An Automated Guitar System. Group Seven Kacey Lorton, BSEE Brian Parkhurst, BSEE Anna Perdue, BSEE

Transcription

1 An Automated Guitar System Group Seven Kacey Lorton, BSEE Brian Parkhurst, BSEE Anna Perdue, BSEE

2 Table of Contents 1. Executive Summary Project Description Project Motivation and Goals Objectives Project Requirements and Specifications String Depression Requirements String Picking Requirements Software Requirements Power Distribution Requirements Research Related to Project Definition Existing Similar Projects and Products Relevant Technologies Electromechanical Devices Linear Motion String Depression Rotational Motion String Picking Directional Motion String Selection Dynamic Control- Picking Depth MIDI Conversion Control System FPGA/Microcontroller Comparison Programming Language Power Supply Serial Communication Strategic Components Stepper Motors Servo Motors Driver belts Solenoids ii

3 3.3.5 Microcontroller Power Supply Project Hardware and Software Design Details Initial Design Diagram Electrical Block Diagram Mechanical Assembly Block Diagram Software/Firmware Block Diagram Stepper Motor Control (Picking System) Servo Motor Control (Pulley System) Dynamic Control String Depression Solenoids and Control Power Supply Design Summary Electrical Design Summary Power Regulation Servo Drivers Stepper Motor Drivers Solenoid Drivers Mechanical Design Summary PCB Enclosure String Selection Assembly Picking Assembly Project Prototype Construction and Coding PCB Design Software and Firmware Summary Software Summary..78 iii

4 6.2.2 Firmware Summary.82 7 Project Prototype Testing Solenoid Subsystem Testing Stepper Motor Subsystem Testing Servo Motor Subsystem Testing Dynamic Control Subsystem Microcontroller Testing Power Distribution Testing MIDI C++ Programming Testing Integrated System Testing Administrative Content Budget and Finances Milestone Timeline...96 A Appendix A.. i A.1 Table of Tables.....ii A.2 Table of Figures. iii A.3 Copyright Permissions..v A.4 Bibliography...vi iv

5 1 Executive Summary Music is nearly ubiquitous in our lives. There are many great musicians, Jimi Hendrix, Slash, Keith Richards, or even Jimmy Page. These musicians each have a different sound and are known for their remarkable talent with the electric guitar. Though these bands have either broken up or the artists have passed away their talent is unique to only them. There have been many cover bands that have tried to reproduce the sound of Led Zeppelin or the Rolling Stones. However some may say that the one component that is missing to these cover bands is a great guitar player. Imagine going to see a Led Zeppelin cover band play. While watching and listening to the music play, all of a sudden you hear Jimmy Page s solo performance in Stairway to Heaven, almost exactly the same as when you first heard him play in For some this could be a beautiful experience, because the one thing that always seems to be missing in a cover band is the true talent that comes from the original electric guitarist. This is where an automated guitar would be useful in the modern era. The idea of an automated guitar is to be able to replicate the sound of the electric guitar without human performance. This document discusses the idea of a design to make this experience possible. This project will be achieved through the efforts of Kacey Lorton (Electrical Engineer), Brian Parkhurst (Electrical Engineer) and Anna Perdue (Electrical Engineer), while satisfying the requirements of ABET (Accreditation Board of Engineering and Technology) for all graduating computer and electrical engineers. This will also provide hands-on opportunity in the field of electrical engineering, prepare us for design opportunities upon graduation, as well as learning to work in a group and understanding the importance of team work. The documentation discusses the design of a controlled electrical circuit and mechanical system that produces reliable playback of audio sound files on guitar. The electrical circuit including control system should be installed on a PCB with interfaces both to a computer USB drive and to the mechanical system. The idea starts with inputting any music file through the computer. The file will be converted, and will talk to the control system. The control system will talk to the different subsystems that we will create to make an automated guitar possible. The electromechanical devices we will create will include string selection and depression. This is what gives the guitar its ability to produce different pitches. Along with that we will create a subsystem for picking. The picking apparatus will also include dynamic control so that the guitar picks will move up and down to create a softer or louder sound, depending on the file chosen. This will give the automated guitar its unique quality of mimicking an electric guitar sound. The focus will be on electromechanical design, power systems, circuit design, programming, and serial communication. We decided to design an automated guitar because it incorporates 1

6 the group s interest in music and engineering. Though combining music and electrical engineering could take a variety of forms, a challenge would be creating an automated musical instrument performance through an electromechanical system. 2 Project Description In section two, we describe our project. This includes the motivation and goals behind the development of the guitar as well as the requirement and specifications in order to complete the design. For our senior design project we have developed a plan to create an automated guitar. The project description will give brief insight on our objectives, and what is required as a starting point to begin development. 2.1 Project Motivation and Goals All three of our group members have a love for music, whether listening or performing. Our group members have multiple guitars and have enjoyed playing them for years. One of us holds a Bachelor s of Music Education. Though now he is working toward the field of Engineering, he thinks the idea of joining the two areas of interest in a project is great way to integrate his past and future experiences. Another project that was considered was a system that aided in skills development for novice guitarists. Both of these ideas illustrate our affinity for music and our desire to share great listening and great performance experiences with others, even if they have no musical background. As an extension of our sheer interest in music, another motivation for this project is the assumption that it can hold our interest for several months. Many groups choose to design similar projects, such as robotic devices or vehicles. Since our project path is much less traveled, and since the expected end result is not just an object maneuvering the way it is intended to, but creating enjoyable and impressive audio playback, we feel this project will keep itself fresh and interesting for each of the few hundred days we work on it. Automated playback of musical instruments is not a new concept. From music boxes to player pianos, this automation technology extends back no earlier than the 19th century. Automation of guitar performance, however, is a much more fledgling pursuit than the same for piano. As is explained further in the Existing Projects section (3.1), several, if not dozens, of guitar automation designs exist. There is room for improvement. Many musicians and music lovers, and seemingly guitar players in particular, have distinct notions of what constitutes good music. Contempt exists in the minds of many that attempts to approximate human instrument performance are either illintentioned or doomed to failure. The reason for this is the nuance with which humans are capable of playing music instruments, with thoughtful variations in tempo, note length, dynamics (volume), pitch, and tone, is extensive. These 2

