I think they mean which vertical axis is reference. I can try and see if I can calculate which is the most neutral (maybe the one with the lowest sum of absolute difference of the SPL vs average SPL).Isn't the horizontally flat Listening window "best" according to @MZKM 's scoring system?
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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
That would be very nice!
Or maybe the one with the lowest sum of difference of squared values of the SPL vs average SPL?
EDIT: What I meant is maybe you should use the formula for standard deviation and choose the one with minimum value as the most neutral
Last edited:
Why squared?
Just did it with the Neumann, and +10 was lower than -10, despite the microphone being higher than the reference axis; this is because the upper treble rolled off more with the -10. There is no +5 or -5, so don't know how those would perform.
I am not sure what @somebodyelse meant but my quote from @NTK 's document wasn't about the vertical axis.
"When it is not provided by the loudspeaker manufacturer, we may define the loudspeaker’s reference axis as the line pointing in the forward direction either from the center point of the tweeter or the halfway point between the tweeter and the mid-range driver. "
Same results, +10 is chosen as better than -10.My English sucks, please see my edited post.
If +/-5 was included then maybe -5 would have been chosen.
Right...reference axis meaning where vertically is the listener supposed to listen, aka which vertical axis is more neutral.
Good for KH80 as in most situations head of the listener will be above, not below the speakers.
Yes. But it also metters where the axis crosses the front baffle plane - at the center of the tweeter or somewhere else.
Does the microphone rotate when going off-axis? If it's always straight, then when it's vertically off-axis, it's still facing the baffle head-on. Meaning, if -5 degree was the Neumann logo, that measurement would be the same as if Amir measured the 0 axis at the Neumann logo.
Unless I'm misunderstanding what we are talking about.
In any regards, the graphs Amir posted for the JBL One shows whether the vertical axis is at 0 or not.
From @NTK 's document: black x-axis vs red x-axis. Angles remains the same but coordinates of measurement points change.
OP
- Thread Starter
- #52
@Dave Zan , @No. 5 : I am not a member of the diyaudio forum, and have no access to the attachments. If you don't mind, can you post them here too?
I thought about the measurement grid some more, and begin to think the proposal in my part 4 report of taking measurements in the space between 2 concentric spherical surfaces may not be the best. The issue is that we need r in our measurement points to be different. I think actually a single spherical surface but not centered to our coordinate system may work too.
That will greatly simplify the robot. My proposal was way too elaborate and excessively complicated. If we can stick with a spherical measurement surface, we will need only 2 rotational axes (instead of 5). This will get us back to live within our shoestring budget.
The coordinate system and the location of its origin are only theoretical concepts. Only the location of the speaker and the measurement points are physical. All we need is to know how to relate the physical space to the theoretical coordinate system. The coordinate system (and the speaker reference axis) can move around freely. We just need to adjust the coordinate transformation to translate physical locations into the coordinate system. I remember Klippel said in it patent that they change (optimize the location of) the acoustical center (the origin) with frequency to help the math to converge faster and better.
I will run some simulations using an offset spherical measurement surface to convince myself that it works. I will try to get it done in a week or less.
No. My proposal was way too complicated. I guess I have some subconscious desire to build my own fully fledged robot, but will probably have to scratch that itch with another project
Edit: Now that we are discussing hardware too, I have adjusted the thread title.
Last edited:
somebodyelse
Major Contributor
- Joined
- Dec 5, 2018
- Messages
- 4,781
- Likes
- 4,260
Image frim @Dave Zan
His suggestion to make the adjustment of R manual (slide between 2 end stops?) cuts it from 3 powered axes to 2 if you want to save a little money/complexity - assuming measurements on concentric spheres are ok. It's more or less yours turned upside down, and with the extra 3 pivots removed.
His suggestion to make the adjustment of R manual (slide between 2 end stops?) cuts it from 3 powered axes to 2 if you want to save a little money/complexity - assuming measurements on concentric spheres are ok. It's more or less yours turned upside down, and with the extra 3 pivots removed.
@somebodyelse reposted the relevant one above. The DIYAudio thread died before anything more concrete came out of it.
OP
- Thread Starter
- #55
Thanks!
The main reason why I had my robot hang from a frame is because I couldn't find any reasonably prices hollow shaft rotary indexing table. The added benefit is that it is easier to move the speaker in and out of the test setup. The material cost for building a frame is probably less than the additional cost of finding a hollow shaft rotary table. I'm beginning to think we may not need the r adjustment, but need to verify first with simulation.
We can use a rotary table (with a stepper or servo motor) for the ϕ axis
and a robot servo or another rotary table for the θ axis.
Last edited:
Dave Zan
Active Member
Yes, the idea was to slide between two stops - to save money and complexity but also to reduce the amount of structure near to the microphone that reflects sound and messes up our time window.
