Proceedings of Meetings on Acoustics Volume 19, 2013 http://acousticalsociety.org/ ICA 2013 Montreal Montreal, Canada 2-7 June 2013 Architectural Acoustics Session 3aAAb: Architectural Acoustics Potpourri 3aAAb3. Design of the new public address (PA) system for the cathedral of Münster, Germany Gottfried K. Behler* and Michael Vorländer *Corresponding author's address: Instritute of Technical Acoustics, RWTH Aachen University, Aachen, 52056, NRW, Germany, gkb@akustik.rwth-aachen.de One of the most renowned cathedrals in Germany, the Dom St. Paul in Münster was completely closed for renovation for almost one year. During this time the entire electro acoustical sound reinforcement system has been renewed. As for most buildings of this type the acoustical situation is far away from optimal. This mainly is due to a huge reverberation time which makes the understanding of spoken words almost impossible. Moreover, the old concept of sound reinforcement by using distributed loudspeaker systems all over the church is not satisfying anymore with respect to nowadays demands for quality and speech intelligibility. Due to the situation that the number of people in the cathedral during service times is varying from only some hundreds to over 2500 a more flexible PA system is required, that takes into account that only occupied areas inside of the cathedral should be supplied with amplified sound. To achieve the target an entirely new concept for the sound reinforcement based on digital signal distribution and modern digitally operated array loudspeakers was developed. The requirement for the speech intelligibility was to reach at least an STI of more than 0.5. The system will be discussed and results will be shown. Published by the Acoustical Society of America through the American Institute of Physics 2013 Acoustical Society of America [DOI: 10.1121/1.4799630] Received 22 Jan 2013; published 2 Jun 2013 Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 1
INTRODUCTION Sound reinforcement systems for churches historically were using small loudspeaker line arrays of passive type mainly. The overall frequency response was limited by the factor that the small transducers (< 4 cm in diameter) typically have not been able to reproduce frequencies below 200 Hz and the upper end of reproduction was set to 8 10 khz. The possible output power as well was quite limited. To achieve an acceptable sound reinforcement for the entire space, a great number of these loudspeaker line arrays were used. This resulted in a distributed sound reinforcement design that was aiming neither to create a correct localization of the source nor to provide a hifi - equivalent reproduction of other content than spoken words. The typical result was a rather diffuse, very soft and thin sounding reproduction of the ceremony. The huge reverberation times found in many of the large Gothic and Romanic style churches create further problems for this kind of loudspeaker design. Reverberation times up to 10 seconds for the empty church at the lower end of the frequency scale are often found. During normal service times when only a limited number of church goers are present the reverberance is not significantly reduced, but still all loudspeakers in use are feeding the volume of the hall with sound energy. As a consequence the reverberant level is quite high and the intelligibility is low. During high days like Christmas, Easter and Pentecost the church is fully occupied and the acoustical conditions can be very different. This is a dissatisfying situation which should be considered when planning a new sound reinforcement system. A technical disadvantage of the passive loudspeaker line array design is its frequency dependent directivity which is wide open at low frequencies and can be very narrow at high frequencies. The main lobe is directed perpendicular from the line array which results in two different placing philosophies: either it should be mounted at ear-level, which causes a progressive sound attenuation for people placed further away due to the people in front, or the loudspeakers are mounted above the audience ear level which then requires a downward tilting of the loudspeakers since the beam at higher frequencies has to reach the audience. In both cases an optimization of the directivity could only be achieved by selection of an appropriate line array length. Since several years loudspeaker arrays with digital sound processing (DSP) units are available that allow a frequency dependent adjustment of the beam of a loudspeaker. This includes the shifting of the main lobe up- or downwards and even the split of the beam into two or more different directions. All this depends on the versatility of the filters provided for each transducer in the loudspeaker line array. As a consequence the length of the loudspeaker line array was not limited but the extension to larger (longer) arrays became meaningful and the steering of sound energy at low frequencies became possible. Loudspeaker line arrays with a length up to 6 m are available from different manufacturers. FIGURE 1: Picture of St. Paul Cathedral in Münster Germany (copyright: Sebastian Engler [2]) Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 2
THE TASK One of the most renowned churches of the Gothic style in Germany is the Cathedral of Cologne (Kölner Dom). A complete renovation of the sound system under supervision of the authors was undertaken in 2009-2010. The result was such an unexpected success (at least for the authorities of the cathedral) [1] that in 2011 the authors have been asked by the authorities of the St. Paul Cathedral in Münster to do a similar consulting prior to a restricted call for tenders. The St. Paul Cathedral in Münster (see FIGURE 1 for an outside view) is an important seat of a bishopric in Germany and is very well accepted by the public. The renovation started in 2011 is probably the biggest after the reconstruction after World War II. The interior was completely renovated and an archeological ground excavation in the western (the oldest) part of the church was undertaken. It was only logical that during this big and expensive renovation the wish for a new loudspeaker system with up to date technique should be fulfilled. The task for the consulting team was to create a text for invitation to tender, that was describing the demands of the new system in such a way that there was no definition of components but only definition of the functionality and the quality requirements. To reach this a simulation model was defined, checked by own simulations and finally distributed to the tenders. To get comparable results the settings for the simulation were well defined and distributed with the model. Creation of a simulation model For the simulation of the new sound reinforcement system a computer model was created to be used in EASE. A very limited number of materials have been used to define the acoustical properties: Wooden chairs, Stone walls, Windows, Marble floor, and Plaster were used. The simulated reverberation time without tuning of the model already was quite satisfying with 10% - 15% deviation. However, the model was adjusted using the measured reverberation time (FIGURE 2). This was accomplished by variation of the absorption of the materials in different frequency bands. For the low frequency adjustment mainly the windows have been used whereas in the mid frequency range the Stone and Plaster properties have been changed. For the high frequencies all materials were used except the windows. FIGURE 2: Reverberation time T30 in the empty cathedral St. Paul. For the satisfaction of the client a quite detailed model was created thus leading to an overall number of 5200 faces. Though this seems to be a number leading to unpleasant calculation times, the calculation of a single situation took about 3-4 hours with good accuracy. A faster simulation for tentative evaluation required about 30 minutes calculation time on a 2.7 GHz I7 8-core PC with 8 GBytes of RAM. FIGURE 3 shows the wire frame sketch of the model. It is due to the curvature of the vaulted Gothic ceiling and a lot of details that the rather high number of faces Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 3
