Design and Implementation of an Exoskeleton Arm Joint

Transcription

1 Design and Implementation of an Exoskeleton Arm Joint Prepared for: ECE 4600 Prepared by: Alex Reimer Colin Peterson Logan Froese Patrick Pagé Advisor: Robert McLeod, Ph.D, P. Eng Department of Electrical and Computer Engineering University of Manitoba Winnipeg, Manitoba, Canada March 2017

2 Abstract This document outlines a project to design and implement a modular exoskeletal arm joint. It provides with a brief introduction on exoskeletons and the purpose of the project to use modularity to make exoskeletons more cost-effective. A project overview is provided to establish the desired performance metrics for the exoskeletal arm joint, and to show how the project was separated into four subsystems: the power system, sensor system, control system and the mechanical (motor) system to accomplish this task. A detailed description of the design of the four subsystems outline how the overall system was designed. The document measures the performance of the integrated systems against established performance metrics for the project. The results of this project showed that system designed met all performance metrics for the exoskeletal arm joint successfully. 1

3 Table of Contents Abstract 1 List of Figures.. 4 List of Tables... 6 Nomenclature... 7 Chapter 1 - Introduction Purpose.. 8 Chapter 2 - Project Overview Project Subsystems Power System Sensor System Control System Mechanical (Motor) System Initial System Design Initial Performance Metrics. 11 Chapter 3 - Project Design Power System Battery Selection High Power Supply Low Power Supply Sensor System Signal Acquisition and User Protection Instrumentation Amplifier Filtration Design Changes Control System EMG Signal to RMS Value Motor Control Code Design Changes Mechanical (Motor) System System Controller Circuit Amplification Stepper Motor

4 3.4.4 Physical Design. 47 Chapter 4 - Results Results of Integration and Final Design Final Design Performance Description of Design Requirements and Results System is Able to Draw a Usable Actuating Signal Signal is Filtered with Desired Frequency Range System is Able to Translate the Analog Signal into a Usable Digital Signal System is Able to Use Digital Signal to Control Movement of Joint Joint Bends Linearly in Occurrence with a Signal as Desired by User Joint Responds to Electrical Signal Within Desired Time Frame Joint Moves with Precision to the Desired Angle Joint Stops Motion in Response to Flex Sensor Joint is Able to Move in Any Direction Within the Range of Motion Joint is Able to Move Arm to Meet Speed Expectations Range of Motion Power Usage Torque Casing of the Unit Fits onto Arm and is Functional Mechanical Brake Inhibits Joint from Overstretching Should include electrical safety features for the circuit and user Budget Possible Design Improvements.. 64 Chapter 5 - Conclusion.. 65 Bibliography Appendix A - Control System Program...69 Appendix B - Detailed Budget...72 List of Figures Figure 2-1: Block Diagram of Initial System Design. 11 Figure 3-1: Basic Boost Circuit Design.. 16 Figure 3-2: Final Boost Circuit Design

5 Figure 3-3: Transient Response of Boost Circuit 18 Figure 3-4: Frequency Range of the EMG Signal...20 Figure 3-5: Electrode Placement on the Arm Figure 3-6: Schematic of Signal Acquisition Area..22 Figure 3-7: EMG Signal Output from the Sensor System...25 Figure 3-8: AD620 Instrumentation Amplifier Pin Diagram..26 Figure 3-9: Schematic of Instrumentation Amplifier Area Figure 3-10: Generic Second-Order Low-Pass Sallen-Key Filter Figure 3-11: OP77 Pin Layout Figure 3-12: Schematic of Filter Design.29 Figure 3-13: Filter Simulation Results...30 Figure 3-14: Schematic of Original Filter and DC Offset Design Showing Negative Voltages Figure 3-15: Schematic of Flex Sensor System Figure 3-16: Block Diagram of the Control System Program Figure 3-17: Block Diagram of Timer Figure 3-18: Clock Output Variation with RMS.37 Figure 3-19: Appearance of Mechanical Design.39 Figure 3-20: Driver Pulse Design Figure 3-21: Arduino Clock Input (Yellow) Compared to Flip Flop A Output (Blue) Figure 3-22: Phases on Motor (Phase A is Yellow and Phase B is Blue)...41 Figure 3-23: Phases on Motor (Phase A is Yellow and Phase C is Blue)...42 Figure 3-24: JK Flip Flop Truth Table 42 Figure 3-25: Amplification Circuit for Signal A.44 Figure 3-26: ULN 2803A: Signal B to C Design 44 Figure 3-27: Amplification Results on Phase A. 45 Figure 3-28: Torque Graph. 46 Figure 3-29: Phases Into Motor.. 47 Figure 3-30: Bicep Bundle.. 48 Figure 3-31: Pulley Design. 49 Figure 3-32: Pulley Ridge

6 Figure 4-1: Final Design Block Diagram...51 Figure 4-2: Usable Actuating EMG signal..54 Figure 4-3: Arduino Serial Monitor Output Displaying Calculated RMS Values..56 Figure 4-4: Signal Response Time..57 Figure 4-5: Joint Triangle...58 Figure 4-6: Test Setup for Power Measurement. 61 Figure 4-7: Shoulder Model Mounted.63 5

7 List of Tables Table 2-1: Initial Performance Metrics. 12 Table 3-1: Power Usage by Component 14 Table 4-1: Final Design Performance...52 Table 4-2: Summary of Power Usage Calculations...61 Table 4-3: Budget Outline. 64 6

8 Nomenclature Symbol Description Value C i, L i, F i, R i, V i Roman upper case letters, Italic font with subscript Scalar Values X Roman upper case letter with Italic font Scalar Length A, B Start of Roman Alphabet upper case with Italic font Scalar Angle v, V, V i Non-Italicized Roman letters and words Scalar Values 7

9 Chapter 1 - Introduction Powered exoskeletons are made in order to assist people with movement via an external electrical device. Such devices have immense potential for rehabilitative care, as well as for disabled individuals. Most of these exoskeletons either assist half or the entire body in one unit. This fact has caused current exoskeleton devices to have extremely high costs, and as a result they are only accessible by wealthy individuals. This means that many people who are in need of such a device are unable to purchase one. Given that many people may only have difficulty moving a single limb, modular devices that assist with the movement of one body part are desirable. A modular design can render the needed assistance to a specific limb while not incurring as prohibitive a cost as full body exoskeletal systems[1]. Therefore, this project was interested in the creation of a modular exoskeleton design in order to reduce the cost in order to have the device be more accessible to the average person. 1.1 Purpose The purpose of this project was to design and implement a modular component of a powered exoskeleton. In particular, the component that this project aimed to create was a modular exoskeletal arm joint which assisted in the flexing of an arm. Another goal of the project was to have the device be affordable, in order to demonstrate the ability of a modular component to reduce cost. Chapter 2 - Project Overview In order to design and implement a modular exoskeletal arm joint, the device was required to accurately detect when and how the user wants to move and assist the user with that motion. Based on this, the project was broken up into subsystems. From these subsystems, it was possible to create performance metrics for every aspect of the design in order to achieve the project goal of creating an exoskeletal arm joint. This section provides an overview of these components and the initial performance metrics. 8

10 2.1 Project Subsystems The project was broken up into four different subsystems: a power system to power the device, a sensor system to detect when the user wanted to move, a control system to assess the desire to move, and a mechanical (motor) system to move the exoskeleton and assist the desired motion. Based on these components, the team produced a variety of requirements for each portion of the project Power System The power system was required to provide mobile electrical power to all the other systems. For the motor, a high DC power line needed to be created. For the other control and sensor systems, a line of lower power was needed. All together, it was decided that this system should be able to run at less than 200W, in order to reflect a standard of power system efficiency. This value was set to be higher than the rated power input of typical stepper motors being assessed at the beginning of this project Sensor System The sensor system was required to accomplish the following tasks for the project. Firstly, the sensor system needed to be able to draw a signal that was usable by the control system to move the arm. Secondly, the system had to remove frequencies above 200 Hz in order to remove high-frequency noise which might interfere with the functionality of the control system. Thirdly, the system was to use a flex sensor initially in order to stop motion. Finally, the sensor system was to include some electrical safety features to protect the circuit and user Control System The control system was required to be able to translate the sensor system s analog signal into a digital signal usable by the motor system to control the movement of the motor joint. The control system was also required to have a response time of less than 500 ms to ensure a quick response for the user. 9

