Quasi anechoic loudspeaker measurements Part 2
How to make loudspeaker measurements using ARTA ?
This is the second part of our loudspeaker measurements tutorial. If you haven’t already, go and read the first part of the article, where we set everything up and properly calibrate the equipment. Now that everything is all set up, we will go directly to the real deal loudspeaker measurements. We are measuring the frequency response of the M-Audio BX5 studio monitor.
Near-field response
We shall start with the near-field measurement. Because our speaker is bass-reflex, we need to take 2 readings : one for the speaker and the other for the port.
- Place the microphone as close as possible to the speaker cone.
- The microphone should be pointing at the center of the diaphragm.
- The distance where you place the microphone should be no more than 11% of the effective radius of the speaker.
- The effective diameter (from the apex of the surround, to the apex of the surround, at the other end) of our speaker is 10.5 cm.
- This means that the radius is 5.25 cm, therefore the distance you place the microphone should be no more than 11% * 5.25 = 0.58 cm.
The size of the speaker also dictates how high in frequency the response measurement is valid.
fmax = 10950 / D
fmax is the maximum frequency and D is the diameter in centimeters. So for our speaker, the upper limit is 10950 / 10.5 = 1043 Hz. So our near-field loudspeaker measurement will generate a valid frequency response up until 1043 Hz. After that point, till 20 kHz, the response doesn’t reflect reality. This formula works well with infinite baffle and sealed enclosures. The maximum valid frequency is a much lower number for finite baffle (almost all scenarios). Even lower, when measuring port response.
That’s why in the first part I told you to aim for 100 Hz maximum, with the near-field loudspeaker measurements. Always try to find an adequate room, to go as low as possible with the free-field gated loudspeaker measurements. That way you don’t have to go very high with the near-field loudspeaker measurements and the overall curve will be more accurate. If not, it is accepted to splice the response higher.
ARTA instructions
Measure speaker nearfield response
Now, after placing the microphone (Dayton audio EMM-6 (Amazon affiliate link) in our case) correctly, go ahead and press the Play button.
Go to the sweep tab, select your preferred input channel (left in our case), then select an appropriate output level so you don’t get any clipping. After you made all of your fine tuning, make sure you keep them the same for all of the measurements to come. Same goes for the knobs on you equipment, leave them in the same position for the whole experiment. Go ahead and press Record.
Your response curve should look something like the one above. One thing to mention, that is very interesting when you measure the speaker, using near-field method, is that the dip in response, where I placed the cursor (yellow line), is the resonant frequency of the box. So fb = 64 Hz. This is more accurate than the one you get from the impedance chart, because it doesn’t get corrupted by voice coil inductance. The response dips, because at resonance, the port violently takes over and the speaker barely moves, therefore the reduced output of the speaker shows on the graph. Make sure you save your response curve : File -> Save as.
How to design loudspeakers - video courses
After that, go ahead and click Overlay -> Set as overlay and close the window. Then, place the microphone in the middle of the port, flush with the baffle.
Measure port nearfield response
Press the Play button again, go to Sweep and press Record. Go back to Smoothed frequency response.
And your graph should look something like this :
Scale port response
Now we must adjust the level of the port, because it has a different size than that of the speaker. To do so, we must first calculate the area of the speaker and the area of the port. Since both are circular : Area = πR2 .
- Aspeaker = 3.14 * 5.252 = 86.59 .
- Aport = 3.14 * 22 = 12.56 .
To scale your port response : Go to Edit -> Scale amplitude and enter the following PIR Scaling : (Aport / Aspeaker)1/2 . It should look something like this :
Then, your graph should look something like this :
You should save your response, maybe you will use it later. Now we want to set this as overlay also. When we combine the 2 responses, it is nice to see all 3 curves and how they add up. This is totally optional. You can delete the overlays if you want.
