FACULTY OF ELECTRONICS, TELECOMMUNICATIONS AND INFORMATION TECHNOLOGY Ing. Norbert Ştefan Toma PhD THESIS SUMMARY CONTRIBUTIONS TO THE EVALUATION AND MODELLING OF THE ROOMS ACOUSTICS Supervisor, Prof. Marina Ţopa PhD thesis evaluation committee: PRESIDENT: - Prof. Tudor PALADE, Vice-dean, Faculty of Electronics, Telecommunications and Information Technology, Technical University of Cluj Napoca MEMBERS: - Prof. Marina ŢOPA Supervisor, Faculty of Electronics, Telecommunications and Information Technology, Technical University of Cluj Napoca - Prof. Liviu GORAŞ - reviewer, Faculty of Electronics, Telecommunications and Information Technologies, The "Gheorghe Asachi" Technical University of Iaşi; - Prof. Alexandru ISAR - reviewer, Electronics and Telecommunications Faculty, Politehnica University of Timişoara; - Prof. Lelia FEŞTILĂ - reviewer, Faculty of Electronics, Telecommunications and Information Technology, Technical University of Cluj Napoca
Acoustics is the science of sounds. It deals with all the aspects regarding the sound as its generation, propagation, interaction with different materials and perturbations, effects on people and animals, perceiving and analysis. Acoustics is a research domain with interdisciplinary applications based on different disciplines like physics, psychology, physiology, transmission of information, materials science and electronics. Reverberation is a common phenomenon in our lives. Closed spaces as offices, auditoria, theaters, operas, concert halls, have this phenomenon due to several reflections of the sounds within them. The acoustics of halls and recording studios is very important to suit the destination: listening to music or speech. The evaluation of the room s acoustical quality is a difficult task if all the details of the room s complex structure (shape and different materials) are taken into account. The hearing assessment is a subjective approach due to the differences that appear from one person to another. For the global acoustic evaluation of the room, the following objective acoustic parameters are used: reverberation time, early decay time, clarity, definition, center time. To analyze some particular locations or their subjective acoustic aspects it is necessary to establish a relationship between the objective and subjective parameters. To each subjective parameter there exists at least an objective one. The thesis tries to connect two very large scientific domains: the acoustics and the digital signal processing. It will try to answer to questions like: Which are the acoustic characteristics of the room?, Which are, from an acoustic point of view, identical locations?, How many reverberators are necessary to model the acoustics of a room?, Which reverberators have the best performance? or Which are the design methods and the structure of these reverberators?. The thesis is organized into 4 chapters: The first chapter The sound propagation in a room, the early and late reverberations presents general aspects regarding the sound propagation, defines the acoustic parameters for the evaluation of real rooms and artificial reverberators. The second chapter Reverberators describes 3 artificial early reverberators and 7 late ones. The design procedures are presented; their starting point is the impulse response of a real room. For each reverberator an improved version was achieved by implementing the frequency dependence of the reverberation time. The third chapter The evaluation of acoustic parameters of rooms using measurements is centered on modeling and evaluation procedures of rooms acoustics. The acoustics of 3 rooms was evaluated: a school hall situated in the building of the Iuliu Maniu Technical College in Carei, the St. Maria Catholic Church located in Satu-Mare and the P03 auditorium of the Technical University of Cluj Napoca. A model of the school hall was established by using the CARACAD program. A procedure for grouping the measured locations was developed based on 2 subjective (perceptive) parameters: the perceived reverberation and clarity. The fourth chapter Conclusions and future work presents the experimental results and conclusions of the thesis, trying to detect and define the future development direction of this generous domain. In the second chapter the following reverberation algorithms were studied: - Schroeder; - Schroeder based on all-pass filters; - Moorer; - Gardner; - Jot; - Absorbent all-pass; - Modified Schroeder. 2
