Understanding Directivity Error in the Measurements

[fineMen]

It has already been established that 3-ways allow you to get the best of all worlds.

Making a 3-way isn't extra development cost or time to market. Making a 3-way is simply more expensive in BOM costs and has a size trade off (you need cabinet volume for a midrange chamber and space on the baffle for the drivers). There is a (large) market for speakers which aren't large enough or expensive enough to be true 3-ways.

Which proves my point?

 

[fluid]

I'm not sure if "the red line" relates to these time values. or how these are taken into account by the Klippel algorithms.

It does not and it is not accounted for. There is a correlation to the early reflections curve being intended to be representative of the first bounce arrival spectrum, but that is a statistical average of the rooms in the Devantier research.

I can hazard a guess that Directiva has a narrower dispersion compared to the Revel but what on the graph depicts the wider dispersion? Sorry if the answer is somehwere above in the comments.

You could hazard that guess but it is not quite right. Look at the beamwidth graphs in the two reviews and you will see that the Directiva is +-60 degrees where the M106 is closer to +- 50, so the Directiva is wider. The Directiva has a Directivity mismatch of 2dB in the DI just as the M106 does.

On this point wide or narrow is a very imprecise way to describe directivity. For example Dennis's speakers are more like +-80 to 90 degrees at least above 2K on a normalized polar plot, horizontally.

It is also important to look at the difference between horizontal and vertical response to see what is happening and how the early reflections and power response have been balanced. Your other lobing thread had some information that crosses over to this from kimmo
https://www.audiosciencereview.com/forum/index.php?threads/some-help-with-lobing.22661/post-755427

Also, does a mismatch in the directivity cause non linear behaviour in the off axis response or different things?

It is a linear effect.

The technique for perfect (!!) dispersion is at hand, really. Don't be fooled by business men who want to sell You a cheap compromise.

"Perfect" is a strong word coaxial speakers do not come without compromise either.

 

[EEE272]

It is a linear effect.

Maybe the linear in this context is very different but I always get confused with linear and dB.
When people talk about an average of the measurements, does it mean you average the dB values, or do you convert to an absolute value, average and then go back?

Another question I have about directivity, is the following.
I never understood why orientation of the speaker is not taken into account. If you don't orient the speaker to the listener, your off-axis measurement becomes your on-axis sound. Should the score that is used in the ranking not be the result of an optimization with respect to speaker orientation?
I do understand why they did not do it in the original study because they listened to mono but for a stereo setup it might matter.

 

[fluid]

Maybe the linear in this context is very different but I always get confused with linear and dB.

dB's are logarithmic. Distortion can be both linear and non linear.
https://www.nti-audio.com/en/suppor...tion is defined,usually meant by "distortion".

When people talk about an average of the measurements, does it mean you average the dB values, or do you convert to an absolute value, average and then go back?

Averages can be mean or RMS, it is not 100% clear which is intended to be used in CTA2034 but the difference between them is very small.
The dB value is used for the average.

Another question I have about directivity, is the following.
I never understood why orientation of the speaker is not taken into account. If you don't orient the speaker to the listener, your off-axis measurement becomes your on-axis sound. Should the score that is used in the ranking not be the result of an optimization with respect to speaker orientation?
I do understand why they did not do it in the original study because they listened to mono but for a stereo setup it might matter.

Many speakers are designed with the straight ahead response as the flattest and the tweeter as the design axis. This is why it is the default. If the listening axis or design axis vertically are specified by the manufacturer these should be used when making a measurement.

The score should only be used to rank speakers into whole number values.

 

[ctrl]

Here I am just a little confused about how a dip in the off axis relates to a the bump in the directivity line?

As you get more involved with the "math", it will seem clearer.
In the beginning, as so often, it is easier to just accept the context.

Maybe a mnemonic will help you to get a rough idea when evaluating a CTA-2034 diagram - it's a little silly, but maybe it makes a very rough evaluation of the diagram easier for you.

To evaluate the CTA-2034 diagrams, I first look for anomalies. Very roughly and simplified one can say that humps generally influence the sound of a loudspeaker more negatively than dips (as others have said) and an "imaginary" mirror line runs through the diagram.

So we draw a "path" in the CTA-2034 diagram and let a cat with a hump*** and its mirror-image equally evil/bad twin wander through the diagram.

All frequency ranges that show a hump or a transition from dip to hump (or vice versa) should be viewed critically, as they can have a negative effect on the sound.
For all curves that represent a directivity index (SPDI, ERDI, hor ERDI, ver ERDI,...), the mirror-image equally evil/bad twin comes into play and we look for dips.

*** What kind of association you have with the "evil/bad" hump is of course irrelevant. It can be a dromedary that bit you as a child or the (sorry, not PC) "The Hunchback of Notre-Dame" (tragic, not "evil/bad", but you know what I mean....).

 

[fineMen]

The technique for perfect (!!) dispersion is at hand, really. Don't be fooled by business men who want to sell You a cheap compromise.

"Perfect" is a strong word coaxial speakers do not come without compromise either.

Perfect in that it wasn't sensible to ask for better. The engineer's finalizing "good enough" 'perfect'. Anyway, if e/g a Neumann speaker is reference (as a studio tool in the studio), then the same Neuman is pretty perfect by definition

I was tempted to contradict the notion that wide dispersion is always in conflict with controlled directivity.