Synergy horn - 3d printing entry?

[Flaesh]

X, right click\open in new window working

 

[xrk971]

Thanks, please post images stored on other servers using the "image" button, then it will appear in-line for all to see.

 

[Nissep]

That is what i did.
But dropboxlinks doesn't work apparently.
Good to know.

But the big differense is higher sensitivity in the horn with 3fe22.
Lower weight, 3fe22 have a bigger hump before i drops off in the low end.

 

[cookiemonster77]

I have started to tune the ABEC3 synergy horn model. To not repeat everything, there are some initial results from the first run here.

I had probably placed my mids too close to the compression driver - at least if the compression driver in ABEC3 (I still don´t know which one it is) is representative. It contains a 22mm waveguide transition prior to the exit to the horn.

That waveguide is part of the LEM model (so you don´t see it in the graphics I linked to), but the sound from the midranges travel through a RadImp element that represents the throat of the horn/exit of the compression driver, into that element, and bounces back, causing a reflection. Interestingly, I tried to shorten the compression driver throat to 1mm, and even turn it off. I thought this should effectively shorten the horn (as seen by the midranges), and move the reflection notch upwards in frequency. It turns out the notch just disappeared.

I didn´t solve the model to a very high frequency, so my hypothesis is that the notch I was wasn´t the reflection of the midranges, but rather a resonance from the LEM waveguide element. Or, maybe I´m mistaken. I could model this part in the BEM model too, to avoid the confusion - but the smaller the items the slower everything solves (and memory requirements grow very fast). ABEC really makes me want to get a faster computer.

Anyhow, I´m moving the midranges out a bit - to 70.5cm axially from the horn entry, and increasing the mesh frequency to see up to 4kHz. Hopefully that would both move the real reflection notch well into view. (7 cm is about 1/4 wavelength at 1200Hz, but I don´t think one can think in terms of waves travelling freely and bouncing back, and I suppose there are pipe resonances too).

it will take a while to solve.. 🙂

In the mean time - could anybody give me an idea of what the ideal axial distance from the throat would be given a 60x60 horn? I am thinking of having the compression driver high-passed at 1300Hz or above.

 

[nc535]

Synergy patent says the circumference of the horn at the mid entry point must be less than a wavelength at the crossover frequency. You can get circum as 4 * sqrt (s2) and you can get s2 from Synergy Calc V5.xls. That will give you the maximum Fxo for a given entry point. Now add the CD acoustic path length and the CD mounting plate to the axial distance to S2 and compute the reflection null frequency based on the sum being equal to 1/4 lambda. (just another calc you can do in the xls). Now decrease the axial distance to s2 until both calculations give a number higher than your desired cross over frequency. I got an Fxo_max of 1211 hz and an Fnull of 1022 hz for a 60x60 horn with an L12 of 1.875". Your sim is no doubt a better way to predict Fnull

 

[nc535]

Had some more time and did HR sim to verify null frequency based on distance from S2 to CD's diaphragm. The calculated null seemed to be much lower than I expected based on sims I've been doing.

There is a big difference! My latest sim is 60x60 horn good down to 400 Hz. At axial distance of 2.25 inches, HR shows a null at 1410 hz. The spreadsheet calculated null, allowing 3 cm for the CD acoustic path length and .25" for a CD mounting plate with a speed of sound of 1127 ft/sec is 918 Hz. HR must know something I don't know!

The horn circumference at L12 = 2.25" allows a max XO of 1048.5 hz. I wanted to find the entry point for an XO of 1 Khz and that seemed to be it.

I have to go back and review how my measurements on my 90x45 horn correlate to HR.

 

[kessito]

Hi cookiemonster

If you divide your model into more subdomains solving will be quicker and memory req. Will be lower.

But a fasster CPU is always good..

 

[cookiemonster77]

Thanks Kessito - quick question - how would you divide it? into horn segments perhaps? so that the parts close to the throat has higher resolution?

 

[xrk971]

Slice it along centerline and use symmetry of quadrants. May need to specify different boundary conditions at slices. Typically set symmetry boundary condition as all derivatives = 0 at slice.

 

[cookiemonster77]

@xrk971: Originally I tried that variant. However, ABEC and symmetry is not quite straight forward. If you are able to cut a RadImp element in two, and mirror it, ABEC will function do what you expect. However, if you have a complete RadImp element on one side, and mirror that side across, you get a mismatch between the lumped element model and the boundary element model. One driver on the LEM side will then drive two RadImp elements, which according to J. Panzer won´t give the expected results. I suppose it could work if you don´t have any coupling, as a way to simulate multiple diaphragms with just one set of driver parameters.

However, I have a feeling that since the high frequencies are controlled at the small side of the horn, and the larger frequencies closer to the mouth, that I could split the horn into segments axially - one segment for the first 10cm, and another for the rest etc.

I am not sure what kind of influence that would have on modes crossing the two domains - this is way beyond my understanding. Perhaps you or Kessito could help explain the tradeoffs?

