Phase RIP – Phase Data for Loudspeaker Measurements

By Dick Campbell

Dick Campbell expresses his concerns about adding phase to the existing standard for SPL measurements.

Something has been bothering me about measuring (and publishing) phase data for loudspeaker measurements. An AES Standards Committee is currently considering adding phase to the existing standard for SPL measurements as exhibited in manufacturer’s published radiation balloons.

The argument is that phase data would allow the user to combine two or more sources and to observe phase-related “bumps” in the far field response curve. Ideally this can be done – but the practically of it is doubtful. I have been making these measurements for a very long time so I know the pitfalls.

A standard is a document that implies certain accuracy and a known repeatability. Inter-laboratory exchange should produce the same data with specified tolerance. In the case of loudspeakers, there are few test facilities and they are all different designs with different “proofs of performance”, especially at low frequencies. Also, these data should extend to 4000Hz to be useful.

  • Issue A. Measurement of phase in the far field requires a lab positioning accuracy that is easily calculated. With a sound speed of 344m/s the quarter-wave at 4000Hz is ~2.2cm. Within the quarter wave (90 degrees) assuming a sine shape, then a 30 degree phase difference occurs at 1.1cm. If a 20 degree phase error is acceptable, then the distance from the center of radiation must be repeatable within 7.5mm. A 10 degree window is 3.5mm. Imagine a loudspeaker hydraulic lift and a wobbly mic stand!
  • Issue B. Inter-laboratory accuracy using impulse responses (IR) requires some standard for adjusting the ‘time of flight’ (TOF) to a standard amount. When I chaired SC-04-01 I suggested truncating the TOF gap to 1000 data points before the rise in the zero-degree impulse begins. This time domain point then marks where the other angles in the IR are “phased”(normalized) to the zero degree axis. Question is: if ten persons, in their opinion, pick the first rise of the impulse, how different will they be? Might be better picking out the top of the first impulse.
  • Issue C. When loudspeakers are brought close together, there is always some interaction between them. One loudspeaker is creating a field so that the other is not radiating into a ‘free-field’, and vice versa. There is no way to accurately model this interaction. It depends upon driver size, cabinet position, and radiation coupling. There are papers on this subject from Chalmers University and others.

One must not jump into publishing phase measurements unless the standard addresses these issues. In addition, the ear can not detect bumps up or down less than 1/6 octave wide and +/- 6dB when listening to program material. We do not normally spend our day listening to colliding sine waves!

Cheers – Dick Campbell