A multi-way speaker design around the EXAR 400 horn

[Vineethkumar01]

On with more things done in the real world...

Prototype measurements

Here is the ugly cabinet made for the MEH using XPS foam and hot glue (I know that it will not win any prizes for looks or functionality.. That thing causes its own sweet resonances )

Measurements were made only out to 90 degrees. Measuremets were gated at 4ms..
Hence the masking of high Q resonances in the passband are also probably there..

Anyway, here is how the highs on Rosso 65CDN-T look like (Normalized 20 degrees off axis)

Now the mids (2xTC9FD in closed box of 10+ litres volume)

And here is a rough crossover that stitches together the two sets of drivers (Off course it can be done infinitely better )

With all the defects in construction, it looks surprisingly good I think..

 

[Vineethkumar01]

I am very encouraged by the results already..
We have a point source (c-t-c spacing of drivers less than 1/4th wavelength at crossover) with very high dynamics capability and nice pattern control (with the MEH alone) down to 500ish Hz..
Add the dual 15-inch woofer module, and we get directivity control down to less than 100 Hz and literally explosive dynamics in the full audio bandwidth from a point source.
Like this..

Time to get serious I think..

Here are the actual drivers I will be using per box on the final MEH
Shiny new things from TLHP are here..

 

[Vineethkumar01]

The edge termination business of the waveguide and connecting it to a rectangular box has been on my mind..
Till now, I have been using circular arcs to join the waveguide to the box that holds it. Inspired by comments from members over at diyaudio, I thought, why not use a super elliptic curve defined over one quadrant as the piece to connect the waveguide to the box instead of a circular arc.
So I took the parameters of the super ellipse (except parameter L) from the ATH profile definition (of this OSSE waveguide) and created a smaller super ellipse of L=44mm to serve as the edge termination.. This is how it looks
The waveguide alone is the yellow surface. The edge termination solid body has been extruded from the smaller (approximation using splines) super ellipse I created..

I haven't simulated the effect of this over the cirrcular arc in ABEC sims though.. I am feeling lazy to check that part..

Here is the waveguide Joined to the box

 

[Vineethkumar01]

Further studies and designs iterated overs hours and days..
My PC tirelessly did the number crunching over days and nights, and I finally found a good-looking waveguide with slightly smaller dimensions.

The main region why I liked this waveguide is because of the relative absence of wide-narrow-wide trends in the horizontal polar pattern and the gently sloping up sound power DI and ERDI..
Vertical polars were clean anyway..

Here are the details of the waveguide.

Device width x height = 346.46 x 287.46 mm (13.640 x 11.317")
Device length = 115.00 mm (4.528")

40mm roundovers on all sides 25cm depth to the box..

 

[Vineethkumar01]

WIth the above device, I may be able to get a system like this in the best case

 

[Vineethkumar01]

Then I tried to make the waveguide slightly larger because we need a certain specific directivity pattern through the upper mids and treble region gently transitioning to meet a cardioidish radiating twin 15inch woofer system.

Here is the tiny bit larger and deeper waveguide

Device width x height = 357.62 x 296.53 mm (14.080 x 11.674")
Device length = 118.00 mm (4.646")

Predicted polars

Assuming the blips around 2.5 kHz are numerical issues... I really liked this waveguide..

 

[Vineethkumar01]

Now we try optimizing the entire structure further to try and get every last bit of performance available..

As a first step, I tried the above waveguide in an elliptical enclosure with edge rounding formed using below shown cubic bezier curve.
Cubic bezier curves are super awesome in that they can literally form all kinds of usable curves with simple parameter adjustments. The curve formed below was trying to minimize the curvature of the transition from the waveguide edge to the back enclosure..

Here is the curve..



And the overall structure for simulation focussing on horizontal polars

Vertical polars are excluded from below directivity measurement due to lack of rounding on top and bottom edges in the sim

The dotted curves below are the DIs with the above enclosure with a cuboidal cabinet (shown in the above posts).
Solid curves overlayed on it are the ones from the elliptical back enclosure

I think I have come to the conclusion for this build.
I will make the MEH using the waveguide and cuboidal enclosure in the previous post.
I am not getting enough returns (in terms of simulation results) with complex enclosures and complex edge termination profiles with this particular waveguide..
Maybe it is time to build one and hear how it all comes together in terms of sound..
I will try to get the waveguide 3D printed first..

 

[Vineethkumar01]

And for that I did the cad design for this MEH waveguide..
Here is a view from the bottom

Here is one from the top

Looks like some alien space ship (or a kitchen sink ) I think..
Honestly speaking, all those "holes"/mid ports near the throat are not very confidence inspiring..
So I will print one more prototype and measure with the drivers I have before I move to final printing of a pair of these waveguides..

 

[Vineethkumar01]

I got the Bigger MEH waveguide printed, attached it to a cardboard box and took some quick and dirty measurements

Here is an on axis response plot of the DE 360 on the waveguide in 3 cases
1) All mid ports open (red, lot of wiggles)
2) Side mid ports closed (green, much less wiggles)
3) All mid ports closed (blue, slightly better than above but not by much)

I ended up using only the two drivers on top and bottom connected in parallel (and not the two on sides) because of response anomalies due to the holes in the sidewalls (mid ports) near the throat

Pic taken before closing side wall ports

DE 360 polars out to 90 degrees (normalization done 20 degrees off axis)

4NDF mid polars out to 90 degrees (normalization done 20 degrees off axis)

A first cut crossover

Impedance plots of the mids in parallel and DE 360, all drivers attached to the horn

Even with the issues due to resonances and other anomalies due to the flimsy cabinet construction, this look like good results I think
It feels like this MEH idea is worth pursuing..

 

[Vineethkumar01]

REW autoEQ'd response of the DE 360 CD on horn and the burst decay

Here is a comparison of the normalised sound power DI between the BEM sims I made and the measured results (only using 0 to 90 degree measurements).

Dotted line below is the SP DI from SIM
solid line is from actual measurements..

Looks really close I think..

 

[Vineethkumar01]

Attempt at a passive crossover just for fun..

 

[Vineethkumar01]

With the above passive crossover, the crossover frequency is somewhere around 1150Hz

For such low crossover, maybe a more robust compression driver is needed I guess.
Below is the distortion profile of the DE 360 I measured on this horn (Probably not very accurate but shows the relative trends)

Anyway, I am happy with this active crossover, where the crossover frequency looks is 1350Hz

In that region the THD has climbed to only 0.2% and the driver is rolling off at 36dB per octave due to the effective high pass roll off

I think I probably need to look into phase related aspects also with the crossover, given that it is an MEH..

 

[Vineethkumar01]

Just applying response inversion tricks for full range FIR filtering and 8th order Linkwitz-Riley filters at 1350Hz..

Now I am done..
Based on the lessons learnt so far, I should probably plan to get the final printing of a pair of these waveguides with two drivers per waveguide done..