If the speaker is rotating then it’s in a fixed location?
(Maybe classic monopole behavior only relative to room modes?)There is also the problem of vertical rotation - especially in relation to full “balloon” polars and particularly with large floorstanding loudspeaker designs.
I’ve been very slowly progressing with the idea of *semi-anechoic in-room with a (horizontal) rotating speaker turntable, but this thread has me somewhat hopeful of Klippel-like results with a lot less fiber-fill in my test-room.
I was also planning on an arc’ed moving mic. vertically with the horizontal speaker rotation (..but not more than about +/- 30 degrees vertical) - shooting for sort of a “stubby” balloon polar result.
The weird thing with fiber is that at lower freq.s it needs to be **less dense the thicker the “paneling”, and you need both characteristics (thicker paneling + less dense) to get lower in freq. with a higher absorption coefficient. Still though, while costs go up for thickness with insulation, they also go down with a loss in density.
*not really “eliminate”, but lower the room’s effects down to at least 60 Hz with an absorption coefficient of .9 and a fast taper-off from there at lower freq.s. (where I’m back to extreme near-field testing again).
**presumably the sound wave at lower freq.s requires thicker (panel) lower density material to more easily vibrate and dissipate as heat.
Last edited:
I wonder about that as well, especially in relation to vertical results close to the floor (or ceiling).
I also wonder about it with respect of loudspeaker depth (as it travels around the rear of the loudspeaker). As I look at Erin’s setup the boom itself seems to need almost 2 feet THEN you add-in the maximum distance required away from the loudspeaker which looks like another 2 feet: what happens if the loudspeaker is 3 feet deep? Does the measuring process account for the change in box proximity and require an extra *2+ feet of distance (so the mic. doesn’t slam into the side of the loudspeaker as it travels around it and still have almost a foot of clearance)?
*over 6 feet x2 = more than 13 foot wide with a 3ft deep loudspeaker
..to be fair though, these are also issues with other testing formats that don’t include your standard extreme near-field single low freq. test (..and adjust for level and splice into your higher freq. test data).
cr@p, I’m talking myself further away from this..
Last edited:
It goes back to my original question.If the speaker is rotating then it’s in a fixed location?
There is also the problem of vertical rotation - especially in relation to full “balloon” polars and particularly with large floorstanding loudspeaker designs.
I’ve been very slowly progressing with the idea of *semi-anechoic in-room with a (horizontal) rotating speaker turntable, but this thread has me somewhat hopeful of Klippel-like results with a lot less fiber-fill in my test-room.
I was also planning on an arc’ed moving mic. vertically with the horizontal speaker rotation (..but not more than about +/- 30 degrees vertical) - shooting for sort of a “stubby” balloon polar result.
The weird thing with fiber is that at lower freq.s it needs to be **less dense the thicker the “paneling”, and you need both characteristics (thicker paneling + less dense) to get lower in freq. with a higher absorption coefficient. Still though, while costs go up for thickness with insulation, they also go down with a loss in density.
*not really “eliminate”, but lower the room’s effects down to at least 60 Hz with an absorption coefficient of .9 and a fast taper-off from there at lower freq.s. (where I’m back to extreme near-field testing again).
**presumably the sound wave at lower freq.s requires thicker (panel) lower density material to more easily vibrate and dissipate as heat.
Because from about 170Hz and upward can be measured in most rooms.
I actually looked it up, the Klippel NFS needs a room of at least 3x3 meter (about 9x9 feet) and 2.5 meter in height (7.5 feet) for the compact version.
Or 4x4 meter by 3.5 meter height for the extended version.
More over, this space needs to be empty.
So if you have a room left, a big garage or an office space left that's great.
Although that makes me wonder, because that only leaves 25cm space left between the microphone and the walls?
For the extended version that's only 20cm.
Here in Europe, a standard ceiling height is 2.5-2.6 meter.
So an extend version will never fit in a normal sized room.
For just a rotating speaker solution, we can make-do with a meter left and right from the source.
As well as a meter between the mic and source.
Ideally a meter behind the speaker and a meter behind the microphone.
Although I have noticed that placing the setup anti-parallel to the walls will help.
Therefore we don't need that much space behind the speaker and microphone.
So in the end we win a meter in width and maybe a tiny bit in length.
But what's a lot more important, is that it's easy to setup and remove again.
If we don't use the setup, we can also use the space for other things, instead of an entire apparatus being in the way.
The pretty offaxis balloon pictures are mostly just useful from about 200Hz and upward.
All I am saying is that an hybrid approach is also still possible.
But also to keep in mind that there is still quite a bit of space needed for this system.
Making a variant that's easier to pull apart again would be a big win.
