the standardized ER curve seems to be a good predictor of what we actually hear.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
This work is based on this paper and research: AES Paper
Characterizing the Amplitude Response of Loudspeaker Systems
Allan Devantier
Harman International Industries Inc., Northridge, CA, 91329, USA

15 different rooms were examined and a set of heuristics determined as to the strongest and most important reflections.

1579465036419.png
 
OP
Dave Zan

Dave Zan

Active Member
Joined
Nov 19, 2019
Messages
168
Likes
486
Location
Canberra, Australia
No, this is not the case. The three points only is correct.

That's what I started the thread to check, either one way or the other. Hopefully the Allan Devantier paper will make it clear.

In any case, that's what the sound power curve comes in.. they would show up in the sound power curve.

Yes, but substantially de-emphasised.
I was only half serious, however it does seem possible, just as anomalies could be pushed into the diagonals.
I do think I will try to push any anomalies in my own speakers into the 20 - 30 zone - should make the spinorama look better:)

It *is* true that in some rooms and setups...

I realise it is based on "typical" rooms and presumably reasonable for them.
The surprise is that it's a discontinuous function, so may not deal gracefully with other rooms.


Does the paper define exactly what is included in the various subsections?
I have access to JAES but not readily to conference papers.

Best wishes
David
 
Last edited:

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Yes, but substantially de-emphasised.
I was only half serious, however it does seem possible, just as anomalies could be pushed into the diagonals.
I do think I will try to push any anomalies in my own speakers into the 20 - 30 zone - should make the spinorama look better:)

Good idea, not only will the spins look better, the speakers will sound better too! :)
 
OP
Dave Zan

Dave Zan

Active Member
Joined
Nov 19, 2019
Messages
168
Likes
486
Location
Canberra, Australia
No, this is not the case. The three points only is correct. The early reflections curve does not include 100,110,120, degrees etc.

Do you have the Devantier paper to confirm this?
Or do you have an independent source?
The CTA 2034 document appears to be based on a direct cut and paste from Floyd Toole's presentation, or perhaps the reverse.
So an independent check would be nice.
But I take your point that this would then make it very similar to the sound power curve, so the difference does tend to support your position.
In which case my question is why it is done this way.

...because for standard speakers the rear reflections don't affect our perception all that much.

For standard speakers the rear reflections won't affect our perception all that much simply because the speaker doesn't put much power into the rear hemisphere anyway.
For speakers that do have substantial rear power why should these reflections be treated differently?
Most speakers are mounted not too far from the back wall so the reflections will be in about the same time window as side reflections.
And a reflection off the rear wall should be as likely to cover the listener as a similar spread of the front or side reflection, say 50° or 60° of arc.
So my initial reaction is that the rear wall reflections should be treated the same way, not just measured at one value.
I suppose the speakers that do have substantial rear power are precisely the ones usually not mounted close to the rear wall.
But it's still not clear to me.

Best wishes
David
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,065
Location
Zg, Cro
For standard speakers the rear reflections won't affect our perception all that much simply because the speaker doesn't put much power into the rear hemisphere anyway.
For speakers that do have substantial rear power why should these reflections be treated differently?

This sounds logical to me. And it is yet another proof how strong influence of bipolar speaker haters is toward those making audio standards. :D
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Do you have the Devantier paper to confirm this?
Or do you have an independent source?
The CTA 2034 document appears to be based on a direct cut and paste from Floyd Toole's presentation, or perhaps the reverse.
So an independent check would be nice.
But I take your point that this would then make it very similar to the sound power curve, so the difference does tend to support your position.
In which case my question is why it is done this way.

Seems weird to think that an international standard would be based on a presentation =] Here you go, straight from Devantier:

Snag_1c277aaf.png


The selection of angles is pretty clear, I think.:)

Edit: More specifically, since it's not evident from the above, the average floor reflection was 32.5 degrees and the average ceiling reflection was 48 degrees.
 
Last edited:

Koeitje

Major Contributor
Joined
Oct 10, 2019
Messages
2,292
Likes
3,880
OP
Dave Zan

Dave Zan

Active Member
Joined
Nov 19, 2019
Messages
168
Likes
486
Location
Canberra, Australia
Seems weird to think that an international standard would be based on a presentation

True, I had more in mind that the standard was based on Toole's work and that this work was also then shown as a presentation.
The point was that the two sources weren't independent.


