Abstract:  Improvements and optimizations are proposed for acoustic design of large auditoriums in this paper. The main difficulty for room acoustic design is the discrepancy between the original theoretical prediction of the sound
field and the real acoustic performance of the auditorium, and this is what we focus on. We discuss the existing inadequacies of acoustic design and find out the key reasons, by analyzing the reliability
of sound field model and simulation, the rational selection of absorption coefficients of acoustic structures, and the correct reverberation measurements in laboratories. Improvements and optimizations are proposed on the accuracy of
the sound field predictions. Firstly, traditional methods of using Sabine and Eyring formulae are not sufficiently accurate due to the hypothesis of ideal diffuse sound field. Another two more practical approaches of ray tracing and
image -source methods are proposed. In the ray-tracing method large quantities of sound rays are sent out from the source and the physical data are stored during the tracing of the rays to synthesize the required parameters in
observation points. In the image -source method image sources are produced due to the presences of room walls, and contributions of the visible images are added on to the reflection patterns based on their reflection paths and the
walls- absorption coefficients. Secondly, the spatial and ensemble deviations are essential facts that determine the accuracy of reverberation measurements of acoustic structures. Base on theoretical and quantitative experimental
statistics, methods of using rotate diffuser and integrated impulse response with maximal length sequences are analyzed and proven effective to improve the spatial and ensemble deviations, making the absorption measurements
more reliable. Further more, the studies above are carried out to guide the practical acoustic design of the main auditorium of Haigeng Conference Center. The characteristics of its acoustic design, the computer-aided sound field modeling and
prediction, and the measured acoustic performances are introduced, and the practical results validate the effectiveness and practicability of our improvements.