Reliable way of measuring tweeters in speaker?

[OllBoll]

Hello there,

I've just measured a bunch of tweeters in a dipole ring array both inside a normal room and outside under semi anechoic conditions and the data I got made me wonder...

I get why anechoic conditions would be a good way to test single drivers but for this speaker the data was almost worhless. I have more than a single tweeter ( 8 per speaker ) and so all I got was a complex interferance pattern with nulls and peaks everywhere so it was almost impossible to read anything from the data.

Inroom they measure very differently, no nulls or peaks at all but a nice and straight response but how do I know if the frequency response was affected by the room or not?

So what I wish to know to enlighten myself Is there a better and more reliable way to measure tweeter responses when you have more than one? Should one use massive smoothing or is this where the microphone arrays come in?

[6.283]

Hi OllBoll,

I am kicking around the idea of a circular dipole array since a while myself now because it theoretically could overcome some typical issues of double-dome-dipole tweeters. However, I was always afraid of the complex interference pattern it could create and endless hours of wood work. But on the other hand it might be possible with the right drivers, correct spacing, correct panel size and appropriate filters.
If it works, this animal should measure fine outdoors. How far was your mic placed away from the array and did you apply any gating ?
Maybe you can post some measurements and pictures and describe what drivers and filters are used.
Would be cool if we could make this thing work !

[OllBoll]

The current test iteration looks like this:



The drivers are Obi T22-oa, a driver designed for carlsson speakers and modern clones of them. What made me find it interesting is that it's a cone tweeter, so it has a much more consistent directivity than normal dones. It's more like a waveguided dome in that aspect but without the large waveguide.

Here is a measurement of the T22 I've shamelessly stolen from another forum



What I'm currently testing is wherever the baffle design works, the idea is to mount two midrange drivers back to back in the middle. What kind of midrange I have not yet decided, currently leaning on 2 6ND430 but the question is if those magnets would let me mount them back to back.

And back to the measurements:

I tested briefly inside to instead of doing a single measurement to do several while moving the microphone slightly, a few cm and then averaging those. I then got a fairly flat frequency curve, now I just need to test if it works as well in anechoic conditions.

If I'm not mistaken that should counter the combing effects and instead let me measure the average sound energy, which is what I want to know.

[CharlieLaub]

You have just discovered why you DON'T want to use a circular array of drivers at high frequencies. I don't think that there is anything wrong with your measurements... just your expectations of this particular physical arrangement of drivers.

Have you thought of playing around with a diffraction simulation program like "The Edge"? It can simulate the response of lots of transducers in whatever physical arrangement you desire, including baffle step diffraction and open baffle responses. Here's the link to it:
Tolvan Data

In general, one reason that "tweeters" are physically small is that the sound wave coming from different parts of the radiator surface can interfere with each other. As the diameter increases or you add multiple tweeters, these issues are only magnified and extend to lower frequencies. Certain properties of linear arrays of drivers can make then useful, but circular arrays are not generally helpful.

-Charlie

[OllBoll]

Yes, but the ears and the brain is not a measurement microphone. I have had earlier speakers where they should comb like mad, but to my ears ( and lots others ) they don't. Even this array doesn't comb in a for me hearable way. What I want is to measure total energy and ignore combing effects, to see if there is any unevenness in the response common over the whole listening area.

[tvrgeek]

You have also found out the room matters. DIY has the advantage of measuring where they will be used. OEM's do not. Tough job that.

You may also find if you look at OEM plots in a chamber, it is for IB, so no diffraction lobes.

I would save the money from 8 tweeters and buy one good one. Just my way of thinking.

[6.283]

Hmmmm, what I have in mind looks completely different. How do you achieve dipole behavior ?

Originally Posted by OllBoll

Yes, but the ears and the brain is not a measurement microphone. I have had earlier speakers where they should comb like mad, but to my ears ( and lots others ) they don't.

Exactly right. Even Toole describes combing as a measurement artifact.

Originally Posted by OllBoll

What I want is to measure total energy and ignore combing effects, to see if there is any unevenness in the response common over the whole listening area.

Maybe Arta could help with the feature of two measuring windows. The latter will then capture the room response which will most likely even out the response of the direct sound (to some extent anayway).
Have you been using that ?

[OllBoll]

Originally Posted by 6.283

Hmmmm, what I have in mind looks completely different. How do you achieve dipole behavior ?

Exactly right. Even Toole describes combing as a measurement artifact.

Maybe Arta could help with the feature of two measuring windows. The latter will then capture the room response which will most likely even out the response of the direct sound (to some extent anayway).
Have you been using that ?

Dipole behaviour is because each half plays out of phase, it's basically like a normal dipole but instead of a single tweeter on each side there are four to improve high frequency off-axis response.

Currently I tested with REW and averaging responses, which worked kinda well but would be nice if I could specify measurement window to ensure that I get the true anechoic response.

[CharlieLaub]

Originally Posted by 6.283

Exactly right. Even Toole describes combing as a measurement artifact.

Well I believe that depends...

At high frequencies (e.g. above 3k Hz) the comb filtering is changing over distances that are smaller than the separation of your ears. The signals will incoherently add IIRC and you only perceive the average soundfield. There have been lots of speaker designs that used this kind of principle, but it seems mostly to have fallen out of favor in the last 20 years...

At lower frequencies you can both measure and hear the nulls quite easily.

-Charlie

[tvrgeek]

"but it seems mostly to have fallen out of favor in the last 20 years..."

Probably because a range of very low distortion wide beam dome tweeters have been developed. It is no longer an advantage.