Klippel Near Field Scanner on a Shoestring

[Kravchenko_Audio]

My old professor for control theory would stand up and applaud you for saying that. haha, no joke.

Yes, don't participate without insight, because then it will become a statement without a vision.
(little hard to translate, but I think it translates pretty well).
The "new" form of discussing/debating seems to be either just out of the blue responding with "can you prove it" or just linking to some kind of (3rd party) article without insight what the article exactly is about.

What I personally miss the most, is just debating with an open mind and curiosity.
You can still have a (strong) opinion, but always keep that little but of curiosity open.

Specifically about the Klippel measurements and in general all measurements. They are all smoothed approximations of what is going on. So to say that anything is precisely anything that we put a mic in front of is silly.

As for Klippel. I have an admiration for the work behind the Klippel company. And a disdain for it in that they have branded themselves as the be all and end all. There are other systems that can make similar measurements. I use the Smith and Larson Speaker Tester Pro. It can do many of these measurements as well. For quick and dirty I use REW and even shudder a UMIC mic.

These are all tools. They point us in a direction that helps us understand what we are doing and how the changes we make are affecting the sound that we are hoping to achieve.

I design drivers and systems, horns and other things that allow us to listen to music. Bottom line I do all of this because I like music!

There was a comment about impedance and what is going on inside the cabinet. Hoorah! To few regard this as important. I spent most of last summer working on exactly this problem. And it was interesting on how we were able to manipulate the sound of a high resolution system. Off topic ? Yes. But it is part of why and what we measure in the first place. What did we use to measure the changes? Phase, impedance. Nothing to fancy. A little thinking, and simple tools can get you a long way.

 

[Joel Wesseling]

Outdoor measurements have been used to confirm accuracy of NFS.
Accurate phase data is important for designing loudspeakers but phase not possible with KNS or at least haven't seen it

 

[b_force]

Specifically about the Klippel measurements and in general all measurements. They are all smoothed approximations of what is going on. So to say that anything is precisely anything that we put a mic in front of is silly.

As for Klippel. I have an admiration for the work behind the Klippel company. And a disdain for it in that they have branded themselves as the be all and end all. There are other systems that can make similar measurements. I use the Smith and Larson Speaker Tester Pro. It can do many of these measurements as well. For quick and dirty I use REW and even shudder a UMIC mic.

These are all tools. They point us in a direction that helps us understand what we are doing and how the changes we make are affecting the sound that we are hoping to achieve.

I design drivers and systems, horns and other things that allow us to listen to music. Bottom line I do all of this because I like music!

There was a comment about impedance and what is going on inside the cabinet. Hoorah! To few regard this as important. I spent most of last summer working on exactly this problem. And it was interesting on how we were able to manipulate the sound of a high resolution system. Off topic ? Yes. But it is part of why and what we measure in the first place. What did we use to measure the changes? Phase, impedance. Nothing to fancy. A little thinking, and simple tools can get you a long way.

Oh yes. There is no doubt about the quality and technical aspect behind Klippel.
On the other hand there is also clearly a price ratio behind it. They aren't cheap (at all). The vast majority of companies I do professional consulting for, changes or develop a new design every 3-5 years. Very often is an iteration on the previous design.

So for many companies a Klippel system doesn't make a lot of sense to invest in. If they really need the best data, they would rather just rent an anechoic room for a day. Tip: Use a super long time window in those cases as well! (In combination with the non-gated measurements).
Or when it's possible find a nice day with good weather and measure outside.
Big line arrays and other big systems won't even fit on a Klippel system.

As for impedance and phase of the impedance. It's one of the first things I look at. Even just from the drivers themselves. It immediately shows cone/motor resonances, surround dips etc. Plus it's something manufacturers don't mess up to often. Unlike the frequency response, which is often total garbage. It sometimes even shows that demodulation rings were being used. Even when this wasn't mentioned in the datasheet itself.

 

[Kravchenko_Audio]

Oh yes. There is no doubt about the quality and technical aspect behind Klippel.
On the other hand there is also clearly a price ratio behind it. They aren't cheap (at all). The vast majority of companies I do professional consulting for, changes or develop a new design every 3-5 years. Very often is an iteration on the previous design.

