After getting used to the basics of in-room measurement with the app note UMIK-1 setup with REW, you may want to perform some more advanced measurements. In this app note, we will explain some techniques that will help with your active loudspeaker design. We will assume that you're already up and running with the UMIK-1 and REW.

What you will need [Top]

  1. A miniDSP UMIK-1.
  2. Room EQ Wizard (REW). Be sure to download the latest version from the Downloads Area for UMIK-1 support.
  3. A microphone stand with boom arm is recommended for loudspeaker measurements.

Gated measurements [Top]

An acoustic measurement with the microphone at the listening position is generally referred to as an in-room measurement. You are measuring the combined effect of the loudspeakers (and/or subwoofers) and the acoustics of the listening room. That is fine for assessing room acoustics, but when doing loudspeaker design, you want to measure the loudspeaker only, without any effects from the room. This figure illustrates the problem:

Illustrating reflections when measuring a loudspeaker

The sound from the speaker arrives at the microphone first, followed by a reflection from the floor, then from the ceiling, and somewhere along the way from the side walls. These reflections continue to bounce around the room for up to several hundred milliseconds. That is what we hear, but it makes it hard to get an accurate measurement of just the loudspeaker! To illustrate, here's a measurement of a 3-inch full-range driver taken with the UMIK-1 and REW in a typical domestic environment:

Raw measurement response

Obviously, without a little work, this is fairly hard to use! One way to address the problem is to use an anechoic chamber, which is a very large room with lots of acoustic absorption in it so there are almost no reflections. Most of us don't have one of those. Another technique is to use a gated measurement, where the reflections are simply removed from the measurement. We can do this in REW.

First, let's understand what we are looking for. The measurement was taken with the driver mid-way between the floor and ceiling at 120 cm from each, and the microphone level with the driver and 1 m distant. According to the floor-bounce calculator at mehlau.net, we would expect the first reflection to arrive at the microphone 4.65 ms after the direct signal.

Let's see if we can see this first reflection in our example measurement. In the main REW window, click on the "Impulse" button, and then hover the mouse over the display and select the "%FS" dropdown (you may need to adjust the graph limits to get a good view of the impulse response):

Impulse
response showing first reflection

Yep! There it is... To eliminate this reflection (and all subsequent reflections), we will gate the signal at 4.4 ms as shown by the marker. Go to Tools→IR Windows, set the "Right Window" to 4.4 ms, and then click Apply Windows. The "window" is the region of time that REW uses to calculate the frequency response. The display will update to show the new window, which reduces the signal so that it is zero by 4.4 ms (in blue):

Impulse response window parameters

Clicking back on the SPL & Phase button will now show the gated measurement — that is, with the reflections removed from the frequency response calculation. Here it is (shown on the Overlay screen) in green:

Gated vs smoothed responses

Another technique that can be used to make the raw measurement graph more usable is smoothing. This function is accessed from the Graph menu when viewing the main window, or from the Control dialog on the Overlays window. For example, the graph above shows, in red, the same measurement with 1/12 octave smoothing (and without gating). While smoothing can sometimes be used to give a similar effect as gating, this example shows that you must be careful: the peaks at 14 and 17 kHz have been smoothed over and now are not accurate.

Gating does have limitations. For one thing, the frequency resolution of the measurement is reduced—the IR Windows dialog shows this resolution. In addition, any frequency below about 1 / (window length) cannot be represented. In the case of the 4.4 ms window, the lowest frequency that can be represented correctly is 1/0.0044 or about 230 Hz - anything on the graph below this frequency must be ignored. So this technique works best at high frequencies and can't be used for low-frequency measurements at all.

Viewing phase [Top]

The phase of the acoustic output from a driver is useful to know in some cases. For example, when looking for the causes of poor crossover integration, or when performing overall phase corrections to obtain a linear-phase loudspeaker (see the rePhase FIR tool app note). To view the phase of a measurement, click on the "Phase" checkbox in the main REW window. (Or alternatively, go to the Overlays window and click on the Phasebutton.)

Once there, use the Limits and Controls dialogs to adjust the graph. Typically, the Unwrap Phase button will be used to remove discontinuities in the phase display, and the +360 and -360 buttons can be used to put the display into a usable range. Here is the result for the measurement above (the phase scale is over on the right):

Phase example graph

Note: you may not get a clean phase reading when the default 500 ms window is used. If so, gate the measurement as described above to remove the reflections, and the display should now show the correct phase of the driver or loudspeaker being measured.

Low-frequency measurement [Top]

As noted above, gated measurements are only relevant at higher frequencies, typically above a few hundred Hz. Even with in-room measurements, you may be able to design and equalize your speaker by judicious use of smoothing and taking care to mentally separate room effects from the loudspeaker. For better accuracy, though, here are some techniques you can use:

Nearfield measurement. The microphone is moved close to the driver being measured so that the relative level of room sound is reduced. With the mic right near the cone, this can be effective for subwoofers. For full-range measurement, the response will change with distance, and so the mic typically needs to be a few times the cone diameter away. Even then, the effect of the cabinet on the response may not be fully visible.

Outdoor measurement. Room modes will be eliminated if measurements are taken outdoors. It is still important to minimize reflections though. This can be done by moving the speaker and microphone well away from any walls or other large objects and raising them both as far off the ground as possible.

Ground plane measurement. This is usually done outdoors as well, to avoid room modes. The microphone is placed right on the ground so that there is no reflection from the ground at all. In practice, the mic capsule is still a small distance from the ground, so this technique can be used up to perhaps a few kHz.

More info [Top]

These app notes will help you with your active speaker project.

Don't forget to ask on the miniDSP forum if you have questions!


 

Related Products - Acoustic measurement