11 2.1.4 Mechanical (Motor) System The mechanical system was required to complete all of the movement and torque performance metrics. Firstly, the movement was required to be precise such that the user could move in at least 5 degree increments within the range of motion. Secondly, the mechanical system was required to have a range of motion of at least 120 degrees for a linear motion of the arm. Thirdly, the torque was required to be a minimum of 5 Newton meter (Nm). Lastly, the arm should be able to move at least 20 degrees per second minimum. The mechanical system was also required to meet all performance metrics relating to the arm itself. These performance metrics included that the casing fit onto the arm of the user and be functional, that the arm bend in accordance with the user s input signal, that the arm be able to move in both directions, as well as that it have a mechanical brake in order to prevent the system from overstretching the user s arm and causing harm Initial System Design These four components detailed in the previous subsections were used to create an initial system design. In Figure 2-1 below, there is a block diagram of the system made during the initial stages of the design process. In this design, the power system, highlighted by the orange blocks, provides power to all other components of the system. The sensor system, highlighted by the green blocks, provides sensory input to the control system, highlighted by the blue blocks, which is then used to drive the motor system, which is highlighted by the yellow blocks. The motion provided by the motor was registered by a flex sensor, which was initially the stopping mechanism for the system. 10

12 Figure 2-1: Block Diagram of Initial System Design 2.2 Initial Performance Metrics The performance metrics which were initially chosen for our design are detailed completely in Table 2-1 below. 11

13 Table 2-1: Initial Performance Metrics Feature Value or Range System is able to draw a usable actuating signal Signal is filtered with desired frequency range System is able to translate the analog signal into a usable digital signal System is able to use digital signal to control movement of joint Joint bends linearly in occurrence with a signal as desired by user Joint responds to electrical signal within desired time frame Joint moves with precision to the desired angle Joint stops motion in response to flex sensor Joint is able to move in any direction within the range of motion Joint is able to move arm to meet speed expectations Range of Motion Power Usage Torque Casing of the unit fits onto arm and is functional Mechanical brake inhibits joint from overstretching Should include electrical safety features for the circuit and user Pass/Fail < 200Hz Pass/Fail Pass/Fail Pass/Fail < 500ms +/- 5 degrees Pass/Fail Pass/Fail 20 degrees/ sec minimum Linear, 120 degrees < 200W 5 Nm minimum Pass/Fail Pass/Fail Pass/Fail Cost < $

14 Chapter 3 - Project Design In order to meet the design criteria laid out for this project, each system was designed using engineering methods and practices. The power system, sensor system, control system and mechanical system were all designed in parallel and then synthesised together to form a cohesive final design. This section provides an overview of the design of each system component. 3.1 Power System The power system was designed to be a portable supply of power for the other systems in the project. A high power line and a low power line were required, along with a common ground, in order to meet the power requirements of the other systems. As a result, the power system design was divided into three different sub-sections: the battery, the high power supply, and the low power supply Battery Selection The battery was selected while taking a number of criteria into consideration. The most important criteria considered was the power requirement. The battery had to provide enough power to the stepper motor, the Arduino and the sensor circuit. Besides providing power to these systems, there were a number of other criteria considered in the battery selection, which will be discussed later in this section. The majority of the power was required to power the motor. In order to meet the required torque of 0.6 Nm, 3.4 A was required [2]. Since the torque of a stepper motor is mostly dependant on current rather than voltage [3], and the internal resistance of the motor was quite low, the voltage required while supplying the current did not need to be very high. As a result, the minimum voltage requirement was set by the low power supply to the sensors and Arduino. The sensors and Arduino required power from a 5 V supply. This meant that while the motor was drawing 3.4 A, the voltage would have to be at least 5 V. The Arduino and sensors were estimated to require at most 1 A of current in total. Initially, it was believed that a supply of V to the stepper motor was desired in order to increase the charging time of the motor coils. (The effectiveness of this design decision is discussed at the end of Section ) Since most feasible batteries did not provide voltage at such a level, a boost circuit 13

15 would be required to boost the voltage to the desired range. A boost circuit with an estimated efficiency of 75% was therefore included in the power estimates of the high power line. The total power required by the battery was calculated to be W minimum. Table 3-1 shows a summary of these power requirement calculations. Table 3-1: Power Usage by Component System Max. Current Min. Voltage Power Stepper Motor 3.4 A 5 V 17 W Power To Stepper Accounting for a 75% Circuit Efficiency W Arduino and Sensors 1 A 5 V 5 W Total Estimate W Besides a power requirement, there were three other criteria that were considered when selecting the battery. Firstly, the battery had to be cost effective to fit into the project budget. Secondly, the battery should be as portable and lightweight as possible. Finally, a rechargable battery was also desired, although not mandatory, to make testing of the prototype simpler and consistent. In order to meet the required criteria, a KT12B-4 Koyo Motorcycle battery was chosen. This battery had a number of advantages. Firstly, it was rated high enough to provide the power required. Testing revealed this particular battery supplied up to 37.6 W. Secondly, it was donated by the project advisor, Robert McLeod, and therefore allowed more budget to be allocated to other systems. Finally, the battery was rechargeable. The main disadvantage relating to this battery was that the battery was bulky and heavy, and likely would not be mountable to the arm. This was acceptable, since only a prototype was desired. It also only provided 12 V, which meant that a boost circuit was necessary if V was to be delivered to the motor. Despite the disadvantages, it was decided to use and design based upon the KT12B-4 Koyo motorcycle battery. The benefits of having a donated battery that was rated for the power criteria outweighed the disadvantages of portability and the requirement of the boost circuit. 14

16 3.1.2 High Power Supply One of the desired outputs to the motor was a high power supply with a voltage of V. This value was based on the rated voltage of the stepper motor. Although the torque of the stepper motor was primarily dependant on the current flowing through it, the high voltage provided a faster response time and slew rate for the system. Since both of these are desirable in a control system, the positive voltage line to the stepper motor was designed to output voltage on the range of V. Since the battery selected was only a 12 V source, a solution was required. The primary solution was to design a boost circuit to increase the voltage to the desired range. As a backup plan, more batteries could be connected in series to boost voltage, but this was undesirable as it added significant weight and cost to the project. Along with increasing the voltage to V, the boost circuit was also required to output up to 3.4 A. While a supply of both 30 V and 3.4 A would provide power over the rated power of the stepper motor, it was decided to overdesign the components of the circuit so that they were rated for both 30 V or 3.4 A. After a brief search through possible boost circuit and DC/DC voltage convertors powerful enough to produce the desired output, it was noted that they were beyond the budget constraints of the project. As a result, it was decided that a custom boost circuit was required to meet the design and cost specifications. There is a basic idea behind boost circuit design. Using an inductor, current is stored and converted into voltage. Figure 3-1 shows the basic circuit [3]. 15

17 Figure 3-1: Basic Boost Circuit Design Initially the switch is closed, causing current to flow through the inductor L1, increasing the voltage across it. Once the switch is opened, this stored current flows through the diode into the load and charges the capacitor. The voltage across the inductor is seen by the load as added to the initial voltage source, providing a voltage boost on the output. When the switch is closed again, the capacitor discharges into the load, allowing the load to maintain a voltage until the switch is opened again. This idea was the basis for the physical boost circuit design. In order to simplify the design process, the LT-1680 chip was selected to act as regulating clock source. The LT-1680 is an high power DC/DC boost circuit controller, and is designed to sense the output voltage and adjust the clock cycles accordingly. It also contains a number of other useful features programmable through the circuitry around it [4]. The features and requirements largely regulated the final design of the boost circuit. The final design is shown below in Figure 3-2: 16

