Thanks! Yeah taps on the peaks gave a considerable improvement in the polars, so that's definitely what I'm sticking with.
Still have another measurement session where I'm going to dial in the DSP presets some more, but here's the on axis FR with the latest prototype + FIR full range preset. As you can see there's still some high q choppiness in the upper midrange from diffraction off the taps, but I think its within an acceptable tolerance and I don't think its audible enough to be of concern. The peak at 15k kind of just naturally happens with this particular compression driver, so not sure if I want to tame that down a bit or leave it as is.
My friend with a 3D printer has suggested we should instead cast them with epoxy granite. That will be fantastic if we are able to do it.
I think it's possible if you're okay with a 1-time (or very few) use mold. I explored urethane vacuum molding for these but ultimately it made more sense to resin print them since the mold life wouldn't have been great with the ports and tricky draft angles.
It' still early in the development cycle, but I've got a first prototype cut out on the CNC so that should be going together shortly. It's a considerable step up from the Solana in terms of output, and is more suited to be used as mains (think audience coverage of 200-450 people). Coverage pattern is 90x60 rather than the 90x90 that the Solana features. I've had the v1 deployed at various events over the last year and have gotten a really good feel for how it performs and what its strengths/weak spots are. Additionally, there are probably a couple of dozen that have been built worldwide at this point, so the feedback from people DIYing them has really been a driver in integrating the improvements into v2.
There's a laundry list of small changes with v2, but really I can summarize it with these 3 goals:
- Optimize assembly and construction methods to ensure greater consistency among builders. Reduce number of bevel cuts and redesign throat adapter to minimize tolerance stacking errors in trouble areas
- Provide more thorough documentation, measurements, and assembly instructions
- Provide a fully parametric Fusion project to allow for customization
#3 is probably the most important. I've been an avid user of @bwaslo's synergy calc spreadsheet ever since I began experimenting with MEHs. I've had the thought for a while that it would be cool to make a fully CAD integrated version of it where you input the desired coverage angles and cutoff frequency, and the full 3d model is generated automatically. Assuming I can get the parametric model working, you'll be able to configure your desired MEH and use the outputs from fusion to generate a throat adapter in Ath4 using a pre-configured script, obtain hornresp inputs, and even import mesh files into AKABAK for BEM validation. Very rough workflow of that below, but essentially transforms the JMOD design into a modular platform with the goal of making MEH designs more accessible and open.
The base model will still retain roughly the same dimensions and 90x60 coverage pattern though. Some previews of that below (not final, some changes still need to be made)
I’m curious to hear more about the throat adapter. What benefits does it offer compared to the throat in the previous design, and how did you arrive at its shape?
Definitely agree it looks cool!
What would the pros/cons be of these for the domestic setting?
That's a bit of a long answer, but I'll try to break it down into 2 main aspects:
Mechanical
The horn flares in v1 are assembled using rebates in the top/bottom panels and a few notches. These panels are not easy to cut and involve making compound miter cuts with a table saw. The tolerance for error here is slim and making a cut .5mm too short here or .5mm there can compound and will cause things to not fit right.
The rear features a removable throat adapter plate (like in Scott's original version, and in most MEH designs I'm familiar with). There are two fasteners that secure the throat adapter in place, with enlarged through holes in the throat adapter that allow you to move the throat adapter and align it into place with the horn before securing it in place.
So as you can imagine, if the cuts on the horn panels aren't made precisely, it can cause the throat adapter to be misaligned and a gap or abrupt change in the transition from the throat adapter to horn. As to whether this matters acoustically or not - maybe? At the very least its probably not going to help things and if there's a reasonable way to address it, then I think its worth pursuing.
So, pivoting to the new & improved throat adapter, its a 3d printed piece that indexes into the horn flares during assembly and is permanently glued into place. The transition gap can be filled and sanded before painting for a (nearly) perfect smooth transition to the conical section. The benefit of this part being 3d printed is also that the complexity of the miter cuts can be transferred to the 3d printer to take care of, meaning the inner horn flares only involve basic 90 degree square cuts that the CNC can easily do with a high level of precision.