Sum up speaker and port responses
Then go to File -> Load and Sum and select the file which corresponds to the near-field measurement of the speaker. You should have 3 response curves on your frequency response plot. After you analyzed them you can go ahead and delete the overlays. You should be left with the combined response of the port + speaker, which looks like this :
Scale near field response to far field
Now that we have combined the 2 curves, all we have to do, is scale to far field. When we will do the far field gated measurement, the microphone will be placed at 1 meter away of the speaker. This means that the magnitude of the far-field measurement will be significantly lower than our near-field magnitude. This means we have to scale the near-field response curve accordingly, so we can splice it later with the far-field response.
Go to Edit -> Scale amplitude and enter the value : R / (2 * d) , where R is the radius of the speaker and d is the distance between the speaker and the microphone. For our example it looks like this :
And after the scaling is done, the response was lowered by about -31 db :
Apply box diffraction compensation
At this point, you might want to save your response curve. Next, we will apply a low frequency compensation. Because the microphone is so close to the speaker, it simulates an infinite baffle. On an infinite baffle, all the frequencies gain a 6 db boost. In reality, the baffle is finite and only some of the high frequencies get the boost. We need to compensate for this. Go to your frequency response window click Edit – > LF box diffraction and enter the dimensions of your baffle. This will decrease the level of the low frequencies, for the reason we just discussed.
Then go to Overlay -> Set as overlay , and this is how the graph looks like :
Now we are ready to take the gated far-field response.
Far-field response
Go ahead an place the microphone 1 meter away from the speaker we are testing. Try to place the speaker in the middle of the room, so you are equally away of any obstacles in any direction. The room I am making the measurement is not the most desirable, but I will use it anyway, just to show how it is done.
The speaker needs to be places on a stand. The height of the stand should be the distance between the ceiling and floor divided by 2. The microphone is also placed on a stand, directly in front of the speaker, 1 meter away from it. Just like with the near-field measurement, press the Play button and press Record with the exact same settings as before. Before we head over to the frequency response window, we need to check out the impulse. After a few zoom-ins, the impulse should look like this :
Gated loudspeaker measurements
Now we have to select the time-span where there are no room reflections, which is between the first impulse and spikes we get from the room reflections. Press right click right before the big impulse, to place a marker for the start position. Then, press left click right before the first room reflections.
This will give us a frequency response that ignores all room reflections. Go ahead and go to the frequency response window, which should look something like this :
As you can see, from the bottom of the graph, there is a yellow bar from 20 Hz to 220 Hz. This shows that the chart from 220 Hz downward is not valid, because we made a gated measurement. If the room was larger, and the first room reflections would have reached the microphone later, the yellow bar would have been shorter.
- The near-field measurement is valid 20 Hz – 1000 Hz.
- The gated far field measurement is valid 220 Hz – 20 kHz.
This means we need to splice the 2 curves somewhere between 220 Hz and 1000 Hz, ideally in a spot where the graphs intersect. This happens at 708 Hz mark. We are going to place the cursor at 708 Hz , click Edit -> Merge overlay above cursor , then go for Overlay -> Delete all overlays , and it should look something like this :
Please note that ARTA sees this graph as an overlay, not the original frequency response curve. So make sure you save the graph (print screen maybe). If you exit the frequency response window, the graph will revert to the original far-field gated measurement.
Congratulations !
You have successfully made a complete, full range, frequency response measurement. Also, you now know the real life frequency response of the M-Audio BX5. Making loudspeaker measurements is not difficult. With the appropriate room and few experiments for practice, you will be good to go in no time.
References
- Image source : link.
Learn loudspeaker design from scratch
17 comments
Very nice article! I was measuring my speakers and I found after 1khz magnitude becomes strange. Why are you using 10950?
Thanks for sharing.
