ECE 4901 - Fall 2016 Team 1714: Smartwatch-like Device and Apps for Continuous Glucose Monitoring Evan Brown - Electrical Engineering Magda Kaczynska - Electrical Engineering Brian Marquis - Electrical and Computer Engineering Ahmed Sugulleh - Electrical Engineering Advisor: John Chandy Biorasis Michail Kastellorizios michail@bio-orasis.com
Introduction The Smartwatch-like Device and Apps for Continuous Glucose Monitoring project is sponsored by Biorasis, located in Storrs, Connecticut. The company is developing a wireless, needle-implantable biosensor, which ultimately will be used for real-time, continuous glucose monitoring. The size of the device is 0.5mm x 0.5mm x5mm and is comparable to a grain of rice, as seen in Figure 1. The purpose of the device is to help patients with diabetes management, by allowing individuals to monitor their glucose level at any time, and respond with treatment when needed. Biorasis wants to develop a user-friendly smartwatch-like device to interface with the biosensor. The device will communicate with the sensor through the patient s skin, sending commands to the sensor and receiving data optically. The user should be able to adjust the settings of the smartwatch such as the frequency of glucose readings. This device should be capable of storing glucose data points, exporting said data when prompted, and plotting glucose levels over time. The smartwatch must be compatible with Biorasis s biosensor and might also collect and display additional user information such as the patient s temperature and hydration level. Similarly, Biorasis needs a bench testing configuration with the at least the same functionality and capabilities as the smartwatch-like device. The test bench configuration should be easily modifiable to allow testing on multiple sensors simultaneously. In addition, the device needs to communicate with an external computer to process the data and possibly interface with an insulin pump to maintain a steady glucose level automatically. Figure 1: An image of Biorasis s glucose sensing device to illustrate its size.
Background As shown below in Figure 2, the glucose sensing device that Biorasis has developed is comprised of four major components: the photovoltaic cell, the CMOS circuitry, the glucose sensor, and the infrared LED. Figure 2: This diagram outlines the interaction between the desired smartwatch device and the glucose sensing module. The smartwatch will provide power to the glucose sensing device when a measurement is to be taken by shining light upon the photovoltaic cell using an LED array. When powered, the device is able to output a frequency-modulated infrared signal through its LED based on the level of glucose in the user s bloodstream. The smartwatch is the responsible for capturing the IR signal and measuring its frequency which can be used to determine the glucose level of the bloodstream. Biorasis has laid out some basic requirements for this project which must be fulfilled. For instance, the design must be able to measure the signal frequency and also analyze/process that same input. In addition, the smartwatch must have some form of wireless communication capability so that it may send data to a computer to eventually interface with an insulin pump. The final design must also be an appropriate size and weight for a smartwatch. Finally, Biorasis has already developed the LED array and IR sensor modules needed for the design. Therefore, the smartwatch needs to be compatible with those existing components. To improve the design process, some soft constraints were imposed on the project. For example, the design should be rechargeable and optimize battery-life as much as possible. Finally, since Biorasis has not given a definite development or per unit cost specifications, the budget goal will thus be to keep the cost below $1000.
Solutions: I: Modification A potential design option would be to modify an existing smartwatch to satisfy the project requirements. However, after some research, it was found that this would not be a feasible option for several reasons. First, the cost of an individual retail smartwatch can range from $250 to $500 which could easily consume much of the self-imposed budget. In addition, the design needs to interface with the LED array and IR sensor modules that Biorasis has already developed. Integrating such elements with these retail smartwatches would not be feasible; the chassis would need to be replaced entirely. Furthermore, existing smartwatches are not designed or meant to be programmed. Therefore, it would be extremely difficult or even impossible to try to load upload code to the microprocessor. As a result, it was decided that modifying an existing smartwatch would not be practical and it would be better to use a custom embedded system instead. II: Custom Embedded System Instead, a custom embedded system will be designed in order to fulfill the design requirements. That is, the system must be able to measure a signal s frequency, drive a display, transmit data to a computer, and log information. To better understand the needs of this design, some early prototyping was done with an Arduino Mega and an Adafruit LCD screen. Figure 3: This picture shows the early prototyping done with an Arduino Mega and the Adafruit screen. This process outlined the importance of having an available graphics library to help expedite the design. Furthermore, this exercise helped to highlight some of the intricacies of measuring a signal s frequency while trying to perform other tasks simultaneously on a single microcontroller. That is, for the most accurate frequency measurements with a typical scheme where one counts
