-
Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
DIY 3D Speaker Scanner - the Mathematics and Everything Else
- Thread starter NTK
- Start date
HARS?
But it is a thought to connect REW somehow.
Tom Kamphuys
Member
- Joined
- Feb 15, 2024
- Messages
- 82
- Likes
- 238
What does that mean, integrate with REW? What function could REW perform for us or what could we do for REW?
somebodyelse
Master Contributor
- Joined
- Dec 5, 2018
- Messages
- 6,653
- Likes
- 6,514
HALS - from #553. Integration into REW might raise patent issues in its paid version.
Valhalla
Active Member
- Joined
- Jan 16, 2021
- Messages
- 118
- Likes
- 181
REW consists of almost two decades of software interface development and it's capabilities are beyond commercial acoustic measurement softwares according to my limited experience. I use Clio from Audiomatica and it is a nice measurement tool but I always export graphs as *.csv into REW to do merge/subtract/compare graphs and REW is the most user-friendly and versatile and graphically sophisticated software I've ever used.
Just imagine REW will be compatible by an Acoustic scanner robot like HALS to do all the communications between a software and hardware and do all the mathematical calculation and including all the standards like CTA-2034, Spinorama, you name it....
Just saying...
Tom Kamphuys
Member
- Joined
- Feb 15, 2024
- Messages
- 82
- Likes
- 238
I've tried imagining what you said. I'm sorry, I don't see the big gains. It might be a blind spot of me, but after reading your post a dozen times trying to distill what you have in mind, I think you might have a wrong understanding of HALS.
HALS is way more than an automated microphone mover that can measure at 70 points for a CEA-2034.
HALS allows you to create an optimized 500-2000 measurement points grid all around the speaker and measure the IR response on each one.
It then uses spherical harmonics fitting and sound field separation to create a virtual model of the speaker that allows virtual measurements to be taken everywhere you like (also at places you haven't measured). The original (real) measurement can be done in an echoic room, but as all echos are removed, it is as if the speaker was measured in an anechoic room.
Although any program can be made to do anything, I think the two solve different problems for the audio enthousiast.
HALS is way more than an automated microphone mover that can measure at 70 points for a CEA-2034.
HALS allows you to create an optimized 500-2000 measurement points grid all around the speaker and measure the IR response on each one.
It then uses spherical harmonics fitting and sound field separation to create a virtual model of the speaker that allows virtual measurements to be taken everywhere you like (also at places you haven't measured). The original (real) measurement can be done in an echoic room, but as all echos are removed, it is as if the speaker was measured in an anechoic room.
Although any program can be made to do anything, I think the two solve different problems for the audio enthousiast.
The post processing exports the files as frd according to how you want them on whatever grid you choose. If you want to import them to REW you already can. REW doesn't offer anything to benefit the measuring mechanism, in fact it would just make everything more complicated.
The FRD's can be put straight into Vituix for instant CTA2034 and polar sonograms. The impulse responses can also be exported if you want to use REW to analyse those.
I like REW, it is good for viewing response. It also has good audio capture system.
When I started to think about how to make HALS (or LAH that was my first project attempt before I joined efforts with Tom and Jan) I did want to use REW as a way to capture impulse response and connect with audio interface. But, API is only available to paid users.
John makes great effort to support audio community, but REW is not open source. If we want to integrate, probably we must keep a lot closed. I don't want to do like that - I hope that after time more users can continue to develop and improve it. All of HALS is made in Python which is always open and human readable, it does not have binary compiled code.
In the past months HALS has made good developments in user experiance. The post processing is now completed in a GUI. Here you can run all the process to compute the spherical harmonic expansion and manage the evaluation of the sound field. You can click one button to calculate and export CTA2034!
You can export as FRD files or as Impulse Response wav. This you can easily load in REW or VituixCAD or another program to view or use for loudspeaker design.
Before post processing you must use the robotic arm to capture the measurement points around the loudspeaker. This is done by a GUI that is inside a web browser. That means you can control from one computer connected to the hardware and also view status from mobile phone or laptop in another room.
This software is three parts - Measurements points generator, Audio Capture and Robot Control.
Here you can move robot to waypoints and that gives the size to generate the measurment points grid. You can view the points and path for the robot in 3D.
