Behind the Scenes at Secrets - Number 2 - September, 2008

This time, I want to discuss one of the more important tasks in bench testing that is often overlooked: calibration of the test microphone.

Many consumers use the old reliable Radio Shack SPL meter for testing volume levels in their home theaters. What's nice about the meter is that it is accurate at 1 kHz, because this is the frequency that it is calibrated at. When the meters are constructed, the last step before packaging is to set the meter to read accurately at 1 kHz. A 1 kHz sine wave is created with a microphone calibrator, and the meter's nose is inserted into the calibrator. A small screw is turned until the meter reads the SLP that the generator is creating (probably around 100 dB).

The problem is that the meter is not accurate at other frequencies. There are some compensation charts out there that tell you how much to add or subtract from the meter's reading at different frequencies to get the "True" reading. This may work in a rough way, but not in a precise way, because each meter has variations in its sensitivity throughout the audible band.

What we do at Secrets is use a very high quality measurement microphone. We have several Earthworks M30BX mics (about $700 each) and a B&K (about $1,400). Each microphone comes with its own frequency response chart that was generated at the factory. Using that chart, we construct a compensation file that adds or subtracts a certain amount from the response when testing a speaker. The compensation is usually on the order of 1 or 2 dB at various frequencies up to 30 kHz. Once that is completed, we can say that the test system using the calibrated microphone is accurate to about ± 0.5 dB.

Here is a factory-generated frequency response for one of the Earthworks microphones that we use.

Frequency Response

In creating the compensation file (*.mic), we follow the response, so at 10 kHz it would say + 2.0 dB. We put in response values all along the curve rather than just one here and there, perhaps 20 values. Once the compensation file is complete, it is activated in the software during measurement of a speaker's performance.

Here is one of the programs we use, called SpectraLab, with the microphone compensation window open. You can see near the bottom that the microphone file for mic 9756 has been activated.

Besides the frequency response, it is critical that the measurement software know what millivolt output from the microphone corresponds to what SPL. As an example, let's look at the interface in the Audio Precision SYS 2722.

Utilizing a special microphone calibrator, which generates a 1 kHz sine wave at a choice of 94 dB or 104 dB, the tip of the microphone is inserted into the calibrator. A photo of the calibrator and microphone is shown below. In use, the tip of the microphone would be inside the opening of the calibrator (it abuts against a small ridge that keeps it at the proper distance inside the calibrator), and an XLR cable would be connected to the other end of the microphone and the Audio Precision.

To calibrate the microphone's output to read 104 dB, I open a 1 kHz SPL test window, turn on the calibrator so the 104 dB sine wave is being fed into the tip of the microphone, and run a script that sets the output of the microphone to be 104 dB as read by the Audio Precision.

Mic & Calibrator
Caption 1

Here is a screen shot of the test window.

The blue arrow shows where the Audio Precision has recorded the output of the microphone and set it to indicate 104 dB as seen in the 1 kHz spectrum on the left. The red arrow shows the 1 kHz peak reading 104 dB on the Y axis (blue arrow).

We try to keep the room at approximately the same temperature from test to test because a cold microphone puts out slightly less voltage than a warm one. It has not turned out to be critical, but we like to eliminate every variable we can to maintain consistency. The Earthworks microphones use a 6 volt battery to provide bias, so when we change the battery, we recalibrate the microphone. Again, the difference in output between batteries is almost negligible, but it's just one more variable that we can eliminate by doing it when the battery is replaced.

In the next newsletter, we will discuss the graphs and their interpretation.