So for many companies a Klippel system doesn't make a lot of sense to invest in. If they really need the best data, they would rather just rent an anechoic room for a day. Tip: Use a super long time window in those cases as well! (In combination with the non-gated measurements).
Or when it's possible find a nice day with good weather and measure outside.
Big line arrays and other big systems won't even fit on a Klippel system.

As for impedance and phase of the impedance. It's one of the first things I look at. Even just from the drivers themselves. It immediately shows cone/motor resonances, surround dips etc. Plus it's something manufacturers don't mess up to often. Unlike the frequency response, which is often total garbage. It sometimes even shows that demodulation rings were being used. Even when this wasn't mentioned in the datasheet itself.

Definitely. As for demodulation rings. Sometimes they are spec'd but not provided! Some manufacturers claim that they can do this but the reality is that a deep draw copper sleave is not as easy to get done as many like to think it is. Aluminum is even more fun.

 

[aslepekis]

My old professor for control theory would stand up and applaud you for saying that. haha, no joke.

Yes, don't participate without insight, because then it will become a statement without a vision.
(little hard to translate, but I think it translates pretty well).
The "new" form of discussing/debating seems to be either just out of the blue responding with "can you prove it" or just linking to some kind of (3rd party) article without insight what the article exactly is about.

What I personally miss the most, is just debating with an open mind and curiosity.
You can still have a (strong) opinion, but always keep that little but of curiosity open.

Sounds like someone I would like. :-D

As I like to say, 'why' is the most important question. And I've been noticing the same decrease of curiosity and open mindedness in discussions too, they just kind of become dismissive augments of "so-and-so says X and that's all there is to it" or "the science says this", and it ends up feeling like you're reading headlines and Cliff-Notes rather than a technical discussion.

Specifically about the Klippel measurements and in general all measurements. They are all smoothed approximations of what is going on. So to say that anything is precisely anything that we put a mic in front of is silly.

As for Klippel. I have an admiration for the work behind the Klippel company. And a disdain for it in that they have branded themselves as the be all and end all. There are other systems that can make similar measurements. I use the Smith and Larson Speaker Tester Pro. It can do many of these measurements as well. For quick and dirty I use REW and even shudder a UMIC mic.

These are all tools. They point us in a direction that helps us understand what we are doing and how the changes we make are affecting the sound that we are hoping to achieve.

As evidenced by this thread, I'm impressed by their engineering, but I do find it a bit sad to read forum posts asking "but was it on the Klippel?" As though there were no other way to gather useful data. Being able to "visualize" the sound field around the microphone will get a person a lot further than any brand named equipment.

I design drivers and systems, horns and other things that allow us to listen to music. Bottom line I do all of this because I like music!

There was a comment about impedance and what is going on inside the cabinet. Hoorah! To few regard this as important. I spent most of last summer working on exactly this problem. And it was interesting on how we were able to manipulate the sound of a high resolution system. Off topic ? Yes. But it is part of why and what we measure in the first place. What did we use to measure the changes? Phase, impedance. Nothing to fancy. A little thinking, and simple tools can get you a long way.

Exactly! It's marring of a love of both music and technical things.

It may be off topic, but it's also interesting. Was it a resonant behavior, or something else that you were investigating?

 

[aslepekis]

Accurate phase data is important for designing loudspeakers but phase not possible with KNS or at least haven't seen it

It is actually. One of the selling points that Klippel advertises with the NFS is that you can obtain more accurate phase information since it's a nearfield measurement that calculates the phase response at any location relative to the speaker and therefore wouldn't be as affected by temperature and humidity as a measurement a couple meters away would be.

 

[Kravchenko_Audio]

It may be off topic, but it's also interesting. Was it a resonant behavior, or something else that you were investigating?

It is a loudspeaker manufacturer that I have been working with for about 10 years. A mature design that I designed the woofer and the tweeter for. Literally we were trying to see how much we could remove internal reflections from the enclosure, and what this did to the overall response. SPL, and Impedance and therefore phase. There were some very distinct gains made. After the point source omni was tarted up we turned our attention using the same principles to seeing what we could wrought out of a subwoofer. Let's say that I learned quite a bit. To little attention is paid to the interior of an enclosure by most engineers as if it is simply a black box.

