Well I suppose if it was easy someone would already have done it
Maybe I'm missing or misunderstanding something. Was your approach able to do something about the lack of frequency resolution in a time gated measurement?
The gain provided can be in the eye of the beholder. You spent the time to work out the math and write your own software for the system you use. Another person could have said "that's a significant expense of time and energy, I'll just use commercially available software and rotate the speaker in smaller increments for finer angular resolution, it's worth the financial expense to me". To get high resolution, far field, anechoic data from an echoic space will absolutely require complexity. I can't imagine a way that it wouldn't. But for me, it's worth it to try and see if it can be done. There are many ways to "see" what's going on at lower frequencies, and I think most of the contributors to this thread already use those methods, but if you could also use anechoic data in your speaker design, wouldn't you want to? For me the answer is yes, but then the next question is: is it worth the complexity and expense? As this project currently sits, for me it is. Certainly there could reach a point when it no longer is (there is a reason I haven't bought a 100k Klippel NFS, after all), but we're not there yet.
In that case it's also assumed that this (Klippel) method is the only way to get this anechoic data.
I think there are plenty of other approaches possible, even more so when you start combining a bunch of these other solutions.
Work smarter, not harder.
It just totally beats me why there is just such a strong focus on this one and only method?
(For some it's even the only and most holy way).
Let alone approaching it the objective way, by just making a (imaginary) table with all the pros and cons and actual benefits.
These other methods have been proposed a couple of times.
Are these just not sexy enough?
I have been thinking a lot about a 100k Klippel system. Even more so with other companies and clients.
100k is a lot of money for a lot of companies to spend on a system that, except for a few options and modules, mostly just gives you convenience.
Unless someone can clearly show me the opposite, I still have doubts if this is the smarter method.
To stay in signal processing puns, it feels pretty convoluted.
I think there are plenty of other approaches possible, even more so when you start combining a bunch of these other solutions.
Work smarter, not harder.
It just totally beats me why there is just such a strong focus on this one and only method?
(For some it's even the only and most holy way).
Let alone approaching it the objective way, by just making a (imaginary) table with all the pros and cons and actual benefits.
These other methods have been proposed a couple of times.
Are these just not sexy enough?
I have been thinking a lot about a 100k Klippel system. Even more so with other companies and clients.
100k is a lot of money for a lot of companies to spend on a system that, except for a few options and modules, mostly just gives you convenience.
Unless someone can clearly show me the opposite, I still have doubts if this is the smarter method.
To stay in signal processing puns, it feels pretty convoluted.
Personally, I think this is one of the most interesting and technically challenging projects currently active on this forum.
After all we are not a company who have to think about how to manage available resources on a project like this or even think about the novelty involved. I feel it is always good to have more tools/methodologies for data collection and processing to arrive at some meaningful observations and take the science/technology involved forwards. In that context, I feel that the existance of a tool like the nearfield scanner or a reduced feature set diy version of it will definitely reduce the amount of manual guesstimation involved in making polar response measurements in the frequency range below 1kHz.
Hence, I feel this project deserves all the encouragement and support it needs to move forward. I wish I could also help with the technical aspects involved, but unfortunately I am not an acoustics person though my area of expertise is related to it in some ways in terms of the math involved.
Hopefully this project moves forward (at full speed) and stays on its subject and not get bogged down/diverted by repeated discussions about "why do we not need a klippel nearfield scanner on a shoestring"
After all we are not a company who have to think about how to manage available resources on a project like this or even think about the novelty involved. I feel it is always good to have more tools/methodologies for data collection and processing to arrive at some meaningful observations and take the science/technology involved forwards. In that context, I feel that the existance of a tool like the nearfield scanner or a reduced feature set diy version of it will definitely reduce the amount of manual guesstimation involved in making polar response measurements in the frequency range below 1kHz.
Hence, I feel this project deserves all the encouragement and support it needs to move forward. I wish I could also help with the technical aspects involved, but unfortunately I am not an acoustics person though my area of expertise is related to it in some ways in terms of the math involved.
