A DSP based Equalizer and Sound Level Control System The Ecudap

A custical Isolation
Figure 2: Block Diagram of the System In order to design the Ecudap , we must analyze the typical transfer functions of all the blocks in figure 2. We also select a suitable function F(x) to transform the sound levels into the perceived levels. Then we will go back to (1), and try to design an algorithm to solve the problem. We start by analyzing the human hearing system in order to select F(x). Although, we have only a partial understanding of it some fundamental characteristics are well established. One important
A udio Signal Source Ecu dap A udio A m plification System R oom R esponse
x(t) Audio Source x1(t) The Ecudap x2(t) x3(t) x4(t) Room Walls
g
Figure 1 : Diagram of the Problem From this figure , we can derive a model in terms of signal processing blocks as shown in figure 2. The perceived level is estimated as a function of the signal level ( we say F(x) ). If Th is the threshold on the perceived level in the adjacent room, the problem can be stated now as to maximize F(x3(t)) with the cቤተ መጻሕፍቲ ባይዱnstraint that F(x4(t)) must be smaller than Th :
Figure 3 : Measured Acoustical Isolation of a Wall From figure 2 we can write :
X 4 ( w) = ( H ( w)∗ L( w)∗ A( w))∗ X ( w) (3) X 3 ( w) = ( H ( w)∗ L( w))∗ X ( w) (4)
Maximise F(x3(t)) The Objective (1) Conditioned to F(x4 (t)) ≤ Th
x(t) x1(t) x2(t) x3(t) Perceived L evel Inside F(x3(t)) x4(t) Perceived L evel in the adjacent room F(x4(t))
A DSP based Equalizer and Sound Level Control System : The Ecudap
P. ReviriegoVasallo, J.M.Santos Suárez ,and Ramón García Gómez e-mail : revirieg@gtd.ssr.upm.es Dep SSR, ETSI Telecomunicación, Universidad Politécnica de Madrid Ciudad Universitaria s/n, CP28040 Madrid Spain J.E. Arias Puga Proceso Digital de Audio S.L. Miguel Iscar, 16 Portal 4, 2 CP 47001 Valladolid Spain ABSTRACT We describe in this paper an application of Audio Signal Processing and its real-time implementation. The device, that from now on we name as Ecudap, monitors the audio signal which feed the loudspeakers and ensures that, on a short term basis, the output power in each 1/3 of octave band is below a specified threshold. static equalization is also provided. The system guarantees that the sound emitted from an audio amplification system equipped with the Ecudap is controlled in order to avoid annoying sound levels in adjacent rooms or buildings. The algorithm that controls the sound level in each band is based on the measure of the short-time energy in such band and in adjacent bands in order to minimize signal distortion and maximize the overall sound level. The Ecudap also monitors the activity of the Audio Amplification System , and keep statistics of the attenuation introduced , the maximum measure sound level , the average measure sound level, and the time and date of each measure. This information is maintained in Non Volatile Memory ( NVRAM ) and can be downloaded to a PC to check the operation of the system. The supporting hardware is based on a couple of Digital Signal Processors ( DSP ) ,one per audio channel. A complete description of the hardware architecture is also included in the paper. INTRODUCTION The main goal of our system is to limit the sound level emitted from an Audio amplification system so that it will not produce annoying disturbances in adjacent rooms or buildings. Additionally , we want that given this restriction, to maximize the perceived sound level inside the room .We start with a diagram of the problem shown in figure 1. In this diagram , we observe the elements that determine the perceived sound level in side the room and in the adjacent ones : the audio signal itself , the amplification chain , the room acoustic response, the acoustical isolation of the walls and finally the hearing system of the person .
aspect is the frequency resolution of the ear , that is the ability of the ear to differentiate signals which are close in the frequency domain. It has been established that the ear´s frequency resolution is not uniform. It is higher at low frequencies and decreases approximately logaritmically with frequency. An standard approximation is to consider the one-third of octave bands starting at 40 Hz. Another characteristic of the hearing system is the variation of its sensitivity with frequency. That means, for example ,that a tone of a given level is perceived with different intensity depending on its frequency. The ear has its higher sensitivity at about 2 Khz and decreases at lower and higher frequencies. The sensitivity is also a function of the level of the tone. These two characteristics of the human hearing, system are taken into account by the models used to evaluate the perceived noise. The perceived level is computed as the sum of the levels in each 1/3 of octave band weighted by the ear´s sensitivity in that band. A frequently used measure of the perceived level is the decibel A [2].. So given a short-time measure of the power of x(t) in each of the 1/3 of octave bands P(i) we can define the function F(x) as :