Speaker acoustic center – How to find it
How to determine speaker acoustic center?
First of all, determining the speaker acoustic center is useful when designing loudspeaker crossovers. The point from where the speaker is radiating sound is somewhere near the voice coil area. As a result, big speakers, like woofers, have the acoustic center somewhat in the back. On the other hand, tweeters, which are more shallow, are in front, acoustically, compared to woofers. This offset, whether expressed in units of time or distance, must be taken into account to have an accurate representation of the phase between the drivers of the loudspeaker.
Prerequisites and device under test
To find where the speaker acoustic center resides, you need to know how to make frequency response measurements. Here are some articles to help you out :
- These show you how to measure to measure frequency response using ARTA :
- This one shows how to measure distortion using REW. While it’s not of interest the distortion measurement, it shows how to make an impulse response measurement using Room EQ Wizard. You will get more familiar with the software :
The device under test is a 3 way floorstander using the following drivers (Amazon affiliate paid links) :
- Tweeter – Dayton Audio DC28F-8
- Midrange – Dayton Audio RS125-8
- Woofer – Dayton Audio SD215A-88
The loudspeaker is placed on a small box to raise it up slightly. This is to place the speaker in the center of the room and reduce early room reflections. The microphone is placed at 1.2 m from the floor, on the same axis with the middle of the midrange driver. Microphone is 1 meter away from the speaker.
What we need to do is to take a far-field measurement of each individual driver. After that, take another far-field measurement of all the drivers wired in parallel.
REW measurements
I’m making the measurement using sweeps from 200 Hz till 20 kHz. Not going down till 20 Hz for 2 reasons. First, my room only accommodates accurate measurement only till 200 Hz. Second, starting from 200 Hz protects the tweeter.
We have 4 measurements in total :
Already there is something weird about the combined response (red line). I’m talking about that dip. No worries. It should look weird. That’s because of the acoustic offset of the speakers. We’ll calculate each speaker acoustic center distance soon enough.
Before anything else, we need to smooth out each response. Click the “cog wheel” on the top-right and apply 1/12 smoothing. Then, we need to apply a window to each response. Click the “IR Windows” button on the top :
I like to set the Ref time just before the initial peak (time 0). So, I go for -0.2 ms. Then, set the right window according to your room size. Or check the impulse response to see when is the 2nd big peak (which corresponds to the first room reflection). In conclusion, I set the right window to 4.7 ms. The left window is irrelevant, but I like to set it low, like 15 ms.
How to design loudspeakers - video courses
This is true for both mid and tweeter. For the bass driver and for the combined response it’s not the same, at least in my case. It would be for a 2 way bookshelf on a stand, for example. But here, the 2nd bass speaker and the port are too close to the floor and room reflection hit the microphone earlier. Here, the window is only 3.4 ms. Check your exact impulse response for an accurate time window. If you are having trouble with this, let me know in the comments and I’ll see if I can clear things up.
Using Xsim
If you haven’t used Xsim before here is an article where I cover some of its functions. But before we get to Xsim, we need to create FRD files for all of those measurements. For each measurement create a FRD file by clicking “File -> Export -> Export measurement as text”. These we will load into Xsim. First, place 3 drivers wired in parallel :
We will load just the frequency response file for each driver. The impedance is not of interest. So, right click the first driver and click Tune. Load the appropriate file.
Hibert Bode transform
After the file has been loaded, we need to click the “derived” check box.
This will fix any errors with the measurements and extend the frequency response measurement by following its trend. Just to make an analogy : Imagine you are measuring a large piece of land using rudimentary measuring tools. And this land is a right angled triangle. After you measured each side you can apply the Pythagorean theorem and make corrections so the theory holds true. That way you know you are closer to the real values. This is what the Hilbert Bode Transform kinda does.
This is how the HBT looks like. We need to set the tails for the ends of the response curve.
We know that the response is not accurate below 200 Hz, because of the width of our window when we made the measurement. In conclusion, we set the lower tail just above 200 Hz. Try to follow the natural roll-off of the tweeter so the blue line matches the red line. In our case, 10.5 dB / Octave slope. As for the other end, 20 kHz, we know that the tweeter will lose amplitude quite quickly, as it’s not an expensive tweeter. In conclusion, 18 dB / oct seems appropriate.
Do the same for the midrange.
Same procedure for the bass driver.
Finding the speaker acoustic center
Next step is to load the combined response file so it shows on our graph. Click “Curves -> Get File” and load the response with all the speakers combined.
We got 2 curves on our chart :
- Blue line – Is the modeled response from our circuit.
- Green line – Is the measured response of all the speakers wired in parallel.
The goal here is to match the two curves by altering the delay of each individual speaker. This will determine the acoustic offset between the drivers.