7 nuances, combined with other intangible human qualities, add to the emotion, the soul of performance. The thought that an electromechanical system can exactly reproduce or closely approximate those qualities, if not threatening, is considered questionable. These concerns are founded in truth, as many automated instrument designs limit their level of nuance to correct pitches or notes at correct times, without even dynamic sensitivity, and their performances can rightly be described as robotic. All of this is another motivation for our project. We do not intend to create the world s first human-sounding guitar automation. We do understand all this room for improvement means room for design. Our project plan addresses pitch, time, and dynamic sensitivity. However, once these goals are achieved, this project affords several avenues for modification and upgrade, to closer approximate human performance. A final motivation is an educational one. Our hope was to cover as many areas of study as were appropriate in the design of this project. Our group members would like to complete their careers at UCF with knowledge and experience in areas as diverse as electromechanical systems, pulse width modulation, embedded systems, and power systems. This project will give us the opportunity to do so. 2.2 Objectives The overall goal of this project is to create an automation system to be used on an electric guitar, essentially making a real-time analog playback device. Our device would be used in place of human guitar playback or digital re-creation of guitar tones from a digital music sequence. To further detail why this would be useful, there are several cases in which such a device is advantageous. In the case of live music performance, bands or individual musicians sometimes use backing tracks or automated musical accompaniments during live performances, usually in the form of MIDI sequences being played back on a synthesizer, keyboard, or simply on a computer. The usual reason for this is because of complexity (or simplicity) of the sequence, or in the case of one-manbands, to add more depth to the performance. We want our guitar system not only to be as functional as one of these options for automated sequence playback, but also to provide the sense of a human playing the guitar, instead of a machine. The typical ways that sequences are created are through either MIDI-enabled keyboard pianos, or by software on a computer. Sequences can be created on either and sent to either. We wish to use a similar method of taking a predetermined music sequence and sending to our instrument hardware, and having it play it back as if it were a person. To give a picture as to what the end product could look like, a user would download or create a MIDI file on a computer, be converted to whatever format for control lines is needed for our system, be sent to the guitar hardware, and at the proper 3

8 time, be played back. The playback would be sensitive to speed, note intensity, durations, and as many inflections and characteristics as possible. The guitar output would be sent to an amplifier or passed through effects pedals, as any other guitar might be in a live performance. Our design can be broken into four main sections, encompassing hardware, firmware, and software. Falling under hardware is the mechanical systems required to press frets, pick strings, and create dynamic elements in the music. The other side of hardware is the electrical components used to control the mechanical parts, both in terms of power distribution and control signals, along with an interface with a computer. For firmware, we will create a control construct based on picking a note, which encompasses picking a string and fret, deciding when to play that note, at the same time potentially simultaneously doing the same for a different fret/string combination. In terms of software, we will be converting a MIDI file format into our firmware format, using a program that we would need to create ourselves in a high-level language. Also included in software would be a GUI controlling when the guitar plays, and potentially other features. 2.3 Project Requirements and Specifications In the section of project requirements and specifications, we discuss the different subsystems of the project. In this section we describe our general idea and what specifics will be needed in order to make the design a reality. This includes the use of different subsystems. The subsystems included in our design are string depression, string selection, string picking, and dynamic control. Our string picking and dynamic control subsystems will be joined in the same apparatus, with one set of motors for string picking and another for dynamic control. Similarly, the string our string depression system will work in tandem. The actuators will be movable, to select different strings. Along with these systems we discuss the requirements for our power supply String Depression Requirements The string depression system is one of the electromechanical subsystems in our project. It will require the use of 12 actuators, one for each fret we choose to be able to depress. This is less than the 22 frets on a standard guitar. We have chosen to focus on 12 frets rather than 22 for cost and feasibility purposes. For each actuator there will be two pulleys, driven by a motor. The pulley subsystem will be controlled to move the actuators back and forth between six strings. The motor driving the pulleys will be located at the bottom of the neck, one for each fret. The actuator will be attached to a drive belt that will be fed through the pulley system. The fret system will require 12 drive belts, one for each actuator. To prevent metal on metal interface 12 small pads, one for each actuator, will be affixed at the bottom of each device. The fret system will then be controlled through a digital electronic control system. Table summarizes the component requirements of this subsystem. 4

9 Components Desired Actuators 12 Pulleys 24 Drive Belts 12 Motors 12 Table : String Depression System Mechanical Components String Picking Requirements The materials and devices we expect to use for string picking include standard guitar picks, motors to control the picks, intermediate structures to interface between the motor shaft and the picks, and a holding structure to captivate each motor and provide stability and support for the whole system. The specifics of the demands for and selection of these parts are given in later sections. One of the more crucial parameters of this system is the frequency of unique notes, or sounds, produced. We expect that each motor in the picking system should cause the pick to produce unique notes at a frequency of 10 Hz, which, in music terms, is equivalent to 16 th notes, at 150 beats per minute (bpm). Modern music tempo generally falls in the range of beats per minute. 16 th notes are considered to be fast notes, and 150 bpm is considered to be fast tempo. So, though we don t expect our system to produce superhuman performance, the upper limits of our system s capabilities would be considered fast. Inherent in the note frequency requirements is the ability for the motor to change direction. We do not expect the motor to use its whole range of motion, but to cause the pick to move back and forth across the string as a human would cause it to do. Not all motors are designed or intended for frequent changes in direction, so this requirement should filter out many potential motors from consideration. The numerical requirement is that the change of direction does not hinder the system from achieving a frequency of 10 Hz. Table , shows the parameters we must work within. Parameters Values Frequency 10 Hz Notes 16 th Beats per Minute 150 Table : String Picking Frequency Parameters Another requirement on the motor behavior is the reliability of position control. The timing and synchronization of all electromechanical parts to produce stable performance is the most important overall goal of the system. As such, we must be able to know and control in great detail, the angular position of the pick and the 5

10 motor. Only a few specific types of motors are designed to control position in the way we require, so this requirement should further narrow our focus in motor selection. The motors we choose must deliver the torque need to vibrate the string as a human would. We intend our system to have dynamic control. As mentioned in the dynamic control section, in order to produce a louder sound the string must oscillate at a higher displacement amplitude. To accomplish this, the pick must displace the string further while picking. This will happen when more of the pick overlaps the string. When this is the case, the length of the effective lever arm for the motor is decreased, so that more torque is required to apply the same force to the string. For the motors we choose, torque will be an important criterion for our design. Our system for string picking will not include one pick as a human would use. To improve performance, and for simplicity of design, our system will include six actual guitar picks whose motion is controlled by motor. As this is the case, our system must fit six motors within the picking area which is 60 mm wide. Because of this, the required width of each motor is less than 10 mm if all motors are placed sideby-side, and 20 mm if the motors are staggered, with two separate sets of three motors each. The height limit for the motor is 60 mm, twice the height of an average pick, as a motor that size would not allow the pick to rest in place to pick the string without itself interfering with the string motion. The length limit of the motor is 50 mm, as a longer length would disallow the potential use of staggered motor sets. Each motor should weigh less than 250 g to avoid placing a design burden on the system. The following three requirements are common to many systems: low noise, low cost, and low power. Motors can be noisy, and while some modifications may be applied to the motor itself, one of the mitigations for this concern is to raise the amplification of the electric guitar sound to a level at which the noise produced by the motors is relatively negligible. The other requirements are loose but considered reasonable to meet. Our goals for these parameters are for each motor cost less than $20, and to dissipate less than 10 W of power each when in regular use. Table summarizes our requirements on the motor for the picking system design. Goals Maximum Price $20.00 Power Dissipation 10 W Weight 250 g Height 60 mm Length 50 mm Motors 6 Table : String Picking System Motor Requirements 6

