Poor man's directivity response measurement technique

[Keith_W]

Hello all. I believe I might have come up with a new technique for measuring the directivity of speakers and I am seeking comments / criticism of my methods. I don't think I have seen this mentioned anywhere else.

My problem is that I own big and heavy speakers. Moving it outside to take measurements is a problem. I don't own a Klippel, nor do I know anybody who owns one.

Method #1
Move your furniture out of the way. Do a moving microphone measurement (MMM) in a vertical line from floor to ceiling at various angles to the speaker, trying to be as precise as possible with your mic movement. Stay the same distance from the speaker, and do not adjust the volume of the speaker in between measurements.

Result

Once done, I loaded up all the curves to compare them. You can see that my speaker is a 3 way consisting of a conventional woofer and two horns for midrange and tweeters. The crossover frequencies are 80Hz, 500Hz, and 5kHz. What we are looking for is drops in volume as the angle increases. The larger the drop in volume, the narrower the directivity.

 

[Keith_W]

Method #2
Place a tape measure on the floor, and do MMM's at various distances from the speaker. Again, do not adjust the volume between measurements.

Result

What we are looking for here is drop off in volume between different drivers. I have again indicated the crossover points. Here, we can see that both horns (800Hz and up) drop off in volume very little as the mic gets further away from the speaker. The conventional woofer drops off in volume much faster.

The more "bunched together" the curves, the tighter the beam has been directed.

You can also see in this measurement the effect of the sofa. All the other measurements were taken in front of the listening sofa, and the MLP was done where I normally would sit.

Discussion

The advantage of this method is that it can be done with the speakers in situ, with some furniture pushed out of the way. It requires minimal equipment, all you need is your existing measurement equipment and appropriate software. It is also pretty quick and easy to do, it took me just less than an hour to perform all the measurements you see here. Maybe a little longer to come up with the pretty diagrams

The disadvantages ... well, that's what this thread is here for. I can see a few:

- room reflections means that the LF is meaningless below Schroder, and maybe below the transition frequency.
- precision and repeatability is poor, but you may be able to improve on this by constructing a jig (e.g. a vertical slide where you can slide your mic up and down) for the first method
- resolution is poor, and it can only identify large differences in directivity.

For those of us without Klippels, and too lazy to move speakers outside, or maybe impossible to move speakers outside (you may live in an apartment, etc) I believe that these methods give you valuable information. It is not the best or most accurate information, but it is the best we can do within our means. It should provide sufficient information in the higher frequencies for us to make some types of decisions.

For me, the motivation was to implement the Magic Beans correction. To do that, I needed to know what the nearfield response of my speaker was. That lead me to measuring the speakers at various distances to find what could be considered "nearfield". Then I realized the pattern I was seeing (bunched together curves) was indicating high directivity for the horns and less directivity for the woofers. This prompted me to do more experiments and I came up with this method.

(Please note, the intention of this thread is to discuss the method, not to discuss my speakers. Yes, I know they are not ideal - no need to tell me that).

 

[kemmler3D]

I think this should work aside from the obvious caveats you mentioned, but maintaining the same exact distance and orientation of mic-to-speaker would be really tricky. So I think doing finer spins than 15deg at a time might not go so well. Especially when you have directivity problems and the lines bunch up, I feel like errors introduced by position/distance could swamp the actual differences between traces.

So I think this is way better than nothing, but won't replace the normal techniques for detailed measuring.

I'm also a little skeptical of the MMM vs. just a simple sweep at each angle. Isn't it sort of hard to sit far away enough from a speaker that the vertical angle doesn't matter at all?

 

[bigjacko]

Are you going to do any post processing for those data? I can't see how a vertical line and on axis line of measurements can represent directivity.

 

[Blumlein 88]

I might eventually be helpful. I have access to 3 speakers that have been measured with the Klippel. Some experimenting with NF MMM might help guide us to a version that gets closer to the Klippel listening window results or shows us we are getting nothing of the sort. Otherwise I think this suffers from the circle of confusion issue. I think that might be the case with the Magic Beans software at least partly. What is it we are really getting out of this? Less influence of the room? Okay, but that hopefully is moving closer to the actual speaker response, because if not then what is it?