 

[b_force]

Sounds like someone I would like. :-D

As I like to say, 'why' is the most important question. And I've been noticing the same decrease of curiosity and open mindedness in discussions too, they just kind of become dismissive augments of "so-and-so says X and that's all there is to it" or "the science says this", and it ends up feeling like you're reading headlines and Cliff-Notes rather than a technical discussion.

For me it all shows the amount of experience people have (or the lack thereof). Back in the day we used those old paper writers to plot our responses, phase, impedance etc (a real pain to even set them up!) as well as using all kind of analog equipment. I have a background in applied physics. Acoustics is pretty rough compared to subjects that require a lot more precision. Even those things worked with 30 year old equipment.

The results were more than adequate.
It just took a lot more time.

This is what people forget. You pay for convenience. I don't see a Klippel system as a revolutionary system but an evolutionary system.
A very very good one and the engineering is very impressive indeed.

But there are more ways to skin a cat.

 

[Kravchenko_Audio]

I learned to plot frequency response in the late 80's with a paper chart recorder, a function generator and my Radioshack SPL meter as a mic. As you say a pain, slow, but it got the job done. Did you ever put to much ink in the pen! Wow what a mess. And the chart paper was expensive! And we have had to think our way through problems. I'm self taught not formally educated. That has had it's pros and cons. I am pleased to have been a young man in the late 80's and lived reasonably near Canada's National Research Council. I spent nearly two years in the CISTI library studying all I could get and photocopy. My math abilities are sorely lacking many times. And I feel it looking through the math to get this method of measurement accomplished. But I do understand the concepts. And that is enough to see the value of what can be accomplished. Living in Canada near to Ottawa means I have winter for 4 months of the year. No out door anechoic chamber for me!

So now that we can Meow meow quickly we should all have much better speakers. Truly nothing other than the marketing departments is in the way of this goal. The telecoms companies have agreed standards. And even the audio engineering society and IEC have reasonably agreed standards. We just need to get more people to understand and value them.

If Klippel, or what we are discussing here gets us further towards that goal then lets chase it. I will make up a simple speaker that can be tested anechoically and with this new method that I will try. Perhaps it will be possible to have truly useful measurements that are doable indoors and characterise a speakers output a little more accurately than nearfield measurements.

 

[aslepekis]

It is a loudspeaker manufacturer that I have been working with for about 10 years. A mature design that I designed the woofer and the tweeter for. Literally we were trying to see how much we could remove internal reflections from the enclosure, and what this did to the overall response. SPL, and Impedance and therefore phase. There were some very distinct gains made. After the point source omni was tarted up we turned our attention using the same principles to seeing what we could wrought out of a subwoofer. Let's say that I learned quite a bit. To little attention is paid to the interior of an enclosure by most engineers as if it is simply a black box.

The first attempt I ever made into loudspeaker design was an attempt at improving a set of large monkey coffins had I bought for cheap at a second hand store, and I remember being surprised by the improvement that filling the box with rockwool had. I haven't been giving that much attention to the enclosure in my recent projects, but your comment and the results of the box shape directivity connection in the "Acoustic Horn Design – The Easy Way (Ath4)" thread are making me rethink that for my future projects.

 

[aslepekis]

I have a background in applied physics. Acoustics is pretty rough compared to subjects that require a lot more precision.

I remember hearing it said that in acoustics, you never get the same measurement twice. Or something like that.

I learned to plot frequency response in the late 80's with a paper chart recorder, a function generator and my Radioshack SPL meter as a mic. As you say a pain, slow, but it got the job done. Did you ever put to much ink in the pen! Wow what a mess. And the chart paper was expensive! And we have had to think our way through problems. I'm self taught not formally educated. That has had it's pros and cons. I am pleased to have been a young man in the late 80's and lived reasonably near Canada's National Research Council. I spent nearly two years in the CISTI library studying all I could get and photocopy. My math abilities are sorely lacking many times. And I feel it looking through the math to get this method of measurement accomplished. But I do understand the concepts. And that is enough to see the value of what can be accomplished. Living in Canada near to Ottawa means I have winter for 4 months of the year. No out door anechoic chamber for me!