Btw, you don't have to completely fibre fill your room.
You only need some damping material between the path between the DUT and microphone and the ceiling/floors/walls.
So basically a tube of damping material hanging/floating in the air between speaker and mic will do great.
Yeah, I’m already aware of all measurement methods - but the Klippel result is just so useful for *all types of designs; but their (physical) implementation is not as useful.
*I’m particularly interested in a cardioid pattern to about 60 Hz.
Anyway,
-I burned some REM cycles going through the mechanical limitations last night and came to a conclusion:
Klippel method or not - the design must have a rotating table for the loudspeaker (despite my preference otherwise).
reason: the resulting rotational diameter is “only“ as wide as the loudspeaker itself. So, even if you have a 4 foot deep loudspeaker you can still ”squeeze“ that into a small room that has a little over 8 feet for its width. The length of that room will then determine just how far out your mic. can be from that (horizontal) rotational area.
From there it’s just:
1. a level platform for all axis’s,
2. alignment for the inward/outward (x) axis relative to the center of the rotational platform,
3. placing the vertical (z) axis ON the ”x” axis and making it almost as high as the room will allow.
Also, unlike the Klippel, the “x” axis should be just about “hugging” the floor and considering that any design could be extremely heavy: the rotational platform should also be “hugging” the floor with a wide stable support. From that literal base you would then add a narrow profile speaker (swapable) stand for whatever height is required for testing the particular design.
*I’m particularly interested in a cardioid pattern to about 60 Hz.
Anyway,
-I burned some REM cycles going through the mechanical limitations last night and came to a conclusion:
Klippel method or not - the design must have a rotating table for the loudspeaker (despite my preference otherwise).
reason: the resulting rotational diameter is “only“ as wide as the loudspeaker itself. So, even if you have a 4 foot deep loudspeaker you can still ”squeeze“ that into a small room that has a little over 8 feet for its width. The length of that room will then determine just how far out your mic. can be from that (horizontal) rotational area.
From there it’s just:
1. a level platform for all axis’s,
2. alignment for the inward/outward (x) axis relative to the center of the rotational platform,
3. placing the vertical (z) axis ON the ”x” axis and making it almost as high as the room will allow.
Also, unlike the Klippel, the “x” axis should be just about “hugging” the floor and considering that any design could be extremely heavy: the rotational platform should also be “hugging” the floor with a wide stable support. From that literal base you would then add a narrow profile speaker (swapable) stand for whatever height is required for testing the particular design.
Is this from the Klippel manual? I noticed Erin’s measurement space was nearly empty, and wondered if that’s required for field separation.
I'm wondering why. If the major sources of reflection are always there, what is the problem?
Many of these suppositions are theoretical.
Same goes for required absorbers. They should not be needed if the system is doing it job. Can it hurt? Nope.
I've done a fair bit of acoustical room treatment work if it will help. But let's do that off of this thread. PM if you want.
Basically my assumption is that we don't know the limitations until a few of us have a system running with which we can do measurements. They need to be controlled measurements in each location. Controlled in that pick a speaker and use it. Wide enough bandwidth to excite room modes so at least down to 60 hertz and then up high enough to test out at least 10khertz. Higher than that is mainly mic placement adjustments anyway.
We need a functioning test software package, even if it is a pre-canned test routine that more than one person is willing to try out. The more tests the more data the more trends to see what is working and what is not.
I too am torn between the rotating speaker and the rotating mic boom. You cannot do a 180 degree test with a rotating speaker. But truthfully that is kind of useless in almost every situation. Plus or minus 90 degrees from the baffle is already at the useful limits. If you have a port, or a P.R. rotate the speaker! Same goes for is you are measuring a dipole.
Another bunch of suppositions thrown into the ring.
If we are going to be guinea pigs and report back there needs to be some kind of a checklist made up that we can all have a common report, otherwise this will be nearly useless.
Room dimensions, system placement etc need to be reported.
The test would need to be run through the same software so that there are no math differences. Believe me, there are differences. How the computations are processed in different test programs are quite interesting and aggravating at the same time.
So, Herr Programmer. Pre-canned MATAA test? Those of us with turntables or patience can quickly test out if sound field separation is possible via rotating the loudspeaker.
A higher order ambisonic microphone, maybe? Those can allegedly do nifty tricks like record a voice but not reflections... direction and angle dependent, of course. Maybe two ambisonic mic's, one in front of the speaker to capture the direct sound from the speaker, and one behind the speaker to sample the reflection from the wall behind the speaker to remove them from the desired signal. But I might be getting off track.
There shouldn't be any problem.
