Acoustic prediction as a matter of fact

Nov 6, 2020

As years pass, not only PA systems become more sophisticated, but also the tools employed for tuning them. The design and optimization process for the sound system that will be installed, for example, in an event, begins in a prediction software. These kinds of software were, after all, mainly developed to predict line-array systems. Such systems are complex and involve a great amount of variables but make unprecedented events  possible. But, at the moment of truth, how can we make sure that our systems perform as predicted?

For instance, let’s take the case of the main PA in a stadium. The Santiago del Estero ‘Mother of Cities’ Stadium is 37 meters tall, approximately 180 meters in diameter, and it is an excellent example to answer this question.

All predictions begin with a model

Fig1. Architectural plans and virtual model.

For such dimensions even a difference of a few degrees can result in considerable error. The first step is to build a virtual model that is representative of the place where the sound system will be installed. To achieve this, architectural plans are thoroughly analysed to recreate all relevant aspects of the place, not only where the audience will be, but also hang points, and such. At this stage, it is also important to define the expected performance from the PA system. For a stadium, speech intelligibility is key,  the system mut be loud and clear, and must be placed in such a way to minimize adverse acoustical effects, if present. With this model as a starting point different strategies are put to the test to find the best results.

An extensive knowledge of equipment

Fig 2. Polar pattern measurement

The main PA comprises 8 arrays of 12 V15i loudspeakers distributed around the  stadium. A curved line-array, as shown below, aims for uniform coverage from the first row to the last.

Fig 3. Side view and simulation.

To avoid destructive interference between arrays, they were placed around the stadium, as shown below, achieving a very homogeneous distribution. No perceptible difference can be found when passing from the coverage of one system to the next.

Fig 4. Plan view and simulation.

Another aspect of the optimization process is carried out practically along with simulations. Exploring the best alternatives for loudspeaker hanging and installation, as well as trying out different processing options to ensure top performance without sacrificing equipment integrity. In this particular case, the challenge was to adjust output limiters to maximize spoken word reproduction for long periods of time. This kind of adjustment is not trivial and requires substantial testing, so that the results can, in the end, be included in the simulations.

A good simulation is only half the job done

Fig 5. Array installation

The goal here is to materialize a project. A project that begins with the loudspeaker design and manufacture. This involves loudspeaker measurement and formatting for simulation data. Nonetheless, it is equally important to pay the same level of attention when installing the systems. Particularly, this application requires a hanging setup very different from what is commonly found in live show applications, for this, a rigging setup was especially designed and built to ensure proper positioning.

Additionally, there is a key step in sound system optimization, connection (in this case Live Audio over IP), configuration and monitoring of everything amplifier related. Just to name and example, at the time of placing the arrays, a speaker showed faulty wiring, this was detected directly on the amplifier software.

Foreseeing venue acoustics

Fig 6. hanging an array

Another simulation aspect that must be taken into account in practice is the effect that different acoustic characteristics inherent to the venue may present. Meaning reverberation time, echoes, atmospheric conditions, ambient noise, etc. Even with a virtually ideal coverage of the audience on the simulation, if the loudspeakers are placed somewhere acoustically unfavourable, little can be done in practice to improve their performance. Usually, prediction software for loudspeakers have no (or reduced) means to simulate the acoustical characteristics of the venue, caution must be taken.  Even when being able to carry out this kind of simulations, given that, usually, only partial information of the materials and their physical disposition is available.



Now back to the stadium, it is a large, relatively open space, with a reverberation time of a couple of seconds in length but very low level since the surfaces are all very far apart from each other. The principal reflection is produced on the tensile fabric that covers the stadium acting as a ceiling. This is the closest surface to the arrays and where the most predominant reflection is produced, for this reason, a reflection analysis was carried out to study the impact those reflections could have on intelligibility.

Fig 7. Reflection simulation for an array.

It can be observed that for the seats at the top rows the reflection arrives close enough in time and level to produce comb filtering, but most of the energy is reflected backwards safely into the game field. Coming across these situations, obviously the acoustic design of the venue exceeds the task at hand, instead of fixing the room acoustics, it is imperative to acknowledge them and find the best placement available, taking into account what problems may arise when interpreting measurement results. For this case, specifically, the most important factor is the noise floor, since 28.000 sport fans can make some noise. This, in combination with the stadium’s reverberation, has a direct effect in speech intelligibility, for mid frequencies in particular.

Fig 8. Engard. D. (2009). September 18th Folsom Field Octave Band Analysis and Overall Leq. Department of Environmental and Radiological Health Sciences. Colorado State University.

This noise level is close to the human ear’s pain threshold so it is important that the system performance is optimized, by means of processing, for maximum intelligibility without wasting power in less critical frequencies. For this, intelligibility studies are carried out simulating different processing options, considering this noise floor and the previously mentioned acoustical conditions. Simulation results show that intelligibility never drops below STI=0.45 which is considered very good for massive PA systems.

Fig 9. Speech Transmission Index prediction.

Simulation as an on-site optimization tool

Fig 10. Simulation side view. 6 sensors on array axis.

The last optimization step involves a performance evaluation with the system completely installed and fully functional. At this stage, it is important to carry out both hearing tests and measurements in strategic points throughout the venue. Measurements can help pinpoint problems that are hard to quantify by ear and vice-versa. In this case, measurement microphones were placed on axis for one array, as shown on the picture above. It is important to consider the audience area as a whole and not end up worsening the coverage in, say, the high seats, trying to improve the coverage on the lower seats.



The following plot displays the simulated frequency response for those measurement positions.

Fig 11. Frequency response over six positions on axis for one array.

Fig 12. Frequency response on axis comparison: simulation (green and purple) VS measurement (red and blue). For clarity only the response at the centre of each level is displayed.

Comparison with the measured response reveals a great match. Less than a 2 dB difference. The mismatch at lower frequencies (below 60 Hz) is due to noise present during measurement. For the frequencies over 8 kHz the wind modified the response from one moment to another making impossible an exact measurement over that frequency range. Those factors aside, measurement was considerably clean due to the little acoustic influence of the venue. Remember, it is a mostly open space.

Fig 13. Frequency response measurement at midpoint between two arrays for the centre of the top and bottom levels.

Also included is a measurement at the midpoint between two arrays for the centre of the top and bottom levels. It can be noted that coverage is practically identical in both positions for either array.







Fig 14. Frequency response measurement between two arrays, either one active, and both simultaneously.

This is verified on the next plot, comparing this response with that of two arrays working together (light blue curve) There are no major differences or interference present for any part of the spectrum.






Having measured multiple points, and validating the close resemblance between the simulation and measurement, processing alternatives can be tried out, obtaining results in a matter of seconds without having to carry out measurements throughout the entire stadium. In this case, the processing between upper and lower tiers was altered slightly to better match coverage. The lower tier extends further into the game field at the West and East sides, this extra distance requires a little of extra level for higher frequencies.

Fig 15. Comparison between extra processing (left) and without extra processing (right).

In conclusion:

The advantage of working with simulations is not only to predict what is going to happen when the system is hanged on the far future, but also to predict what will happen in the immediate future. This means working with simulations on-site practically while the system tuning is carried out. For this, every design stage must be closely followed, this encompasses the loudspeaker manufacture and the development of particular configurations unique to each case.

Lic. Lucas Landini

Engineering Department – Equaphon

If you are interested in knowing how loudspeakers are measured to produce simulation data you can watch a video about it here.