Ps. I’m using REW
My question is can I generate frd file with USB interface microphone and without audio interface soundcard
If it is possible which one is good like in your article what method is you used or direct USB mic interface without preamp interface
You can use the miniDSP UMIK-1 which is a USB measurement microphone and you can skip the audio interface. The process is quite similar, however you have to install ASIO4ALL in that case. The Focusrite scarlet audio interface uses ASIO drivers but for the UMIK-1 you need to setup this audio plugin. I talk about both variants in detail in my acoustics 201 course. If you want to learn more about it, you can find it here
One more question please help me . For passive speaker we have to send test signal to speaker can I directly sent earphone Jack to amplifier to speaker or any interface pre amp to amp to speaker or directly earphone jack to speaker or interface to speaker directly (without amp).. In india no measurement tools are selling in market , my bro is comming from USA next week so i am planning to bring from there i have no time to see your video right now but definitely I will see your lecture in comming days
Thank you
You need an amplifier to make the measurement. Connecting directly to headphone jack, there is not enough voltage to make an accurate measurement.
Thanks a lot for this great article, and also many others. I do have a question though. In part 1 you mention that the nearfield measurement should be considered accurate upto about 100Hz. But you splice the near and far field responses at around 700Hz. Doesn’t this make the range between 100 … 700Hz less reliable?
It’s better to splice as low as possible. I was probably thinking of a large subwoofer, as the larger the speaker, the lower this point has to be. It really depends on the size of the speaker. For the speaker in this example I would recommend splicing 250 – 600 Hz. However, in this example I splice at 700 Hz because that’s where the curves overlapped. To be honest, now I use a different technique with Room EQ Wizard.
When we measure a 2 way speaker, what the height of the microphone should be? You are saying it sould be 1 meter away from the speaker, but not telling the height of the speaker. Should it be on tweeter axis? or between the tweeter and midwoofer?
Normally you place the mic at tweeter level. Sometimes, when I measure tall floorstanders, and the tweeter is closer to the ceiling, I place the microphone just as you mention. A bit lower, between the tweeter and midrange, so I can delay the earliest reflection.
Thanks a lot
Not a very flat frequency response for a studio monitor..
Hello Marius, I bought your lesson about measurement from REW, is it same result with ARTA here?
and also for sealed enclosure, is it same procedure? and do we need to apply baffle step correction too?
Yep, it’s the same procedure. REW doesn’t have baffle step correction option, but I included an excel spreadsheet in the course which does the same thing.
thanks for your reply, I sent you an e-mail about problem in excel spreadsheet from udemy course, please kindly check it, and I hope for your reply and solution.
What’s Up, Marius !!
Excellent content, as usual. But I’m a bit confused.
To load frequency responces of the individual drivers into XSim to model the crossover, I assume we should do
1m far field measure of the Tweeter
and then for the Mid/woofer (on a 2-way) do we just do a Far field @1mor should we do the combined Far + Near field of Driver And Port with baffle correction using just the Mid/woofer driver on its own and use that as the Woofer responce in XSim?
You do all this at 2.83V for 1W @1metre?
Keep up the great work.
Cheers Steve
Placing the mic at 1 meter seems to be the norm. But there is nothing wrong if you place it a bit closer or a bit further away. You will most likely get identical results. In terms of output voltage of the amp, I just eye-ball it. When you do the measurement if it’s too loud, turn it down, if it’s too low, turn it up. The exact voltage doesn’t really matter, just make sure you get a good signal to noise ratio and you don’t clip the mic.
oops i just found this in the three way crossover deesign post:
“Now that I have refreshed your memory, all you need to do is to create your FRD and ZMA files for each driver and calculate the acoustic offset between the drivers. When you are doing the frequency measurements, just do a gated far field measurement. You don’t need to do nearfield measurements. Why? Because it introduces phase errors when splicing nearfield with far field response. In addition, when you are designing a crossover, you don’t manipulate the lowest octaves. You just let the bass driver naturally roll off. It’s nice to have the whole response on the chart, but you can do it just as well without. As a result, we don’t need that portion of the frequency response (20 – 200 Hz), because we don’t alter it in any way in our 3 way passive crossover design”