was reached. The degree of details can be estimated from FIGURE 4 showing an architectural rendering of the interior. FIGURE 3: Wire frame model of the cathedral St. Paul. FIGURE 4: Architectural view of the cathedral St. Paul. The positions of the loudspeakers are visible, though very small. The model finally was distributed to the companies called to send in a tender. The requirement was to calculate different typical situations with well defined settings so that the results of the different tender offers would become comparable. The tenders were instructed to provide a certain set of results for the STI, the overall level of direct sound and total sound. For each result the resolution of the color maps were defined and for the STI the distribution of the achieved values was of particular interest. The settings listed in Table 1 were binding far the calculation: Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 4
TABLE 1. Parameters defined for the simulation of the loudspeaker systems using EASE Aura module parameter settings Patch size in audience area 2.5 m Number of particles 250000 / loudspeaker Default scattering 20% Length of impulse response 8000 ms Parameters to be calculated: Direct SPL without diffuse field Total SPL incl. diffuse field A-weighted from 100 Hz to 10 khz Distribution of STI values (color plot) Distribution of STI values (statistics) The calculation had to be performed for three different situations: One with a small number of church goers so that only the central nave and the transept to both sides of the main intersection on the east side is in use. The second with all three naves in use and the transept. The third Situation should respect a typical high day situation, when the entire interior of the Cathedral will be in use by up to 3000 people. As one improvement compared to the old loudspeaker system the three situations of occupation should be taken into account for the loudspeaker set-up. Only the loudspeakers required to reach the present audience had to be used during the simulation. This would definitively result in an advantageous situation compared to the prior system use, where all loudspeakers - regardless if needed or not - have been in use. However, it was not the aim to show that we can spoil the sound system by a certain misuse, but it was the aim to show that with a proper adjustment and setting of the system a much better sound reinforcement and a good intelligibility can be achieved. The positions of the loudspeakers were defined and could not be changed by the companies. However, the type of loudspeaker (except that it was a requirement to use an active DSP-type line array), the signal processing and the aiming was to be chosen by the bidding companies. FIGURE 5 shows one possible loudspeaker arrangement used in the calculation of one bidding tender. FIGURE 5: Position and aiming of the loudspeakers. Results The high claims defined in the text for invitation to tender finally caused several of the invited companies to deny offering, either because they did not feel happy with the quite complicated system definition or because of the limited chance to be taken into account after having spent a lot of effort (and money) for the delivery of a comprehensive tender. However, four companies made the effort to submit a complete calculation and a well Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 5
designed proposal for the entire system. The selection was made by the church authorities after an extensive study of the four tenders by the consulting team which resulted in a ranking for the four offers. Simulation Results To give an impression of the simulation results achieved with the new system the results of the winning tender shall be shown (anonymous). The most interesting property for the comparison of the systems is the speech transmission index STI. It shows directly how well the aiming of the loudspeakers and the time alignment is adjusted. In the following pictures the results for the fully occupied cathedral will be shown. It is most likely not the most difficult situation since the absorption due to people will basically result in a much lower reverberation time. It is, though, a realistic and often found situation. FIGURE 6: Distribution of the STI in the entire Cathedral for the audience areas for full occupancy. (Only one half of the ground plane is shown) FIGURE 6 shows the STI mapping and the distribution for the places calculated. The results show values between 0.5 (only very few below) and up to 0.7 (very few even above). The mapping definitively pictures a disadvantage of the placement of the loudspeakers resulting in a mixing zone between the big columns caused by the sound incidence from two equally loud but probably misaligned loudspeaker system (the one from the central nave and the second one from the side nave). However, in this area the visibility conditions are as well not very good and acceptable STI values still achieved. The worst places are in the corners of the transept, where obviously shadowing by the big columns combined with an offset position relative to the loudspeaker can be found. In FIGURE 7 the direct SPL is shown, both mapping and level distribution. It is clearly shown that the sound level behind the big columns is very low compared to other positions. In comparison to FIGURE 6 the level in between the big columns is not the reason for the lower STI values in this area. As a result it can be stated, that the overall level distribution is fairly equal and only a few position (which are not the very best with respect to visibility as well) are questionable. Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 6
FIGURE 7: Direct SPL - mapping and distribution - for the fully occupied Cathedral. THE INSTALLED SYSTEM It would be nice to show that the simulation results are in agreement with the measurements of the finally installed system. Unfortunately, as often found with these kind of projects, several delays in different subsections of the restorations projects occurred, so that a final test of the sound reinforcement system could not be performed until the end of the submission date for this paper. We therefore have to apologize for not being able to present the most interesting data in the paper, however, we are confident to have all results together for the aural presentation at ICA in Montreal. REFERENCES 1. D. Michel, Neue Beschallungstechnik für den Kölner Dom, ProSound Magazin 2.2011. 30-41 (2011). 2 http://www.facebook.com/pages/sebastian-engler/300106360008021 Proceedings of Meetings on Acoustics, Vol. 19, 015082 (2013) Page 7