Page 14 - Summer2017

Acoustical Measurements with Smartphones
may also include its own algorithms for manipulating the input audio signal to optimize it for telephony or automatic speech recognition. If an app developer has no way to cir- cumvent OS-level signal-processing schemes or ensure the desired behavior of the various hardware components in the signal path, then it may be impossible to provide a reliably accurate or precise measurement solution by relying on the existing audio signal path. Fortunately, some OS-level APIs include options to bypass any extra signal pro-
cessing that would significantly corrupt or dis- tort the incoming audio signal (whether from the built-in microphone or some other audio input source). Of course, a solution that by- passes the well-established audio signal path in favor of a proprietary one could be devel- oped. This kind of proprietary solution would require additional external hardware and would necessarily come at a higher cost.
A real-world example of the impact an OS can have on the signal path is shown in Figures 2 and 3. In some iPhone OS (iOS) versions be- fore iOS 6, a high-pass filter was applied to both the built-in microphone signal and the headset microphone signal. The low-frequen- cy effects of this filter can be seen in Figure 2. For measurement purposes, the microphone inputs of the iPhone were of limited value for frequencies below about 200 Hz. When iOS 6 was introduced in 2012, an API was added that allowed app developers to enable a so-called “measurement mode.” The effect of enabling measurement mode, which also disabled AGC for the same input signals, is shown in Figure 3.
Calibration
An accurate sound level measurement re- quires that the smartphone solution be cali- brated. Calibration refers to the comparison of a sound level measured by the smartphone- based meter and a properly qualified SLM with a known degree of accuracy. Once such a com- parison is made, input sensitivity values with a smartphone app can be adjusted to produce results that match those of the reference SLM.
1 SignalScope Pro is a product of Faber Acoustical, LLC, which is owned by the author.
A digitized signal is represented by a series of numerical val- ues that are less than or equal to some full-scale value (the maximum numerical value that can be represented by the ADC or the computing platform in which the digital val- ues exist; FS). When calibrating the built-in microphone of a smartphone, the sensitivity of the microphone can be de- termined in pascals relative to the FS or Pa/FS. Once that sensitivity is known, it should be simple for a smartphone
Figure 2. One-third octave frequency response of the iPhone 4S headset microphone input, with measurement mode disabled (as it was in earlier versions of iOS). This measurement was made with SignalScope Pro.1 Leq, equivalent (nonexponential) time weighting used to calculate the root-mean-square (rms) level of the signal in each one-third octave frequency band. Republished from Faber Acoustical (2012), with permission.
Figure 3. One-third octave frequency response of the iPhone 4S headset micro- phone input, with measurement mode enabled. This measurement was made with SignalScope Pro. Republished from Faber Acoustical (2012), with permission.
12 | Acoustics Today | Summer 2017