For each reverberator a Simulink program was developed. Experiments with different acoustic signals were carried out, their advantages and disadvantages were pointed out. Nowadays, the majority of reverberators are implemented with recursive filters. The first reverberator using digital filters was built by Schroeder in the 60s and it is based on comb and all-pass filters. This reverberator has a lot of drawbacks as: The damping acts as beating or fluttery; For large reverberation times the sound is metallic; The echoes density is insufficient and does not increase with the time; One cannot specify a relation between the reverberation time and frequency. Moorer made some performance improvements to his reverberator. Gardner built reverberators based on simple and nested all-pass filters. These reverberators have a good performance, but one cannot specify a relation between the reverberation time and frequency. Furthermore, the design of these reverberators is entirely empirical. Jot built reverberators that can be designed using the energy decay relief of a real room. Figure 1 presents the structure of the Jot reverberator. It can be considered as a generalization of the Schroeder s parallel comb filters with a diagonal feedback matrix. This device has a larger echo density than the parallel comb filter, due to the large number of non-zero feedback coefficients. The feedback matrix enables the delay cell output to be recirculated at the cell input, where the coefficients matrix controls the weights of the feedback taps. The low-pass filters h 1 (z), h 2 (z), h 3 (z) and h 4 (z) are simulating the sound absorption phenomenon. a11 a12 a13 a14 a 21 a 22 a 23 a 24 a 31 a 32 a 33 a 34 a41 a42 a43 a44 b 1 b 2 b 3 b 4 z -m 1 z -m 2 z -m 3 z -m 4 h 1 (z) h 2 (z) h 3 (z) h 4 (z) s 1 [n] s 2 [n] s 3 [n] s 4 [n] c 1 c 2 c 3 c 4 X(z) d Y(z) Figure 1 The Jot reverberator structure The last two reverberators (absorbent all-pass and modified Schroeder) have a good performance, due to the use of absorption filters, which model the absorption phenomenon. The reverberator algorithms were implemented in Matlab s Simulink and the obtained data was analyzed with WINMLS. The Simulink program is a high speed data processing tool and can offer a real-time data processing due to their optimized blocks and a frame based sample processing. The WINMLS software is used for performing and evaluating audio, acoustic and vibration measurements. In the near future a reverberator equipped with a Simulink interface will generate the instructions for programming DSP s, simply by drawing the reverberator scheme. 3
The third chapter presents the acoustic properties of the rooms as well as the evaluation methods. 3 rooms with different destinations were evaluated and different equipment was used in the evaluation procedures. The excitations were: sweep sine signals obtained at the output of some conventional loudspeakers, impulses using balloons and firecrackers, sweep sine signals obtained at the output of an omnidirectional sound source. In one of the rooms, the reverberation time was measured also by a sonometer. The advantages and drawbacks of every method were also analyzed. The suitable destination of every analyzed room was determined using the values of the reverberation time. For the first room acoustic evaluation (School hall of the Iuliu Maniu Technical College in Carei) two excitation sources were used: sweep sine signal obtained at the output of some conventional loudspeakers and an impulsive one generated by balloons inflated at the same pressure. In the first case, due to the excessive directivity of the conventional loudspeakers (reflected by the EDT/T30 early decay time/reverberation time ratio) a large difference was detected between the acoustic parameters measured by the two methods. Due to their directivity the evaluation of room s acoustics using conventional loudspeakers was abandoned. Next the acoustics of the school hall was analyzed after introducing 70 chairs in the room. The reverberation time was reduced and the diffusion increased. The measured reverberation time was compared to the values obtained by modeling the room in the Caracad program. The Dirac and Winmls programs were used to process the measured samples when the room was empty or contained 70 chairs. Figure 2 presents the measured and simulated reverberation times. One can observe that, there are no significant differences between the processing programs Dirac and Winmls. Noticeable differences appear between the measured and simulated values below 500Hz. The errors in Caracad were due to rough approximations as: impossibility of modeling the complex form of all the objects in the room, very rough approximation of some objects shape (circles were approximated by hexagons) as well as in component materials in the room. a) b) Figure 2 The measured and simulated reverberation times for the in school hall case: a) empty; b) with 70 chairs. Using the computed reverberation time, the possible destinations of the room were detected: the listening to chamber music and opera. Using two subjective criteria (the perceived reverberation time using EDT and the perceived clarity by C80) some similar acoustic locations were identified. By this procedure the number of reverberators which could be used to model the room s acoustics was reduced with 61.5 %. For the second room (St. Maria Catholic Church) 2 types of excitation sources were compared: impulses generated by balloons and firecrackers. Due to the placement of firecrackers into the upper part of a pipe, a directional tendency was observed for the impulses generated by them. Consequently, the evaluation results were processed only for the balloons case. According to the computed reverberation time, the best destination of the room was determined: audition of 4