And to reduce the mass/moment of inertia of the movable parts to simplify a fast scan.
We don't need to duplicate the whole scan at different R to produce the field separation, we only need a comparatively few points.
So the idea was to start the scan with the low resolution outer field separation, check that there is no interference with walls or room objects, measurements are ok etc.
Once finished a complete pass around the speaker then move the microphone radially and let it work away on the hi-resolution scan.
If this is very slow then it could be even be unattended, if safe.
Sorry about that, the attachments showed for me, of course, so I didn't realise the restriction, thanks to SomebodyElse for the fix.
We need to measure radial variation for field separation, intuitively I think increased radial distance should improve sensitivity, up to the point we have alias problems.
Field separation is restricted to low frequencies, so for a practical sized system, more is better and little point to measure intermediate radial distances, as far as I can see.
I have a turret mill and lathe and expected I could make up a stepper motor drive to pivot around the shaft.
But I did think about the way to avoid that work.
If we do more or less the same system but with the axis not on the speaker support but out the back of the support table.
So a counter balanced stepper motor on a horizontal axis shaft behind the speaker, call it the polar axis.
This positions the second counter balanced stepper motor at the "equator" (a vertical circle in the frontal plane) that scans the "latitude", similar to the first sketch.
Best wishes
David
Last edited:
maxxevv
Major Contributor
- Joined
- Apr 12, 2018
- Messages
- 1,873
- Likes
- 1,974
Through-axis turn tables can be found reasonably priced if you know where to search.
But they are not the solution.
From what I understand of what's needed from the various discussions on this forum, its basically an extendable arm that can do polar plots of location and an additional vertical Z axis movement .
Am I right to say that ?
If that's the case, then your original idea of using standard extrusions to build 90% of the structure still holds. Just need some custom made sections to fill in the functionality.
Last edited:
OP
- Thread Starter
- #58
My thought is to offset the center of the measurement sphere from the origin of the coordinate system. Say, if the offset is x, at one end, the minimum r will be radius_of_sphere - x, and the maximum r will be radius_of_sphere + x, and all the other r's will be in between.
So we can just scan one sphere and not having to adjust the mic at all. Of course I want to run a few simulations to see if it will work.
Yes. I believe for low frequencies, a larger range of r will be better. However, from Amir's post, Klippel only used ~5 cm. The one thing (I think) we need to avoid with double concentric surfaces is to hit one of the zeros of the spherical Hankel and Bessel functions, i.e. h(1)_n(kr) = 0 or j_n(kr) = 0. Because k depends on frequency, we need to stay away from the frequencies that are close to the zeros. This may not be a problem for non-concentric measurement surfaces (which means r vary continuously).
On the hardware design, I am only making a suggestion. I am usually conservative, and try to err on the over-engineered side. The only thing I don't like about having the speaker stand next (offset) to the ϕ rotary table axis is that the axis can no longer freely rotate 360 degrees. This will place some limits on the measurement positions range, but may not necessarily be an actual problem.
On the Speaker Reference Axis
I would like emphasize again that the speaker reference axis is only relevant in the presentation of the results. It doesn't affect the measurements. We can reconstruct the sound pressure level anywhere. All we need to know is where the locations of reconstruction points are when expressed in the coordinate system used in our math.
That also means the physical location of the origin of our coordinate system does not need to match that of CTA-2034. So what is the "best" location of our coordinate system origin? Our mathematics of using spherical wave expansion functions is that we approximate the speaker as spherical sound source. The approximation can be pretty good when we are further away from the speaker than the measurement surface. But because of the spherical source approximation, the closer our origin is to the acoustic center of the speaker, the better off we are. So I propose we instead try to place our origin as close to our estimate of the acoustic center as possible. I'd rather not use mathematics to calculate it, as this will be violating Klippel's patent. I think the center point of tweeter or midpoint between tweeter and midrange is a reasonably good choice.
Last edited:
OP
- Thread Starter
- #59
Both David and I are proposing a system that rotates around both the ϕ and the θ axes, thus the measurement surface is spherical. This is different from Klippel's, which scans in 2 concentric cylinders. I don't know why Klippel chose to do it this way. It is a mystery to me as they are also using spherical wave expansion functions.
maxxevv
Major Contributor
- Joined
- Apr 12, 2018
- Messages
- 1,873
- Likes
- 1,974
Based on this, its still possible to do with an additional rotation axis for the microphone then.
So essentially 4-degrees of freedom.
1 : Polar rotation axis , 2 Lateral (r) extension, 3 Z vertical extension, 4: microphone angle rotation.
This will allow you to do a spherical plot of measurements while keeping construction very simple.
Similar threads