11 2.3.3 Software Requirements Our software will be required to properly decompose the MIDI file format into its components of note, duration, volume, and other parameters available within the file and rearrange the information to the form that we will use on our control system to control the hardware. The software must also include a user interface to allow the selection of a MIDI file, a button to begin the conversion process, various data outputs indicating status and estimated time to completion of conversion. A button should also exist to allow the user to signal the guitar to begin playback. Our hardware should also be a recognizable USB-pluggable device with the proper serial communication protocols in place to allow the transfer of data and commands Power Distribution Requirements The power will need to be distributed so that the supply will be provided to motors, actuators, controller(s) and driver circuits. In order for this to be achieved power supply will need to be centralized. Along with a centralized distribution, the power supply will require power regulators so that enough current is being passed to the actuator, motors and circuit. The power supply will need to have enough current that can be provided to the different systems on the board simultaneously. The power supply we wish to use should be simple and plugged into a wall, where we can simply flip the power on and off. Table is a chart of the expected voltage and current values we will need. We will need 24 VDC, because we are regulating it to many different sub-systems. This is also why we desire a higher output current than more common, low-current devices. VDC OUTPUT CURRENT Table Ideal Voltage and Current Relationship 3 Research Related to Project Definition In section three we describe our research process. First, we discuss existing products or similar projects related to our project, the automated guitar. This is important to research because it offers a way to learn how projects similar to ours have been created. We may need to recall this and it will prove to be very useful if we need to troubleshoot in the building process. In section three will also discuss research done on the types of devices that will help us meet our requirements. We have spent time gathering information on electromechanical parts, MIDI conversion, software language, power supplies, serial communication, and 7

12 controller parts. In order to make this design a reality we need to research and consider many potential solutions to our design concerns. This is why we have used the research to discuss our ideas and propose alternatives for each subsystem, as can be seen in the documentation. Along with the research of the various types of components, a purpose of this section is to determine strategic components we will need. This is where we have recorded our shopping. This will include tables created from component specifications, to compare and contrast parts. We will also be discussing which parts we have picked and why we have decided to choose the component for the automated guitar. It discusses the relative strengths and weaknesses for each strategic component needed for the system and how we learned that one part component may offer better technology for our specific project than another part. 3.1 Existing Similar Projects and Products Few documented automated guitar projects exist, relative to the large number of quadcopter projects, for example. Despite the relative lack of precedent and documentation, what is available has been useful to inform the creation of our project.the first project that comes up as a search result for automated guitar Is one created by Ken Caulkins. His device, along with a majority of the other projects we were able to observe, makes use of linearly moving parts for string picking. The underside of the part has a projection that comes in contact with the string as the part moves across the body of the guitar. Our original conception of this function was of rotary motion from the shaft of a motor to control the picking. However, we can perceive the benefits of a linear system potentially including high speed of operation. In addition, his fret system was an array of parts, each one above a unique fret location, that would depress the string when pulled by a wire from below the guitar. To accomplish this, Mr. Caulkins ran the wire through holes he created in the guitar neck, a process which we are not inclined to emulate. Another process we are not inclined to emulate is that used by Clippard Instrument Laboratory in the design of an automated guitar; pneumatic actuation. Though pneumatics is an interesting and somewhat unknown area of study for us, we project that it would be expensive and beyond the scope of the knowledge and skills we desire to acquire in this project. Though Clippard employed pneumatic solenoids to depress the strings, the majority of the projects we researched used electromechanical solenoids. Gregg Bizier, whose automated guitar project was for a Senior Design course at the New England Institute of Technology, also used solenoids to strike the strings. This, an alternative to picking, is the same kind of sound production used in pianos, and as such, produces a correspondingly different tone. It is our thought that, as much as possible, we ought to design our system so that it plays the guitar in as human-like an approach as possible. Because of this, we do not plan to incorporate solenoids for the sound production subsystem. Before considering more elements of Mr. Bizier s design, we note that another project used solenoids differently in sound production. A system created by Demin 8

13 Vladimir aimed solenoids, parallel to the body of the guitar, at the sides of the strings. The result was still a piano-like, striking motion, but in doing so, the edge, rather than the middle, of the solenoid shaft attachment brushed the string. The sound quality was improved over Mr. Bizier s design. However, since the attachment was not a flexible plastic pick, but a thick metal plate, the sound was still overbearing and crass. This solidified our confidence in the design idea of including real guitar picks in the sound production subsystem. Mr. Bizier had his guitar system supported by a stand underneath the neck of the guitar and by two rails along the neck. The wire harness ran along the rails, so that the wires were kept from obstructing the movement of any electromechanical parts. This structural subsystem is similar to what we would like to implement in our design. It would provide stability to the guitar itself and the system built around the guitar. Also the upward force applied on the neck serves to lessen the force required from the solenoids, which loosens the specifications on parts we are considering. This will drive down cost and give us more flexibility to focus on other important parameters, including size. Finally, we were able to get a sense of the software/firmware side of Mr. Bizier s project. He created a GUI in which he could energize and test each solenoid individually. This sort of functionality is not necessary for final performance, but can be useful during development. It is worthwhile to consider not only writing programs only for performance, but intermediate programs that may expedite troubleshooting. At one point in his project, he disconnected the system from the controlling computer and allowed the system to perform music by pulling its data from built-in system memory. While his project memory held only two songs, it should be noted that he created his project in 1998, and on-board memory systems for controlling devices are presently much larger and more powerful. Because of this, we could import a bank of music for the guitar to play, as a jukebox would. Above, we mentioned our interest in sound production through rotary actuated picking. Two projects we researched included rotary picking. Both were Senior projects, one at the US Naval Academy and one at Saginaw Valley State University. What was notable about the performance of both projects was the poor timing of the electromechanical devices actions. We cannot verify that this is correlated to the use of rotary picking, but it does cause us to be more aware of potential design flaws and limitations that may arise. One potentially significant difference between the use of rotary motion in these projects and our concept is the length of the arm that extends perpendicularly from the shaft of the motor, to which the pick connects. We estimated the length from the shaft to the point in which the arms connects to the top of the pick at around mm. This unnecessary extra distance makes precision movements more difficult to accomplish and may lead to timing issues. Our plan is to connect the arm to the shaft so that the arm acts as an extension of the shaft, in line with it. It would then run horizontally across the top of the pick, fastening to it. This reduced moment of inertia of the system will require less torque for the motor and should enhance the 9