Hopefully this project moves forward (at full speed) and stays on its subject and not get bogged down/diverted by repeated discussions about "why do we not need a klippel nearfield scanner on a shoestring"
What is the estimated space requirement for such a homebuilt system? I would be super happy with some sort of device, that would allow to measure/calculate anechoic response on one axis in a relatively small room (2 m ceiling, roughly 2x3 m in front of the speaker). That would be good enough for me to tune the DSP crossover on reference axis and check on two or three different angles.
That's also one of the things I'm interested in.
My current understanding is that you ( @gedlee ) basically do a (horizontal only) quasi anechoic measurement. You've optimized the nr. of (manual) measurements you have to take by putting some physics in the processing. With that processing you can deduce the sound at any position in a horizontal plane.
Is this an analogy?
People are measuring something that has to be a parabola. People take lot's of measurements spread along the x-axis, measuring the y value. They don't realise it has to be a parabola.
You use the physics of the problem at hand and use the fact that is has to be a parabola and say you only need 3 measurements to fully describe the parabola. You can use more measurements and do a (least square) fit. From these measurements and the formula for a parabola (y = ax^2 + bx + c) you end up with a, b and c. After that you can evaluate that function (the specific parabola) everywhere.
I'm not trying to divert from the original idea of this thread, but it would be nice to have some intermediate goals (and hopefully successes) along the way to the ambitious goal of this thread.
I feel that as well, but it's still important to not overshoot the goal as well.
To me it feels the goal is now designing a system, not designing loudspeakers.
Which is obviously totally fine and great, but it's important to understand the reference point to move further with the project.
Otherwise it will result in a never ending loop of discussion.
Well, the topic of this thread is "Klippel near field scanner on a shoestring" and it's explicitly stated purpose is to try and DIY a near field scanner, therefore one should expect a strong focus on it. This is kind of like someone posting on a Ford Model T forum and asking "why are you always talking about Model T's?? Don't you know modern Honda's exist??" Yes, I know other methods exist, but I didn't start this thread to talk about them.
Just to be clear, I use those other methods and see the value in them, but you do understand that if this thread was "let's talk about the mic in box method", there would be way more talk about the mic in box method than about ground plane or IR windows, right?
It's not the smarter method for you, and that's fine. I have exactly zero interest in trying to convince you otherwise, and I respect that you have not reached your conclusion arbitrarily. But I would appreciate if you extend the same respect to me when I say that there is value in it to me.
In case it was overlooked, the purpose of this specific thread has always been designing a system.
Agreed.
As I have said before, the two reasons big to have a near field scanner are high angular resolution polar measurements and anechoic data from echoic rooms. The second point is certainly ambitious and will undoubtedly be complicated, but the first appears to be far more obtainable. If memory serves, much earlier in this thread, the recommendation for the first goal was software/data processing to calculate radiation modes in 2D (sort of an open source version of what Earl is using); then (2D) robotics; then the ambitious 3D stuff.
I don't see lack of respect, all I do is asking questions?
Is asking questions not allowed anymore?
I would appreciate if people would ask and answer questions as well, instead of misjudging people's intentions.
Because I never said anywhere that you're not allowed to do things your way.
If you would cobble things together with ducttape, I couldn't care less about whatever makes you happy 👍
I am not interested in my personal smarter way of working. (Or anyone else for that matter).
If you have been reading my posts that way, you have been misreading a lot.
I am interested in an OBJECTIVE way of working smarter.
Meaning, as soon as people can show that something else will take less time, less effort with better results, I am sold.
That's not to bash down an idea, but to maybe be aware that there others roads leading to better results for less effort.
Also known as brainstorming and making sure that this leads to the goal.
Which I have doubts about.
Doubts with good intentions, not to bash as a grumpy person.
Yes , I can read and understand topic titles.
You can call it potatoes, the goal is still designing loudspeakers.
What is the basis is our being able to do measurements in a controlled fashion with repeatability. We can do this manually and we can do this automatically. The Klippel method being the only way, not so. If I remember correctly the Thesis behind the gentleman who works for Klippel has been impossible to find. And I am good at finding information online. At least it was not possible last year when NTK and I were looking. So, the math is from another adventurer in Belgium if I remember correctly. Prof, that there are other ways. Earls reference to Mr. Wienrich is another example.