Material properties (durability, heat resistance, UV protection) are of course important since this is a PA design. I do think that additive manufacturing has made leaps and bounds of advancements in the last few years with more DIY accessible printers and high performance materials that are easy to print even on budget machines. I've made sure the throat adapter can be printed without supports to make it a little easier. Below is a test print done by @dsl1 in PLA with no overhangs.
PLA wouldn't be great beyond prototyping, but I've got several filaments on the way to do some testing (PC, PETG, PETG-CF, PC-blend)
Acoustic
The acoustic benefits of the new throat adapter are still to be verified, since I don't have measurements yet. So the below is me working off of some assumptions based on data from prior designs, and cobbled together models. The gist of it is that I wanted to design a throat adapter that was informed by data from simulations on the frontend and minimize the time spent doing trial and error physical prototyping.
Starting with the v1 throat adapter, its a basic round to square loft profile. It doesn't particularly match the exit angle of the DCX464 well, and is a somewhat abrupt transition. What I wanted to do with the new design is test a smooth transition profile. At first I tried to make a quadratic throat adapter profile work, but the geometry just doesn't work out in a 90x60 horn like this without extending the throat adapter considerably into the horn.
So instead what I did was create an akabak project and imported the conical horn flares as well as the internal geometry of the DCX464 (huge thanks to @alex6679 & @Olombo for the scans). There is a 90mm gap between the exit of the DCX and the start of the horn section for the throat adapter. This section is occupied by a mesh that is imported from Ath4 using a script to morph to a rectangular exit. This mesh can quickly and easily be swapped out with various throat adapters by simply reloading the mesh file.
I then used a script to automatically generate various profiles by randomizing the Ath parameters, which then could be simulated, scored, and iterated using a genetic algorithm to arrive at a more optimal shape. Based on what I observed, the directivity from 5kHz upward is largely determined by the throat adapter, with very little difference in those frequencies between simulating just the throat adapter alone versus with the throat adapter + conical walls.
The other area to optimize on was how I could leverage the throat adapter geometry to minimize interactions of the compression driver wavefront with the woofer ports. I'm making a couple of assumptions here:
1. That setting up a spherical shell field indicating pressure, located at a distance of approximately where the taps are can help inform tap placement - lower pressure zones are more ideal and will have less harmful diffraction effects. (See below, red circles indicate lower pressure zones, 4-8kHz is shown)
2. That diffraction from the ports will typically result in off-axis lobes relating to the wavelength distance corresponding with the below 3 distances:
So for the iterative algorithm, obtaining lower pressure in the corners was a goal. You may notice that there are subtle ridges inside the throat adapter itself. Those provide a very marginal improvement in that metric versus the two control throat adapters I simulated against (basic loft, default OS-SE formula)
That's pretty much the sum of it. I'm very hesitant to say at this point in the project whether the changes will result in a definitive improvement without real world measurements, but I am very optimistic that they will.
I think they're extremely well suited for domestic settings. Listening to a pair in my room as I write this. The ample headroom and ability to drive them hard with minimal distortion makes them a lot of fun to listen to.
Deciding the coverage pattern was something I put a lot of thought into in the beginning, going back and forth between a few options. Symmetrical coverage was important as I don't like the whole pattern flip thing with horns. I know a lot of people run 50x50 or 60x60 MEHs in domestic rooms, but for me that can be maybe a little too focused? The strong phantom center with narrower pattern MEHs can be really nice (and even disarming), but personally I like to hear some room and soundstage in what I'm listening to without going overly wide with it.
So 90x90 seemed like a good balance that would be pretty happy in most rooms and fit my own preference. I personally really like the DI that you get with a lot of the Genelec/Nuemann monitors - fairly wide with a slight narrowing, but not to the point of beaming in the top 1.5 octaves. Makes for a nice wide sweet spot that sounds good anywhere in the room. Especially noticeable is how sitting or standing makes no difference in the sound. So that definitely influenced what my target directivity looked like with this design.