14 maximum precision and frequency of picking. One final note about the US Naval Academy project is that it employed two PCBs and one breadboard. In our projected design we expect to integrate the central electrical system on one PCB. The final project to consider is that of a Senior Design team from ITT Technical Institute. Their approach was to remove the body of the guitar altogether, leaving just strings to pick and to depress at precise intervals where frets would be. Though we will not in any way emulate the guitar-less guitar system, their design did include elements worth consideration. Their control system was centered on an Arduino microcontroller. Though this information was not readily available for most of the projects, no FPGAs are known to have been used for any of the projects we researched, while at least this one employed microcontroller technology. We will discuss the choice between FPGA and microcontroller technology in our Relevant Technologies section (3.2). The ITT Tech team s electromechanical system included solenoids, as did most, but also included relays to control the current, determining the state of the solenoids. We intend to consider relays in our design. What was more helpful was that the team called out the specific voltages used for both types of components. This gives us a reference point when considering and comparing parts. In summary, researching this projects gave us several examples and nonexamples of good design choices to incorporate, including solenoid depressing of strings, computer interfaces, pneumatic actuation, and rotary picking. One common element in all projects was the fret array. No team or individual chose to create an automated guitar system that attempted to emulate the movement of the human hand across the fret board in selecting fret locations. We believe that, as we are interested in creating a sliding string selection system that moves solenoids between strings to select correct fret locations, this and other aspects of our design come together to create a concept that is both derivative and original. 3.2 Relevant Technologies In the relevant technologies subsection we discuss the technology that will help us create the automated guitar. This includes parts such as motors, actuators, digital electronic control devices, and computer language. Relevant technologies include electromechanical devices, MIDI conversion, serial communication, and power distribution Electromechanical Devices The design of an automated guitar has many subsystems, which require the use of electromechanical devices. Electromechanical device combine electrical design and mechanical theory, which means these devices carry out electrical operations to control mechanical movement. For our purposes this includes linear motion of the actuators, rotational motion of motors, and the linear motion of the belt and pulleys. All of these devices will be required to work synchronously to produce the guitar sound we require. 10

15 Linear Motion - String Depression The string depression system in our design requires the use of linear actuators. The linear actuators we have researched and found to be satisfactory for our design, are solenoids. Solenoids produce mechanical force by the induction of a magnetic field through a coil. This mechanical force acts upon a moveable plunger. We plan to apply the force of this plunger to the guitar string, pressing it to the guitar neck so that the string vibrates at a certain frequency, producing a unique sound. Many solenoids include a spring that, when the power in the circuit is cut off, the plunger returns to its neutral off position. Solenoid plungers can be switched in between an on or off position by a number of electrical components, including transistors. Solenoids act as inductive loads on a circuit, whose voltage drop is proportional to change in current so a protective component is required to be placed in parallel with the solenoid coil to prevent high voltage from damaging the semiconductor switching device. In this case we will need a freewheeling diode. In the design or testing part of the solenoid we may also use a small value resistor in the circuit. Figure shows how a DC solenoid can be controlled using a transistor to switch. Figure Transistor Controlled circuit A freewheeling diode is used to conduct current when the transistor turns off. Adding resistors from the control system, output to ground and from the controller to the transistor gate/base adds resistive isolation between power switch and the control system and prevents the transistor from overheating. It is important for the impedance of the solenoid will fall to the DC resistance after the solenoid core saturates. The transistor needs to be able to handle the current load or increase the resistance by adding resistor in series between the solenoid and the transistor to reduce coil flow. Adding a capacitor in parallel with the resistor will provide an extra start up kick to the solenoid. 24V applied voltage power supply would be needed. The information is to help aid us in our circuit design, if we have any issues 11

16 during the simulation period. The way of pulsing a solenoid would be to have a pulse duty cycle. Usually 10% to 25%, this can generate a large force for a short time but will quickly overheat if it is running for a long time. The duty cycle can be expressed as Figure Duty Cycle Formula Cycle % = [( Rotational Motion String Picking Duty For the string picking subsystem, it would require the use of a motor that would be able to provide a rotational motion. The best way to create a rotational motion would be through using servo or stepper DC motors. Servo motors require a controlled circuit and a position sensor. Power will need to be constantly applied, with the control circuit regulating the power to drive the motor. The angle of rotation is fixed at 180 degrees so the shaft moves back and forth. This is very fast, the micro servo high- torque motor is small enough and has enough torque for the string picking. The motor then would be connected to the control system to control the motion and synchronize it with the MIDI conversion. Table shows specifications for a sample servo motor. Modulation: Analog 4.8V: 0.10 sec/60 Speed: 6.0V: 0.09 sec/60 Weight: 0.30 oz (8.5 g) Length: 0.87 in (22.1 mm) Width: Dimensions: 0.43 in (10.9 mm) Height: 0.91 in (23.1 mm) Table : Data for Servo Motor A stepper motor is similar to servo motors, require the use of an external circuit to energize each electromagnet and make the motors shaft run. Each rotation from one electromagnet to the next is called a step and thus the motor can rotate and full 360 degrees. Stepper motors will provide a constant torque even without the motor being powered. Positioning errors don t usually occur in stepper motors since they physically have pre-defined stations. Stepper motors have very poor 12

17 torque characteristics at higher speeds, compared to the servo motors. The maximum speed of the string picking system would be 10 Hz. Since the tension of the string is not a large load on the motor, the relatively poor torque should have no effect in picking a stepper motor for the design. Table shows specifications of a sample stepper motor. Modulation: Analog Speed: 5 V: 1 step/45 Weight: 0.30 oz (8.5 g) Length: Dimensions: 0.79 in (20 mm) Width: in (28 mm) Table : Data for Stepper Motor Above is information from a small reduction stepper motor, running at 5VDC found at the Adafruit website. There are only 32 steps which is degree per revolution. Inside is a 1/16 reduction gear set so in reality there are 513 steps. The shaft for the stepper motor is flat and easy to attach stuff such as the pick. For this specific stepper motor at 5V in order to keep the motor to run smoothly it was recommended to keep the 5V stepper motor under 25 rpm. However this will not work for our project because we require at least 100 rpm, with a rated voltage from 5-20 volts. We have chosen to use stepper motors in our design as their open-loop position control will be advantageous in standardizing the string picking action. The stepper motors would communicate with the control system to simultaneously pick the string while the solenoids press down on the string. In order to do this an external circuit providing voltage and current to the motor would need to be designed on a PCB board and connected to the output pins of a control system Linear Motion - String Selection The string selection is part of the string depression subsystem. We plan to use 12 solenoids, one for each fret. The solenoid will need to move up and down the individual fret in a directional motion. For linear displacement of the solenoid between strings to take place we have determined that the best subsystem would be of a pulley subsystem. The idea is to have two pulleys that would be located at the end of each fret. The pulleys then would be connected through a driver belt that would run through either a stepper motor or a servo motor. The driver belt would then meet at the solenoid, where it will be connected. The motor will need to rotate fast enough so that the solenoid activation and the belt displacement would happen with a minimum frequency of 10 Hz. 13