I'm self taught too (this and my professional career), I'm no where near the level of actual driver design, but I personally believe that curiosity and the right resources can get a person almost anywhere. And as a resident of north-east Wisconsin, I sympathize with the summer/winter ratio.

So now that we can Meow meow quickly we should all have much better speakers. Truly nothing other than the marketing departments is in the way of this goal. The telecoms companies have agreed standards. And even the audio engineering society and IEC have reasonably agreed standards. We just need to get more people to understand and value them.

If Klippel, or what we are discussing here gets us further towards that goal then lets chase it. I will make up a simple speaker that can be tested anechoically and with this new method that I will try. Perhaps it will be possible to have truly useful measurements that are doable indoors and characterise a speakers output a little more accurately than nearfield measurements.

Yeah! This is one of the big things I love about DIY audio, not just getting the sound that you want, but working to understand the why of what is happening and developing methods of making something better.

 

[aslepekis]

Another update:

Based on the suggestion by Kravchenko_Audio, I did some more tests using the same speaker and microphone. This time to compare the effects of speaker elevation on the data, and this method to ground plane.

Blue, red, and black are all 15 data points taken from 36" away from the speaker out to 84" at (approximate) elevations of 48", 66", and 88" respectively. Green is ground plane in my driveway with the microphone 36" away from the speaker. Nothing is smoothed or IR windowed, and alignment of ground plane measurement is eyeballed.

I think for the next test, I want to do some more investigation into spacing of data points and see what more distance does.

Also, a bonus indoor measurement with this method:

This is only 29 data points, 30" of elevation, and a whole lot of very nearby reflective surfaces (and a different speaker than the above graph), but I think there is potential.

Any observations or suggestions for ways to improve this method are welcome!

 

[aslepekis]

Another update:
Red is the sum of the all 45 data points from the 48", 66", and 88" elevation measurements, and black is ground plane. And as always, no smoothing or windowing has been applied. I'm not sure if this suggests that a collection of heights is useful, but it does reinforce to me that more data points give a better result.

Another thing to share from the usability of the software side of this is that you may be able to do all of this from inside REW. Using the "Align SPL" and "Vector Average" tools get you almost exactly the same results as summing all the IR's in Audacity.

 

[Kravchenko_Audio]

Another update:
Red is the sum of the all 45 data points from the 48", 66", and 88" elevation measurements, and black is ground plane. And as always, no smoothing or windowing has been applied. I'm not sure if this suggests that a collection of heights is useful, but it does reinforce to me that more data points give a better result.
View attachment 1062961

Another thing to share from the usability of the software side of this is that you may be able to do all of this from inside REW. Using the "Align SPL" and "Vector Average" tools get you almost exactly the same results as summing all the IR's in Audacity.

Regarding the control, your ground plane measurement. Are you tilting the speaker box forward about 10 degree's? Mic is 2 metres away from the baffle of the loudspeaker? The resolution is great. From papers that I have you can use a ground plane up to 13Khertz so this is a very useful method often overlooked. The low end resolution is limited by surrounding barriers. I find the discrepancy from 630 hertz to about 1200 hertz interesting. Could be floor or ceiling bounce.

So now you are comparing an acknowledged equivalent of an anechoic test to the multiple point test. My question is how many test points are required to remove those areas that are different?

I spent all last weekend cutting hay, err grass. Confluence of rain when I had time to cut it and 3 weeks of perfect growing for the grass. So now I have a reasonable area to work in. I'll see what I can rig up this weekend. I might try and message you to determine your test method so we are doing apples to apples testing.

 

[aslepekis]

Are you tilting the speaker box forward about 10 degree's?

I was under the impression that that was something that was done if you wanted results at higher frequency, so since I was only thinking about comparing below a couple hundred hertz I didn't tilt the box. I can try another test with the box tilted and see what kind of a difference it makes.

Mic is 2 metres away from the baffle of the loudspeaker?

I tried 6 feet at first since that was double what I was using for the "free field" distance, but 3 feet ended up giving a cleaner result.

