1374 Chapter 35
triangle spans multiple voxels. Thus a ray-triangle inter-
section is only considered if it occurs within the bound-
aries of the current cell. The grid traversal continues
until a hit point is found or the ray exits the simulation
box.
The software implementation in AURA was care-
fully designed to facilitate vector optimization on
SIMD-capable processors—i.e., to minimize branching
and optimize instruction scheduling—in particular it is
easily transferable to programmable graphics processing
units. Using the Intel C++ optimizing compiler on a
Pentium4 processor, the grid algorithm applied to real-
istic test cases with more than 10,000 triangles can be
up to five times faster than a typical HBV method and,
of course, orders of magnitude faster than the linear
search method.^54
35.3.2.7 Features of All These Methods
All of the ray-tracing or image modeling methods that
calculate impulse responses have to take into account
the directivity of the sound sources and the absorptive
and scattering characteristics of the surfaces encoun-
tered en route from the source to the receiving point.
The design program must allow the user to designate
specific surfaces/planes as being reflective or non-
reflective. This will make it possible to simulate not
only sound-reflecting walls, but also simplified floor
planes—i.e., which in reality are complex shapes such
as seating areas or orchestra stages or pits. At present
these methods use statistical absorption factors that are
readily available instead of angle-dependent ones (for
the latter no sources are available in textbooks), as well
as some diffusion factors estimated by rule of thumb
and/or specially measured diffusion factors. The diffrac-
tion behavior is still in the academic stage and some
program approaches are using FEM or BEM
methods26,35, see Section 35.1.4.2. and Fig. 7-46. Addi-
tionally the dissipation of sound energy in air—i.e., the
frequency-dependent air attenuation—must be consid-
ered too.
A library of potential natural sound sources must be
available, such as the human voice and various orches-
tral instruments/sections to go along with the electroa-
coustic sources/loudspeakers that should include the
sound power level and directivity of these sources/loud-
speakers. As a result of all these calculations you get
impulse responses or energy-time curves as shown in
the following figures.
The program CATT-Acoustic^36 shows the complete
echogram with all input data (room, loudspeaker,
listener position, frequency) and presents all resulting
room acoustic measures this way, Fig. 35-41. With
EASE and AURA it looks different, Fig. 35-42.
The calculated energy-time curve should be able to
be stepped through reflection by reflection, with the
appropriate rays and surfaces being highlighted to indi-
cate the ray’s path and the surfaces it encounters en
route from source to receiver, Fig. 35.43.The software
should indicate median/lateral/horizontal positioning of
energy arrivals (and relative magnitude as well) at the
receiver’s location, Fig. 35-44.
Additionally, a simulation program should provide
the capability to calculate early/late energy ratios. It is
important to be able to set the early/late transition time
and also to select the cutoff time for the late energy inte-
gral, Fig. 35-45.
The software’s ray-tracing or image modeling
method of deriving an energy-time curve should
provide the ability to indicate interaural cross-correla-
tion (IACC) as well as lateral energy coefficient predic-
tions at specified listener positions.35.3.3 AuralizationThe simulation program must have the ability to transfer
the calculated impulse response curve to a postprocess-
ing routine that will be used to auralize the room
time/energy data with anechoic music or speech source
material. Of course the routine must generate a binaural
data file in WAV-format or other computer sound file
format in common use, Fig. 35-46.35.3.4 Sound Design35.3.4.1 AimingAiming the individual loudspeakers is an important
operation insuring the proper spatial arrangement and
orientation of the sound reinforcement systems. Once
the corresponding room or open-air model is at hand
and the mechanical and acoustical data of the loud-
speaker systems is exactly known, these systems are
approximately positioned and then one may begin with
the fine tuning of the same. A modern simulation pro-
gram uses a kind of isobeam/isobar method to initially
aim the loudspeakers, preferably utilizing the –3 dB,
–6 dB or –9 dB contours.
Fig. 35-47 shows various types of projection of the
–3, –6, and –9 dB curves into the room. On audience
areas one can then also see superposed aiming curves
for multiple loudspeakers, Fig. 35.48.