18 We researched both servo motors and stepper motors for the belt driver. The research shows that a servo motor runs faster than a stepper motor. Since the directional motion needs to work fast to synchronize all the subsystems, we chose to use a servo motor. A servo motor then will be connected to a driver board, and then a controlled circuit to run the motor. The top of the band will be connected to a part that we will be able to place on a solenoid. If we are able to design this part, we can 3D print this small part for an affordable price. Servo motors are a prepackaged combination of several components. They include a motor, gear ratios used to increase torque and decrease angular velocity, and a stator which feeds the position of the motor back to motor. This prepackaged system makes servos easy to use for many applications in hobby projects as well as industrial design Dynamic Control - Picking Depth To enhance the dynamics of our guitar playing, we wish to be able to control how loud or soft the notes are being played. To do this, we must consider how this is accomplished when a user is playing a guitar. Upon inspection, the guitar players among our group concluded that it is a combination of several factors: how far down the pick is pushed past the string when it is being forced into it, how abruptly the pick is pressed to the string, and also whether or not one s picking hand s side is resting on the strings, very closely to the bridge of the guitar, which is where the strings are attached, in a sense, providing some dampening to whatever inputs are coming from the picking. Essentially, picking the guitar string is displacing the string a certain distance per the stroke, and then releasing it. The more one displaces the string in the perpendicular axis, the higher the amplitude of the sine wave that is generated. Therefore, we conclude that if one lowers the pick further down so that as one picks the string, the string will have more distance to travel with the pick still forcing on it, and therefore will be forced to oscillate with a greater amplitude, ergo, louder. As for the techniques for accomplishing this dynamic feature, we have yet to find any existing projects that have attempted this. Hence, we have come up with our own solution, which is to have the apparatus on which the picks are attached be capable of raising and lowering a short distance to a very fine degree, so as to change how far the picks would be displacing each string when they are used. Our plan is to incorporate all picking motors on the same frame, which could be raised and lowered using a motor, with a certain gear ratio enabling the movement to be incremental. To accomplish this, it must be determined how to use gear ratios to our advantage, and what kind of forces are involved that could be limiting factors. The weight of six motors used for picking, as well as the weight of the apparatus that is holding them in place, cannot be ignored. Figure is a concept sketch of the apparatus design that could be used. 14

19 Figure : Sketch for apparatus design An alternative design configuration has been considered more recently while looking for possible mechanical parts to be used. The part that was noticed that inspired a different, possibly more easily implementable design, is a worm screw, pictured below in Figure Figure : Alternative Design The worm gear removes the inherent issue of the weight of the pick motors wanting to force the apparatus to a resting position, as a worm gear can turn a load gear, but a load gear cannot turn a worm gear. This means gravity cannot cause the frame on which the motors are mounted to slide unwillingly. See Figure Figure : Apparatus including worm gear design 15

20 A potential candidate for the stepper motors to be used in the picking system has a weight of 60 grams. Since there are six, that is 360 grams of weight, or 4/5 of a pound. Accounting for whatever material the framework is made up of, it would be reasonable to assume that the sum total of the weight being lifted is between 1 and 2 pounds MIDI Conversion As one of the objectives of our automated guitar project is to play theoretically any song that an end user desired, it is important to use a commonplace data format already in existence. One of the first music file types was the Musical Instrument Digital Interface, or MIDI. As a brief history, shortly after the personal computer industry took off, Instrument companies collaboratively created the MIDI standard in 1983, as a means for instruments and computers to communicate and control one another, allowing musicians to change how music was performed and recorded. For our guitar to be able to play a MIDI file, it is first necessary to observe the MIDI file structure, so as to decode it to our own purposes. MIDI data is stored in bigendian, variable-length format, with the upper one bit of each byte indicating whether another byte in the same datum follows, the maximum variable length of datum being 4 bytes. There is a header chunk for each MIDI file, consisting of a chunk ID, which is identical to every other MIDI file. Next are fields indicating size of chunk, whether the file contains one or several tracks, and the total number of tracks. The final field, and of most use to us, is Time Division. The MSB indicates whether the tracks will be played in terms of ticks per beat (bit mask 0x8000) or frames per second (bit mask 0x7FFF). Frames per second uses the SMPTE timecode, a music/film industry standard. For individual tracks within a file, there is a header indicating how many sequences are encompassed within the track, after which comes the actual musical sequence data. Several types of events exist, with track title and beats-per-minute normally being encoded at the beginning of the sequence. If BPM is not specified, a default of 120 BPM is selected. Events are encoded in chronological order, with a field indicating the time delay from the previous event, with the lowest value being zero, meaning the event should occur simultaneously with the previous event. There are three types of MIDI events, but the bulk are MIDI Channel Events. There are eight Channel Events: Note Off, Note On, Note Aftertouch, Controller, Program Change, Channel Aftertouch, and Pitch Bend. Each Channel Event has two parameter byte fields, specifying which of the 128 notes available is being played, and note velocity, which is known to most as the volume of the note. For Pitch Bend, the fields indicate the value of modulation, with 0x2000 being the position of no modulation occurring. Note Aftertouch is a change in note volume on a note already pressed down. Channel Aftertouch Event is similar but applies to all keys pressed on a channel. Controller Events signal a change in channel state. The byte fields are for specifying what control is changing and defining the 16

21 new value. Program Change Event changes the instrument on a channel. Meta Events are not sent to the instrument, such as text and track name, copyright notice, lyrics, and the end of the track. Time Signature sets a sequence s time signature. The fields for Time Signature include a numerator and a denominator, as if it were written on sheet music. Key Signature has fields to specify how many sharps and flats are used in the music scales used, and for stating whether the sequence is in a major or minor key. There are a few other specialized events that could in theory be ignored for the sake of our project Control System The electromechanical devices used in sound production, including solenoids and motors, need to be governed by a central control unit. The main end-requirement of the control system is that it must be able to send commands to the electromechanical devices in such a way the string picking system is synchronized to the string depressing system. This will bring about musical performance that is cohesive and does not include poor or unexpected sounds. The control system must have a processing speed that is fast enough to send all of its commands to the devices, so that the performance appears to the human to be completely synchronous, even if it is not. That is, the resolution of the deviations from synchronization must be finer than human sound perception is able to sense. Early estimates of the number of electromechanical devices we plan to use in our design are around 30. We would like to control each electromechanical system with separate outputs from the control system. Some devices may require just one control line, while others may require two or more. Because of this, we expect to use a large number of output pins, greater than 40, in our control system. We estimate our memory requirements to be relatively low, as we expect the RAM to contain variables for the use of running the electromechanical device control, and the ROM to contain the actual code and the MIDI files. MIDI files are exceptionally small, with the file size for an average piece of music to be on the order of tens of kb. Because of these considerations, the minimum memory requirement for both RAM and ROM is in the kb range, rather than MB and beyond. Beyond performance and general interface resources, we are also interested in power dissipation, cost, and specific interfaces and functionalities. As with the rest of our design, a general goal is to minimize power dissipation. We do not expect any potential control system candidate to have poor performance in this area, but we do consider it, nonetheless. We expect the cost of our control system to be less than $20, as we are looking for good performance, but understand that the scope of functionality we are expecting it deliver is relatively narrow. One type of electromechanical device we expect to use is meant to be controlled using pulse width modulation (PWM). The control system technology we should be able to produce this output. As for specific interfaces, we want to use a USB connection to connect to a computer. 17

