by Pat Brown
In this article, Pat Brown addresses the limiting factors in room modeling.
It’s been about two decades since the the first use of computer ray tracing to approximate the room impulse response (RIR).
Here is an excellent overview of the history.
RIR prediction was a natural extension of previous work in the field of optics – a very important point that I will come back to.
It’s been about three decades since the first use of measured coverage isobars to predict the SPL over an audience area. The loudspeaker was the Bose 802, and the room modeling “program” was the Prohs Global Mapping program. This later became PHD (Prohs-Harris Design) program for the DOS operating system.
These two fields naturally merged for use in sound system design, where isobar mapping could determine the direct field coverage of a loudspeaker, and ray tracing could estimate the RIR. Decades of continued development have given us the powerful, highly graphical modeling tools that we have today.
It is logical that the loudspeaker data itself would advance along with the modeling tools. Both angle and frequency resolution have reached their practical limits, and have “settled” to the generally used 5 degree,1/3-octave format with higher resolutions possible for special cases. For an overview of loudspeaker polar measurements, see Principles of Audio Online Training – Modules 100_45a and b – Loudspeaker Fundamentals.
Before we revel in our modern capabilities, it should be remembered that the limitations of the algorithms used in ray tracing predictions have not changed fundamentally since their first use in the 1960s. The major limitation is that these techniques stemmed from the field of optics, and the sound wave behavior is considered as a ray of light. None of the advancements made since the 1960s have changed this. In fact, as time has passed, the limitations of geometric acoustic assumptions have largely been forgotten and today many fail to realize that this puts a fundamantal limit on the accuracy possible in computer room modeling with regard to acoustics. Here is a recent paper on the topic.
There is a growing demand for complex (both magnitude and phase) loudspeaker data that fully resolves the low frequency response of the loudspeaker. While I am not against either of these, I am quick to remind users that these are valid only for the direct field which represents only a small part of what a listener actually hears in a room.
In summary, while computing power has increased dramatically since the early days of computer ray tracing, the underlying limitations are still in place. For the direct field, data and predictions below 100 Hz and above 10 kHz should be met with skepticism, with frequency resolution being the major limiting factor at low frequencies and component variations and diffraction effects the limitations at high frequencies. The limitations for predicting reflected sound are even more constrained, being confined to the 125 Hz – 8 kHz octave bands, and a number of caveats regarding room detail vs. frequency.
The reader is reminded that when a process involving several factors is limited by one of them, advancements in the others do not produce improved results from the process. Know your limitations and work within them. PB