romantic classic music. Regarding the just noticeable difference (JND) analysis, it was possible to identify some points with the same acoustics using 2 subjective criteria: the perceived reverberation and clarity. It means that, it is possible to optimize the number of reverberators to model the rooms acoustics. The result was the reduction of the necessary reverberators number to 60%. The 2 impulsive methods have several drawbacks as: poor repeatability, questionable omnidirectionality, danger of fire (in firecrackers case) and ecological inefficiency (for the broken balloons case). Due to the neighboring areas protection the acoustics can suffer distortions. The church was evaluated also using a B&K 4295 omnidirectional sound source. The acoustics of the room was analyzed using 2 different points for the excitation source. No important differences between evaluated reverberation times using the impulsive sound sources and the omnidirectional one were detected. The advantages of the noise source are: ecology, good repeatability and omnidirectionality according to the ISO 3382 standard. Due to JND analysis, it was possible to identify some points with the same acoustics using 2 subjective criteria (the perceived reverberation time using EDT and the perceived clarity C80) and the locations of the excitations sources. Thus, the number of reverberators necessary for modeling the room s acoustics was optimized. The attained reduction was of 46.66 % and 53.33 % respectively when having one source. The reduction became 33.33%, when two sources were considered. For the evaluation of the third room s acoustics (the P03 auditorium of the Technical University of Cluj Napoca) the B&K 4295 omnidirectional source, firecrackers and the HD 2010 sonometer were used. According to the computed reverberation time, the ideal destination of the room was determined: chamber music concerts and opera. The acoustics of the room was analyzed using 2 different points for the excitation source. The analysis of the EDT/T30 ratio shows that the hall does not have directional locations, but there are some excessively diffuse locations especially at the end of the hall when the source is located in the central front of the hall. According to JND analysis, it was possible to identify some points with the same acoustics using 2 subjective criteria (the perceived reverberation time EDT and the perceived clarity C80) and the locations of the excitations sources. Thus, for individual locations of the sources, the optimized number of reverberators (necessary for modeling the room s acoustics) was attained. The reduction was of 26.66 % and 40 % respectively for one source and of 20% for two sources. The auditorium has poor acoustic properties for speech events (like lectures, courses), so acoustical improvements are necessary to be applied. The thesis original contributions are: For Moorer s early reverberator the design of early reverberator starting from the early impulse response (early energy) of the real room, implementation of the early reverberator in the Simulink program [TOM07a]; For Schroeder s late reverberator, Schroeder s all-pass late reverberator, Moorer s late reverberator the design of the late reverberator with the reverberation time of 1.757 s, improvement of the reverberators performance by inserting absorbent filters using the generalized Jot design method, implementation of these late reverberators in the Simulink program, acoustic parameters analysis using the sound files obtained from simulations and the Winmls program [TOP04a], [TOP06], [TOP07]; For Gardner s late reverberator highlighting the reverberation time dependence with respect to the feedback gain, choice of the optimal gain in order to obtain a simulated reverberation time closer to the real room s one, improvement of the reverberator s performance by inserting an absorbent filter with every delay cell, 5
Future work: implementation of the late reverberator in the Simulink program, acoustic parameters analysis using the sound files obtained from simulations and the Winmls program [TOP04b], [TOM07a], [TOP07]; For Jot s late reverberator, modified Schroeder late reverberator, absorbent allpass late reverberator - the design of the late reverberator starting from the frequency dependent reverberation time, absorbent filters design with Jot s method, improvement of the reverberators performance using the generalized Jot design method, implementation of the late reverberator in Simulink, acoustic parameters analysis using the sound files obtained from simulations and the Winmls program [TOP04a], [TOM05d], [TOM05e], [TOM06a]; For the first room (School hall) comparison of the acoustic properties of the hall for sweep sine excitation obtained from conventional loudspeakers with ones obtained from impulsive sound sources using balloons inflated at the same pressure (50 mm Hg), comparison between the acoustics of the empty hall with the one filled in with 70 chairs, identification of points with the same acoustics using two subjective criteria (the perceived reverberation time