22 FPGA/Microcontroller Comparison The technologies that can be used to accomplish this behavior include FPGAs and microcontrollers. The potential benefits and drawbacks of both are considered below. After brief research into available FPGAs and microcontrollers focusing on those that fit our desired price range, we found that a large number of models of both FPGAs and microcontrollers were available that satisfied our internal clock speed, I/O, memory, and cost requirements. As such, we had to narrow our search based on the secondary and other criteria. One main difference between FPGAs and microcontrollers is the way in which they are implemented. FPGAs, being vast arrays of logic gates, are able to emulate large digital electronics systems. If multiple modules are desired to run simultaneously, more gates are used, and the modules are able to run in parallel. On the other hand, microcontrollers, which use digital logic differently to form memories that interact with a central processing unit, are limited by the performance of the processor. That is, the processor executes a set of instructions sequentially, so the amount of parallel operations is limited by the word width of the architecture. However, as the processing speed increases, the microcontroller can closer approximate the completely parallel behavior of the FPGA. The FPGA becomes more powerful by being large. The microcontroller becomes more powerful by being fast, with wide word widths. The significance of this comparison is that FPGAs often have greater capacity of functionality than microcontrollers. FPGAs can generally do more than microcontrollers can. They can even create systems that imitate microcontrollers. They are also often more expensive, though we have determined that cost is not likely to be a problem with either device. One concern, though, is of using a device that has much more power and functionality available than is needed for the design. That likely describes our situation. An FPGA may be overqualified to control our system, whereas a microcontroller may be simpler, yet may be more appropriate. After some research we have found some microcontrollers to have designated PWM functionality and outputs. Both microcontrollers and FPGAs can certainly implement PWM functionality, but to have existing functionality that is designed specifically for PWM is a benefit to the design of our control system. It may simplify the corresponding design of the motor control. The remaining comparisons are of a practical, rather than performance, nature. All of the group members have had lab experience working with both microcontrollers and FPGAs, as part of required classes. The Computer Organization and Embedded Systems classes have required the use of microcontrollers, specifically the MSP430 from Texas Instruments. The Digital Systems class required the use of FPGAs, specifically the BASYS2 board, which used the Spartan-3E series of FPGA from Xilinx. From this consideration, our group members have had slightly more experience with microcontrollers than with FPGAs, two classes to one. 18

23 In working toward a design project it is important to find a balance between familiarity with certain subject matter and the benefit of cultivating new skills. The use of either of these technologies will allows us to do both. FPGAs must be configured using hardware description languages (HDLs), including Verilog and VHDL, while microcontrollers can be programmed using more commonly used languages like C. This presents an advantage in using microcontrollers. Also, we have found the development software associated with FPGAs to be less helpful and less intuitive to use than that of microcontrollers. Though we have had experience with HDLs and FPGA development software, and though the benefit of cultivating new skills is significant, the potential project delay associated with the lead time of learning to use these tools in the way our project requires, is even more of a concern. When troubleshooting a problem with the function of the control system, it exacerbates the issue to keep in consideration the possibility that the improper use of the language, not the functionality itself, could be responsible for the problem. Considering all these factors, we have chosen to use microcontroller technology for our control system Programing Language An important part of the design is picking the programing language to use. The programming languages in consideration were Assembly versus a higher level language such as C. We all have experience with controlling a microcontroller with both assembly and C programming. Programing in Assembly can execute the program faster, C is easier to develop and debug. It basically came down to the key advantage, when using assembly language it requires us to spend extra time to understand the architecture of the microprocessor being used. Another consideration we needed to discuss when choosing a language was how complex and large our project is. In theory we will have a few microcontrollers that will each have to be programmed in accordance with the sub-system. The software language that we write in will also be more complex because of the midi file conversion. Since the project is large and intricate C programming offers an easier way to control the microcontroller. Since each of the members of the group have experience with Code Composer Studio and controlling a microcontroller through C programing we have decided to also use this development tool for our project. In conclusion we have decided due to the complexity of our project we will write the software in C language using the Code Composer Studio development tool. For the MIDI conversion application, the C++ language will be used because of its versatility in visual and data structure applications Power Supply When researching power supply, a consideration for a device was its ability to be plugged into the wall. This will require AC to DC conversion of power. When the strings are being pressed down, the pulley system that includes the motors will be off. We will not have to take into account the current draw for these motors at the 19

24 same time as current draw for the solenoids. At most 6 solenoids will be in the on position. If they typically draw up to 1 A each, the fret subsystem will draw roughly around 6 A when all solenoids are activated. When the solenoids are on, the string picking subsystem will be working. There are 6 motors for the string picking system, and if a motor is in use, it will draw around 0.4 A. We expect the entire system for string picking will draw around 2.4 A from the power supply. The control system will draw a negligible amount of current. There may be other parts that may be later added to the design that will draw power. When the solenoids are in the off position, meaning that they are not pressing down on the fret board they will moving up and down each fret. When this occurs the stepper or servo motors at the bottom of the neck are being activated. In the solenoid off position, the maximum about of the stepper or servo motors moving will be roughly 0.4 amps for each motor, so 0.4*12=4.8 amps, for the total fret system. While the solenoids are moving down and up the fret, the guitar may still be strumming. This means that the motors at the string picking system may always be on they will still draw there 2.4 amps from before. The microcontroller will still be drawing the 1 amp, and there also may be miscellaneous current being drawn on the guitar. When the solenoids are in the off position it will roughly be using around 10 A of current. When picking a power supply, it was important to calculate the amps being drawn at each maximum time. It is never a good idea to over draw amps, and therefore it important to have a power supply that will provide at least 12 amps of current to the guitar. The power supply, will also need to plug into a wall. Along with a power supply, it will need voltage regulators that will be connected on the PCB board. There are different types of voltage regulators such as linear regulation, switching regulation, and ferroresonant regulation. Linear regulation uses a transistor that controls voltage and reduces ripple. For low-power situations that demand a clean power, linear regulation is the most recommended in comparison to the others. Switching regulation is best used in situations that require an efficient power conversion. With the power supply, we would want one that converts AC to DC efficiently, through a wall plug. Once the power supply has been chosen the next thing to consider is how the power is going to be distributed. There are many ways to distribute power. Since the power being delivered is filtered DC power from power conductors to circuits, motors, and actuators, a centralized supply is required. Figure shows an illustrated centralized power supply. The figure above is a reference design that was taken from Kepco Power Solutions and redesigned to show the basic use of regulated power converter. This figure is included as reference circuit, and shows how one power regulator can be used to distribute power through multiple loads using a simple resistor. This is also a basic circuit design on how power supply can be centralized. Since the automated guitar 20