The resolution is great. From papers that I have you can use a ground plane up to 13Khertz so this is a very useful method often overlooked. The low end resolution is limited by surrounding barriers. I find the discrepancy from 630 hertz to about 1200 hertz interesting. Could be floor or ceiling bounce.

And it's a pretty accessible method too: I have my house, garage, and a neighbor's shed nearby and still got pretty good results.

I was thinking that the 630-1200Hz discrepancy may be the result of the larger virtual baffle from the ground reflection, so I ran a simulation of what the baffle-step might look like between free field and ground plane for the speaker I was using, and there's only about a 1dB difference... so back to the drawing board to figure out what's causing that.

So now you are comparing an acknowledged equivalent of an anechoic test to the multiple point test. My question is how many test points are required to remove those areas that are different?

That is the big question! I'm quite pleased with how close they track down to 200Hz, but worried that the long wavelengths below that will make getting lower easy. Since there needs to be enough randomization between reflections in the data points, it might be possible to apply some simple math/geometry to come up with ballpark number of what would be needed to get full range. Or at least down to 100Hz.

I spent all last weekend cutting hay, err grass. Confluence of rain when I had time to cut it and 3 weeks of perfect growing for the grass. So now I have a reasonable area to work in. I'll see what I can rig up this weekend. I might try and message you to determine your test method so we are doing apples to apples testing.

One of the downsides to country living.

I look forward to seeing your results! And I'm more than happy to share details in how I set up my test.

 

[Kravchenko_Audio]

I was under the impression that that was something that was done if you wanted results at higher frequency, so since I was only thinking about comparing below a couple hundred hertz I didn't tilt the box. I can try another test with the box tilted and see what kind of a difference it makes.

I tried 6 feet at first since that was double what I was using for the "free field" distance, but 3 feet ended up giving a cleaner result.
View attachment 1063557

And it's a pretty accessible method too: I have my house, garage, and a neighbor's shed nearby and still got pretty good results.

I was thinking that the 630-1200Hz discrepancy may be the result of the larger virtual baffle from the ground reflection, so I ran a simulation of what the baffle-step might look like between free field and ground plane for the speaker I was using, and there's only about a 1dB difference... so back to the drawing board to figure out what's causing that.


That is the big question! I'm quite pleased with how close they track down to 200Hz, but worried that the long wavelengths below that will make getting lower easy. Since there needs to be enough randomization between reflections in the data points, it might be possible to apply some simple math/geometry to come up with ballpark number of what would be needed to get full range. Or at least down to 100Hz.

One of the downsides to country living.

I look forward to seeing your results! And I'm more than happy to share details in how I set up my test.

I see great progress. But on my end we had rain both days off and on all day.

And I have a full week. So lets see what happens next weekend! Rain go away.

 

[aslepekis]

Hopefully bright and sunny! That give extra time to let ideas percolate.

 

[kyleneuron]

Regarding the control, your ground plane measurement. Are you tilting the speaker box forward about 10 degree's? Mic is 2 metres away from the baffle of the loudspeaker? The resolution is great. From papers that I have you can use a ground plane up to 13Khertz so this is a very useful method often overlooked. The low end resolution is limited by surrounding barriers. I find the discrepancy from 630 hertz to about 1200 hertz interesting. Could be floor or ceiling bounce.

You can get over 13kHz with good accuracy if the ground plane between the source and receiver is very smooth and reflective. A piece of glass is good for this, albeit heavy and awkward and possibly dangerous if the area isn’t clearly marked.

You might also want to put some rubber/sorbothane damping pads under it to prevent vibrations from LF making it vibrate sympathetically for a full range sweep.

I actually carried around a vinyl sample roll of fake laminate wood flooring for this purpose for a little while, but it proved cumbersome

Only just stumbled across this thread but it’s very interesting and there’s been some great progress. I might have missed these things while catching up, so forgive me if anyone is repeating themselves:

• Did @gedlee share the code anywhere? GitHub would be great for this, and is easy enough to use. I'd be very interested to take a look.
• How are you maintaining the measurement distance currently @aslepekis ? Also, what is your choice for the point of rotation?

