How bad is it? What are the potential solutions? How do we quantify those "solutions?" This blog is going to answer the three questions in detail, and attempt to provide you with an outline of an approach that removes the guesswork and you risking your good reputation.
Our friendly SI has a good deal of experience in these types of projects, and they did do a loudspeaker demonstration at a facility function for this client, as both a service and a proof of concept. The demonstration was successful in showing that speech intelligibility would be substantially improved by using a digital steerable column loudspeaker. Then the SI built an EASE model to investigate where a good location for the column would be in terms of providing good coverage for the specified listener area. Next, I received a call re. possible recommendations on types of and amount of acoustical treatment. They had already provided a rough amount and cost, so this was more of an exercise of confirmation for the SI. My response to the questions provided was a series of questions. What, where, why, and how? The how was, how much is budgeted for acquiring acoustical data? Why guess or simply trust a simulation, when the space exists? Knowing the room volume, and types of surfaces (absorption coefficients), I could use a Sabine or Eyring equation, but that still leaves the variables of how accurate is the data on the surface materials? There is also the case of occupancy, where those equations simply fall apart and become inaccurate, due to non-homgenity of surfaces. Most important, we can quickly acquire some acoustical data and have most of the answers we need, without guessing! Let's measure it!
Acquiring Acoustical Data - Get Our Yer Balloons & Starter Pistols
The truth is, acquiring acoustical data is a simple process. For this project we decided to get multiple data sets that would provide for answers on "how bad is it", while also possibly creating more questions. Here we go!
Let's start with some pictures of the space. In order, the first three pictures are the receive locations for our measurements, named positions R01, R02, and R03, with R03 being closest to our (test speakers) energy sources. Our energy sources are a Dodecahedron loudspeaker (farthest right), a small powered 8" two-way/low Q (middle unit), and a 56" column speaker (farthest to left). All three were augmented by a 15" subwoofer that was placed on the floor. We added the existing loudspeakers to our group of energy sources, and you can see two of three units in the fourth slide, near the the ceiling.
The first thing we can do with this data is do the needed post measurement processing to have all the recordings converted to Room Impulse Response (RIR) files in a .wav format. Since we were in the space, we took the time to simply listen with the best measurement device we can access, our ears. To experience the room again, we can use our RIR's and a convolution software to listen to dry speech, with the room "added in." To better match the on site listening experience, the following convolutions are of the stereo X-Y mic capturing data from the energy sources. Our receive location is what we label as R02, which is the center of the room.
First up, the column speaker, then it's the 8" two-way, and finally the existing system.
What other data do we get from the measurements? More than this forum allows for showing, but lets look at the IEC 60268-16 metric more commonly know as the Speech Transmission Index (STI). This metric is used internationally to determine the suitability (intelligibility of) for sound systems in various applications. For Emergency Communication Systems (ECS), a minimum STI is "the law" in regard to a sound system meeting the requirements as determined by the local AHJ. Let's look at the STI scores for both the existing and column loudspeakers.
Room Decay (Reverberation)
The last piece of measured data to look at for this post, is the room decay time, often referred to at RT60, T60, or T30. These are all one in the same, as being the rooms statistical reverb time. This needs to be considered, as do other room acoustical properties, because not all communications in the space will be reinforced by a high Q column loudspeaker. In this space, two people can't have a conversation once they move more than say 10' apart, without saying "what?" often. Add some other energy into the space, like another pair of people having a different conversation, or many others having conversations …. .well, you get the picture. Here's the room decay times below.
In EASE you can view the statistical room decay (reverberation time) in the form of a graph, as shown below. It is important to note that this data is derived from either the Sabine or Eyring equations, and is ONLY accurate for rooms with homogenous surfaces. If this room had a carpeted floor, or a "soft lid", or any other large surface that wan't a highly reflective surface, the equations would be less than accurate.
If you compare the measured room decay with the modeled room decay graph, you will see some correlation. I could average the data of the octave bands centered at 500 Hz, 1 kHz, 2 kHz, and 4 kHz for the modeled space and compare to the same average for the measured space. Modeled 500 Hz - 4 kHz average is 3.45s. Measured average is 3.51. They are almost the same, but if you look at the data at 250 Hz, you'll see the modeled room has a decay time of 4.4s, and the measured room decay at 250 Hz is at 3.7s. At 8 kHz, the modeled room shows a room decay of 1.25s, when in reality it is 1.7s, and that's a substantial difference!
At this point we could altar the room surface material to get the model closer to the measured data, but that would only be useful for gathering data for the room without any treatment, and/or without a large amount of people. The model will once again become less accurate as we change the surfaces. The answer to this problem is to move from looking at the simulated statistical data that is based on the geometric model, and to move to a more advanced acoustical study by using either ray tracing or particle emissions in the modeling program.
Stay tuned as we cover that in the beginning of our next post. We will then also introduce the proposed loudspeakers into the equation, and derive new data sets!
Originally published on: 7/22/2014