Getting bogged down in the minutiae Is worthless. As soon as my shop is cleared from what we are working on now, I will knock up a as simple measurement system as I can engineer. It may inspire others. I'm a cabinet maker. It will be out of wood! BUt that is a well understood engineering material if there ever was one.
I have a few ideas for this, and when I get this up and running I will welcome any and all ideas, and criticisms. Enough typing and lets build something!
The technique has a marginal effect on the frequency resolution limitations due to gating - being near field the gating times can get a little longer than one would have taking the measurements in the far field.
But its resolution enhancement for polar responses is significant. This is because interpolating polar data directly yields very poor results. This was initially the problem that I was trying to solve (at the time, I wasn't even aware of Prof. Weinreich's work - even though I knew him!!!) I wanted a better means of showing polar data accurately than simply interpolating a fixed number of angular measurements. As I have said before, my 13 angles yields a 2 degree angular resolution - that's a huge improvement in this regard. But in the frequency domain the improvement is only marginal, although still useful.
I believe, although I never implemented this, that a few very near field points could be used to further enhance the LF resolution, i.e. kind of like Keele's approach. This is because at LFs only a couple of radiation modes exist - monopole and dipole (but a dipole exists only if the system is ported, monopole systems (like all mine) have only a single mode and hence can be described by a single data point.)
That is what I thought but it is good to have confirmation.
This would be interesting to explore if you have any thoughts on how to approach it. NTK has confirmed that he is interested in making an open working version of your approach, when I know what he needs I'll let you know.
I think this can be viewed as an intermediate step towards improved measurement accuracy and for many, may be as far as they wish to go.
I don't believe that this would be a distraction from anyone wishing to purse a full robotic system.
This might also be usefull https://github.com/AppliedAcoustics...es/tree/master/loudspeakers/loudspeakers_3D3A
The database also has the data for the @gedlee Nathan.
The database also has the data for the @gedlee Nathan.
As stated before, for each additional point we can add another mode. Two ideas come to mind:
1) rotate the speaker on its stand and do polar measurements in two normal planes. Combine this data into a program to generate a "best fit" to the data. This will not be accurate for systems with no symmetry at all, but should be accurate enough for almost any DIY approach.
2) move the mic up a small amount, then up again and down a couple of points. Add this data in to the horizontal model to generate a compromised 3D plot.
As I said before, cancelling room reflections requires a full 3D model and I just do not see doing that without an automated system, meaning that it would be last on a list of tasks.
I firmly believe that taking smaller steps, starting at the simplest approach and then moving on to the more complex ones is the only way that anything will happen. People can then just jump off at the stop that most suites them, and hopefully the train will continue it's journey to the end with enough interest to be completed.
I've been playing with NTK's Octave code. I hacked some stuff in and out. I was trying to get a feel for what was going on, so I tried to make my own script that measured in a circle on the horizontal plane.
This is what I get with 21 measurements:
stars : point sources
dots : measurement positions (color is amplitude)
crosses : points for comparison in plot just right of blue-red colorbar
asterisks : measurement positions
As you might have noticed, I didn't measure on a perfect circle. If I do that, I get this:
It looks like it might be fitting the circle ok, but it doesn't have enough knowledge of the rest of the space and the results their are wild.
Now I need to think what I just did....
This is what I get with 21 measurements:
stars : point sources
dots : measurement positions (color is amplitude)
crosses : points for comparison in plot just right of blue-red colorbar
asterisks : measurement positions
As you might have noticed, I didn't measure on a perfect circle. If I do that, I get this:
It looks like it might be fitting the circle ok, but it doesn't have enough knowledge of the rest of the space and the results their are wild.
Now I need to think what I just did....
Earl, if a person wanted to put in the time this is totally possible manually. And that needs to be accomplished with the mechanics in the first place. Then as you say, we can try and add CNC control. One follows the other.
What I was thinking for repeatability is a old time saw tooth shaped gear and a pawel :
This can get easily repeatable results on the horizontal rotation. We just need to know how many are required for a good measurement. This same mechanism can be made in a linear fashion for the vertical height adjustment for the mic boom, and the forward and backward mic boom adjustments:
Both can be cut on a table saw, or with a reciprocating saw (jig saw) or a scroll saw. The spring can literally be a wire wound spring or a thin piece of wood.
Brain storming. Any other ideas?