EDT and the perceived clarity C80), reducing the number of reverberators which could be used to model the room s acoustics by 61.5 % [TOM10], [TOP10]; For the second room (St. Maria Catholic Church) comparison between the acoustic properties of the hall for 2 types of impulsive excitation sources: balloons and firecrackers, reducing the number of reverberators which could be used to model the room s acoustics by 60%, the JND analysis for the B&K 4295 omnidirectional sound source for other locations in the room, for 2 individual positions of the source obtaining a reduction of 46.66% and 53.33% respectively and a reduction of 33.33 % when all the sources are considered [TOP10], [TOP11]. For the P03 auditorium - according to JND analysis, it was possible to identify some points with the same acoustics using two subjective criteria (the perceived reverberation time and clarity) and the locations of the excitations sources. Thus, for individual locations of the sources, the reverberators (necessary for modeling the room s acoustics) number optimization was attained. The reduction was of 26.66 % and 40 % respectively. The reduction became 20%, when all the sources are considered [TOM09a], [TOM09b]. - Computation of JND values for the acoustic parameters for low and high frequencies. Using these values, one could better compare the reverberators modeled room s performance with the real room s one; - Development of a software to determine the optimal grouping of the measurement locations in order to determine the minimum reverberators number to model the rooms acoustics; - Development of new methods to improve the rooms acoustics (architecturally and electronically) based on the JND approach ; - Development of a method for rooms acoustics improvement using reverberators. 6
REFERENCES [TOM05d] [TOM05e] [TOM06a] [TOM07a] [TOM09a] [TOM09b] [TOM10] [TOP04a] [TOP04b] [TOP06] [TOP07] Norbert Toma, Marina Ţopa, Erwin Szopos, On Improved Reverberation Algorithms, Proceedings of the 47 th International Symposium ELMAR-2005, ISBN 953-7044-04-1, 08-10 June 2005, Zadar, Croatia, pp. 217-220. Norbert Toma, Marina Ţopa, Erwin Szopos, Aspects of Reverberation Algorithms, Proceedings of the International Symposium on Signals, Circuits and Systems ISSCS 2005, ISBN 0-7803-9029-6, 14-15 July 2005, Iasi, România, pp. 577-580. Norbert Toma, Marina Dana Ţopa, Victor Popescu, Erwin Szopos, Comparative Performance Analysis of Artificial Reverberation Algorithms, Proceedings of the 2006 IEEE-TTTC International Conference on Automation, Quality and Testing, Robotics AQTR, ISBN 1-4244-0360-X, May 25-28, 2006, Cluj-Napoca, Romania, pp.138-142. Norbert Toma, Marina Dana Topa, Erwin Szopos, Design and Performance Analysis of Reverberation Algorithms, Acta Technica Napocensis, ISSN 1221-6542, Volume 48, Number 1, pp. 35-43, 2007. Norbert Toma, Marina Ţopa, Modelarea şi optimizarea acusticii încăperilor, Volumul Simpozionului naţional Edmond Nicolau, ISBN 978-973-53-0017-3, Vol. I, pp.1530-156, 2009. Norbert Toma, Marina Ţopa, Irina Mureşan, Botond Sandor Kirei, Marius Neag, Albert Fazakas, Acoustic Modelling and Optimization of a Room, Acta Technica Napocensis - Electronics and Telecommunications, ISSN 1221-6542, Volume 50, Number 2, pp. 25-30, 2009. Norbert Toma, Marina Dana Ţopa, Botond Sandor Kirei, Ioana Homănă, Acoustic Analysis of a Room, Acta Technica Napocensis - Electronics and Telecommunications, ISSN 1221-6542, Volume 51, Number 2, pp. 14-19, 2010. Marina Ţopa, Norbert Toma, Erwin Szopos, Design and Simulation of Reverberation Algorithms, Proceedings of the 6 th Cost 276 Workshop on Information and Knowledge Management for Integrated Media Communication, ISBN 960-88136-0-3, May 6-7 2004, Thessaloniki, Greece, pp. 139-144. Marina Ţopa, Norbert Toma, Erwin Szopos, Efficient Reverberation Algorithms, Cel de al XXXV-lea Simpozion de Comunicări Ştiinţifice al Agenţiei de Cercetare pentru Tehnică şi Tehnologii militare, ISBN 973-0-03501-6, May 27-28 2004, Bucharest, Romania, on CD. Marina Dana Ţopa, Norbert Toma, Erwin Szopos, Performance Analysis of Some Artificial Reverberators, Proceedings of the 6 th IEEE International Conference Communications 2006, ISBN (10) 973-718-479-3, ISBN (13) 978-973-718-479-5, June 8-10, 2006, Bucharest, Romania, pp.73-76. Marina Dana Ţopa, Norbert Toma, Victor Popescu, Vasile Ţopa, Evaluation of All-Pass Reverberators, Proceedings of the 14 th IEEE International Conference 7
on Electronics,Circuits and Systems ICECS 2007, ISBN 1-4244-1387-8, December 11-14, 2007, Marrakech, Morocco, pp. 339-342. [TOP10] [TOP11] Marina Dana Ţopa, Norbert Toma, Botond Sandor Kirei, Ioana Crişan, Evaluation of acoustic parameters in a room, SIP'10 Proceedings of the 9th WSEAS international conference on Signal processing, ISBN: 978-954-92600-4- 5, Catania, Sicily, Italy, May 29-31, 2010, pp. 41-44. Marina Dana Ţopa, Norbert Toma, Botond Sandor Kirei, Ioana Homănă, Marius Neag, Gilbert De Mey, Comparison of Different Experimental Methods for the Assessment of the Room s Acoustics, Acoustical Physics, ISSN 1063-7710, e- ISSN: 1562-6865, Volume 57, Number 2, 2011, pp. 199 207. 8