Just thinking about it, I think combining the multiple vertical movements you have with horizontal movements done at different heights might give a more useful combination. Seems about right, but fuzzy in my mind. So it needs a little clarification with actual experimentation. In the OPs case, maybe horizontally between the two 90 degree points at horn upper edge, middle, lower edge, tweeter center and below tweeter.

It also strikes me that restricting the pink noise with filters matching crossover points, and doing NF MMM for each driver individually might have more clarity about what this measurement shows us. I know this is all adding more time and bother to the process. I think it is necessary to figure this out. At that point one can probably/possibly figure out which simpler approach gives important info.

 

[Keith_W]

I think this should work aside from the obvious caveats you mentioned, but maintaining the same exact distance and orientation of mic-to-speaker would be really tricky. So I think doing finer spins than 15deg at a time might not go so well. Especially when you have directivity problems and the lines bunch up, I feel like errors introduced by position/distance could swamp the actual differences between traces.

So I think this is way better than nothing, but won't replace the normal techniques for detailed measuring.

It's not intended as a replacement for normal techniques of measuring. My intention was "this is the best I can do under my circumstances".

I'm also a little skeptical of the MMM vs. just a simple sweep at each angle. Isn't it sort of hard to sit far away enough from a speaker that the vertical angle doesn't matter at all?

I was wondering about that, too. The reason I did MMM's instead of a single sweep is because I was concerned that for a large speaker mine, the vertical directivity will affect the measurement. However it would be advantageous in that placement can be more precise. I am sure that would work well for small speakers.

Are you going to do any post processing for those data? I can't see how a vertical line and on axis line of measurements can represent directivity.

You have to read between the lines, so to speak The rate of change between the lines represents directivity. For example, when taking progressively off-axis measurements, the further off axis you go, the more the volume drops. Very directive speakers will show sharp volume drops. More omnidirectional shows less volume drop.

(EDIT) @joentell just told me how to process these curves into something easier to read. Wait, I will go do them.

 

[Trdat]

It's not intended as a replacement for normal techniques of measuring. My intention was "this is the best I can do under my circumstances".

Yes but is this method enough to see if our directivity match between CD and woofer is decent enough? And do we still need to convert these measurements to polar plots?

 

[HarmonicTHD]

Method #2
Place a tape measure on the floor, and do MMM's at various distances from the speaker. Again, do not adjust the volume between measurements.

Result

What we are looking for here is drop off in volume between different drivers. I have again indicated the crossover points. Here, we can see that both horns (800Hz and up) drop off in volume very little as the mic gets further away from the speaker. The conventional woofer drops off in volume much faster.

The more "bunched together" the curves, the tighter the beam has been directed.

You can also see in this measurement the effect of the sofa. All the other measurements were taken in front of the listening sofa, and the MLP was done where I normally would sit.

Discussion

The advantage of this method is that it can be done with the speakers in situ, with some furniture pushed out of the way. It requires minimal equipment, all you need is your existing measurement equipment and appropriate software. It is also pretty quick and easy to do, it took me just less than an hour to perform all the measurements you see here. Maybe a little longer to come up with the pretty diagrams

The disadvantages ... well, that's what this thread is here for. I can see a few:

- room reflections means that the LF is meaningless below Schroder, and maybe below the transition frequency.
- precision and repeatability is poor, but you may be able to improve on this by constructing a jig (e.g. a vertical slide where you can slide your mic up and down) for the first method
- resolution is poor, and it can only identify large differences in directivity.

For those of us without Klippels, and too lazy to move speakers outside, or maybe impossible to move speakers outside (you may live in an apartment, etc) I believe that these methods give you valuable information. It is not the best or most accurate information, but it is the best we can do within our means. It should provide sufficient information in the higher frequencies for us to make some types of decisions.

For me, the motivation was to implement the Magic Beans correction. To do that, I needed to know what the nearfield response of my speaker was. That lead me to measuring the speakers at various distances to find what could be considered "nearfield". Then I realized the pattern I was seeing (bunched together curves) was indicating high directivity for the horns and less directivity for the woofers. This prompted me to do more experiments and I came up with this method.

(Please note, the intention of this thread is to discuss the method, not to discuss my speakers. Yes, I know they are not ideal - no need to tell me that).

Why not use the established poor man method of sweeps and proper gating (to minimize room effect for higher frequencies)?

 

[Keith_W]

I was told how to convert my measurements into listening window, sound power, etc. Here is how, step by step.