That clearly presents the data for their decision to omit the 30° from the vertical reflections, thank you.
I didn't suspect any typo in that, my interest was more in the methodology.
The issue is that a particular angle is either selected or not, it's not a smooth function and is a bit arbitrary.
For instance -20 is included while +70 is omitted despite the fact they occur equally frequently.
Or that -20 has an equal contribution to the early reflections as -30 despite the fact it occurs far less frequently.

However, my questions mainly, and post #8 entirely, have been about the rear data.
Is there comparable data for the rear reflections?

Best wishes
David
 
Last edited:

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
However, my questions mainly, and post #8 entirely, have been about the rear data.
Is there comparable data for the rear reflections?

Ugh. I'm now having doubts and starting to think you may have been correct about how much rear data should be included. I see your logic better now, and upon rereading the Devantier paper more carefully, I'm more confused about calculating the Early Reflections curve than ever! It's a bit hard to summarize this portion of the paper, but the relevant bits are:

...In this plane we want to include all the horizontal measurements in our approximation of the early arrivals. After some experimentation and debate the following three spatial averages were defined: 1.) “Front” is the average of seven measurements at 0 degrees +/- 30 degrees. Careful readers will note this is very similar to the “direct sound”. 2.) “Side” is defined as the average of ten measurements at 60 degrees off the main-axis +/- 20 degrees to either side. 3.) “Rear” is defined as the average of 19 measurements at 180 degrees off the main-axis +/- 90 degrees (i.e.: the horizontal part of the rear hemisphere).

Equally important, the average appears to be described more explicitly in Devantier than in CTA-2034A

Snag_152a8962.png


So if I'm understanding correctly, it's not just a simple average of all the listed angles, as I previously assumed, but an average of five other averaged curves, each made up of several other curves. That certainly weighs the Rear measurements much more heavily than I'd believed. In addition to this, there is the confusion @Dave Zan originally pointed out in the Rear/Back curve, where at one point it appears to be written so as to mean an average of three (±90 and 180) measurements, and another to mean 19 measurements (±90 through 180). Reading the text of Devantier, it appears to be 19 measurements for rear reflections is the correct answer.

All this certainly weighs the rear data more heavily than I'd thought, leading to a slightly steeper slope in the ER curve. About1-2 dB down on some of my previous measurements. The shape is basically the same, so it's not a huge deal, but still.

Adding some personal frustration, previously my manual Early reflections calculations closely tracked those automatically made when I import the data into VituixCAD. Now the data is quite a bit off, so I'm wondering whether it's possible VituixCad is making the same mistake I was. Or maybe I'm just still misunderstanding the process.

I now think we need someone who works or has worked at Harman to chime in and settle the score on the procedure. Either way, it's clear that this needs to be clarified in the document
 
Last edited:

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,381
Location
Seattle Area
I now think we need someone who works or has worked at Harman to chime in and settle the score on the procedure. Either way, it's clear that this needs to be clarified in the document
Can you bring me up to date on what the question is? I have not kept track of what all you are discussing. :)I know Alan (Devantier) and of course I can ask Klippel folks what they are computing. I also have the documentation although it is very sparse.
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Can you bring me up to date on what the question is? I have not kept track of what all you are discussing. :)I know Alan (Devantier) and of course I can ask Klippel folks what they are computing. I also have the documentation although it is very sparse.

Thank you @amirm!

Pretty simply, we're trying to figure out the process of calculating the early reflections curve. It's not 100 percent clear from CTA-2034A or Toole's book.

CTA-2034-A says:

Snag_156c464d.png


There are two issues that are unclear to me and Dave Zan.
  1. Is the ER curve calculated as a simple average of every angle listed above weighted equally? Or is it an average of five separate averages? Originally I thought the former and the separation in the list above was cosmetic, but upon second thought and re-reading the devantier paper, it appears to be the average of five averages, which would weigh certain regions differently.
  2. Is the rear data included in the calculation just the 3 angles shown above (H180, H+90, and H-90), or is it 19 measurements (H±90, H±100, H±110...H180)? This part is confusing because the CTA-2034A document only lists 3 angles above, but it later goes on to define the individual "rear" curve as including the entire horizontal rear hemisphere.
The curves are not dramatically different any way I calculate them on the speakers I have full 360 data for - just off by a dB or two - but I would like to adhere to the standard as much as possible for the sake of consistency. Moreover, omnidirectional and bipole speakers might benefit from more rear reflection data. Vertical reflections also are weighed a bit more strongly depending on the process used.