 

[cookiemonster77]

BTW: I did do one thing though. I replaced the outer shell with an infinite baffle, so the model now don´t have to solve the outside of the horn. So I now get "in-wall" (really big in-wall with no floor and ceiling) results. This should underestimate the low-end, so at some point I need to lower the resolution, place some boundaries in there, and see how what kind of in-room bass response I can expect. Sounds rather complex. I know Akabak calculates total power radiated too - I haven´t found a way to do this in ABEC yet - although it has polar diagrams etc.

 

[cookiemonster77]

x: Perhaps I misunderstood you. Did you mean that I keep the full 3D model but create four individual sections that are connected via boundary elements, or did you mean that I mirror four identical quarter-horns?

 

[cookiemonster77]

Nissep: Thanks! For now, the cone volume will do - don´t want you to start cutting paper until I´m sure I´ll use it. 🙂

 

[xrk971]

x: Perhaps I misunderstood you. Did you mean that I keep the full 3D model but create four individual sections that are connected via boundary elements, or did you mean that I mirror four identical quarter-horns?

I meant slice it into a quadrant so only 1/4th the number of elements. However to do this the boundary at a slice needs to be applied to simulate a symmetry condition. Symmetry is mathematically defined as point where the derivative d/dx=0. This is frequently done in CFD. I don't know how ABEC3 handles this - I would imagine it exists. You will need to read manual.

 

[msibilia]

By symmetry, the velocity perpendicular to the axis of the horn will be zero at the symmetry planes, which should be the same as adding a planar wall there. Since viscosity and boundary layers aren't being calculated, it seems that just adding planar walls and solving one quadrant should give the same answer.

You already have at least one solution. Try cutting the model and re-running it (it should be about 10x faster) and see if it gives you the same answer.

Marc

 

[cookiemonster77]

Ok, since I started this thread, I´m comfortable spamming the thread. 🙂

First: @ xrk971 and msibilia: Yes - I agree. But I have not come across a solution that works with a quarter radiating element for the compression driver and full radiating elements for the other drivers. I did try what you propose, but was told by the author of the software that it won´t give the right result. But I have tried another variant:

I split the model into four subdomains instead of two I had originally:
1: The CD entry and QTWG
2: The main flare
3: The extension flare
4: The exterior

I had moved the entry out to 7.5 cm from the CD, measured axially.

Simming only the compression driver, this is the result:

There is a huge dip for the compression driver at 1500Hz because of resonance from the midrange frustum and cavity.

Simming only the midranges, shows a similar story:

Keep in mind, that the other drivers are "off" in these examples, so what you see should be the resonances and reflections with a "dead" diaphragm (no passive radiator effects included).

So it appears that at the distance of the mids chosen (7.5cm), the mids reflect off the ports in a negative way at two high a frequency.

I tried setting the frustum depth for the mids to close to zero, to simulate a phase plug:

Much better. Now lets look at the mid response:

The dip on the left-hand size is due to the resonance of the woofer ports. Disconnecting the woofer ports:

It seems that there are two ways of going about a Synergy horn. One is to try to avoid these resonances, using phase plugs, or the second is to tune the resonances by placement of the ports and shape/size of the cavities.

Back to the mids position. Its possible to tune remove much negative effect of the mid cavity from the compression driver, or at least tune the notch higher in frequency (beyond the crossover). That leaves the entry position.

The following chart shows the compression driver response along with the mids. (Sorry, I cant highlight both curves at once). Here I used the fake phase plug for the mids, but forgot to turn off the woofer entry port (so disregard the left side notch).

To my untrained eyes, it seems like it should be possible to make a crossover from compression driver to mids with this distance - or do you recommend further changes?

 

[nc535]

Well, I think you're mid entry point is too far down the horn to meet one of the Synergy criteria. I have no idea how audible this would be; perhaps more so off axis than on axis. If you think about it, if the 4 mids enter the horn within 1/4 Lamba of each other, then there is no forward angle from which they can be viewed in which their outputs don't add constructively. If they don't enter that close together, I don't think you can tell sufficiently close to on-axis. So maybe I should try this "cheat" myself as every fraction of an inch eases the problems due to the mid entry ports being off center from the cones of the drivers.

As you say, a crossover around 1 Khz looks very doable.

 

[cookiemonster77]

Good point. I also want the crossover higher in frequency. So i will readjust. Its difficult when you cant immediately say which effect youre looking at. Thanks.


Sent from my iPhone using Tapatalk

 

[nc535]

I probable should have said "looks doable around 1200 Hz"; I'm not good at interpolating on a log scale so I didn't try...

are you still modelling Visaton M10? If so, I'm surprised to see it going so low but then you aren't showing excursion.

 

[cookiemonster77]

Yes not showing excursion - it wont go that low at all.. think 350-400 at best. I will try the faital mids next, as they also look easier to mount. Then try to produce new sims and 3d printable mounting adapter.


Sent from my iPhone using Tapatalk