There's a cool little paper and set of files for a teeny open source & low-cost motorised turntable from the CLIO guys. That might be useful for further work:
https://www.audiomatica.com/wp/?page_id=3024

I think it comes to about $150 for the extrusion, 3D printed parts and other components. Working on a small and simple source is probably a good idea to develop the concept anyway - a 3" cone in a sealed or vented box has good analytical solutions to verify the measured results.

I'm also a big fan of taking a photo of each measurement setup, and either marking that up or making a quick notebook sketch of the configuration and relative reference points. When going back over archival data, I find it quicker to assess stuff than trying to visualise from text. This becomes more important when working with datasets of several hundred measurements or more!

 

[Kravchenko_Audio]

You can get over 13kHz with good accuracy if the ground plane between the source and receiver is very smooth and reflective. A piece of glass is good for this, albeit heavy and awkward and possibly dangerous if the area isn’t clearly marked.

You might also want to put some rubber/sorbothane damping pads under it to prevent vibrations from LF making it vibrate sympathetically for a full range sweep.

I actually carried around a vinyl sample roll of fake laminate wood flooring for this purpose for a little while, but it proved cumbersome

Only just stumbled across this thread but it’s very interesting and there’s been some great progress. I might have missed these things while catching up, so forgive me if anyone is repeating themselves:

• Did @gedlee share the code anywhere? GitHub would be great for this, and is easy enough to use. I'd be very interested to take a look.
• How are you maintaining the measurement distance currently @aslepekis ? Also, what is your choice for the point of rotation?

There's a cool little paper and set of files for a teeny open source & low-cost motorised turntable from the CLIO guys. That might be useful for further work:
https://www.audiomatica.com/wp/?page_id=3024

I think it comes to about $150 for the extrusion, 3D printed parts and other components. Working on a small and simple source is probably a good idea to develop the concept anyway - a 3" cone in a sealed or vented box has good analytical solutions to verify the measured results.

I'm also a big fan of taking a photo of each measurement setup, and either marking that up or making a quick notebook sketch of the configuration and relative reference points. When going back over archival data, I find it quicker to assess stuff than trying to visualise from text. This becomes more important when working with datasets of several hundred measurements or more!

Definitely. Good advice. I use a sheet of plywood or OSB on the ground and that is exactly the idea that is in mind to get the best measurements.

Thanks for the CLIO link I will look. Right now I use a motorised remote control photography turntable. It is nice. But it does require my attention. Being able to do the measurements unattended would be awesome!

 

[aslepekis]

I might have missed these things while catching up, so forgive me if anyone is repeating themselves:

• Did @gedlee share the code anywhere? GitHub would be great for this, and is easy enough to use. I'd be very interested to take a look.

Gedlee posted his original Mathcad files for radiation mode analysis here, and the polar map Python program that @3ll3d00d created is here.

• How are you maintaining the measurement distance currently @aslepekis ? Also, what is your choice for the point of rotation?

The most imprecise way possible: I lay a tape measure on the ground and move the mic stand in steps along it after each measurement sweep. The starting location is measured between the speaker baffle and microphone tip.

I'm working on plans for a sliding track for the microphone and a turntable turntable for the speaker. The rotation axis for the speaker is intended to be coincident with the front baffle, but it might not end up being there.

There's a cool little paper and set of files for a teeny open source & low-cost motorised turntable from the CLIO guys. That might be useful for further work:
https://www.audiomatica.com/wp/?page_id=3024

I think it comes to about $150 for the extrusion, 3D printed parts and other components.

That's nifty. My current measurement turntable uses a bearing like that, I was surprised at how wobbly it ended up being. Although different vendors may have tighter construction tolerances.

Working on a small and simple source is probably a good idea to develop the concept anyway - a 3" cone in a sealed or vented box has good analytical solutions to verify the measured results.

As a matter of fact, I have just such a speaker I can use for this!

I'm also a big fan of taking a photo of each measurement setup, and either marking that up or making a quick notebook sketch of the configuration and relative reference points. When going back over archival data, I find it quicker to assess stuff than trying to visualise from text. This becomes more important when working with datasets of several hundred measurements or more!

Good suggestion. Plus having a body of data that can be more easily passed onto someone else.