by Pat Brown
The ECS workshop included an acoustical room survey of an historical church.
Our recent Emergency Communication Systems – Design course included an evening site visit to a local historic church for an acoustical room survey. Everything is historic in Alexandria, VA, but St. Paul’s Episcopal traces its roots to 1817 when it was designed by Benjamin Henry Latrobe. Other projects to his credit were the U. S. Capitol and the White House.
Most of the facility has been preserved in its original state, so we couldn’t pass up an opportunity to experience the same acoustics as Thomas Jefferson and others. We were fortunate to have the system designer – Keith Dorset – in the seminar. Keith made the arrangements for the after-hours visit and gave us a great tour of the space.
Photo 1 – System designer and tour guide Keith Dorset.
Photo 2 – Sander van Wijngaarden (right) provides some instruction on STIPA testing.
The objectives for the evening were established in advance:
1. Collect room impulse responses (RIR) from a low directivity source at three log-spaced test positions.
2. Use this data to determine the reverberation times and clarity scores at the test positions.
3. Use these measures to create a reasonably accurate room model for acoustics simulations and sound system design.
The ECS event provided the opportunity to add an interesting additional step – perform STIPA measurements using hand-held instrumentation for comparison with the RIR data. ECS staffer Sander van Wijngaarden brought (20) Bedrock SM50/SM90 STIPA meters with him from his headquarters in the Netherlands.
Two low Q sources were used – a balloon, and a dodecahedron augmented by a LF box (Photo 3). Figure 1 is an ETC of the balloon pop at 8 m. Figure 2 is the same for the dodec source, using a log-sweep stimulus and GratisVolver for post-processing.
Photo 3 – The room was excited from the same position using a balloon pop and a dodec loudspeaker.
Figure 1 – Energy-Time Curve of the recorded balloon pop. I intentionally made the time span extend beyond the room decay to make the noise floor visible (ARTA).
Figure 2 – ETC made from a log sweep at TP2 (8 meters from dodec source). The 30 dB increase in SNR is obvious. The room has fully decayed at just under 2 sec.
A comparison of the two plots reminds us that
1. A balloon pop is better than nothing. This was a very quiet space, and we barely achieved sufficient SNR to determine the T30 as per ISO3382. The link provides a nice overview.
2. Some grunt work is required to drag a dodec, lf box, and DSP/amplifier into a space. But, the payoff is MUCH better data than can be achieved by an impulsive source.
3. Ironically, in this quiet space the balloon pop yielded pretty good RT data (Figure 3). If that were the only objective for the test, then a 25 cent balloon was the fastest way to get there.
Figure 3 – The balloon pop (left) slightly under-resolved the reverberation time as compared to the log sweep through the dodec (right).
I would normally use the B-Format rig for acoustics studies such as this. This did not really fit the context of an ECS workshop, so I used a hand-held recorder for data collection. The Roland R26 produces omni and XY-Stereo files from the same recording.
Photo 4 – Recording the log sweep at TP3 using the hand-held recorder. The sweep recordings are post-processed in the freeware GratisVolver to produce the RIRs.
The STIPA meter is a purpose-built, self-contained instrument that can collect the needed room measures in real time. The modules used were
1. RT Timer – Provides RT data by timing the decay from an interrupted noise source – in this case pink noise played over the dodec rig.
2. Noise Analysis – Provides noise data at 1/1 or 1/3-oct resolution.
3. STIPA – Provides an intelligibility score that includes the effects of noise and reverberation.
Photo 5 – Sander collects a STIPA measurement using the SM90 meter.
The Test Positions
My standard practice is to place the source on the stage, and strategically select 3 or 4 test positions in the audience area (Figure 4). The use of a non-directional source assures that any position selected will be in the coverage pattern of the source. I used the main aisle for convenience, and selected the longest distance within the audience plane – 16 m for Test Position 3. Halving this distance put TP2 at 8 m, and halving this distance put TP1 at 4 m. The log spacing produces a somewhat predictable direct field level that falls 6 dB at each successive position. This relationship facilitates interpolation between the positions, as well as assessments as to the “quality” and diffuseness of the reverberant field.
The Room Model
A simple wireframe was generated in Sketchup using the major measured dimensions of the room. A simple model can be sufficient absorption coefficient selection based on the measured RT data. A spherical source is placed in the virtual model, along with the specific test positions used on-site. This allows measured vs. modeled comparisons that guide the remainder of the modeling process. Details can be added as needed, and the process is complete when you are satisfied that you have considered the important room and system details.
Figure 4 – A very simple model (CATT-Acoustic) of the space is adequate for considering direct field coverage, direct-to-reverberant (statistical) ratio, and signal-to-noise-ratio. It also documents the source location and test positions. Additional detail can be added for specific geometric acoustics simulations.
The RT and noise data only serve to quantify the room. They tell us nothing about the actual performance of a source. Below is a matrix that shows the Clarity score (C50) and STI for each test position as calculated from the RIR data. Since this is for a non-direction source, it represents the worst-case performance for a loudspeaker at this location. The objective of the design process would be to select and place a loudspeaker(s) that improves the C50 and STI at all seating positions. This could be done “in situ” by auditioning actual devices in the room, or it can be done in the virtual world of acoustics modeling.
Figure 5 – Performance data matrix for the dodec source, measured at 4, 8, and 16 m. Post-processing was done in ARTA.
The RIRs can be downloaded for your own evaluation. These are XY-Cardioid files (stereo). If you just want to have a listen, I have convolved each with a short speech track. You can play them below.
The design of a sound system must begin with information about the room. For emergency communication systems (ECS) we begin with reverberation times and noise data at 1/1-octave resolution. These can be collected during a site survey using hand-held instrumentation as described above. RT data requires that the room be stimulated by pink noise played from a low-directivity loudspeaker capable of producing sufficient level to dominate the ambient room noise. If there is no time for that, some balloon pops can be recorded, saved as WAV files, and post-processed to get the RT in an acoustics measurement program.
The most complete documentation of the room’s acoustics is from the room impulse response, as collected using log sweep recordings as described above or using an acoustics measurement software/hardware platform such as ARTA, EASERA, RoomCapture, Smaart, or others. Armed with RT and noise data, a sound system design program can be employed to select and place loudspeakers.
This represents a minimalist approach to ECS design, as might be used by the fire alarm industry. The sound reinforcement industry offers more advanced tools to take the process further. These include full-blown acoustics measurement platforms as listed above, and acoustics modeling programs that perform geometric acoustics simulations. These include CATT-Acoustic, EASE, Odeon, Modeler, and others. The SR industry also offers a broad assortment of directional loudspeakers that can outperform fire alarm loudspeakers with regard to speech intelligibility. These are now allowed in difficult spaces as per recent changes to the NFPA72 code. pb
Click here to add your own text