Notice the coherence trace (red line) remaining near the top of the display up until around 700 Hz. It then shows poorer coherence in that gap. We see another trend in poorer coherence in the 4.5 kHz region. Hmmmm, we have poorer coherence in those areas showing a dip in the transfer function. Think about it. What could cause poorer coherence? If it were in the crossover region, the phase shift caused by the crossover could be the culprit. However, it is not likely the HF is crossed over at either of these frequencies. It is very likely we are seeing some cancellation caused by reflections, and if that is the case these are non-equalizable events. Note: in general equalization practice we will never boost, only cut.

Are we ready to equalize? Not yet. Let’s move the microphone and attempt to get a better measurement. First, save this trace by selecting reference register “A” at the bottom of the display and pushing the Capture button (“Capt”) to store the active trace in register A1. You may label the reference by clicking in the comment box to the left of the register buttons.

Figure 7 displays the live trace vs. the stored reference trace. Notice the microphone has been moved only 1.06 feet and yet the transfer functions differ significantly in the 700 Hz area, and the coherence is now much better than before in this area. Why? The author’s guess is again cancellation caused by a reflection. We can see these reflections in the impulse response, but this is beyond the scope of this article.

Be aware of the environment of the room/system under test. Reflections can cause poor coherence and contaminate your measurements. When choosing your measurement location try to set up in the near field.

Equalization

Let us return to our current measurement. Hide the stored reference trace in register A1 by pressing “hide” on the reference trace bar. See figure 8.

Look for trends. Our transfer function is displayed with 5 point smoothing, so we can see some trends here without worrying about minute variations.

The jagged hump centered near 170 Hz looks like a good candidate. Use a single filter to even the response of this area. The two peaky components of this hump are probably too close together to equalize separately, and the peaks differ by only about 2 dB.

There are a couple of other areas in which we can see trends. In order to establish a “baseline”, use the “dB +/-“ control to raise the transfer function 1 dB on the trace. By doing this we have caused the area from 250 Hz up to about 1 kHz to reside just above the 0 dB line and “exposed” more of the non-linearities at 1.5 kHz and about 12.75 kHz.
This allows us to equalize down to the 0 dB line. See Figure 9.

This technique requires some subjective evaluation on the user’s part. Depending upon the transfer function, one may be required to “raise” the transfer function considerably more in order to establish this baseline.

In a multi-way PA system it is good practice to measure each band, store each as a reference trace and use this to adjust individual band levels before attempting equalization.

As a practical example of what can happen when excessive reflections contaminate you measurement, let us place a magazine just beneath the measurement microphone. Let us capture the current trace to register B1, place the magazine beneath the microphone and compare the traces. See Figure 10. Note the dips near 3.3 kHz and 8.5 kHz and the associated poorer coherence.

All that has changed is the addition of a reflective surface. When equalizing a system consider getting the measurement microphone within the near field as much as possible/practical.

A useful technique for getting better measurements is to place the microphone on the floor or console lid. Coil the cable a couple of times and place it under the microphone barrel aiming the capsule down toward the reflective surface. The “business end” of the microphone should be very close to the reflective surface. This technique emulates pressure zone technology and can be very useful for reducing reflections in the measurement.

In general, try to use broad gentle filters in order to address trends in response. Narrow filters can become “misplaced” if conditions change, such as the addition of a large number of water bags to the room/system under test (i.e. the audience enters the building).

Summary

1. ALWAYS LET YOUR EARS DECIDE
2. Be aware you will probably never get the system under test “flat” on the display.
3. Learn to tell whether or not your measurement is good.
4. Look for trends and try to address them, rather than using a surgical approach.
5. Never boost – always cut.
6. Try to measure in the near field.

In the next and final installment we will perform complimentary equalization on this loudspeaker by overlaying a stored reference trace of the loudspeaker under test with the transfer function of the equalizer’s input and output. By inverting the EQ transfer function we can set filters which match the stored reference trace, which should accurately remedy the non-linearities present in the loudspeaker’s response.