How to make full range frequency response measurements without a true anechoic chamber ?

You are probably asking yourself, is that even possible, and what is this quasi anechoic loudspeaker measurements ? Well it actually is. If you have access to a true anechoic chamber, you can get accurate measurements with little effort. When you are making quasi anechoic loudspeaker measurements, you will have to get creative, make additional measurements and with a little bit of math, you will get results with high accuracy.

This article will explain how to do this, with real-world examples, using ARTA software to generate the response curves. To limit the size of this article, we are going to split it into 2 parts. In the first part, we will cover what gear is necessary, how to set it up, adjustments and calibration that needs to be made prior to the actual measurement. When we will get to the second part, we will cover how to actually measure, and the steps are needed to make quasi anechoic loudspeaker measurements, for in the end, to get a full range frequency response curve.

In the following explanation, I could be using terms that you might not be familiar with, regarding measurement techniques. If you feel that is the case, please read this article before continuing.

Techniques for quasi anechoic loudspeaker measurements

We will use 2 techniques, to get an accurate full range response. We will make a gated free-field measurement to obtain the mid and high frequency response, and a near-field measurement to get the low frequency response. After a bit of number crunching and software sorcery, we combine the 2 response curves, and get the full frequency response curve.

The gated free-field measurement 

The loudspeaker is placed on a stand and the microphone is placed directly in front of it, 1 meter away. The room needs to have certain qualities :

  • It needs to be quiet. This is done by either making a moderate sound insulation, or by making the measurements at night.
  • It needs to be sufficiently large. The larger the room, the lower in frequency we can go with the measurement.

How this works ? We are sending an impulse through the speaker, and take a reading using the ARTA software. We can determine after how many milliseconds the first sound wave reached the microphone, and after how many milliseconds the 2nd wave have reached the microphone. The second wave is sound reflected from the nearest boundary (floor or ceiling etc), and it will be picked up by the microphone few milliseconds later.

Between these two, there is a time window (that’s why it is called gated), where the microphone picks up only the sound generated by the loudspeaker, therefore ignoring room reflections. This is all fine and dandy, but there is a catch. Depending on the size of the time window, the frequency response will be accurate until a certain point. The larger the distance, between the loudspeaker and the first boundary, the lower in frequency you can plot the response curve.

Generally, you would want to make the gated free-field measurement down to around 100 Hz. This would mean there should be no obstacles for at least 1.5 meters in all directions. This means the room needs to be at least 3 meters high, to satisfy this condition. By using this technique we can get an accurate response from 100 Hz to 20 kHz , assuming we have an appropriate room. For the lower octaves, we will use the near-field method.

The near-field measurement

This type of measurement needs the microphone to be as close as possible to the driver. Depending on the type of loudspeaker, we could need to do more than one measurement, and need to do some extra calculations to make the appropriate scaling.

Let’s make some scenarios :

  • If you got a sealed enclosure with only one speaker, you will make only one measurement.
  • In case it has multiple speakers, you will need to measure each individual speaker, and add up the response curves. If the speakers are of the same type and you know they receive the exact same signal (not 2 ½ way or 3 ½ way system), you can measure only one speaker, and scale the response by +6 db for each additional speaker.
  • For a bass reflex loudspeaker, you will have to measure the speaker and the port individually. Because the port will probably have a different size than the size of the speaker, its response needs to be scaled accordingly (this is the math I’ve been talking about). After the response has been scaled, you can add up the response of the speaker with the response of the port.
  • If you got multiple ports, you can measure only one and scale the response by +6 db for each additional port.
  • Same rules as with the sealed enclosure applies to bass reflex, if there are more than one speaker.

Microphone positioning for near-field measurement

When you are making the measurement, the microphone needs to be as close as possible to the speaker. Take into consideration that the speaker will move slightly when making the actual quasi anechoic loudspeaker measurements, so don’t get it too close, but no further than 1 cm. The microphone will be pointing directly in the middle of the speaker diaphragm. As for port measurement,  place the microphone in the middle and flush with the port (don’t stick it in). Because the microphone is so close, it only measures the speaker. No diffraction, no baffle gains, no room reflections etc. However, this type of measurement has certain limitations when it comes to going higher in the frequency spectrum, but it should give very accurate results up to 100 Hz or slightly higher.

The near-field measurement is a little bit more complicated, but we shall see all the exact steps we need to take, in the 2nd part of this article, where we do some actual quasi anechoic loudspeaker measurements.

After we have made the two measurements, we scale the near-field measurement. Since the microphone is so close to the speaker, the magnitude will be higher than the free-field measurement, where the microphone is 1 meter away. After the near-field curve was scaled by the appropriate amount of db, we splice the 2 response curves at the appropriate frequency, and we get a full range frequency response curve for the loudspeaker under testing.

