Do you want ear-splitting noise and migraines with that order?
Last month, I discussed how a number of manufacturers are using dedicated in-house software to not only predict the performance of their line array products, but also to steer the aiming of an array and so manipulate its overall performance. This enables, for example, sound to be kept off balcony fronts (or at least be substantially attenuated) or overspill at outdoor events to be reduced. AFMG, the authors of EASE, the well known electroacoustic prediction program, and EASERA, the complementary measurement software, recently announced the development of a new program called “FIRmaker,” a non-product-specific utility that takes the concept a step further.
FIRmaker aims to optimize the performance either of a system or, of more interest perhaps, the performance of a system in a space or venue. That is, the program will automatically manipulate the system to optimize its coverage and in-situ performance. Current conventional design procedures usually involve several iterations of a concept design as parameters are repetitively adjusted to produce the desired result. The time this takes depends, to a large extent, on the skill and experience of the designer but, for large or difficult spaces, even for the experienced, it can be an extremely time consuming process. As I designer I typically am trying to simultaneously optimize several system design parameters at once, such as:
- speech intelligibility
- evenness of coverage
- maximum achievable SPL
- smooth frequency response at all listening positions
- avoidance of sound radiation onto particular surfaces or areas.
It’s a bit like trying to squeeze Jell-O: As soon as you get one bit where—you want it, another bit pops out somewhere else. So it can be with sound system design: Fixing a dead spot in one place instantaneously can produce a hotspot somewhere else or screw up the predicted intelligibility. With a modern line array or steered array, there can be just too many variable parameters to try and grapple with. Most of us will try things for a while, but then fatigue or boredom sets in or we become disheartened by the repetitive process.
Computers, on the other hand, thrive on such repetitive work and can try out thousands of trial-and-error settings until they (well, the program) find a convergent or optimized solution. The program just requires some target values to operate with. Now this is fine when it comes to parameters, such as evenness of coverage, as a target range can be readily set and beams then adjusted as many times as required until either the target is met or the variation is minimized.
Figure 1 shows a typical example of the results that can be obtained with computer optimization of a conventional music line array. In order to measure the potential improvement, a statistical approach often has to be adopted. For the simple plots shown in Figure 1 and Figure 2, the improvement can immediately be seen as the SPL variation reduces from ±2.3dB to just ±0.2dB over the area of interest, although, in this particular case, it was at the slight expense of the maximum sound level.
Array optimization, of course, is not just confined to optimizing spatial coverage, but can also be applied to array directivity. An example of this is shown in Figure 3, where the measured vertical directivity of a 1.3m column loudspeaker array is shown. The upper plot shows the natural directivity of the array. As can be seen, the directivity gets progressively narrower as the frequency increases, varying from about 90° down to 20°. By applying suitable optimization, the array is transformed into a “constant directivity” device, maintaining the required directivity across the band.
Whereas, at the moment, AFMG and several array manufacturers can optimize for spatial uniformity of coverage, optimizing for speech intelligibility and uniformity of intelligibility adds another order of magnitude to the complexity of the task. I, therefore, think it will be awhile before we see this as a part of the automated optimization process…except perhaps for the fairly straightforward systems.
However, that does not mean to say that SPL optimization does not affect intelligibility. Because the Direct Sound Field plays an important part in our perception of speech intelligibility, optimizing this, which is effectively what the systems are doing, might therefore be expected to affect the resultant intelligibility. Some preliminary work published recently by Feistel and Ahnert (AFMG, “Improving Speech Intelligibility using Numerical Sound System Optimization,” Wolfgang Ahnert & Stefan Feistel Proc IOA Vol. 35 Pt. 2. 21), shows exactly this effect.
Figure 4 shows the change in Direct Sound levels brought about by optimizing the coverage of a line array system in a large reverberant space (a 10dB variation reduced to a 2dB spread in SPL). Figure 5 shows the resultant change in intelligibility as indicated by the Speech Transmission Index (STI). As Figure 5 indicates, the variation in STI reduces noticeably with the optimization of the direct sound coverage. However, for more than the half of the space, this resulted in a reduction in the STI, but with an improvement being experienced elsewhere. As I said earlier, you squeeze the Jell-O in one place and it changes somewhere else. Perhaps I should invent the JSI (Jell-O Squeeze Index) as a measure of parameter uniformity and sensitivity to change!