Here is the audio setup and audio plots. Live capture shows audio plots and also current measurement points position during measurements set run.
You can see audio devices and support ASIO. There is a sine generator that can help for setting test level and SPL calibration.
With this I have also set a high pass filter for driver protection and there is option to apply inverse filter compensation on the measurment.
I like VituixCAD for viewing the full data because it makes a nice sonogram. I do not think REW does sonogram, which is a sad thing.
Dimitri
- Joined
- May 1, 2021
- Messages
- 503
- Likes
- 1,597
I will go through all of this at some point in time, as I want to do the Klippel NFS in COMSOL Multiphysics (if possible). Then we can easily see which effect the room has, and how close the sound field from the decomposed sources comes to the real output from loudspeaker.
Can you explain a little more about your thinking please?
mwmkravchenko
Senior Member
How? You do realise that 10% difference is no longer a copy of a current patented design. And you can't patent spinning on a central point with a microphone. The software is bespoke, arguably more advanced in some ways than Klippel NFS. And the design for the stand is easily 40% different than the Klippel NFS.
If you are going to try and raise fears, first try to understand patent law, and secondly both designs in comparison. Then you will see that the idea of patent infringement is comical.
Mark
Last edited:
mwmkravchenko
Senior Member
This will be interesting. May raise many questions. But will certainly be interesting.
- Joined
- May 1, 2021
- Messages
- 503
- Likes
- 1,597
If you have a finite element setup with a loudspeaker, you will know the 'true' output from the loudspeaker. If you then on top of this implement the NFS system, you will also get the output from its fitted sources. You can now remove the room, add the room, change the room, and see if you get the same results from the NFS regardsless of the acoustic environment (which I doubt, looking at @NKTs brilliant work here). Perhaps it is close, but then you will still be able to compare the free-field FEM field with the NFS results.
Is that not already done in the pa world, so to use a 3D sound radiation in a large hall? It is not my business but i remember having read something.
somebodyelse
Master Contributor
- Joined
- Dec 5, 2018
- Messages
- 6,653
- Likes
- 6,514
My memory was that at some point @NTK had referred to a Klippel patent when investigating the analysis side, but that was several years ago now and my memory of those threads may be wrong. If there's no patent then great - no problem. If I were the author of commercial software attempting to implement the functionality, I would first want to check what patents may cover the area, and exactly what their claims were. It's the claims that matter, at least in my non-lawyer understanding. Sometimes it's easy to sidestep a claim by avoiding one of its necessary elements. Sometimes it's not. It doesn't matter how different your product is if you're implementing all the necessary elements of a claim. There are of course international variations, particularly around whether or to what degree you can patent software.
Field separation will never be perfect, that is sure. Already there are contradicting requirement for the measurement points grid. Pure SHE can use a Lebedev pattern to maintain orthogonality to a defined harmonic degree (order_N) . For sound field separation we MUST have various radius points to capture field decay, it can not be a surface of a sphere. Now we loose orthogonality and we are struggling with spatial aliasing, mode leakage, radial-nulls and ill-conditioning. It is not analytic integration it is an inverse problem.
For my mind, the most interesting part for your simulation is if you can run optimizer to find ideal measurement points grid design.
For example here is some synthetic tests for two cylinder grids. You can see the clear affects of points spacing.
1) Dual layer cylinder (like klippel documents draw). It has two cylinder surfaces seperated by 50mm. Internal cylinder surface is wrapped with clockwise Fibonacci spiral. Outside cylinder is wrapped by counter-clockwise spiral. Most interesting is the internal / external field separation ratio.
2) Variable density wall cylinder (wall is like cloud of points). I design this to spread the issues across frequency smoothly as possible.
Before it is asked - I do not think Klippel use the dual layer cylinder, they are smart. But, I think they do not share the secrete sauce.
And I am new to posting here so I want to say I am not an expert. I made many mistakes and had wrong assumption. I will keep making mistakes. But, I can share what I found and we can all discover together.
And I am new to posting here so I want to say I am not an expert. I made many mistakes and had wrong assumption. I will keep making mistakes. But, I can share what I found and we can all discover together.
Tom Kamphuys
Member
- Joined
- Feb 15, 2024
- Messages
- 82
- Likes
- 238
Yes you are
mwmkravchenko
Senior Member
Ahem. I agree with Tom.
