1" throat - a long version. Still doesn't beam
As for the subdomains boundaries, notice that the back side element of the waveguide can even cross the channels and nothing happens. I think it's also shown somewhere in the ABEC example projects. That element is simply a part of different domain (mathemtically, not physically).
As for the subdomains boundaries, notice that the back side element of the waveguide can even cross the channels and nothing happens. I think it's also shown somewhere in the ABEC example projects. That element is simply a part of different domain (mathemtically, not physically).
Attachments
It's not yet very pretty but the trend is indisputableHolygrailing
BTW, this is the long 1" throat including the driver LE model, the same as for all the previous simulations ("generic25") -
Attachments
Progress is good. It's interesting that the real part of Z_ar (radiation impedance) doesn't reflect the loudest parts of the on-axis pressure response. You're using three separately driven elements here now though, right? Which element is plotted?
Perhaps it's worth updating the observations to plot the self radiation impedance for all three driven elements, and also the mutual impedance between each of them? That may help optimise the transitions.
Alternatively, the vanes could extend close to a new, initial subdomain containing a single driven element and interfaces could be added at the entry points to each channel. I think it's more interesting to observe the individual impedance of each channel at this point though.
I know this was the last page, but I redid the rigid termination plane wave tube sims as an axisymmetrical model in both platforms. The driven element is on the left in both cases, still just 1.32 kHz.
Akabak:
COMSOL:
Curiously, the pattern of cancellations is very similar but inverted. In Akabak, the 'leading' channel is nearest to the axial plane of rotation, but in COMSOL it's on the outer edge.
I'm not sure that means the data in either is 'untrustworthy' but given the simplicity of the model in both cases, it does make me curious about putting a lot of faith on BEM (at least as implemented in ABEC and Akabak) for internal acoustics problems.
The good news is that shouldn't make much if any difference for external acoustic problems, which covers the majority of what people are interested in - output level, polar responses and the like. But perhaps we need to take a pinch of salt when looking at the internal 'weirdness' inside these channels
As for the comparison sim of the DXF you kindly shared, I think I need to tweak it a little to make it import more neatly before I can report much of anything. The model template I was using as a starting point was based on normal acceleration at the driven element rather than velocity, so that's another thing to look into. I'd like to be sure we're comparing apples to apples, and it's good to have to set things up more from scratch so I learn.
Perhaps it's worth updating the observations to plot the self radiation impedance for all three driven elements, and also the mutual impedance between each of them? That may help optimise the transitions.
Alternatively, the vanes could extend close to a new, initial subdomain containing a single driven element and interfaces could be added at the entry points to each channel. I think it's more interesting to observe the individual impedance of each channel at this point though.
I know this was the last page, but I redid the rigid termination plane wave tube sims as an axisymmetrical model in both platforms. The driven element is on the left in both cases, still just 1.32 kHz.
Akabak:
COMSOL:
Curiously, the pattern of cancellations is very similar but inverted. In Akabak, the 'leading' channel is nearest to the axial plane of rotation, but in COMSOL it's on the outer edge.
I'm not sure that means the data in either is 'untrustworthy' but given the simplicity of the model in both cases, it does make me curious about putting a lot of faith on BEM (at least as implemented in ABEC and Akabak) for internal acoustics problems.
The good news is that shouldn't make much if any difference for external acoustic problems, which covers the majority of what people are interested in - output level, polar responses and the like. But perhaps we need to take a pinch of salt when looking at the internal 'weirdness' inside these channels
As for the comparison sim of the DXF you kindly shared, I think I need to tweak it a little to make it import more neatly before I can report much of anything. The model template I was using as a starting point was based on normal acceleration at the driven element rather than velocity, so that's another thing to look into. I'd like to be sure we're comparing apples to apples, and it's good to have to set things up more from scratch so I learn.
It's the whole group, as if it was a single driving element. It's possible to define the driving this way.You're using three separately driven elements here now though, right? Which element is plotted?
Below is this simulated, for three different gaps between the driving element and the vanes (pictured mesh is for 1 mm). I'd say it's pretty consistent (see #8481).Alternatively, the vanes could extend close to a new, initial subdomain containing a single driven element and interfaces could be added at the entry points to each channel.
Attachments
Last edited:
BTW, this is the "singularity" at 1.2 kHz - Phase and LeveldB. Is it possible that it's something real? Could it help to explain what's happening there? It seems strange to me.
It seems to me that the Q of this peak is too high for this to be a real physical process. Looks like some kind of stability / convergence issue of BEM (small gap in geometry, The Non-Uniqueness Problem ?!).
This is the throat impedance for three different "meander amplitudes" of the channels (see the picture), 1.0 being the full amplitude required for the same path lengths.
(Around 1.2k and 2.7k there are just some numerical problems, IMO).
(Around 1.2k and 2.7k there are just some numerical problems, IMO).
Attachments
4 channels, 1.4" driver. Does anyone have an explanation why is the impedance as it is? It's obviously a propetry of the bended channels, it's a pattern that will probably stay there.
The local minimum on the impedance curve seems to correspond somehow to the length of the channels (wavelength of 1800 Hz ~ 2x channel length).
The local minimum on the impedance curve seems to correspond somehow to the length of the channels (wavelength of 1800 Hz ~ 2x channel length).
Attachments
Last edited:
I don't have an explanation but I see something very similar in a simulation I have been running trying to model a Yuichi A-290, when I take the fins out. A peak at a lower frequency which trends back up.Does anyone have an explanation why is the impedance as it is?
Attachments
- Home
- Loudspeakers
- Multi-Way
- Acoustic Horn Design – The Easy Way (Ath4)