Furthermore, since I've been using VituixCAD to generate spins... the app appears to calculate the ER curve as a simple average of the angles above, as when I compute the different options manually, that's the one that tracks most closely. Would like to contact the developer if a change needs to happen.

Edit: Confirmed that VirtuixCAD uses the simple average of all the angles listed above.
 
Last edited:

DS23MAN

Active Member
Joined
Sep 18, 2019
Messages
161
Likes
190
Location
Hypex doorstep
Some food for thought:

The spinorama is based on generated reflections in a room and are based on anechoic measurements. But a room is not all about reflections. Below the Schroeder frequency the standing waves take over:

https://en.wikipedia.org/wiki/Room_acoustics

How predictive are the estimated in room respons below the Schroeder frequency and in the transition zone?
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,240
Likes
11,462
Location
Land O’ Lakes, FL
Some food for thought:

The spinorama is based on generated reflections in a room and are based on anechoic measurements. But a room is not all about reflections. Below the Schroeder frequency the standing waves take over:

https://en.wikipedia.org/wiki/Room_acoustics

How predictive are the estimated in room respons below the Schroeder frequency and in the transition zone?
Time stamped:
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Thank you @amirm!

Pretty simply, we're trying to figure out the process of calculating the early reflections curve. It's not 100 percent clear from CTA-2034A or Toole's book.

CTA-2034-A says:

View attachment 47164

There are two issues that are unclear to me and Dave Zan.
  1. Is the ER curve calculated as a simple average of every angle listed above weighted equally? Or is it an average of five separate averages? Originally I thought the former and the separation in the list above was cosmetic, but upon second thought and re-reading the devantier paper, it appears to be the average of five averages, which would weigh certain regions differently.
  2. Is the rear data included in the calculation just the 3 angles shown above (H180, H+90, and H-90), or is it 19 measurements (H±90, H±100, H±110...H180)? This part is confusing because the CTA-2034A document only lists 3 angles above, but it later goes on to define the individual "rear" curve as including the entire horizontal rear hemisphere.
The curves are not dramatically different any way I calculate them on the speakers I have full 360 data for - just off by a dB or two - but I would like to adhere to the standard as much as possible for the sake of consistency. Moreover, omnidirectional and bipole speakers might benefit from more rear reflection data. Vertical reflections also are weighed a bit more strongly depending on the process used.

Furthermore, since I've been using VituixCAD to generate spins... the app appears to calculate the ER curve as a simple average of the angles above, as when I compute the different options manually, that's the one that tracks most closely. Would like to contact the developer if a change needs to happen.

Edit: Confirmed that VirtuixCAD uses the simple average of all the angles listed above.

@twelti , Since we have you here, would you be able to skim through the above and provide some clarity?

It's been bugging me for a couple of days =] Thank you in advance!!
 

napilopez

Major Contributor
Forum Donor
Joined
Oct 17, 2018
Messages
2,109
Likes
8,420
Location
NYC
Going to bump the above incase @twelti didn't see it :)

Another thing that I'm not clear about CTA-2034-A that is unclear in the text: Aspect ratios. Someone help me if I'm being a dolt and am missing something.

The standard defines the aspect ratio of the spinorama as such:

"Data shall be reported in graphical form as illustrated in Figure 4. This chart follows the guidance of IEC 60263, which stipulates that the horizontal length for a 10:1 frequency ratio be equal to 50 dB on the vertical scale. For the reported data to be useful it is important that the X-Y proportions and the horizontal and vertical scaling of the graphs be standardized in this manner. "

So a 10:1 frequency ratio = 50dB. Clear enough. But here is the chart they refer to (and all the other charts in the document use the same ratio)

Snag_36b22679.png


IEC60263 has ratios of 10 db/decade, 25 db/decade, and 50/dB per decade. The image above looks to me a 10:1 frequency ratio is equal to 25dB, not 50 dB.

For example (using Amir's Revel C52 measurement because I'm not on my usual REW PC). Here's an image exported with REW's 25 db/decade setting:

25db.jpg


It's a perfect match. Compare that to the 50 dB/decade setting:

50db.jpg


Which doesn't look at all like the images in the document, and obviously flattens and flatters curves too much, even if the above is already a good starting point.

Seems like the document does need some proofing... Either it should be 100:1 = 50dB or 10:1 = 25dB. Right? Am I crazy?
 
Last edited:
Top Bottom