18 Figure 3-2: Final Boost Circuit Design As can be seen in Figure 3-2, several external components were used to tune the functionality of the LT-1680 and the overall boost circuit. Firstly, the 16.3 μh inductor (L1) was selected to balance response time and output ripple voltage. A lower inductor value gives a slower response but less ripple at the steady state. Since the stepper motor was not sensitive to ripples in the voltage, the inductance was increased slightly to increase response time. Secondly, the Ω sense resistor (R1) is used by the LT-1680 to sense and limit the output current. For the purposes of the project the current limit was set high to 24 A. This allowed the system to run at the maximum power available in the battery. Thirdly, the 1000 pf capacitor (C3) and 15 kω resistor (R3) combination was used to program the operating frequency of the LT Using these values, the operating frequency was set to 110 khz. This was well above the sensor circuit filter range (as discussed in Section 3.2.3) and minimised chances for interference with the EMG signal. A frequency of 110 khz also allows the LT-1680 to use enough of the duty cycle to boost the voltage to the desired value. Finally, the 47 kω and 1.8 kω resistors (R4 and R5 respectively) were used to program the output voltage according to Equation 3.1. V out = 1.25 V x (1 + R 4 / R 5 ) (3.1) V out = 1.25 V x ( kω / 1.8 kω ) = 32.6 V All other component values were based on suggestions in the LT-1680 datasheet and did not particularly affect the output of the system. 17

19 The boost circuit met all the design requirements in simulation, but failed to provide much value in practice. In simulation, an output voltage of 33.8 V was calculated, as can be seen in the simulated transient response of Figure 3-3 below. The circuit produced a regulated voltage regardless of load. As a result, the current in Figure 3-3 is a function of load resistance. It is measured here as 3.4 A, as desired. This output current and voltage were used to do all power and thermal rating calculations of the components. However, since the output current only depends on load resistance, care was required to keep current levels outputting correctly as the resistance in the coils of the stepper motor changed. Figure 3-3: Transient Response of Boost Circuit In practice, the boost circuit was not necessary, nor did it provide much of a benefit. This was largely due to the fact that battery supply decreased in voltage much more than anticipated. The boost circuit did provide a V supply unloaded, depending on how much the battery was charged. However, the current requirements of the stepper motor caused the voltage of the battery to be pulled below the voltage required to operate the LT This rendered the boost circuit inactive when a large current was drawn by the circuit. However, if the motor ever stopped drawing current, the voltage across the battery would recover and the boost circuit would increase the voltage again, providing a boost in coil charging time the next time the motor would be turned on. In actual operation this boost is negligible. Despite these imperfections, the high power line still provided enough current to the stepper motor for it to fully function. All of the design requirements relating to the power system were still met. 18

20 3.1.3 Low Power Supply A supply of 5 V was also desired to provide power to the Arduino and the sensor circuit. A system had to be developed to take a varying 5-12 V supply from the battery and output a stable 5 V. It also had to be able to supply the 1 A of current expected to be required by the Arduino and sensor circuit. Finally, high efficiency and low cost were also desirable. While various options were looked at and designed, the final design simply took advantage of features already present on the Arduino. The Arduino Uno used in this project contains an internal voltage regulator that converts anywhere from 12-6 V into a regulated voltage for itself. Below this voltage, it simply uses the supply voltage unregulated as long as it is above 5 V. It also outputs a 5 V supply powerful enough to power the sensor circuit. By plugging the battery directly into the Arduino Uno, all of the low power design criteria were met. The solution was efficient and cost effective. It also provided an isolated power line to the sensor circuit, which helped reduce noise in that system. 3.2 Sensor System The sensor system for this project was required to fulfil several design requirements. It needed to acquire a signal that would be usable by the control system (an Arduino Uno) to control the arm design, to filter out all frequencies above 200 Hz, implement electrical safety features for the circuit and user, and include a flex sensor to stop the system. A signal source was required that indicated the user s desire to move their arm. Ultimately, a design which measured electromyographic (or EMG) signals from the bicep muscle was chosen to provide the signal source. This was chosen over alternatives such as elastic actuation[5], which would incorporate an elastic band in-line with the motor or stepper to mimic the tendons of the human body. However, this methodology required the user to have a sufficient degree of control over their ability to move already without the device. As such, given the design s intent as a device to be used for assistive care, it was decided to use electromyography to maximize its utility. Without any need for substantial physical movement the sensor system can still acquire a usable signal. 19

21 An EMG signal is an electrical signal which is produced by a muscle when its motor units are activated. The signal was acquired using surface electrodes[6], which were chosen due to their availability. Typically, three electrodes are required to acquire the EMG signal. One electrode was used as a grounding electrode to reduce noise and the other two were used to measure the signal of a muscle from two different positions in order to cancel out ambient signals with an instrumentation amplifier[6]. Typically, an EMG signal s frequency range is from around 15 Hz to 500 Hz[6], though the dominant frequencies of this range are from 50 Hz to 150 Hz[6], as shown in Figure 3-4. The amplitude of the signal prior to any amplification is typically in the 0-10 mv range (peak to peak)[6]. If acquired, amplified and filtered appropriately, an EMG signal can be used to indicate the actuation of the muscle [6], and as such fit the purposes of the design. Figure 3-4: Frequency Range of the EMG Signal[6] The sensor design consisted of three key areas. Firstly, there was a signal acquisition area, which involved the appropriate placement of electrodes to acquire the signal, as well as safety features that needed to be put in place for the user and the circuit. Secondly, there was an instrumentation amplifier area to remove ambient signals and amplify the EMG signal for use by the control system. Lastly, there was a filtration area to meet the frequency requirements laid out in the performance metrics. As well, a 20

22 flex sensor system was designed in order to meet the initial performance metrics, but that metric was removed from the final design and this is discussed in Section Signal Acquisition and User Protection In order to acquire the EMG signal, the sensor system design made use of 3M Red Dot electrodes[7], which utilized a gel in order to improve signal acquisition. Before placement of these electrodes on the skin, the skin was treated with alcohol swabs in order to reduce the impedance of the skin, improving signal acquisition further. The electrodes were not meant for extended use, meaning that they had to be reapplied each time the system was disconnected to ensure the best connection possible. The electrodes were placed in the configuration shown in Figure 3-5 below, and were placed in order to measure the desired motion of the biceps brachii muscle. The two measurement electrodes were placed closely together at the apex of the muscle in order to minimize the ambient signal difference between them, and the grounding electrode was placed on the lower part of the upper arm. The placement of the grounding electrode was beneficial despite other options being further away from the active muscle, because it allowed for shorter electrode cables from the user s arm to the system, reducing the noise acquired along the wires. Figure 3-5: Electrode Placement on the Arm When the electrode cables reach the circuit board, they were physically looped through the board. This was in order to prevent the stress of movement by the user breaking the electrode cables, which 21

23 presented serious problems during testing due to large amounts of ambient noise not being cancelled by the instrumentation amplifier. The cables themselves had a built-in resistance of 10 k Ω. After the electrode cables, the grounding electrode was grounded, and the two measurement electrodes were connected to a pair of 10 nf isolation capacitors, which were placed in series with the inputs of the instrumentation amplifier. These capacitors were rated for high DC voltages[8]. This was the electrical safety feature for the circuit, as it prevents inadvertent high DC currents (which may occur due to some hypothetical short circuit failure in the system) from impacting the user negatively. Alternative methods that were considered for safety features included the use of transformers and optoisolators as these have been effective in other EMG system designs[6]. The former was considered too large for use in this system and capacitors were chosen over optoisolators for reasons of design simplicity. A schematic showing the signal acquisition area of the design is shown below in Figure 3-6. Figure 3-6: Schematic of Signal Acquisition Area Instrumentation Amplifier An instrumentation amplifier was required by the design in order to take in the two signals acquired from the muscle, take the difference of these two signals to remove ambient signals, and amplify the difference to acquire the EMG signal. The typical method for removing ambient noise, a notch filter, was undesired as the usual noise frequency of 60 Hz fell within the dominant frequency range of the EMG signal[6]. In choosing an instrumentation amplifier for the design of the sensor system, a few features were key requirements in ensuring that the signal obtained was of high quality. Firstly, a high common-mode rejection ratio (CMRR) was required in order to thoroughly reject ambient signals that were acquired on the two measurement electrodes. Secondly, the instrumentation amplifier was required to have a high gain capability in order to amplify the relatively small EMG signal. 22

