Her fortsætter Barbara Ohlenforst, Marco Ottink og Gerad Encina Llamas.

In the previous post, an introduction to the project and some basic acoustic concepts such as reverberation time were explained. It was also said that a 3D computer model was built in order to simulate the acoustic conditions in the Cosmopol stage. In this post, the work done since now will be presented by going further into some technical details (but easy!!).

Finally, a simplified model was built. The first model (see figure 2 in the previous post; which was drawn by the architectural engineer Kristian Mouridsen, is too complex to be imported in Odeon (the software simulation software - www.odeon.dk).

Odeon uses the image-source method combined with ray tracing. These methods are based on the analogy between acoustics and optics, where sound waves are considered as light rays.By this, the result in the simulation depend strongly on the number of surfaces. For example, in the image source method, a new sound source is created to simulate the reflected ray. A very complex model with very small surfaces, might produce errors in the low frequencies calculations, and will need a long computational time.

The following figure 1 shows the simplified model in Odeon software. The image method can be seen here. The red circle in the tent named P1 is the real position of the omnidirectional sound source. The blue circle is the receiver position number 2. The red circle suspended on the air is the image source for a single surface. The ray coming from the image source has the same length as the whole reflected ray from the real source. The energy loss due to the propagation itself and the reflection is calculated.

Figure 1: Simplified model of the Cosmopol Stage and example of Image source method in Odeon

As it was said in the previous post, a measurement session was performed last Thursday 28th of June. The aim of this measurement session was to get some reliable data in the empty room conditions. Only the metallic structure and the main canvas was build then. This data will be used to be compared with the simulated results in Odeon. By this, the model will be fitted to the real case.

The measurement were performed using an omnidirectional source; which is a source that emit the same "amount" of sound energy in all directions within a frequency band. Two omnidirectional microphones (which similarly capture the same sound energy from all angles) were used to measure the room response. A woofer were used to generate sound in the lower frequency octave bands.

Figure 2 shows a picture during the measurement setup (obviously, the car wasn't there during the measurements).

Figure 2: Acoustic measurement setup at Cosmopol Stage

The measured parameter was the \emph{impulse response}. The room (in this case the Cosmopol tent) is understood as a system, and depending on the geometry and the acoustic properties of the material (basically the sound absorption coefficient) it has a a different response depending on the frequency on each position. The impulse response can be understood as the output of the system when the input is a very short signal. Because very short signals in time excite all frequencies, the response of the system is obtained.

In our case, the signal used to measure was an \emph{exponential sweep}. This is a tone that grows exponentially in frequency. By this, the impulse response can be measured; and post-processing the impulse response, the reverberation time and other acoustic parameters can be calculated. Next audio clip shows an exponential sweep.

The impulse response, which is measured in each microphone receiver position, can also be simulated in Odeon. Next figure 3 shows the simulated impulse response in receiver position 2. The first peak is the direct sound and the second strong peaks are the reflections coming some milliseconds after.

Figure 3: Simulated impulse response using Odeon at receiver point 2

The impulse response can be heard next (it is recommended the use of headphones).

The impulse response is a very powerful information in the simulations because shows the response of the system in a given position. From signal theory, if the impulse response is convolved by another signal, the response of this signal after the system is obtained. Using this, auralisation processing can be done in order to hear how the room sounds.

The signal used for the auralisation must be an anechoic signal (recorded in an anechoic chamber, where there are not reflections, and therefore is like "there is not room"). This makes sense because if we would have used a signal (i.e. a guitar) recorded in a normal room, we would be listening the signal through two systems: the first room where the signal was recorded and the simulated room.

Next audio clips shows two auralisation at the receiver point 2 in Cosmopol Stage. The first clip is clapping where the strong reflections can be perceived. The second clip is a pop-rock song. In both cases, either the original recording and the auralised signal are shown to perceive the effect of the room.

Clapping clip

Clapping - anechoic recording Clapping - auralised sound

Pop-Rock song

Pop-Rock song - anechoic recording Pop-Rock song - auralised sound

As it can be heard, there is room for some improvement. Hopefully, in reality it will sound much better because of the work of the sound engineers at the mix control, and also because of the use of the PA system (Public Address loudspeakers) that will send the sound much directly to the audience area, avoiding sound propagation towards the ceiling.

Once the measured data is obtained, as it has been said before, it will be used to fit the model to the real situation. There are some uncertainties in the model that have been assumed. The main one is the exact absorption properties of the ground. For the preliminary simulation, an absorption between grass and sand has been considered; but it must be adjusted. The measurement data will be really useful for that.

In any case, it seems that the assumption are not that unrealistic because without any fitting adjustment, the simulated and the measured reverberation times are very similar. Figure 4 shows the average results in all the measuring points both for the simulation (red) and the real measurements (blue). Obviously, the points in the simulation are placed in the same position where they were measured in the real tent.

Figure 4: Comparison between the measured reverberation time and the simulated

Once the model will be tuned, it is considered to be validated. Then, it can be used to perform simulation with the tent in other conditions, such as with full audience.

In our case, we will use the model to try to improve the general acoustic conditions. As it can be seen in figure 4, the reverberation time is quite uneven (not flat) in the frequency bands range. One of our goals is to try to made it as flat as possible. We are considering to suggest to install some membrane absorbers, which will reduce the peak around 500 Hz and 1000 Hz bands. However, it can be limited by practical issues such as the weight the structure can hold, or not covering the lights, etc.

Another objective is to evaluate the echoes and to try to avoid the annoying ones.

Finally, on Thursday 5th of July another measuring session will be performed together with the people from Delta (http://www.delta.dk/). Once again the impulse response will be measured but using the sound system used during the concerts; and with the stage already built. This will help us to understand the effect of the sound system in the acoustic quality by comparing the two measurement session data.

This project will be continued after the festival, where all the improvement part will be done. Unfortunately, this is the last post in this blog but if somebody is really interested to know how it will end, just pass by the acoustic department at DTU building 352; or write us an e-mail at gerard.encinagmail.com