1. My measurements were done in Acourate, so the first step is to export the measurements into REW (in Acourate, save Mono .WAV; in REW import Audio File, choose .WAV)
2. For all of these measurements, click on "All SPL" and in the control knob in the top right, use RMS Average
3. For Listening Window: Average of 0 deg and 23 deg
4. For Sound Power: Average of all of-axis measurements (I also did 180deg and 360 MMM's - these were taken by moving the mic all around the speaker)
5. For On Axis Directivity index, Trace Arithmetic 0-deg Axis / Sound Power
6. For Listening Window Directivity index, Trace Arithmetic Listening Window / Sound Power

 

[Keith_W]

Why not use the established poor man method of sweeps and proper gating (to minimize room effect for higher frequencies)?

As indicated in a previous post, I was concerned that the vertical directivity might matter for a large speaker. I don't know how much it matters, but I suppose the only way to find out is to repeat the experiment with sweeps. I will try that later. You bring up a valid point which I do not know the answer to ... yet!

 

[LIΟN]

Anyone trying a new approach should be respected. I'm rooting for it.

 

[Juhazi]

Nice, but what is "new" here? Impact of distance was interesting and new evidence to me, because "normal" speakers and dipoles behave much better.

I have been using REW since 2012 and overlaying measurements is very basic. Indoor measurements present a challenge from reflections, gating is necessary. Smoothing to 1/3 or 1/2 helps too to find tonal expression. Seems like you use RTA or long gating and 1/48 smoothing.

VituixCAD can download measurements and create spectrograms and indexes. VCAD help files and videos tell more and also important general info.

 

[Keith_W]

Nice, but what is "new" here? Impact of distance was interesting and new evidence to me, because "normal" speakers and dipoles behave much better.

I have been using REW since 2012 and overlaying measurements is very basic. Indoor measurements present a challenge from reflections, gating is necessary. Smoothing to 1/3 or 1/2 helps too to find tonal expression. Seems like you use RTA or long gating and 1/48 smoothing.

VituixCAD can download measurements and create spectrograms and indexes. VCAD help files and videos tell more and also important general info.

That is fair. It is nothing more than a conventional outside measurement, but done inside and with MMM's.

 

[HarmonicTHD]

As indicated in a previous post, I was concerned that the vertical directivity might matter for a large speaker. I don't know how much it matters, but I suppose the only way to find out is to repeat the experiment with sweeps. I will try that later. You bring up a valid point which I do not know the answer to ... yet!

The gating is the essential part. Never done it myself so maybe some of the experts can chime in about what value to choose. I suppose the gating time is as short as to “cut away” the reflection from the nearest boundary.

 

[Sokel]

The best way to design your speaker (I feel that that is the intention) is to use Vitruix Cad at the beginning.
That will save you a lot of trouble and time as these measurements won't have to repeat in every single stage right from the beginning.

There are a gazzilion of such attempts around (have a look at diyaudio) but they all eventually end up outside,on trees and cranes as there's no other way around it.
That,if you want accuracy.If is just for fun and a rough estimation gated measurements can show what happens over certain frequencies but even those demand a fair free space around the speaker.

 

[LIΟN]

Isn't the reason trying to use MMM in this thread right now because not going to move the speaker or rotate it? Did I get it wrong somethin?
I understood that the writer was trying to find a simple way to do it with a speaker that was already made, and a speaker that was already set up and placed, rather than a speaker design.

 

[ernestcarl]

Well, everything above 500 Hz seems “easy” to fit into a target given how smooth and flattish it already is — the transition with the conventional woofer looks rather tricky as the change in directivity control appears quite rapid.

 

[Keith_W]

The best way to design your speaker (I feel that that is the intention) is to use Vitruix Cad at the beginning.
That will save you a lot of trouble and time as these measurements won't have to repeat in every single stage right from the beginning.

There are a gazzilion of such attempts around (have a look at diyaudio) but they all eventually end up outside,on trees and cranes as there's no other way around it.
That,if you want accuracy.If is just for fun and a rough estimation gated measurements can show what happens over certain frequencies but even those demand a fair free space around the speaker.

I anticipated that I would get a response like this

Doctors have a different understanding of measurements to engineers. As a doctor, I know that even imperfect or inaccurate measurements have value. It tells me where I should look, and what investigation I should arrange to give me a better picture. I never start off with the "gold standard" investigation unless the pre-test probability strongly indicates that the test will give me useful and actionable information. I don't order PET scans as a screening test for cancer, I only order it when I already know the patient has cancer.