24 The instrumentation amplifier chosen for the sensor system was an AD620. The AD620 had a number of advantages to it. First of all, it met the instrumentation amplifier requirements by having a CMRR of 100 db minimum[9] and being capable of a gain up to 10000[9]. This was more than sufficient for our purposes. The AD620 also had a reference voltage pin in order to assign a different ground to the system. This was useful in solving an issue with integrating between the power system, the sensor system and the control system. The power system provided only a 5 V source and ground for use by the sensor system, and the control system required a signal in the 0-5 V range due to the analog-to-digital converter of the Arduino Uno. However, the input signal from the electrodes average around 0V, and therefore, to acquire a clean signal two key things were required: the ability to set a different reference voltage for the system so as to avoid clipping on the much larger signal output and the addition of pull-up resistors to elevate the input signals to a level above ground to avoid clipping on the rails at the input of the amplifier. This reference voltage was decided to be 2.5 V in order to maximize the signal s distance from both of the rails. Both the 5V supply and the 2.5 V reference voltage use 100 μf decoupling capacitors to ground in order to prevent noise on these lines from impacting the instrumentation amplifier. This reference voltage is also used for our pull-up resistors. The pull-up resistors are connected after the isolation capacitors from the 2.5 V to both inputs. The two resistors used were 2.7 M Ω and were found experimentally to both elevate the input voltage by a DC voltage value of around 1V. As there is slight variation between these two resistors for manufacturing reasons, there is a very small difference between the offsets on both lines, which became important when utilizing extremely gain values in the final system. The gain of the AD620 is determined by a resistance value R G. The resistance R G is related to gain by Equation 3.2[9]: Equation 3.2 can be rewritten as Equation 3.3 to obtain : R G = 49.4 kω / (G 1 ) (3.2) G = kω / RG (3.3) The resistance used for gain was decided to be a 10Ω resistor in series with a 0-100Ω trimmer potentiometer. This was in order to provide the option of adjusting the gain based on the user, as tests with a fixed resistance had too small of a gain in some cases, even if the resistance was effective in other cases. Therefore, using Equation 3.3, the maximum gain of this design was: G = kω / 10 Ω =

25 Furthermore, the minimum gain of this design was: G = kω / 110 Ω = 450 Given the range of voltages for the raw EMG signal of 0-10 mv, which was noted in Section 3.2, a theoretical average voltage of 5 mv was taken during design to determine the relative effectiveness of the gain at obtaining a usable signal. This assumption was applied to the final sensor system, resulting in the following peak-to-peak voltage limits found in accordance with Equation 3.4: V out = G V in (3.4) mv = 24.7 V maximum mv = 2.25 V minimum These voltages have a very wide range, considering that we want a voltage between 0-5V for the Arduino Uno s analog-to-digital converter. However, while testing, it was found that these numbers are much larger than the actual signals found, and that a more reasonable assumed EMG signal was one around 1 mv. Using Equation 3.4 again, it was found that the peak-to-peak voltage limits were: mv = V maximum mv = V minimum Utilizing this gain range, the sensor system was able to produce signals such as the one in Figure 3-7 below, which was sizable enough to be detected by the control system: 24

26 Figure 3-7: EMG Signal Output from the Sensor System It is important to note that the signal shown above was put through filtration, and is only shown here to demonstrate the gain. Something further that should be noted about this design was that due to the very small DC voltage difference between the two inputs from the pull-up resistors, there were issues with the DC offset of the output of the AD620 decreasing at extremely high gains, leading to clipping. Signals at these high gains are still usable by the control system, though they are of lower quality. The pin layout of the AD620 is noted in Figure 3-8 below. The positive rail of the AD620, which is applied on Pin 7, was set to 5 V. The negative rail of the AD620, which is applied on the Pin 4, was set to ground. The reference pin 5 provided a 2.5 V baseline for the system. The 10 Ω resistor and Ω potentiometer were placed in series between Pin 1 and 8. The output on Pin 6 was connected to the filter, which is discussed in Section

27 Figure 3-8: AD620 Instrumentation Amplifier Pin Diagram[9] Using this configuration, the schematic for the final implementation of the instrumentation amplifier of the sensor system is shown in Figure 3-9.: Figure 3-9: Schematic of Instrumentation Amplifier Area 26

28 3.2.3 Filtration After the instrumentation amplifier, the sensor system required a filter. This was to meet the performance metric that the high frequencies above 200 Hz be filtered out. Though such a filter does remove some high-frequency components of the EMG signal itself, this was considered negligible due to the relatively low magnitude of this portion. The sensor system made use of a single second-order low-pass Sallen-Key filter for the purpose of filtering the EMG signal. A Sallen-Key filter was chosen for this application due to the consistency of the gain of the circuit before the cutoff frequency of the filter. This was to prevent any extreme distortion of the EMG signal. The second-order version was chosen for its sharp cutoff of -40 db / decade[10]. The cutoff frequency for a second-order low-pass Sallen-Key filter is -6 db[10]. The second-order low-pass Sallen-Key filter is defined by the following schematic[10] in Figure 3-10, drawn using a generic operational amplifier: Figure 3-10: Generic Second-Order Low-Pass Sallen-Key Filter[10] The cutoff frequency of a second-order low-pass Sallen-Key filter is defined by Equation 3.5[11]: F c = 1 / 2 π R 1 R 2 C 1 C 2 (3.5) 27

29 For the sake of simplicity and the bulk purchase of components, it was decided that C 1 = C 2 = C and R = R 1 = R 2. Therefore, Equation 3.5 becomes Equation 3.6: F c = 1 / ( 2 π R C ) (3.6) For this filter design, R = 10 k Ω was chosen due to similar resistors already being used in the design. Therefore, it was possible to find C for the desired Fc of 200 Hz by rearranging Equation 3.6 into Equation 3.7: C = 1 / ( 2 π R F c ) (3.7) C = 1 / ( 2 π 10 kω 200 Hz ) = 80 nf The desired gain for this filter was 1, as by design the gain was controlled by the instrumentation amplifier. The gain for the filter was controlled by the two resistors, R A and R B, via Equation 3.8[10][11]: G = 1 + R A / RB (3.8) Using an open circuit in place of R B, R B becomes infinite, making G = 1 regardless of the value of R A. This was the case for the final design. R A was set to 10 k Ω due to some minor changes to the filter s gain during the circuit s testing, but this value does not impact the function of the system. The operational amplifier used for the filter was an OP77[12]. It was chosen due to its availability, its high accuracy and suitability for instrumentation applications such as this one[12]. The pin layout for the OP77 is shown in Figure 3-11 below: Figure 3-11: OP77 Pin Layout[12] 28

30 Similar to the instrumentation amplifier, the ground of the system was at 2.5 V. The positive rail on Pin 7 was set to the 5 V supply and the negative rail on Pin 4 was set to ground. Pin 6 was the output, and was connected to the control system. The final schematic of the filter circuit is shown in Figure 3-12 below: Figure 3-12: Schematic of Filter Design From this filter design, a graph was simulated as presented in Figure 3-13 showing the magnitude of the signal relative to the frequency. This demonstrates the cutoff frequency of 199 Hz at -6 db. 29