25 Figure Centralized Power Supply will be using centralized power, we have included the circuit design purely as a reference for the research of distributing power supply. We wanted to show how a power conversion can work, and how to distribute through a PCB board. Software such as Webench offered from TI is in consideration to be used for the power design. We do not intend to use the specific power design above. However it is included as a reference of how multiple loads can be used by one power converter Serial Communication To get control data to our guitar hardware, we need a form of communication between a computer and our control box. A viable way to do this is to use serial communication. The most common form of serial communication in use today is via a Universal Serial Bus, or USB. USB was created by the computer industry to replace the several other standard serial and parallel communication protocols in place by the early 1990 s. The protocol has undergone several revisions, with backwards compatibility. As an overview of the USB protocol, the layout consists of a host device controller and up to 127 peripheral devices simultaneously on the individual controller. On the Physical layer, the link consists of a 4-wire connection: +5V Power, twisted pair Data+ and Data-, and Ground. Self-powered peripheral devices do not draw power from the 5V line and Bus-powered devices do, up to 500mA. There are three USB Speeds: low speed (1.5 Mbps), full speed (12 Mbps), and high speed (480 Mbps). Transmitters consider a logic 1 as D+ being over 2.8V with a resistor pulled to ground, and D- being less than 0.3V with a resistor pulled to +3.3V. Logic zero is the reverse. Receivers consider D+ 200mV greater than D- as differential 1 and D- 200mV greater than D+ as Differential 0. Above the physical layer, there are four different USB packet types. Token packets are sent by the host to let a peripheral device know it wants to send or request data. Data packets, up to 1kb, are sent by both host and peripheral device. 21

26 Handshake packets include acknowledgements, non-acknowledgements, and stalls. Start of Frame packets are sent in 500ns intervals from host to peripheral. When a USB device connects, it is noticed by the host machine, and the host asks the peripheral for its general information and for drivers to make it usable to the application layer. For our custom hardware to be recognized by a computer as a device, it would be logical to find a part that is pre-configured to interface and be recognized by a computer, and have serial or parallel data out, since we should only need to send our control data from the computer to the guitar hardware box. One such DIP packaged chip is the FT245BM. Figure : Typical FT245R Circuit Configuration It has I/O for the USB, several control lines, and 8 parallel data lines. This could then directly talk to our FPGA/MCU. There are boards pre-made to be used although we could most likely incorporate it into our PCB. 3.3 Strategic Components In the strategic components section, we discuss that parts that we decided to pick for each subsystem and why we decided to pick the certain parts. We also recorded our shopping, the parts that we will buy, and the vendor from which we will purchase them Stepper Motors Our research of potential stepper motors to be procured for the string picking system focused primarily on the overall goal of high speed and reliable position control performance. Specifically, our requirements were that the motors would cause the picks to produce stable note production at a frequency of no less than 10 Hz, which is equivalent 600 beats per minute. 22

27 Torque performance was of secondary consideration, not because of any lesser importance, but because of a wider target value range of 1.42 in-oz to 14.2 in-oz. We did not have prior knowledge of common torque specifications of small stepper motors, so our approach was to ensure the speed requirements were met, then to examine the corresponding torque values of the motor. Rated voltage, current, and power dissipation were important for power supply planning, but were not give the most consideration. From our requirements, we were looking for a low-power motor that consumed less than 10 W of power. The non-performance specifications we considered were length, width, and height dimensions, weight, and cost. Weight and cost were specifications we desired to minimize as much as possible, less than 250 g and less than $20, respectively, but were not considered as having as strict thresholds as speed of performance. Length and width were required to be less than 20mm each, and height was desired to be less than 50 mm. Length and width were considered to be hard requirements, as values greater than spec would prevent the picks to be spaced correctly with the strings. Because of all this, our hierarchy of importance of specifications was as follows, in order of decreasing importance: speed of motor, length and width, torque, electrical specifications of voltage, current, and power, cost, and weight. In researching stepper motor datasheets, it was clear that overall expected speed was not a commonly delineated specification. Instead we were able to use equations involving voltage, current, and coil inductance, specifications that were regularly given for devices, to arrive at expected speed behavior. The time it takes for the stepper motor to complete one step is the time it takes to build up the magnetic field in the coils to take the step, plus the time it takes to return the field back to its neutral state, to complete the step cycle. This time equation is shown below, as t 2[2IL /V ] 4IL /V where t is the step time in milliseconds, I is the maximum rated current in Amperes, L is the coil inductance is millihenries, and V is the maximum rated voltage in Volts. So, this equation determines the amount of time one step will take. From this value above the maximum speed in revolutions per minute (rpm) is given by RPM 6*10 4 /(SPR * t) where SPR is steps per revolution, a given specification for each stepper motor, and t is the time per step, determined above. The division is out of 60,000 because there are 60,000 ms per minute. Bearing in mind that our design requirement for the stepper motor is 200 rpm, we examine the specifications of several stepper motors and their corresponding speeds below, in Table

28 Vendor PN Voltage (V) Current (A) L (mh) Steps/rev Final RPM Omega OMHT Omega OMHT Sparkfun SM-42BYG Omega OM Wantai 28BYGH Circuit Specialists 28BYG Wantai 39BYG Circuit Specialists 28BYG Pololu/SOYO SY20STH A Adafruit N/A Table : Stepper Motor Component Comparison From the table it is evident that most of the stepper motors researched did not meet our speed requirement of 200 rpm. Each of the three stepper motors whose speed met our requirements were bipolar stepper motors with a step resolution of 200 steps per revolution. Their unique specifications we will examine more closely. The first stepper motor we considered was Adafruit Product ID No It is a 12V NEMA 17 size, square-bodied stepper motor. It had the highest speed of nearly 600 rpm, which would be more than adequate for our design. Its rated current was fairly low at 350 ma, which is good for keeping power down. An estimate of maximum power dissipation of 12V * 0,35A = 4.2W. Its holding torque was high, at 200 mn*m, which was higher than the majority of the components we considered. That value of torque exceeds what we predict will be necessary to cause the pick to pluck the string without disturbing the expected motion of the motor.the cost fell within our desired range, below $20, at $14. This cost was the third-lowest of the 10 motors we considered. This potential savings is a benefit to keeping our budget on target. The one specification of importance that was not met was the width of the motor. We desired a motor of no more than 20 mm wide. This stepper motor was 42 mm 24

29 wide, more than twice our specification. To accommodate this, we would have to create several separate, staggered structures to hold these motors without them overlapping each other. Other than causing a redesign, this may not even be feasible, considering the length limitations of the body area. However, if this problem could be mitigated, the Adafruit stepper motor would be a prime candidate. In Table are the specifications of the Adafruit motor. Specifications Specifications Step angle ( ) 1.8 Rated Voltage (V) 12 Temperature Rise ( C) 70 MAX Rated Current (A) 0.35 Ambient Temperature ( C) 0 ~ +50 Resistance per phase (Ω) 34 Number of Phases 2 Inductance per phase 4.3 (mh) Insulation Resistance (MΩ) 100 Min (500VDC) Holding Torque (mn*m) 200 Insulation Class Class E Detent Torque (mn*m) 11.8 Length*Width*Height (mm) 42*42*34 Rotor Inertia (g.cm 2 ) 38 Shaft length 24 Weight (g) 200 Table : Adafruit Motor Specifications The second motor we considered was 28BYG201, sold by Circuit Specialists. It is a 4.2V NEMA 11 size, square bodied stepper motor. The estimated speed was lower than the Adafruit stepper motor, at 220 rpm. This still meets our specifications, but does give cause for concern, because we want to build the largest margin as we can. The rated voltage was much lower than that of the Adafruit motor, but the current was much higher, at 950 ma. Ideally, our design would not have to deliver that high of a current, so that is a specification to monitor. The estimated maximum power dissipation was 4.18V * 0.95A = 3.971W, marginally less than for the other motor, but still less. Its holding torque was 54 mn*m, a fourth of that of the Adafruit motor, but still meeting our estimation of what will be required. The cost of the 28BYG201 was comparable to that of the Adafruit motor, at $13.95, again providing potential savings. A specification of note that may not affect the overall design but is different is the use of 6 wires in this configuration, rather than the 4 wires used for the Adafruit motor. This would require a different design of the motor driver circuit, and may potentially increase the number of required output pins from the control system. Additionally, like the Adafruit motor, yet not as far off, was the width of the motor, which came in at 28.2 mm. Again, this is 14 mm closer to our requirement than the Adafruit motor, yet still 8 mm over the required 20 mm. However, the use of this motor would not put as much of a burden of design change on the system. That is, the use of this motor would require three staggered rows of two motors each, rather than the expected two rows of three. We do not consider this to be ideal, but we believe it is within the space constraints of the guitar body to implement this, if needed. In Table are the specifications of the Circuit Specialists motor. 25