The theoretical development of the method, as best I understand it, seems to indicate that the method relies on the reflections being constant for the method to work. Rotating a loudspeaker changes the excitation of the reflected field for each selected rotation angle. This is particularly important and noticeable at frequencies above about 150Hz or so, as the sound source is no longer even approximately omnidirectional. Some NFS measurements that I have seen appear to indicate that the NFS method is capable of resolving relatively small phase differences between separated sound sources at very low frequencies, such as those that are present when a port is on the rear of a loudspeaker enclosure.
Going back to some of the early publications related to the applications of the field separation method, the paper by Melon, Langrenne and Garcia (2012) is informative. They noted that acoustic pressure and velocity fields can be measured on a sphere of a given radius and expanded using spherical harmonic functions. In this formulation, they described that the relevant equations can be rewritten in terms of outgoing and standing waves, where the spherical Bessel functions represent the standing wave field while the spherical Hankel functions represent the outgoing field. Rotating the loudspeaker will change the standing wave field, owing to the varying geometrical and absorption characteristics as a function of where the loudspeaker points during any given measurement.
Nope. Not a solution.A higher order ambisonic microphone, maybe? Those can allegedly do nifty tricks like record a voice but not reflections... direction and angle dependent, of course. Maybe two ambisonic mic's, one in front of the speaker to capture the direct sound from the speaker, and one behind the speaker to sample the reflection from the wall behind the speaker to remove them from the desired signal. But I might be getting off track.
We would be father along to have a vertical line of MEMS mics on a nice stiff circuit board that would have a minimal reflection pattern. And have the computer cycle through them to speed up the measurement. But that is way to far ahead of where we are now. I'd call that a serious enhancement over Klippel.
That pretty much nails the coffin shut on the idea that we can rotate the speaker. I completely agree. The standing waves in the room will be different.
Any chance you have online references to those papers? I can get them if they are JAES papers.
This one?
I don't think that is entirely true. There are papers where spherical microphone arrays using two radii function in a similar way to Weinrich's method using spherical harmonics and plane wave decomposition. With 32 mics some were getting up to n=5 which may be enough for lower frequencies.
https://www.sciencedirect.com/science/article/pii/S0003682X21006174
https://pubs.aip.org/asa/jasa/artic...ne-wave-decomposition?redirectedFrom=fulltext
Last edited:
You are right in the strictest sense. Is it possible. The proper question is it adaptable to the existing software worked out by NTK?
I would hazard a guess that it may not be so compatible.
Not in a drop in sense as that is not how it was envisioned to be used. There is already some open source code for it though.
https://pypi.org/project/sound-field-analysis/
There are some DIY spherical / high order ambisonic microphone builds so it is doable, whether it is easier than making a robot will depend on the individual skill set I suppose.
https://www.researchgate.net/publication/327142959_Modular_Design_for_Spherical_Microphone_Arrays
Well, my typing out loud earlier wasn't intended as suggestion to adapt Weinreich and Klippel's methods (and by extension, NTK's code), but rather a completely alternate idea for someone who wanted to avoid moving the mic around the speaker. Because as has been established, for Sound Field Separation to work (at least as defined as what Weinreich and Klippel do), the speaker must be stationary and the microphone must be the only moving component so as to map the sound field in the room.
The reason I think the mic array has some value in exploring is because the speaker can rotate. The spherical fields are around the mic not around the speaker. If a mic array can be used to separate the incoming and outgoing fields the mic doesn’t have to move. There could well be signal to noise issues at far off axis angles but they exist with current stitching methods anyway.
The beam forming approach shown earlier shows potential, but simple delay sum beamformers introduce their own quirks. Maybe some of the comprehensive adaptive filtering algorithms from other fields might be a solution to these issues.
The beam forming approach shown earlier shows potential, but simple delay sum beamformers introduce their own quirks. Maybe some of the comprehensive adaptive filtering algorithms from other fields might be a solution to these issues.
bummer, that limits use quite a bit - requiring a larger room for larger/deep loudspeakers (..the kind I like to make). Plus, I wasted my REM cycles last night.Ironically I’ve got the floor space in my garage, but not only is it quite noisy with plenty of noise “courtesy” of cars at all time driving by my home on a busy street, but it also has a garage door opener at a height that would also be problematic for larger taller loudspeakers.
It “sounds” like cepstral editing with more data.
Last edited:
The article I referred to was this one:
Measurement of subwoofers with the field separation method: comparison of p-p and p-v formulations.
M. Melon, C. Langrenne and A. Garcia.
https://hal.science/hal-00810930v1/file/hal-00810930.pdf
Building such thing is the easy part.
Getting it properly calibrated is something else.
If we are going that route anyway, having cardioid or dipole MEMS would also solve a bunch of reflections.