31 Figure 3-13: Filter Simulation Results Following filtration, the EMG signal was ready to be used by the control system, and as such the sensor system fulfilled all design requirements Design Changes Throughout the design of the sensor system, a variety of changes took place from the initial proposal for the system. These changes took place in three key areas. The first area of change was that the original design assumed that the power system would provide a negative voltage. This impacted the filtration area of the design immensely, as can be seen from the original filter design shown in Figure 3-14 below. 30

32 Figure 3-14: Schematic of Original Filter and DC Offset Design Showing Negative Voltages This system was designed with a second-order high-pass Sallen-Key filter with a cutoff frequency of 15Hz, in order to reduce lower frequency noise. Also, a DC offset was designed in order to properly offset the signal to interface with the Arduino Uno s analog-to-digital converter. After changing the system to accommodate the lack of a negative voltage, using the reference pin of the AD620 instrumentation amplifier became seen as a much simpler way of implementing a DC offset, and as such that circuit was removed. Due to using the instrumentation amplifier to provide the DC offset, the high-pass filter was also removed. This was to ensure there was no risk of filtering out the DC offset prior to 2.5V being used as the ground for the whole system, and the high-pass filter was never implemented again. The second area of change was to the resistance which defined the gain for the instrumentation amplifier. Originally this gain was fixed, and was designed with an assumed 5 mv signal and desired 0-5 V range in mind. Given this, a desired peak-to-peak voltage of 2.5 V was chosen. The following gain was required according to Equation 3.9, which is a rearrangement of Equation 3.4: G = V out / V in (3.9) G = 2.5 V / 5 mv =

33 This gain was then put into Equation 3.2: R G = 49.4 kω / (G 1 ) = 49.4 kω / (500 1 ) = 100 Ω While testing, it was found that this gain was insufficient. A 33 Ω resistor was used as a replacement until the design was changed to have a variable resistance for the gain. The final area of change was to the flex sensor. The flex sensor was initially required to stop the system by the initial performance metrics. The flex sensor system was designed as a basic voltage divider, as the flex sensor acts as a variable resistor in practice as it bends[x9], and the voltage output was designed to have been connected to the control system in order to indicate the position of the arm, as the flex sensor was meant to have bent with the arm. The flex sensor varied, on average, from 10 kω when unbent to 20 kω when fully bent [x9], and a 47 kω was chosen as the other resistor in the voltage divider. The flex sensor system configuration is shown in Figure 3-15 below, with 10 kω representing the flex sensor: Figure 3-15: Schematic of Flex Sensor System As such, the design varied between the following voltages, found by Equation 3.10, the equation for the voltage output of a simple voltage divider: 32

34 V o = 5 V ( R 2 / ( R 1 + R 2 ) ) (3.10) V o = 5 V ( 47 kω / ( 10 kω + 47 kω ) ) = 4.12 V maximum V o = 5 V ( 47 kω / ( 20 kω + 47 kω ) ) = 3.51 V minimum However, as the design of the control and mechanical systems progressed, the flex sensor was no longer required to stop the arm s movement. As such, it was negotiated to have it removed as a performance metric, and the flex sensor was removed from the sensor system. 3.3 Control System The control system s goals necessitated a computational platform that was able to take in the filtered EMG signal from the sensor system and transform it to a usable digital signal for the motor system. The platform also needed to be able to respond to the signal input in less than 500ms. For this purpose, an Arduino Uno was used. In order to achieve the control system s performance metrics, the EMG signal output from sensor system was input into the Arduino Uno of the control system to be processed into a digital signal. The Arduino Uno modified this input to achieve a Root Mean Square (RMS) value of the EMG signal. Using the Root Mean Squared value, the code was identify if the system needs to run clockwise or counterclockwise. After the identification of the direction of movement, a signal was produced by the control system to be used by the mechanical (motor) system to cause the stepper motor to rotate in the desired direction. Through this process, the control system needed to output a specific clock without being interrupted by the constant calculation of the RMS of the sensor data or the constant manipulation of the direction pins in order to meet response time requirements EMG Signal to RMS Value The control system achieved a useful RMS value from the EMG signal by using the Arduino Uno. It implemented both system hardware and code to derive the desired results. Initially, the EMG signal was inputted to the analog detector pins built into the hardware of the Arduino. The valid input values ranged from 0-5 V, which the sensor system was designed to match. With 33

35 the analog signal from the sensor system inputted into the Arduino Uno, the Arduino Uno was able to convert the EMG signal to an RMS value. The Arduino Uno calculated a RMS value of the EMG signal using the formula in Equation 3.11 below [13]. The formula shows the summation of many voltage samples, divided by the amount of samples, and then square-rooted in order to obtain the RMS of the input signal. (3.11) However, the Arduino Uno lacked the memory capacity to store N values in order to use such mathematical functions like summation. Therefore, to achieve the same task the code implemented a for loop as shown in the following subroutine: float mycollectdata(float SensorValue, float Sensor, float Offset, float OffsetValue) { float result = 0.0; float n = 0.0; float A = 0.0; for (n =0; n <= 150; n++) // For a sampling window of 150 data points { SensorValue = analogread(sensor); // Read value from the EMG OffsetValue = analogread(offset); // Read value of offset. SensorValue = SensorValue * float(5.0/1023.0); // Convert value from ADC back to a voltage OffsetValue = OffsetValue * float(5.0/1023.0); A = sq(sensorvalue-offsetvalue); // Start the RMS calculation by squaring the value result = float(result + A); // Sum these squared values together } return result; } When the program returns to the main loop, the result is divided by the number of samples in the window and the square root is taken to determine the final RMS value. This value, which linearly corresponds to the force a user is attempting to extract from their muscle, is used to differentiate between flexing and relaxing. 34

36 3.3.2 Motor Control Code The motor control code was designed to take the calculated RMS and run the required routines to identify a flex. In response to the result of the RMS value, the Arduino Uno produced an output to control the mechanical system and the stepper motor. Figure 3-16 fully outlines the program used by the Arduino Uno as a flow chart to highlight the way in which the code made changes in order to produce a signal for use by the mechanical system. Figure 3-16: Block Diagram of the Control System Program Step-by step, the program needed to: Set up the input/output pins as well as the clock frequency to be outputted to stepper motor as required by the main program. Acquire a signal from a window of 150 data points and subsequently calculate an RMS. If that RMS is greater than 2, there was a muscle contraction, and the stepper should rotate clockwise (CW) to close the elbow. Otherwise, there was muscle relaxation, and the stepper should rotate counter-clockwise (CCW) to open the elbow. 35

37 If muscle contraction had occurred, the Arduino Uno needed to set the pins for CW direction, and set flag CW to 0 and CCW to 1 to prevent pin setup on subsequent loops. If muscle relaxation had occurred, the Arduino Uno needed to set the pins for CCW direction, set flag CW to 1 and CCW to 0 to prevent pin setup on subsequent loops. Given this output, the stepper motor rotated with the next clock in the direction currently set up. Repeat. Apart from pin setup and the motor movement, the flow chart adequately demonstrates how the system worked. The Arduino Uno only needed to set the required pins to High or Low in the right order to set the direction of motor. The control system outputted to the mechanical system using the standard Timer2 (TMR2) clock that is built into the Arduino Uno in order to produce the clocks used to control motor movement. The TMR2 clock used a system of multiplication sets and comparators within the Arduino Uno to toggle a pin and create a modifiable clock that runs at a constant frequency. Figure 3-17 outlines the way the timer is implemented. In this design, both Prescaler and and Postscaler values are set to 16. Figure 3-17: Block Diagram of Timer2[14] 36