the number of rising edges in an allotted time period, it would be best to have the smallest time increment possible. However, such a routine essentially always occupies the microcontroller so other processes would never get a chance to execute. Therefore, some research and testing will need to be performed to find an adequately accurate frequency calculation algorithm in which the other desired processes can also run. Figure 4: This high-level block diagram shows the different major modules of the proposed smartwatch design and how they interact with each other. Microcontroller Selection In order to select the proper microcontroller for this design, some preliminary filtering was performed to find the components with the necessary features. For instance, in order to interface efficiently with most display modules, SPI (Serial Peripheral Interface) is used. Thus, only microcontrollers with SPI capability were considered. In addition, to interface with some temperature sensors and RTC (Real Time Clock) modules I2C, or Two-Wire Interface, is necessary. Likewise, UART/USART (Universal Asynchronous Receiver Transmitter/ Universal Synchronous-Asynchronous Receiver Transmitter) was needed in order to interface with some serial wireless communication protocols such as Bluetooth. Since measuring the battery state of charge would be a desired feature, ADC (Analog to Digital Conversion) was important. For the input voltage to the device, 3.3 volts maintains a proper balance between performance, size, and power. That is, although lower voltage chips are available, one cannot clock them as quickly with a lower input voltage. Also, higher voltage batteries would be larger in physical size and, thus, not appropriate for a watch device. Finally, since board space is likely to be a major constraint in this design, a small device surface-mount package of 32-TQFP was selected. This chip is often 7mm x 7mm in size and can be soldered by hand.
After narrowing the field down to only microcontrollers which have the aforementioned capabilities, the selection was made on the basis of several optional features which distinguished devices from one another. For example, a major consideration was the clock speed at 3.3 volts; higher clock speeds can help improve the frequency and, therefore, glucose measurement accuracy. In addition, in the interest of maximizing battery life, a device which pulls the less current would be more desirable. Finally, some of the physical memory specifications were compared to see which components had more RAM, program memory space, and even EEPROM (Electronically Erasable Programmable Read-Only Memory). Accounting for all of these factors, the ATMEGA8A was chosen because it was the only remaining microcontroller which could be clocked at 16MHz with an input voltage of 3.3 volts. Biorasis has stated that the design should optimize performance over battery-life so having a slightly higher current draw for this increase in performance is fine. Wireless Communication Selection Some form of wireless communication will be necessary to allow data to be transmitted to a computer from the smartwatch in order to eventually interface with an insulin pump. As outlined in Figure 4 below, several different wireless communication protocols were considered. Figure 5: This table outlines the major differences between some popular communication protocols on the basis of data rate, range, and power consumption (Source: https://learn.sparkfun.com/tutorials/bluetooth-basics) The most important protocol features to this design were the power consumption and the range. Since the smartwatch is battery-powered, a low-power protocol is needed to transmit data without draining the battery too quickly. In addition, the physical range of the transmitter does not need to be incredibly large. If a user is in or around his/her house, the smartwatch should be able to reach the computer that interfaces with his/her insulin pump. These parameters narrowed the selection down to Bluetooth Low Energy (BLE) and ZigBee. However, since the ZigBee interface would require additional hardware on the receiving end as well, Bluetooth Low Energy was chosen.
Display Selection For this smartwatch-like device design, some sort of screen was needed to display data collected from the biosensor: both numerical values of glucose readings and a graphical representation of that data over time. Biorasis has also requested that the device be capable of accepting user input in order to change glucose reading settings. Some of the methods for accepting user input include buttons, a touchscreen, or even Bluetooth. Buttons were not feasible as a user input option because they increase the physical size of the design making it too bulky. Touchscreen and Bluetooth were other options to accept user input and will be implemented in the design. There are three main options for displays: resistive touchscreen, capacitive touchscreen, or no touchscreen. A resistive touchscreen is the most common type available. It consists of two thin layers, and applying pressure to the screen touches the two layers, passing voltage in a specific location on the screen. Similarly, a capacitive touch screen is made from one layer, usually glass, coated with conductive material. A human body is conductive, therefore touching a capacitive touchscreen changes its electrical field, marking a location on the screen. Capacitive touchscreens are more accurate, and are used in smartphones, as well as tablets, and are also more expensive. Resistive touchscreens are still precise, and would work as well as a capacitive touchscreen in this application. The size of the screen was an important factor in the selection process. Biorasis did not set a specific requirement for a display size, but the dimensions of the display should be comparable in size to smartwatches currently on the market, which range from 1.5-1.7 diagonal. Power consumption was another major consideration in the selection process. The display will draw the most current out of all of the design components, and since the smartwatch-like device will be powered by a battery, it should last as long as possible. The display selected for the design should have an available graphics library because creating a custom graphics library will be difficult, extensive, and