Let’s start with the bass driver. Since it’s a woofer, the sound will arrive the latest. Right click the speaker and click Tune. Now fiddle with the delay until you get the curves to overlap.
At 1.6″ the 2 curves overlap nicely. Still there is a problem for the upper frequencies. Now let’s switch to the midrange driver. After a bit of fiddling, the curves started to overlap for a negative delay number. This means that the midrange is in front compared to the tweeter. Usually this does not happen. However, since the microphone is placed on axis with the midrange driver, the sound coming from the mid arrives first, in our case. You want to leave the delay of the front most speaker set to 0 (zero). As a result, leave the midrange driver delay to 0 and move on to the next driver. As a side note : If you were to measure a 2-way speaker, place the microphone on-axis with the tweeter.
Now let’s switch to the tweeter :
As you can see, this time, the speaker acoustic center is 0.34″ further back. As a side note, this is not the absolute speaker acoustic center. It’s the relative acoustic center of the tweeter compared to the midrange, and relative to the position of the microphone.
Conclusion
Now we have found the acoustic offset of each individual driver relative to the microphone. This is particularly useful when you are designing crossovers. It’s a mandatory step, to have an accurate phase representation. Next time you are designing a crossover using XSim, make sure you do this step first before you start designing your circuit.
References
- Image source : link.
You May also Like
Learn loudspeaker design from scratch
29 comments
Another great post. Thank-you!
Any suggestions how to measure an absolute speaker acoustic center? I have an ECM8000 microphone, but as I don’t know where the diaphragm is at, it makes measuring an absolute distance difficult.
If I’m using ARTA to measure FR response and do not choose minphase but instead just use the measured phase do I still need to find the AC? Can I use measured phase in xsim directly without any manipulation of the .fed file?
Does make sense what you are asking, and my first instinct is to say yes. You should verify your results though, as I haven’t done it that way and can’t confirm.
I’ve now confirmed it. Using measured phase in xsim works very well. From now on I will only be using measured phase and xsim to design. Using minimum phase and splicing near field does not work so well…this tilts phase and harms the xover design.
I have to agree that adding the nearfield response skews up the phase a bit. Now I still measure the nearfield to have an idea on how the low-end looks like. Use only the far-field in XSim, and apply Hilbert-Bode transform more accurately (since I know how to low-end response looks like, and set the tails better). The low-end response is mostly given by the enclosure and not by the crossover anyway (except for specific applications), so the crossover design doesn’t change much (if anything) in that area.
Should I also add “tails” to both ends of all my frd’s when doing crossover design? Or just when finding the acoustic center?
Thank you for all the info here. Great site!
Yes, you should add tails when doing the crossover design as well.
I’ve read on forums that when preforming this task using ARTA and Xsim you extract minimum phase from ONLY the individual woofer, mid, and tweeter measurements. You then add the tails. But the combined measurement (drivers connected in parallel) should be entered into Xsim with the ARTA measured phase (not extracted minimum phase). Do you agree with this? I’m still new to this so want to understand the best ways. Thank you.
Yes, I agree. The combined response you simply upload to Xsim, with no extra processing.
Hi,
From Your article, the woofer has positive delay, which, as XSim instructions say, means that it should be moved away from the mic, not closer. This means that its leading the signal, wheras negative delay for the mid mean that it laggs and should be brought closer to the mic. I got similar results when measuring right between the mid and tweeter, curves overlap at positive delay for mid, which means I should move it away from the mic. But its obvious that its voice coil is further from mic. His is that correct?
Regards,
Maciek
Well, on a straight baffle the acoustical center of a woofer is clearly behind a tweeter. Furthermore, the microphone is placed on axis with the tweeter, so the tweeter signal should reach the microphone first. Remember, we are not calculating the offset between the drivers on the horizontal axis, but rather the acoustical offset relative to the microphone. To be honest, it doesn’t matter how you view it, as long as you match the curves, so everything is ready for starting the crossover design. For example, in a 2-way, let’s say you apply a +0.5″ delay to the bass (to align the curves). You can apply -0.5″ delay on the tweeter instead and it will have the same effect.
I understand that its all relative to the mic and I suppose this is the design axis is. What I don’t get is why this method suggests moving woofers deeper into the baffle or tweeters closer to listener, on a flat baffle. I would expect the opposite, hence my inquiry.
It’s not that we are delaying the woofer, but rather, we are telling the software how much the woofer signal is delayed after the tweeter signal hits the microphone.
Oh OK, now I finally got it, thanks!
