In-situ free-field system calibration

Ssource measurement

Before the publication of the ISO 3383-3 standard for measurements in open plan offices, the system calibration was only used for the measurement of the Strength (G) parameter, and related parameters such as the early and late strength (G80 and GL), and the lateral sound levels (GEL and LG). Since the introduction of ISO 3382-3, the use of system calibrations has increased significantly.

Performing an accurate system calibration can be a challenging task. For the best results, an anechoic room or reverberation chamber is required, and even there it is difficult to get results with an error margin close to 1 dB. See [2]
Not everyone has easy access to an anechoic room or a reverberation chamber, and therefore the in-situ free-field system calibration method is often used. The following are three important concerns with this type of calibration.

Non-uniform directivity pattern of the omni-directional source
Typical dodecahedron shaped omnidirectional sources used in acoustic measurements have a non-uniform directivity pattern, particularly in the higher frequencies. The example below shows the directivity of an OmniPower Type 4292-L in the 250 and 4000 Hz octave bands. Placing the microphone directly in front of one of the speakers in the cabinet will result in an overestimation of the source's sound power level. Research [1] has shown that this can be solved by rotating the source over 5, 7 or 9 equal-angle steps of 360°/n, while keeping the microphone position fixed.

Source directivity

Windowing to remove reflections
Contrary to measurements in an anechoic room, in-situ measurements will contain reflections from the floor, ceiling, walls and possibly other objects. The floor reflection is often the first reflection to arrive at the microphone, and the delay of this reflection from the direct sound determines the length of the window needed to remove the reflections from the impulse response. In the impulse response displayed below, the first floor reflection can be seen at 5.5 ms after the start of the impulse response. A window size of 5 ms will remove this and all subsequent reflections from the impulse response. Note that this means frequencies below 200 Hz will have very poor results.

Impulse response window

Extending the time window, and thereby improving the low frequency results is possible by increasing the height of the source and the microphone from the floor. It may also be helpful to place highly absorbing material on the floor area between the source and the microphone.

Source-receiver distance
The omni-directional sound source used in measurements is in fact a rather poor model of an ideal point source. Earlier, the far from ideal directivity pattern was discussed. Another deviation from the ideal is the physical size of the source. This can cause some confusion as to what the actual source-receiver distance is in a measurement.

Source-receiver distance

With an ideal point source centered on axis C in the drawing above, the source-receiver distance would be obvious. With a real source we also have the distance to the acoustical center of the source (A) as seen by the microphone. The exact position of the (measured) acoustical center is dependent on the exact orientation of the source w.r.t. the microphone.

In the configuration displayed above, the start of the impulse response as measured by DIRAC will correspond to the acoustical center (A) of the source. For an OmniPower Type 4292-L this will typically be at 84 cm from the microphone, or 16 cm in front of the geometric center. This difference between the source-receiver distance as measured by DIRAC and the distance between the geometric center and the microphone, results in a corresponding misrepresentation of the SPL at a given distance from the source (center). The graph below depicts the error in dB SPL for a 15 cm difference between acoustical and geometric center as a function of the source-receiver distance.

SPL error versus distance

For measurements close to the source the error can be significant. At a typical distance of 1 m the error is more than 1 dB. For distances over 7 m the error becomes negligible (< 0.2 dB).

In in-situ free-field measurements, the microphone is usually positioned relatively close to the source in order to avoid any unwanted reflections, and the source-receiver distance as measured by DIRAC should be corrected for accurate results. DIRAC offers several options to handle this.
The most straightforward option is to edit the measurements such that the start of the IR is positioned at a point corresponding to the correct source receiver distance. This can easily be done using the 'Set source-receiver distance' option on the Edit - Rotate menu.
Starting with DIRAC 6.0.5952, the system calibration dialog was changed such that you can now enter a source-receiver distance to be used in the calculations.

SPL error versus distance

Note that this option only works correctly if all measurements were taken at the exact same distance from the source. This would be the case if the microphone has a fixed position and the source is rotated over several steps as described earlier.
Leaving the 'Source-receiver distance' field blank, causes DIRAC to calculate the distance from the start of the impulse response, as in earlier versions.

Further reading

[1] Constant C.J.M. Hak, Remy H.C. Wenmaekers, Jan P.M. Hak, Renz C.J. van Luxemburg - The Source Directivity of a Dodecahedron Sound Source determined by Stepwise Rotation - Presented at the Forum Acusticum 2011, in Aalborg, Denmark.
[2] R.H.C. Wenmaekers, C.C.J.M. Hak - The Sound Power as a Reference for Sound Strength (G), Speech Level (L) and Support (ST): Uncertainty of Laboratory and In-Situ Calibration - September 2015 · Acta Acustica united with Acustica 09/2015; 101(5). DOI:10.3813/AAA.918884