beyond the scope of the project. The display should also be capable of communicating via SPI and I2C. In addition, the cost of the display was another factor that had to be considered in the selection process. Ultimately, the 2478-Adafruit display satisfied all of the design specifications, the screen itself is 2.4 diagonal, it has a resistive touchscreen, a built-in microsd card reader/writer, low power consumption and an input voltage of 2.5-3.3V. Battery Selection Based on the specifications given, a power supply was needed to produce a voltage of 3.7V with a capacity of approximately 315 mah. Furthermore, the physical size of the power supply was considered when selecting the appropriate battery. After filtering by the above criteria, only three viable options remained. One such option was the Samsung smartwatch battery supply. This battery was desirable because of its small physical size and large capacity. However, this option is also the most expensive. Another battery option was a product called OBBEY. This battery would be sufficient for a breadboarded design, but it was probably too big
to incorporate into a compact smartwatch. The last option, which was found to be unfeasible, was rechargable coin cell batteries. These batteries had a very low capacity of only 40 mah which would require multiple of these batteries to be used to provide a comparable battery-life to the aforementioned items. Using multiple of these batteries would take up too much valuable board/enclosure space to be a viable option. Considering these different power supply options, the Samsung smartwatch battery was selected because of its small physical size and high capacity. Power Stage Selection Once the battery selection was complete, an appropriate charger was needed to be able to recharge the smartwatch after extended use. Rather than designing the battery charger from scratch, which would be impractical due to potential safety issues and development time, it was more practical to purchase a pre-existing charger. After doing some research, four options were of a small enough size and sufficiently low cost to be considered. One such option was the corresponding battery case cover to the Samsung watch. This would be a favorable option due to its small size as well as its compatibility with Samsung smartwatch battery. Another option which was considered was the basic LiPo charger from Sparkfun. This charger would be a viable option, however, one would need to configure this chip to interface with the Samsung smartwatch battery. As for the other two options that were found, they could be used in this design, but they are far more expensive. Therefore, the charger corresponding to the Samsung smartwatch battery will be used in this design. The purpose of the DC-DC converter in the design is to maintain a steady input voltage of 3.3V for the components from the battery. The main characteristic examined to choose the converter was the amount of space it would occupy on the board because the design will not need to pull much current from the converter. Five different products were considered for the DC-DC converter. That list was narrowed down to the LM317. It was decided that the LM317 was best suited for our design because this component required the least amount of board space (just the chip, two resistors, and perhaps a capacitor) compared to a standard buck converter which would require a large noisy inductor component. Enclosure Once the breadboarded prototype is behaving properly, the design should be moved to a more compact PCB solution so that it can fit within a customized smartwatch enclosure. The enclosure must be able to contain all of the different components such as the display, the microcontroller, the battery, the IR sensor, and the LED array. However, it is important that the chassis does not obstruct the LED array so that it can operate properly. In order to design such an enclosure, Solidworks, a three-dimensional modelling tool, will be used to fashion an appropriate chassis for the smartwatch. Once this is complete, Biorasis has a 3D printer that can be used to print out the design.
Budget Totalling the approximate per unit cost of each of the components yields the following budget sheet shown below in Figure 5. Figure 6: This table shows the approximate per unit cost of each of the selected components in this design as well as the total cost. This total cost is about $100 which is well under the self-imposed $1000 budget. Of course, there are other costs that will need to be accounted for such as shipping, PCBs, passive components, solder, etc. This budget is relative to change as certain components may be switched for equivalent, most cost-effective alternatives. Conclusion In summary, the ATMEGA8A was chosen as the microcontroller for this design because of its higher clock speed when powered at 3.3 volts. For the wireless communication protocol, Bluetooth Low Energy was selected because of its low power consumption and compatibility with devices. Furthermore, the 2478-Adafruit Screen was favored for the display option because of its available touchscreen, microsd card reader/writer, small size (2.4 diagonal), lower power consumption, and input voltage range (2.5-3.3V). The Samsung smartwatch battery was also chosen because of its small physical size and high capacity; this battery s corresponding charger was used as well due to its small size and compatibility. In addition, the LM317 device was selected to maintain a steady input voltage of 3.3 volts to the different components in this design because of the small board space it would occupy and lack of large, noisy inductors. The enclosure for this device will be designed in SolidWorks and 3D printed using Biorasis s facilities. The anticipated per unit cost of the smartwatch will be about $100 according to the budget table shown above. This is relative to change as certain components may be switched for equivalent, most cost-effective alternatives.