38 The TMR2 value was where the count was started and for this it was 0, therefore the clock was only dependent on the value of the PR2 register, which varied from a value of Equation 3.12 outlines the output clock of our Timer2, by which Equation 3.13 was created: Fosc O utput Clock = 4(Prescaler)(PR2 TMR2)(Postscaler) 16 MHz O utput Clock = 4(16)(16)(PR2 + 1) = KHz PR2 + 1 With Equation 3.14, the motor movement speed can be changed by altering PR2, fulfilling the (3.12) (3.13) requirements of the control system to produce a digital signal that can be used by the mechanical system to control the motor. The full control system program is located in Appendix A of this document Design Changes Originally, the control system program was meant to vary the clock frequency along with the calculated RMS. The results of the program attempting this is shown in Figure 3-18, where the blue signal is the input from the sensor system and the yellow signal is the output from the sensor system: Figure 3-18: Clock Output Variation with RMS 37

39 There were two issues that led to this feature being removed. Firstly, upon testing, it appeared that the stronger contractions were being mapped to the slower clock cycle. Since the clock pulse made the stepper rotate, the opposite was desired. Secondly, and more importantly, the program was also sampling too often within the EMG signal. During the one contraction, due to the continuous nature of the looping program, the software's predetermined window size of data points was small enough to make multiple RMS calculations. If attached to the stepper at the time, the resulting changes to the clock resulted in a less than smooth motion of the arm. The first issue was quickly remedied by the application of a transfer function that mapped the stronger contractions to the faster clock rates. However, a solution to the second issue proved to be far more problematic. Lengthening the sample window meant the system updated the clock less often during one contraction, but the resulting loss of resolution, as more data gets averaged together with noise, rendered the system inoperable. Shortening the window allowed the system to update the clock even more often. In order to correct this issue, a voltage comparator circuit was required to compare the EMG signal to the background noise and detect muscle contraction based on the difference between the two voltages. This was unable to be implemented within the time frame of the project. 3.4 Mechanical (Motor) System The mechanical system was designed to meet all requirements relating to the arm and it s motion, including range and precision of motion, the torque and speed, the casing of the arm, the bending of the arm, the ability of the arm to move in both directions and a mechanical brake for safety. The major features through which these requirements were met were the system controller area, the circuit amplification area, the stepper motor and the physical design. When the arm was operated, the system controller area was responsible to take the signals sent by the Arduino Uno to the amplification system in the correct voltage and pulse design in order to control the motor to bend the arm and move in both directions. The circuit amplification increased the current and voltage received by the stepper motor in order to run it effectively. The stepper motor then responded to incoming pulses and outputted the desired direction and speed of motion to the physical design, a pulley system. The rotation of the pulley pulled the string and intrinsically moved the forearm or releases tension. The 3D shoulder piece acted as the system casing and was used to contain the control, sensor and motor systems in a functional package as well as anchor the system to the user. 38

40 The full system design is shown in Figure 3-19 below, and demonstrates how all the parts connect within the mechanical design. The sensor and control system as well as most of the mechanical (motor) system were contained within the shoulder model. Figure 3-19: Appearance of Mechanical Design 39

41 3.4.2 System Controller The system controller received all the Arduino outputs to ensure a stable and functional stepper motor. The design was setup to have two parts. The first part was a flip flop that controlled the edge set of the next series of flip flops. The second flip flops controlled the signal going to the individual stepper motor phases. These together allowed the direction and speed of the stepper motor to be changed. Pins 3 and 4 controlled the flip flop that ran the edge control; Pins 2 and 5 controlled the clock going to the motor. By using this cascading design as seen below in Figure 3-20, it allowed the system to be under the full control of the control system, the Arduino Uno. Figure 3-20: Driver Pulse Design In the implementation of the system controller design, the clock from the Arduino was responsible for controlling all the individual parts in a cascading design. This means that the clock going into the first flip flop controlled the clock in the next two flips flops as shown. Lastly, the output going to phase A-D varied depending on all clocks produced by the system controller. In Figure 3-21, the result of the cascading clocks is shown through the results of the controller system on the clock cycle going to phase A. Phase A operated at a quarter of the frequency than that of the initial Arduino clock. 40

42 Figure 3-21: Arduino Clock Input (Yellow) Compared to Flip Flop A Output (Blue) The following figures below show what the output of the system controller, with the Arduino clock as input, at each stage. A and B phase, shown in Figure 3-22, are opposite of one another and C, shown in Figure 3-23, is a quarter phase ahead of A. C and D are, similar to A and B, opposite of one another. This setup allowed for one coil of the motor to always be powered, maintaining torque to meet design requirements. Figure 3-22: Phases on Motor (Phase A is Yellow and Phase B is Blue) 41

43 Figure 3-23: Phases on Motor (Phase A is Yellow and Phase C is Blue) With the use of the flip flop truth table shown in Figure 3-24 below and the connections to Pins 2-5 on the Arduino, the system was able to set the phase A and C high by setting Pins 2 and 5 Low. Pins 3 and 4 are used to set the edge on the clock coming from the first flip flop going to the next two flip flops. Setting pins 3 and 4 determined the next flip flop trigger order. If the flip flop controlling A was triggered first then a signal for a clockwise rotation of the stepper was output. For this result, Pin 3 was set low and 4 was set high. If the flip flop controlling C is triggered first, then counter-clockwise rotation resulted at the stepper. The direction was then controlled through the use of the Arduino and the system controller. To run the stepper motor continuously in one direction, all pins were set high enabling the outputs of the flip flops to change. This design allowed the arm to move in two directions as required by the performance metrics. Figure 3-24: JK Flip Flop Truth Table[15] 42

44 3.4.2 Circuit Amplification Circuit amplification was used to control the high voltage in the motor in response to the signal received from the system controller. A-D were amplified individually through identical designs. The ULN 2803A controlled the logic and power going to the motor, and the diodes and capacitors in the system improve performance of the design. This role of the circuit amplification area was to increase the voltage and current going to each motor coil in order to have the motor function at a high torque and speed. For ease of understanding, only phase A will be mapped as it completes a full circuit but B-D are identical in functionality and circuit design for the amplification part. The schematic in Figure 3-25 below shows that the ULN 2803A received its pulse from the Flip Flop setup. The input on the ULN 2803A controlled the voltage that the motor is controlled by. When the flip flop input is high the motor will receive voltage as defined by Equation 3.14: V oltage Across Motor = High V oltage 1.4 V (3.14) When the flip flop input is low the voltage across the motor was defined by Equation 3.15: V oltage Across Motor = 0 V (3.15) 43

45 Figure 3-25: Amplification Circuit for Signal A The ULN 2803A has the circuit shown in Figure 3-26 below between the input of B to the output of C. This Darlington transistor setup worked as a gate that is driven by the input value of B. Figure 3-26: ULN 2803A: Signal B to C Design 44

46 The zener diodes in the amplification circuit was used to boost the response of the motor when a phase is energized or deenergized. When a phase was initially charged, there was a spike in voltage given to motor phase due to the change in current flow seen by the zener diode. This spike increased the rate at which the inductor is charged as can be seen in Equation 3.16: where V is voltage, L is inductance and voltage, the faster the current will change[16][17]. V = Ldi dt (3.16) di dt is the change in current. Given this equation, the greater the To improve the overall stability of our design, we also used a capacitor connected close to the ULN 2803A. Connecting high voltage to the ground allowed the driver to produce a smoother function and made the driver less susceptible to unwanted power surges throughout operation. The results of amplification for phase A is shown in Figure Figure 3-27: Amplification Results on Phase A The major issue with this driver setup was heat generation. To improve the driver s heat tolerance, the use of two ULN2803A was implemented and the high current parts were positioned at the top of the physical design in order to maximize heat dissipation. There can be up to 1.7 A going through an individual amplification step up at a time, and as such there was a lot of excess heat produced by the design. Heat is mainly produced by the ULN 2803A chips, due to their high current and voltage values. Each chip can dissipate about 2.56 W[18] with no heatsink and ambient temperatures. This means that with two chips without a heat sink they can dissipate 5.12 W. However, since the system operated at a 50% duty cycle and low frequencies under 500 Hz, the power dissipation is doubled to W. The 45