With this method, I know I am looking at a fuzzy picture, kind of like a low-res photograph that has been blurred a few times and then the print gone through a car wash. But even low-res photographs have shapes. As long as I can discern the shape, it gives me information that the manufacturer did not provide, so it has value. I suspected that the engineer types on ASR might scoff and sneer at how low-res this picture is, but remember: it can be done by anybody with commonly available tools, and for no money if you already own the tools.

As for DIYAudio, speaker designers have different needs to hobbyists. I can use this information to roughly tell me what to do. For example, that sudden change in directivity going from woofer to midrange horn. I could: (1) reconfigure the crossover to a gentler slope, (2) construct a crude waveguide, (3) change the woofer cabinet to a horn, (4) decide if I want to put up with it, how much it really bothers me, and buy a new speaker. And then once the changes are made, I can verify the difference easily using the same method.

 

[Ellebob]

Your measurements have a lot of merit and have more real world usefulness than people think. I measure speakers where I work and we take similar type of measurements. We use a TEF system which is a type of gated measurement. It is an older system but works fine for our purposes and we routinely get mics calibrated and tested. But, when measuring speakers I also measure outside if possible. Usually, not a big difference outside to inside except low end where room modes dominate and the gated measurements help. I would love a Klippel but I am sure they are not going to buy one where I work or give up the space needed for it. It is really overkill for our purposes.

Our purpose for doing multiple measurements in different degree increments is for making an acoustic design of a room. We want to know at what degree the speaker exhibits anomalies and how that anomalie might affect the sound. A spinorama doesn't tell you that. So when we do an acoustic design knowing this information and what the response will be at reflection points and helps us determine what type of acoustic treatment to use or if any at all. Unfortunately, the blanket advice of putting absorption at first reflection points is bad practice and very dated technique in the acoustic world. At least 30 years ago they stopped recommending that as routine practice. However, many companies that make acoustic panels still recommend it, to sell more panels. We see how the speaker response is at the reflection point and make a determination on what should be as well as the goals for that room. You also factor tow in with a design as that affects off axis response and reflection points.

So, these type of measurements are very useful and you can get an idea of how your speaker performs even indoors. Gated measurements will help. But as you stated you will get a good idea of how you can make changes to your design.
I rarely post measurements here of the speakers we measure for the same reason. It would get scrutinized and it isn't in spinorama format, etc.

 

[Sokel]

I anticipated that I would get a response like this

Doctors have a different understanding of measurements to engineers. As a doctor, I know that even imperfect or inaccurate measurements have value. It tells me where I should look, and what investigation I should arrange to give me a better picture. I never start off with the "gold standard" investigation unless the pre-test probability strongly indicates that the test will give me useful and actionable information. I don't order PET scans as a screening test for cancer, I only order it when I already know the patient has cancer.

With this method, I know I am looking at a fuzzy picture, kind of like a low-res photograph that has been blurred a few times and then the print gone through a car wash. But even low-res photographs have shapes. As long as I can discern the shape, it gives me information that the manufacturer did not provide, so it has value. I suspected that the engineer types on ASR might scoff and sneer at how low-res this picture is, but remember: it can be done by anybody with commonly available tools, and for no money if you already own the tools.

As for DIYAudio, speaker designers have different needs to hobbyists. I can use this information to roughly tell me what to do. For example, that sudden change in directivity going from woofer to midrange horn. I could: (1) reconfigure the crossover to a gentler slope, (2) construct a crude waveguide, (3) change the woofer cabinet to a horn, (4) decide if I want to put up with it, how much it really bothers me, and buy a new speaker. And then once the changes are made, I can verify the difference easily using the same method.

Most in diyaudio are hobbyists along with designers too of course.
They have done anything imaginable to get over it,you can read threads as long as books in there the last 15years.

Even using elaborate turntables and conditions sometimes funny the results were never useful below some certain freqs and most certainly not in the area that makes the difference,sub integration and woofer-mid integration,specially with unconventional designs,as with woofer far apart (where the xover has to be low due to wavelength) or on the sides,etc.

Find a way to take them out and something to rotate them on,there's no other way around when it comes to design.
Or else your picture will not only be low resolution but it will also have missing pieces.