A Study on Sound Level Calibration for Listening Tests Performed with Headphones in Architectural Acoustics

Abstract

Subjective evaluations are frequently used in design, renovation or improvement stages in architectural acoustics. Listening tests are one of the frequently used methods in subjective acoustic evaluations. These studies are generally carried out in existing halls or the auralizations digitally created in computer modellings. Listening tests can be conducted via speakers or headphones. Considering the relationship between the sound level and intelligibility, it is expected that the sound level of the recording transmitted may affect the evaluations. It is seen that the necessary importance is not given to the issue of giving sound energy to the listener at the level obtained in the simulations in the listening tests. The aim of this study is to reveal the method that will eliminate the deficiency of sound level calibration in listening tests performed with headphones. It is aimed that the study sets an example for listening tests performed with headphones, so to be able to get accurate evaluations via subjective evaluation tests. For sound level calibration, the sound pressure level transmitted to the listener must be equal to the sound pressure level measured or calculated in the computer models. This also applies to presentations employing both speakers and headphones. Within the scope of the study, a hall is modelled and a listening record was created over the acoustic simulations on this model. The sound pressure level, which is the target value to be measured at the headphone output, is calculated for the listener point through the simulation. Sound level measurements are performed using a head & torso simulator in order to provide sound energy output equal to the sound pressure level calculated in the computer. Thus, the level settings of the system, which are determined to be used in listening tests, can be fixed. Head & torso simulators can simulate the human hearing character and are used for accurate measurements. Using methods without simulator, such as placing a microphone between headphones or placing headphones on the headsets can affect the measurement accuracy. The sound level calibration study for a prepared hall model is shared. In this context, acoustic modelling of the hall is prepared and objective room acoustics parameters, including sound pressure level, are calculated for a specified audience point. Since the real listening conditions in the hall will be conveyed to the subjects, auralizations are carried out using raw speech signals (recorded in an anechoic room) to the simulation. Listening recordings to be used in listening tests are obtained through these auralizations. As a stable signal is required for level calibration, it is not a correct approach to use speech signals in sound level calibration stage. Therefore, white noise is used in this stage. Listening recordings prepared with white noise signal are transmitted through headphones and level measurements are made using a head & torso simulator. The sound level setting of the system is adjusted and fixed until the measured sound level is equal to the target level. Although the details are not included in the study, calibration studies are repeated for different models. The sound pressure level at the headphone output is measured using the head & torso simulator. It is ensured that the sound pressure level obtained in the measurement and calculation in the modelling corresponds to each other. With this method, the target sound pressure level can be read at the headphone output. This will ensure that subjective evaluations are compiled in accordance with the actual acoustic situation and accurate subjective evaluations to be collected. In repeated calibration studies for different hall models, it is observed that the sound level setting determined for one specific model is not suitable for other models. This reveals the necessity of the study. The sound level calculated in the modelling is transferred to the subjects correctly. Sound level control, which is generally not considered in listening tests performed within the scope of subjective evaluations in architectural acoustics, is important for accurate evaluations. It is important to give the correct sound pressure level to the subjects in order to get accurate evaluations in architectural acoustics. The present study emphasizes the importance of sound level calibration in listening tests. As the necessary importance is not given to the calibration stage, the study will contribute to the elimination of this deficiency in subjective evaluations.