47 system sufficiently dissipated the excess heat at the peak power. The peak power of the system was defined by Equation 3.17: V I = P (3.17) 6 V (1.7 A) = 10.2 W Stepper Motor The operation of the stepper motor required a series of pulses sent to the phases A-D. The order in which the pulses are sent to the stepper motor controlled the direction. As the stepper motor was a component that is a pre-built system, there was no design with regards to the motor itself. However, effective integration of the motor with other parts of the motor system was required to meet performance metrics. First, the stepper motor that was used, Powermax 2 M21NRXA, was no longer in production, and was provided for use in the project. Therefore, a datasheet for a similar model was used in order to design the circuit [2].The reference model used was M21NXXA. It allowed the starting max torque value of 1 Nm. The use of a unipolar design meant that system operated at a phase current of 1.7 A. As the maximum phase current on the driver is 2.8 A, it was then found by Figure 3-28 below that the actual torque output is about 0.6 Nm. Figure 3-28: Torque Graph[2] 46

48 For the speed operation of the stepper motor, the motor was a 1.8 motor such that for each change in phase, there is rotor motion change of 1.8. As such, from the operation of the driver, the pulse output from the flip flops determined the rate at which the stepper motor moved. For our system, the maximum speed was close to 190 Hz, which can be written via Equation 3.18 as a rotor speed of: frequency of phase change degrees per phase change = r otor speed (rotations per second) (3.18) degrees per rotation = 0.95 rps 360 Lastly, regarding the phase A-D setup, each phase affected an individual coil within the stepper motor. The coils A and B, when charged, pushed the rotor in the opposite direction. Therefore, in operation, only one was powered at a time to create a continuous rotation rate. The 90 difference between A and B to C and D allowed for the rotation of the motor, by pushing the rotor at different points. These different phases of the motor are shown in Figure Figure 3-29: Phases Into Motor Physical Design The physical design held all the other parts as well as allowed conversion of electrical power into mechanical movement. The physical design had three pieces: the shoulder piece, the pulley, and string. The shoulder piece was the part that contained the control system, sensor system, driver and stepper motor. The pulley was the part that allowed the motor s torque to be converted to mechanical power at the desired torque and speed values to meet our design requirements. To transfer the changes of the pulley to the forearm, the string was used. 47

49 The shoulder piece contained most of the electrical circuitry of the final project. To keep the system compact some electrical systems like the driver, Arduino and sensor system were stacked on top of one another. The stepper motor itself was held to the frame with tight 3D printed sleeve and a velcro strap. The entire design, shown in Figure 3-30, was strapped to the shoulder and kept in place with velcro. Figure 3-30: Bicep Bundle The pulley s role of converting torque to movement was the driving force in the implementation of the design. To connect with the stepper motor, a shaft coupler was used between the pulley and rotor on the stepper motor. Also, in the pulley, there was a place to attach the string without winding it around the whole pulley. The pulley had a groove within it to allow the rope to fit into the pulley. The pulley design is shown in Figure 3-31 below. 48

50 Figure 3-31: Pulley Design The maximum radius of the pulley was 4.5 cm and the minimum was 3.6 cm. These were chosen to allow certain deliverables to be met, and to allow the system to remain functional to the user. The ridge of the pulley is shown in Figure 3-32 below. 49

51 Figure 3-32: Pulley Ridge The final part of the physical design is the string that connected the pulley to the forearm. For safety of the user, a knot in the string was made, and by making this string go through a hole that was fixed to the shoulder piece joint, it was inhibited from overextending. The required mechanical stop was placed 18 cm away from the forearm, in order to maintain the stop at a safe joint distance. The string was wound around the pulley to draw the arm closer to the user and is unwound to release the tension within the string. By doing this, the system safely converted the rotation of the stepper motor shaft into the joint motion that the mechanical system was required to create. Chapter 4 - Results The full integration of the power system, sensor system, control system and mechanical (motor) system was required in order to complete the project. After integration, the combined system was tested against the required performance metrics. This section provides an overview of the integration of the design, results of the tests and calculations used to prove the system met design requirements with justification, as well as the final budget and possible design improvements. 50

52 4.1 Results of Integration and Final Design After building all four systems in parallel, the systems were integrated together successfully. Several design changes had taken place during the design process, and these were reflected in the final integration. Figure 4-1 shows the final block diagram, and outlines the way in which each subsection connected to the overall design. Figure 4-1: Final Design Block Diagram In operation of the exoskeletal joint, each system correctly performed their individual tasks. The integration of the system allowed the project to contract the forearm in response to the flex of the bicep muscle. Along with this, the arm extended if the bicep relaxed. 4.2 Final Design Performance The final design was tested in order to determine the overall functionality of the device and to demonstrate that all of the design requirements were met. Table 4-1 summarises the results of these tests. Section 4.3 describes each design requirement and the corresponding results in more detail. 51

53 Table 4-1: Final Design Performance Feature Desired Value or Range Measured Value System is able to draw a usable actuating signal Pass Pass Signal is filtered with desired frequency range < 200 Hz cutoff frequency -3dB cutoff frequency: 149 Hz -6dB cutoff frequency: 199 Hz System is able to translate the analog signal into a usable digital signal Pass Pass System is able to use digital signal to control movement of joint Pass Pass Joint bends linearly in occurrence with a signal as desired by user Pass Pass Joint responds to electrical signal within desired time frame Joint moves with precision to the desired angle Joint stops motion in response to flex sensor < 500 ms <40 ms +/- 5 degrees +/ No longer applicable Joint is able to move in any direction within the range of motion Pass Pass Joint is able to move arm to meet speed expectations 20 /sec minimum /sec Range of Motion Linear, 120 minimum Linear,

54 Power Usage < 200 W W Torque 5 Nm minimum 5.5 Nm Casing of the unit fits onto arm and is functional Pass Pass Mechanical brake inhibits joint from overstretching Pass Pass Should include electrical safety features for the circuit and user Pass Pass Cost < $1500 $ Description of Design Requirements and Results The following subsections elaborate on how the integrated system met the design requirements in Table 4-1 and the procedures that went into determining those results. The cost design requirement is discussed with the budget in Section System is Able to Draw a Usable Actuating Signal The sensor system was able to capture the EMG signal of the user s bicep and translate it into an electrical signal readable by the Arduino microcontroller and used in controlling the actuation of the motor. Figure 4-2 shows this signal as the output of the sensor system to the Arduino. 53

55 Figure 4-2: Usable Actuating EMG signal Signal is Filtered with Desired Frequency Range The EMG signal from the user s body was filtered so as to cutoff frequencies above 200 Hz. The cutoff frequency of the filter in the sensor system was measured using a sine wave input at the electrodes and monitoring the resulting wave at the output of the filter. This test was first conducted by applying a sine wave at 40 Hz, a frequency much lower than the expected cutoff. This resulted in an output waveform with a peak-to-peak voltage value of 180 mv. After this, a 200 Hz signal was applied, which resulted in a 64 mv peak-to-peak voltage at the output. The 200 Hz output signal amplitude was 50% that of the 40 Hz output signal, which is reflective of the -6 db difference indicative of the cutoff frequency for a second-order Sallen-Key filter[10]. The -3 db frequency of the system was also important, as it is the cutoff frequency for first-order filters and it is a value to which circuits are typically normalized[10]. The -3 db cutoff frequency of the system was determined in the same manner as the -6 db cutoff frequency. It was determined to be 149 Hz, which is on the higher edge of the dominant frequency range of the EMG signal[6], guaranteeing minimal impact on the desired output. A value of 85 mv peak-to-peak output voltage was found at this 54

56 frequency, which corresponds to 70% of the amplitude measured at 40 Hz and a -3 db drop in magnitude. Both the -3 db and the -6 db cutoff frequencies were less than the required 200 Hz, and as such the filter met the design requirement System is Able to Translate the Analog Signal into a Usable Digital Signal The Arduino was able to read the analog signal from the sensor system and convert it into a digital signal to control actuation. The voltage value of the analog signal was read with the analog-to-digital converter built into the Arduino. This value was then used to calculate a RMS value used to control the motor. Figure 4-3 shows the serial monitor output of the Arduino receiving the EMG signal and outputting an RMS value. 55

