Half-cardioid active high-output 3 ways

[Fledermaus]

Hi there,

Just wanted to share, for the fun of it, my current attempt at a relatively healthy-output, decent-sensitivity, proper-bass extension, 3-way active DIY speaker with resistive enclosure cardioid midrange.
This is a yet unachieved project but I'd consider it as settled as for drivers, bass cabinet and treble horn and enclosure - only the midrange cardioid enclosure still needs good bit of work (see below).
For convenience of design and handling, I made it 2-boxes (prototype at this stage) :

After a few trials and errors, I retained the Nero-15SW800 woofer from SB Audience, B&C 8MBX51 for midrange, and BMS 4550 plus AudioHorn X-25 horn for the highs.
X-Over and EQ are done with a Xilica Solaro QR1 DSP.

This is more or less where I'm at with the BR-loaded woofer (near-field measurements of port and speaker, summed) :

As for the midrange, I had a hard time learning the basics of Akabak and simulating the following (polar map @ 300, 600 and 1200 Hz) :

Simulated heatmap :

Where the fun really begins is with the measured heatmap (mind the +/- 90° scale) :

With near-perfect 90° coverage at -6dB, the horn works as advertised.
Midrange seems to be contained within 150°, with kinks, and a sharp step due to 48dB/oct crossover (switching to 24 dB should soften it a bit) and directivity discrepancy between drivers.

The fun goes on :

Midrange doesn't look so pretty... you ain't seen it all :

There are a couple of resonances in there, but even more cancellations... Why so ?? After brainstorming with competent people, it looks like some kind of comb filtering is taking place between the driver membrane's back wave and what emanates from the cooling ports in the magnet, see :

This makes for multiple sources not in-phase, hence not so nice cancelling effect.
This is what I'm working on now, with the idea of isolating/absorbing those magnet holes' emissions, without hampering VC cooling nor speaker working according to specs, by reducing box depth and adding absorption in the back.

Thanks for reading, hopefully I'll be able to report on some progress soon !

 

[Halbknapp]

Following as I plan on doing something very similar in the near future (but with more hifiish drivers as mid and tweeter)

 

[nc535]

Nice response except for the combing, just inside the 90 degree half beamwidth mark

You don't need the grief of dealing with that combing problem. Try a different driver to see if the problem is unique to the one you started with. e.g. FaitalPro 8PR200 looks like it fits just inside your baffle cutout. To my knowledge, no one else has reported a similar problem, so it could indeed be unique to that driver.

I've been working a similar design in ABEC simulation only so far, using an 8" coax instead of separate midwoofer and waveguide. I've simulated varying thicknesses of absorber starting just behind the side slots. With no absorber, I see a broad peak at the low end of the passband, ostensibly since reflections coming back out through either the slots or the cone will be in phase. If I simulate enough absorber to absorb that backwave, I get a nice flat response with smooth gradual roll off. Leaving the absorber just behind the slot out doesn't result in anything like the combing you have but then reflections from the driver frame aren't modelled.

I don't think you can fix that BMS's combing with absorption. The reflections you are dealing with are likely to be between the frame and the cone or between the cone and the surround, where there isn't space for enough of it. For cardioid as for OB, we want a driver with the most open frame that interferes least with the backwave. The BMS surely seems to fit that bill, so the source of that problem will remain a mystery to me until I see the cure. The Faital's frame is also quite open. I haven't used it but I have simulated with it.

 

[bmc0]

There are a couple of resonances in there, but even more cancellations... Why so ??

Reflections from somewhere. Are you measuring in a room?
Sharp notches like that can only be from sufficiently delayed sound. The notches are not present in the nearfield measurements, so must be from external reflections. They get worse as you go further off-axis because of the directivity—direct sound gets quieter, but reflections stay about the same.

Do ground plane or free field measurements outdoors with an appropriate window/gate (should cut just before the first reflection), not a frequency-dependent window.

 

[Fledermaus]

Thank you guys for the tips and suggestions
I didn't observe this effect in Akabak simulations either, because I wasn't smart enough to include an out-of-phase emissive domain at the back of the driver magnet, but someone much more knowledgeable than me did an in-depth FEA sim and got the same main cancellation near 500 Hz, among others (data not available to this day for communication, sorry).
OTOH, I already measured speakers in the same setting (about centre of a large workshop room, mic at mid-height between floor and ceiling, measurements gated) with no signs of such rearward cancellations.
I might try other gating options but I'm quite confident, following past experience and FEA results, that the hypothesis above holds water.
Anyway, proof of the pudding being what it is, I'll show new results ASAP with back driver loading tuned in accordance with the model.