Developing A Sound Level Measurement Application for the Amulet Platform

Amulet Device Internal View

What is Amulet? The Amulet is a wearable device platform that focusses on utilizing a low power hardware architecture along with an efficient operating system that runs on an event-driven programming model. A brief overview of its characteristics are given below:

▪ The Amulet device is capable of using multiple sensors located on the device, along with off device inputs through Bluetooth Low Energy technology.

▪ In order remain an efficiently low powered, the Amulet platform emphasizes robust app isolation mechanisms, and developer tools for optimization.

▪ At the heart of the Amulet device, it has a TI MSP430 microprocessor with 2kB of SRAM and 128 kB of FRAM. It uses a Nordic chip as a modem for communicating with Bluetooth Low Energy peripherals.

▪ The Amulet platform is open source, both in terms of software and hardware. Its slated to be released at SENSYS in November.

P. Yerabati Venkata S. Hearndon J. Hester J. Sorber

The Amulet Platform Overview

MIC MSP430 ADC The main goal behind the sound level application is to develop the capability to distinguish between different sound levels. In order to simply the hardware testing, a separate microphone breakout board connected to Launchpad was used to experiment and gather data. The idea remains the same however, which is to determine the sound levels in a range of decibels. This can be understood be observing the decibel range chart, which shows some examples that can be recognized through daily life applications.

As the sound compression and rarefaction waves travel and reach the mic, a voltage vs time wave is created much like the graph on the left. This is done through various methods, but usually the microphone uses a thin membrane that responds to the waves to induce a voltage using electromagnetic induction.

UsesThe capability to determine the sound level of a particular environment can be in many ways. For example ▪ To alarm the user if they are situated in an environment with high sound levels that can cause

damage.▪ The information can be used to understand what type of surrounding the user is located in, and

how it effects various aspects of their health.▪ This can also be a very good way to test the microphone sensor, and understand it’s limitations in a

low power consumption platform.

Approach Results

The MSP430 utilizes a 12bit ADC, and therefore represents the voltage proportionally between a range of 0 to 4095. The supply voltage for the microphone is 3.3V. Therefore the ADC splits the range 0V – 3.3V into 4096 parts, and a single part represents the resolution. The quantity we are trying to obtain using this information however is the sound pressure level

Sound Pressure LevelSound Pressure Level (SPL) is a quantity often used to measure sound. It is a relative quantity that is measured between a sound pressure and a fixed reference pressure. The sound pressure is proportional to the voltage, and hence the ADC value.The ADC value was first calibrated with a commercial SPL meter to be used as the reference ADC value. The following formula was used to calculate the SPL

𝑆𝑃𝐿 = 20 ∗ log𝐴𝐷𝐶

𝐴𝐷𝐶𝑟𝑒𝑓+ 𝑆𝑃𝐿𝑟𝑒𝑓

The sound levels calculated from the different methods seemed to differ in varying quantities depending on the sample size, noise level and the type of noise. However, the sound level calculated from the RMS value seemed to be the most consistent till about 80dB.Also, there seemed to be some offset value depending on the sound level range (which seemed to be around 50 – 60 dB). Below which the offset would differ compared to the range above. There are still some issues that need to be addressed in order for this application to work properly on the Amulet:▪ The Amulet supply voltage is 2.8V instead of 3.3V. This reduces the overall range in decibels

that would be covered. ▪ The application for the Amulet can only take 10-15 ADC samples. Therefore, the application

would need to take fewer samples while being accurate.

Analyzing ADC Data

The individual ADC values did indeed go up or down whenever there was an increase/decrease of sound. However, calculating the sound level in decibels proved to be harder as these values fluctuated frequently. Therefore a set of ADC values were analyzed instead of individual ADC values, and this was done in a few ways:

▪ Individual ADC Valueso Fluctuated too often, and most of the data were not reliable for an

accurate calculation of sound level.

▪ Averageo Seemed more reliable. However did not account for large and sudden

changes in sound accurately.

▪ Peako Reliable most of the time, but could not account for large and sudden

changes all the time.

▪ RMSo Seemed to be the best fit. This method would result in a decently

accurate sound level that stayed consistent.

Acknowledgements I would like to thank my advisor Dr. Sorber, and my mentors Steven and Josiah for helping me with this project. I would also like to thank Dr. Brian Dean and the NSF for giving me this opportunity.

TI MSP430 Launchpad

ADMP401

SPL Meter