57 Figure 4-3: Arduino Serial Monitor Output Displaying Calculated RMS Values System is Able to Use Digital Signal to Control Movement of Joint The Arduino used the calculated RMS digital signal to control the motor actuation. As described in Section 3.3.2, the Arduino used the RMS signal to output a variable clock that controlled the motor speed. It also controlled the direction of the motor through digital pins connected to the driver circuit. These controls have been verified through physical testing. 56

58 4.3.5 Joint Bends Linearly in Occurrence with a Signal as Desired by User The elbow joint of the user was assisted in moving at a constant speed in both directions of motion. The direction of motion was controlled via the EMG signal controlled by the user. This also has been verified through physical testing Joint Responds to Electrical Signal Within Desired Time Frame The system response to an input signal in less than 500 ms is demonstrated in Figure 4-4, where the blue signal was the input and the yellow signal was the output. The blue input signal shows a flex in the form of a signal increase. Using the timestamp of 10ms and the approximation of 4 frames till the output frequency changed, the final system response was determined to be <40 ms. The voltage scale are 2 V/frame for blue and 5 V/frame for yellow. Figure 4-4: Signal Response Time 57

59 4.3.7 Joint Moves with Precision to the Desired Angle In order to determine whether the system met this design requirement, a few assumptions were required. Figure 4-5 highlights the design of the system. The elbow hinge point to the string attachment is called the forearm distance and given a value of 33 cm, because for the operator used to test the system, the distance was about 33 cm. The distance between the center of the pulley to the elbow joint was called the bicep + shoulder mount distance and this value was also 33 cm based on the distance with the operator used to test the system. The string was the X component on the triangle and A was its corresponding angle. Figure 4-5: Joint Triangle The precision of the joint was dependent on the precision of the stepper and correlated to the pulley radius according to Equation With a radius of 3.6 cm and the precision of the stepper motor in being 1.8 the precision of the pulley was determined to be: 2π(r)(Precision of Motor) P recision of Pulley = 360 (4.1) 2π(3.6 cm)(1.8 ) = P recision of Pulley = cm The degree change resulting at the elbow joint as the string is pulled by the pulley was able to be determined by Equation 4.2, which finds the opposite angle from X. Given this, the following substitution was made: 2 2 2(33) X 2(33) 2 A= cos 1( ) (4.2) 58

60 2 2 2(33) (Precision of Pulley) 2(33) 2 Final Precision = cos 1( ) Final Precision = This value ensured that the design was capable of attaining a precision more precise than the deliverable of Joint Stops Motion in Response to Flex Sensor The flex sensor was removed from the final design of the project, as discussed in Section Therefore, it was no longer a required deliverable Joint is Able to Move in Any Direction Within the Range of Motion The motion of joint was required to extend and contract depending on system inputs. The full integration of the design achieved the ability to rotate the pulley both clockwise and counterclockwise via the control system thereby fulfilling movement direction requirements Joint is Able to Move Arm to Meet Speed Expectations The maximum speed was correlated to the size of the pulley, the location of the string on the forearm and the max speed of the stepper motor. In order to determine whether the system met the performance metric for speed, the same assumptions outlined in Section were used. The size of the pulley has a radius that varied between cm, and the maximum rotor speed was 0.95 rotations/sec. To ensure a change of over 20 /sec, the pulley had to move the string a distance greater than what is described in Equation 4.3: sin(a) 33 cm D istance change > sin(b) This equation is the sine rule and in the case of this system A was 20 and B was 80. Therefore the distance change had to be greater than 11.5 cm. The maximum distance was found using Equation 4.4, which uses the circumference of the pulley multiplied by the max speed of the rotor to determine the maximum distance changed per second. (4.3) 2 π(max Radius of P ulley )(Max Rotor Speed) = D istance changed per second (4.4) 59

61 2 π(4.5 cm)(0.95 rps) = cm/sec Given that cm > 11.5 cm, the maximum speed of the system met the design requirement for speed. Equation 4.2 was used to find the speed at the elbow: 2 2 2(33) (26.68) 2(33) 2 Speed at Elbow = cos 1( ) Speed at Elbow = 48 /sec Given the speed of 48 /sec at the elbow, the design requirement for a speed of 20 /sec at minimum was met Range of Motion The range of motion in the joint was a function of the length of string. Using the assumptions outlined in Section 4.3.7, note that the range of motion had to be 120 or A = 120 to meet performance requirement. If X is set to the knot length of 18 cm then A is 32 using Equation 4.2. The final angle corresponding to the length of string was defined by Equation 4.5: A ngle for F inal String Length = Angle for Knot Length (4.5) A ngle for F inal String Length = = 152 Then Equation 4.2 was rearranged into Equation 4.6 to solve for X with and A = 152 : X = 2(33) 2 2 2(33) cos(152) (4.6) X = 65 cm The length of X = 65 cm guarantees that the range of motion is greater the 120. By using this length of string, this design requirement was met Power Usage The entire integrated system used W during full operation, which was well below the design requirement for a system using less than 200 W. This value was measured while the motor was running at full speed in order to obtain the maximum power required during normal operation. Figure 4-6 shows the test setup for this test. 60

62 Figure 4-6: Test Setup for Power Measurement A 1 Ω resistor was added in series to the supply to the Arduino in order to measure the current in that supply. The current supplied to the boost circuit and motor was measured using the voltage across Rsense in the boost circuit. The total voltage supplied was measured across the battery. Table 4-2 shows the measured voltages and calculated currents. Table 4-2: Summary of Power Usage Calculations Variable V1 V2 V3 Current to Motor Current to Arduino and Sensors Total System Power Value Measured or Calculated 21.1 mv (Measured) V (Measured) 6.05 V (Measured) 4.22 A (Calculated) 94 ma (Calculated) W This test verified that the power usage requirement was met. 61

63 Torque The torque requirement was measured by finding both the stepper motor output as well as the total force exerted on the arm. Derived from the stepper motor, the torque output at the rotor was 0.6Nm, this in combination with the size of the pulley resulted in an output force defined by Equation 4.7: F orce at P ulley = T orque of Stepper Motor /P ulley Radius (4.7) F orce at P ulley = 0.6 Nm /0.036 m = N Equation 4.8 was then used to determine the final torque value at the elbow, using the same forearm length assumption from Section 4.3.7: was met.. T orque at Elbow = F orce at P ulley L ength of F orearm (4.8) T orque at Elbow = N 0.33 m = 5.5 Nm This value of 5.5 Nm exceeded the desired value of 5 Nm and as such the performance metric Casing of the Unit Fits onto Arm and is Functional The casing was the shoulder model that contains most of the electrical circuitry. Figure 4-7 shows the model mounted to the shoulder. Given that the device demonstrably fit onto the user, and that the device was functional while on the user, this design requirement was met. 62

64 Figure 4-7: Shoulder Model Mounted Mechanical Brake Inhibits Joint from Overstretching The mechanical safety brake specification was proven through the design of the string system. Due to the fact that the string only had enough length to allow range of motion about the joint of 120 and was limited to this amount by a knot in the string, the system was constrained to this range unless the string broke, at which point the system is unable to apply any force to the arm. As well, the knot placed on the string limits the string from overwinding and as such prevents any excess stress on the elbow joint Should include electrical safety features for the circuit and user The electrical safety of the design pertained to every electrical component. One part of the safety design implemented was the shoulder mount which isolated the electrical components from the user.. Another safety feature was the capacitors placed between the electrodes connected to the body and the sensor system, which helped protect the user from inadvertent DC currents. Lastly, the power system had a switch connected to the battery to allow for the device to be quickly turned off. Therefore, the final design had many built-in electrical safety features. 63