Equipment used for quasi anechoic loudspeaker measurements

We will enumerate the equipment needed and what gear we will be using for our experiment :

  • A measurement microphone. We will be using a Dayton Audio EMM-6. This is an affordable mic, with included calibration file, to compensate for the non-linearities in the frequency response. Simply go to their website, type in your serial number, and download the file. We will use this file later.Dayton microphone
  • A microphone preamplifier. We will use the Focusrite Scarlet 2i2. It has two inputs / outputs, and most important, it can provide phantom power for the mic. Make sure that your preamp has this option. Another advantage for our Focusrite preamp is that it draws power from USB. This means you can make in-car measurements with a laptop and this little guy, without connecting to a wall socket.
  • An external sound card. Our Focusrite Scarlet 2i2 doubles up as a sound card, so no extra gear here. But you can use any sound card you wish. Avoid on-board solutions as they can have bad frequency response and / or total harmonic distortion.Focusrite front and back
  • A microphone stand.
  • A speaker stand.
  • Various cables depending on you gear.
  • Device for testing. We will be testing the M-Audio BX5 studio monitors. Because they are active loudspeakers, there is no need for a separate amplifier.M-Audio BX5 front and back

Calibration

Before we start our quasi anechoic loudspeaker measurements, we need to do some calibration and check if everything is correctly set up and working properly.

Windows settings

First, let’s start by doing some adjustments in Windows. My operating system is Windows 10, and it should be similar to Windows 8 / 8.1.  Go to the speaker looking icon, near the clock and right click on it. Then click recording devices.

recording devices

After that, select the device you will be using for recording. In our case, we will be using the line in of our Focusrite Scarlet. Right click the device and click properties.

line-in

Then go to the levels tab and take the slider to a very low value, 1 in our case.

levels

Click OK, and then do the same for the playback devices. Right click the speaker icon near the clock, click playback devices. Then select your output device. In our case, it is the line out of our Focusrite Scarlet. Right click on it and select properties.

line out

Then go to the levels tab and select the maximum value. This is probably set by default as maximum, but it is good to check.

levels output

ARTA Settings

Now the windows adjustments are complete and we shall start ARTA, and do some more fiddling over there. When you start ARTA, your interface should look like this :

quasi anechoic loudspeaker measurements using arta

Click on the Audio devices icon.

audio devices

Now select your driver and input and output devices. These depend on the equipment that you have. Since I selected the Focusrite driver, I am left with no choice on the input and output sections, and they are chosen for me : the 1/2 channels which are available as inputs and outputs. Most of the time, this should show up like the devices we adjusted the volume in windows : the recording and playback devices. I’m sure you will get it right anyway.

devices

After that, we will have to check if the sound card has a flat response. To do that, we must do a loop-back. This means I will connect a cable from the left output to the left input. For my particular audio interface I am using an TRS – XLR cable, but you should use one with whatever connectors you have on your interface. Here is how it should look like :

loopback

Equipment sanity check

Now click on the Smooth FR button

frequency measurement

Then click the Record button and go for the Sweep tab. The preferred input channel is left, because that is the channel I’m using. You will then click the record button. If you find that you are getting peak signals (the cursor below starts to get yellow / red), adjust the output volume (in my case is -20 db, the lowest), and make the measurement again. If you are still getting peak signals, turn the knob of your audio interface to a lower value (mine is half way).

sweep

After the sweep is over you should get a response like this :

quasi anechoic loudspeaker measurements

The response should be ruler-flat. If you get something that is almost flat, with little dips and peaks, it means that your sound card is of poor quality. If you get a response that is nowhere near flat and all over the place, it means that you have selected the wrong audio devices, or have selected the wrong preferred input channel, mismatched connections, the wires are bad, your interface is bad etc. In other words, if your response is not flat, there is no point in continuing with the quasi anechoic loudspeaker measurements. When you are doing this measurement, make sure the phantom power (48 V) on your audio interface is not on, and no microphone calibration files are in use.

After that, we need to load the microphone calibration file into ARTA. The file downloaded from the Dayton website has txt extension. Simply rename the file and change the extension from txt to mic. Now press the FR compensation button :

mic calibration

Press load and select the compensation file. Now press the Use frequency response compensation button and then click OK.

FR compensation

End of Part 1

In this part, we covered up what techniques we will use for our quasi anechoic loudspeaker measurements, the equipment we will use, and the calibration we need to do before we take any measurements. Now that everything is properly setup, we can go ahead and start making some measurements. In the 2nd part of this article, we will cover, step by step, how to make a complete full range frequency response measurement, using ARTA software.


References

  1. Image source : link.