Hi thank you very much for the great article you definitely have the best info on the web for designing loudspeakers. I’m using rew to do my measurements and I want to know should I use any loop back or timing reference in my rew measurements to align the acoustic centers and then design crossover. What I understand from your article is that it’s not necessary because you refer to your article about taking distortion measurements using rew and haven’t seen in your soundcard settings that you selected any loopback or timing reference channels. I’m still new to this your help will be greatly appreciated. Thanks again.
The loopback reference is most important when you do nearfield measurements of the speaker and port and you add them together. If you don’t have a loopback you can get weird results.
Hi, wondering if the acoustic center measurement somehow factor the slanted baffle’s angle?
Or the z axis need extra compensation to allow the baffle tilt angle? Thanks in advance.
It is the relative distance from the acoustical center to the mic. Even though it takes into consideration all the axis, but you cannot tell how much is each axis individually.
This person –> https://www.youtube.com/watch?v=xoJjeK4rKNM&t=0s is using Xsim to also find the acoustic center of Z offset. But he is doing it in a different way. He’s not using “tails”. for one. And doing a few other things different. Will his method be accurate? It seems to work for him but I don’t understand without the tails how.
His method will work, however, it won’t be as accurate. When he finishes the crossover and measures the response … if he would compare the measured response to XSim response, most likely the 2 curves will be slightly offset and not perfectly overlapping.
It’s a really good article. I got a lot of help in understanding basic measuring and how to make a passive network.
The thing is .. your explanation is based on the Xim where you can put only the delay value of the drivers. But nowadays, I’m using VituixCad which I think is much easier and versatile.
However, with the VituixCad, I canput the X, Y, values of the drivers(the horizontal and vertical position of the drivers). What I wonder is, when I put the delay values of the drivers, should I put the X, Y values of the drivers as well? or should I leave them as 0 because the delay values must have been reflected in the driven data of the measurement?
Sorry, I can’t help you there. I’m not familiar with VituixCad.
OK.. I think you would come to love VituixCad if you try. It’s pretty easy to learn, and a lot of functions are included even though it is a freeware. VituixCad has a time-align and other convenient functions like volume calculations and so on. But it seems it shows a different simulation result from Xim, so, I’m trying to figure out why it’s like that. Anyway, your articles and youtube videos are a lot of help in learning speaker DIY, especially for me, as a Korean, there are no resources like yours. Only some videos about clippel which is way too much expensive to buy as a DIYer. I hope you keep doing this and hoping to watch more youtube videos.
Biggest thanks with all this info.
I bought a diy speakers and what started with fabricating a new box for the elements, now kind of got out from me hands.
I now need to get new speaker elements, same brand three way speaker with two 8″ woofers 4ohms in series, a 8ohm 5″ mid and a 8 ohm horn tweeter.
I have usb mic and self made rig for impedance measurement.
First crossover ever with three way speaker, no worries, I have time and patient. I use both xsim and vituixcad (actually that means fu**edupCAD)
I will do the measurements all of them but impedance, outside.
But do I measure woofers separate or together?
Common sense here would say together…
You measure them together and in Xsim you consider the bass section as a single speaker.
Another question
I was brushing up on your article once again. What I wonder is, when we get the aliment offset of threeway drivers, you said that is a relative one. Would I be able to use the offset data even when I simulate the crossover with the data measured with the microphone on the tweeter axis? I wonder why you didn’t put the microphone on the tweeter axis while most measurements of the speakers(including 3 way) ore measured 1m away on the tweeter axis.
On 3-way design you can place the microphone on tweeter axis, on midrange axis or between the tweeter and midrange. Any of them are correct, and indeed most popular is on tweeter axis.
Marius,
Great article again. Took your all classes on Udemy btw. Learned a lot from you, appriciate that!
First question is why are you setting the ref time -0.2ms? I always use the default “0” value
Second question is, during designing a 3 way speaker, it’s very very tough to match the 3 way combined measurements with the xsim modeled one. Once I firstly match tweeter & midbass and then midbass and woofer (spontaneously). Do you think it’ll work?
Last question: It’s challenging for me to find the first reflection in speakers that have 2 woofers. I didn’t choose to measure them one by one. Is there a way to determine first reflections while 2 woofers are connected?
All the best for you!
Cheers
Yeah, this was some time ago, and I was used to ARTA rather then REW. In REW you can leave the reference to 0 and set the left window to -0.2 or -0.3, something like that. There is some information in the impulse response right before the initial spike and it’s good to include it in the window.
When you measure a 3 way speaker, or a dual woofer or a speaker which has a port on the lower part of the box it’s difficult to get an anechoic response indoors. The bass drivers will be quite close to the floor and reflections will come too early and you won’t be able to tell where the first reflection is, or that the reflection-free window might be too small. In this case, you either find a larger room, to place the speaker higher-up from the floor (but not too close to the ceiling) or you simply accept that this type of measurement is not accurate and try to match the curves as best you can.