Thank you, this is really valuable. What I noticed now, what was not obvious before, are the little spikes at roughly 1.7 - 2.7 - 3.9 kHz. These seem to be common to both cases, which suggests it's actually the waveguide (or something else, common to both). They should disappear even from the spectrograms if EQed out, though. In fact, I can imagine this could be ringing of the horn. Maybe this could be confirmed somehow.
Regarding distortion, larger diaphragms should always have an edge towards the lower frequencies. That's to be expected.
Regarding distortion, larger diaphragms should always have an edge towards the lower frequencies. That's to be expected.
Last edited:
No, I really don't. I'm still only planning to do these kinds of measurements. I really value your data immensely.
And there is no reason not to use both, some passive parts to manipulate impedance and affect distortion performance, and independently optimize frequency response (and even phase) with DSP. I find it really archaic to limit oneself to some certain technology as everything seems to have pros and cons it's then logical to mix and match to get top performance.
This is wild 
https://www.researchgate.net/publication/375090339_Iterative_metric-based_waveguide_optimisation
Interesting times.
The only pity might be that this all falls apart when you connect a real compression driver...
https://www.researchgate.net/publication/375090339_Iterative_metric-based_waveguide_optimisation
Interesting times.
The only pity might be that this all falls apart when you connect a real compression driver...
Last edited:
Does anyone know what mesh frequency to use?
I've been using 1000Hz for years, and noticed that many of you are using a much higher frequency. In increased things to 20khz, and the sim never finished. I've backed it down to 16khz and so far, it's taken nearly 24 hours for the sim to reach 67%
I have an eighteen core CPU and 64GB of ram and it's still taking over a day to finish a sim...
There's a 3+ year old post with some guidelines here: https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-6234490
That post is a bit confusing to me, as it recommends a mesh frequency of 1khz, but I see people using frequencies that are 20khz and even higher?
I've been using 1000Hz for years, and noticed that many of you are using a much higher frequency. In increased things to 20khz, and the sim never finished. I've backed it down to 16khz and so far, it's taken nearly 24 hours for the sim to reach 67%
I have an eighteen core CPU and 64GB of ram and it's still taking over a day to finish a sim...
There's a 3+ year old post with some guidelines here: https://www.diyaudio.com/community/...-design-the-easy-way-ath4.338806/post-6234490
That post is a bit confusing to me, as it recommends a mesh frequency of 1khz, but I see people using frequencies that are 20khz and even higher?
It depends on a model used. With a full 3D mesh you're typically limited to several thousand elements in ABEC to work smoothly (I always try not to exceed 5000). But this is best controlled by the mesh resolution parameters in the Ath scripts, not mesh frequency (which is kept low, e.g. 1 kHz, so that ABEC doesn't subdivide the mesh further itself). In CircSym mode (axial symmetry) however, the usable limit for mesh frequency is much higher - several tens of kHz are no problem. And even this is possible to "overwrite" by a large number of segments in the script (always the best approach), so the mesh frequency itself has no actual relevance.
It could be a nice approach for doing multi driver hornsThis is wild
https://www.researchgate.net/publication/375090339_Iterative_metric-based_waveguide_optimisation
Interesting times.
The only pity might be that this all falls apart when you connect a real compression driver...
I've been playing around with a dual driver horn but didn't obtain anything spectacular yet...
Both the bends required to fit two drivers and the combining part are tricky.
I don't see those results as bad. It's a wiggle of a couple of decibels.
And you can probably clean it up with some clever positioning and hole shape.
I don't know Somehow I still don't quite like the idea of making holes to waveguides, but there are two very different questions related: Is it the best what we can do? I still have my doubts. But is it good enough to bring a lot of pleasure listening music? Absolutely, I have no doubts about it, if you like it, go for it.
Somehow I still don't quite like the idea of making holes to waveguides, but there are two very different questions related: Is it the best what we can do? I still have my doubts. But is it good enough to bring a lot of pleasure listening music? Absolutely, I have no doubts about it, if you like it, go for it.I agree, it might be maskable etc, and maybe with careful hole placement, it could be minimised. but its only going to 'smear' the result.
to me, it probably makes more sense to try integrating the driver into the waveguide, a-la genelec or this martin speaker, this one is full range. but if you limit the low end like the genelec, you should be able to the the excursion below 1mm, allowing you to have a smooth interface with the horn. I'll put up some tests with that in soon. but using the axial mode, it seems you can displace the surfaces by that much and not do too much to the response at all.
to me, it probably makes more sense to try integrating the driver into the waveguide, a-la genelec or this martin speaker, this one is full range. but if you limit the low end like the genelec, you should be able to the the excursion below 1mm, allowing you to have a smooth interface with the horn. I'll put up some tests with that in soon. but using the axial mode, it seems you can displace the surfaces by that much and not do too much to the response at all.
It weights 1135 g, 550 mm mouth diameter - I made this one very light.
I haven't tested the center of gravity yet but I guess it will be somewhere near the mounting point with a driver attached.
(The gap between the base and the petals is intentional. One of many lessons learned.)
Last edited:
Marcel: are you going to offer the Exar550 files via cults3D? It is now sooo long I follow this thread and also made various sims on my own...but never built something 😱
As I have hf1440 here as well as dcx464...that would be a nice project with little time effort as its ready to.print/use!
As I have hf1440 here as well as dcx464...that would be a nice project with little time effort as its ready to.print/use!