Directivity Reference for System Design

by Pat Brown

The key to working in the field of audio and acoustics is to understand the need to establish a reference condition. This allows changes to be assessed objectively. You need a “known” in order to determine unknowns.

An effective site survey is driven by getting the most information in the least amount of time and with the least amount of gear. If one is clever, they can acquire a meaningful set of room data, with very little site time. That’s important, because there’s never enough time.The measurement and design projects over the last year have reinforced to me the usefulness of a dodecahedron loudspeaker for room surveys (Photo 1). The photo was taken at Kingwood Church in Houston, TX. The site visit was part of the Sound System Design in-person seminar.


Dodec Loudspeaker
Photo 1 – A full-range, omnidirectional sound source for room testing.


The Source
The Outline Globe Source is a 12-loudspeaker omnidirectional device. If the loudspeaker baffles were flat, the shape would be that of a dodecahedron. I built a custom woofer to extend the low frequency response. The size was based on a road case that I already owned. A MiniDSP PWR-ICE125 plate amplifier provides the necessary equalization curve and amplifier channels.
I placed the stack at center stage and collected room impulse responses (RIR) at three positions on the main floor:

– Furthest seat (Test Position 3) – 60 ft.

– TP3/2 (which is TP2) – 30 ft.

– TP2/2 (which is TP1) – 15 ft.



Figure 1 – Log-spaced test positions are related by the inverse-square law. This establishes a predictable relationship between the Direct/Reverberant ratio and distance.

These RIRs are produced from log sweep recordings made on a portable recorder (Photo 2). This eliminates the need for a computer  and microphone cables on site. There is never enough time, and sweep recordings yield an RIR with the best achievable signal-to-noise ratio, in the least amount of time. The dynamic range of the RIRs collected in this manner can exceed that produced by a full-blown measurement system. You still need your measurement software to post-process the data. The sweep recording method just provides an alternate way to get the RIR with less gear and setup time. Also, it is not a viable method for system tuning. But for site surveys, it rocks!




Photo 2 – The Roland R26 allows both mono and stereo recordings to be made simultaneously. Mono is preferred for room metrics. Stereo is preferred for convolutions (listening).


The sweeps are processed into room impulse responses (RIR) using the freeware GratisVolver™. The RIRs can be post-processed by any mainstream room measurement program to yield the reverberation time RT, the Clarity C50, and the Speech Transmission Index STI. Here is a link to the RIR for each test position. You can unzip the file to a directory of your choice, and open the WAV files in your favorite measurement program.

What Does It Sound Like?
Here is what TP2 sounds like with the the dodec source. The lack of directivity produces maximum room excitation. The C50 for this position is about 0 dB A-weighted. The STI is 0.56.



Here is what TP2 sounds like with a medium-format line array (Photo 3). The tight vertical pattern greatly minimizes room excitation – even more so if an audience were present. The C50 is 15 dB A-weighted and the STI is 0.93.



I now have an objective reference for the Clarity/STI as a function of distance for a non-directional source (Q=1, DI = 0 dB). The selection of the source position shows how the room is responding to sound energy emitted from the stage. It’s up to the sound system designer to use the data to determine the room’s response to directional loudspeakers strategically placed into the space.



Photo 3 – A balloon pop produces maximum room excitation. Using a dodec instead produces a better signal-to-noise ratio and improved repeatability. The SLS line array is on stage (just to the right of the balloon). The receive microphone is in the audience (not the one near the balloon!).

Why a Dodec?
The advantages of a dodec over a directional source for room measurements include:- The Q (DI) is fairly consistent with frequency (there is none). A directional source will always have a decreasing Q with decreasing frequency, and (probably) an increasing Q with increasing frequency. This behavior would need to be factored into the room model by the system designer. It’s a lot easier to just start with an omni source in the model. Using an omni source makes the worst case condition the reference.- A dodec does not require aiming. It hits everything with sound and only one measurement axis is required. With directional loudspeakers a single measurement axis may miss some important details about the room, and the room may not be fully excited by the source. You will get a higher reverberation level (not a longer reverberation time) with a low Q source, which is a good thing for a site survey.

– The Q (DI) is similar to a balloon pop. If there is no time for RIR collection using sweeps, a simple balloon pop recording can yield the RT, C50, and STI for the three test positions, albeit with much poorer signal-to-noise ratio (SNR) than the sweep recording RIRs. Balloons are cheap, don’t take up much room, and make a good backup plan in lieu of formal room measurements.

– A dodec is small, even with a subwoofer. Mine is packaged “to go,” with road cases, casters, and the amplifier/DSP built into the road case (Photo 4).

– A knowledge of the direct-to-reverberant ratio (DRR) produced by an omni source allows for simple mental calculations of what could be achieved with a directional source. For example, at Kingwood the omni source produced a C50 = 0 dB at 30 ft. Since the C50 increased to 15 dB with the line array, the DI of the line array must be about 15 dB. Caution: The relationship is not always so simple. There are several variables that affect the accuracy so while the concept is always true, your mileage may vary.

– Dodecs and balloon pops produce much more even room excitation than directional sources. This is what you want when determining the RT of the space. Light it up!



Photo 4 – The dodec, sub, and amplifier requires only one trip from truck to room. I can even stack stuff on top!

Handoff to Room Modeling
My three reference RIRs are exactly what are needed for refining my room model, which should produce approximately the same RT, C50, and STI as the measurements at the reference positions. This gives confidence that the room model is a meaningful representation of the actual space. Now I can try my design ideas in a virtual environment, placing loudspeakers and listeners anywhere I please with no heavy lifting. All modeling programs provide an omni reference source for use in simulations. Place it, along with your listener positions, into the model and calculate the C50 and STI for each position. Tweak the surface coverings (and resultant RT) until you get reasonable correlation with your measured data.


 Figure 2 – The source and test positions shown in the room model. The model was drawn in Sketchup™ and imported into CATT-Acoustic™. The room surface absorption and scattering coefficients are tweaked until the RT and Clarity match the measured data. Sound system design takes place in this virtual environment.

The use of an omni source gives a worst case reference for designing a sound system. You can build your own dodec. Here are some past articles on the subject from the SynAudCon Library. And, you can always just pop a balloon and record the room decay in a pinch.

The sound system design objective is to utilize loudspeaker directivity to improve the Clarity and STI. Starting with “no directivity” yields the worst-case scenario as a reference condition. The room modeling program provides a virtual environment to determine the benefits of increasing the source directivity. You can also add more sources to achieve coverage, and assess the determimental effects of the increased reverberant energy.

Geometric acoustics is far from an exact science. By using measured room data to qualify the model, we can have more confidence in the prediction results.  pb