30 Specifications Specifications Step angle ( ) 1.8 Rated Voltage (V) 4.18 Temperature Rise ( C) 80 MAX Rated Current (A) 0.95 Ambient Temperature ( C) -20 ~ +50 Resistance per phase (Ω) 4.4 Number of Phases 2 Inductance per phase (mh) 1.5 Insulation Resistance (MΩ) 100 Min (500VDC) Holding Torque (mn*m) 54 Insulation Class N/A Detent Torque (mn*m) N/A Length*Width*Height (mm) 28.2*28.2*30 Rotor Inertia (g.cm 2 ) N/A Shaft length 10 Weight (g) N/A Table : Circuit Specialists Motor Specifications The final stepper motor we considered was the Changzhou Songyang Machinery & Electronics S20STH A stepper motor, sold by Pololu. It is a 3.9V NEMA 8 size square-bodied stepper motor. Its estimated speed was between the speeds of the other two, at rpm, giving more margin than the Circuit Specialists motor. The rated voltage was slightly lower than that of the second motor, but the rated current was significantly less, at 600 ma. The estimated maximum power dissipation was 3.9V * 0.6A = 2.34W, nearly half that of the other two, which is very good. The most remarkable feature of this motor was its dimensions. Its width was 20 mm, exactly at the top of our requirements. This is a major advantage over the other two motors, as it would allow us to use our desired two-rows-of-three picking system, rather than requiring a redesign. This motor was unique in that, of the 10 motors we researched, it was the only one that met our width specifications. Other NEMA 8 (0.8 inch, or 20 mm) motors were available, but as far as we researched, we were not able to find any others that could meet our speed performance requirement. There are two potentially significant drawbacks of this stepper motor, cost and torque. The cost of this motor was $17.95, four dollars more than the other two, or $24 more for the six that we need. While this is disadvantageous, if this motor meets our design needs in ways the others cannot, the extra cost would be a reasonable expense. The second drawback is more significant. The holding torque of this motor was 17.7 mn*m, a third of the Circuit Specialists motor, and a tenth of the Adafruit motor. This torque difference is significant, but if the torque performance of this motor were able to meet our requirement of string picking, it would not be an issue. However, testing would likely be required to determine not only whether the motor could produce enough torque to cause the pick to rotate, but whether its timing would be hindered by the impulsive load of displacing the string. The presence of any actual timing delay would be difficult to predict and would require testing. In Table we have listed the specifications of the Pololu motor. 26

31 Specifications Specifications Step angle ( ) 1.8 Rated Voltage (V) 3.9 Temperature Rise ( C) 80 MAX Rated Current (A) 0.6 Ambient Temperature ( C) -20 ~ +50 Resistance per phase (Ω) 6.5 Number of Phases 2 Inductance per phase (mh) 1.7 Insulation Resistance (MΩ) 100 Min (500VDC) Holding Torque (mn*m) 17.7 Insulation Class Class B Detent Torque (mn*m) N/A Length*Width*Height (mm) 20.2*20.2*30 Rotor Inertia (g.cm 2 ) 2 Shaft length 15 Weight (g) 60 Table : Pololu Motor Specifications After considering the three motors, we have decided to use the Pololu motor for our string picking subsystem design. We intend to fit the motors in two rows of three, so the width meeting specifications is essential. We found that the motor speed of rpm would be more than adequate. Also, not mentioned before because we did not have the weight data of all three stepper motors. The light weight of the motor (60g versus 200g of the Adafruit motor) will serve to ease the motor requirements of the dynamic control system, at a total load savings of 140g * 6 = 840g, which is significant to ease the burden on dynamic control displacement speed. As mentioned before, the low power dissipation of the Pololu motor will also provide load savings for the power supply system of 11.2 W. Our concern with the low torque performance of the Pololu motor is significant. We plan to determine during testing the viability of the torque performance of this motor. If we determine that the torque performance of this motor does not meet our needs, our second choice would be to use the Circuit Specialists motor, which would cause us to reconfigure our motor apparatus and dynamic control system, but would improve the torque performance to well above adequate to meet the requirements of the string picking subsystem Servo Motors There are several requirements to be bet in regards to our servo motors being used to move the solenoid assembly. The total distance that the belt needs to travel is the distance is from the bottom E string to the top E string, which, at the first fret of the guitar, is 38 millimeters and at the twelfth fret is 43 millimeters. The maximum size allotted to the servo stem attachment is 9 millimeters or less in radius, although this would change depending on the actual dimensions of the chosen servo. To rotate from one side to the other would require (180*28)/(2*pi*9) degrees of rotation, or from the center position, degrees of rotation. If however the radius requirement of the servo fixture were able to be larger, then the required maximum rotation angle would decrease. There are servo motors with various degree limits of rotation as well as continuous rotation servos. Using a servo that can continuously rotate would circumvent that requirement. 27

32 Another requirement of the servo is to have the ability to go travel from one outlying string to the other outlaying string in a given amount of time. We have chosen to have a maximum of ten notes being played per second. It is unlikely that we would be required to play two notes on the same fret on opposite ends of the fret board with only one tenth of a second between them, but for posterity, it would require an angular velocity, assuming a radius of 9 millimeters, of 240*10 = 2400 degrees per second. This is likely beyond the scope of low cost or small size servos, which are both constraints as we need 12 servos. A more realistic goal is to take a second to travel to the farthest string, which requires 240 degrees per second. This is not taking into account the time required to de-activate the solenoid, then move, then re-activate the solenoid. A final requirement of the servo motor is to fit into a tight space to allow 12 staggered servos, 2 parallel rows of 6 series servos. Based on the model simulating possible dimensions requirements, we allotted 20x20 millimeter square face with a 25 millimeter length. This is not the typical servo shape, but the dimensions can be comparable. An ideal design of our system would have the solenoid assembly slide freely from side to side, with rollers for the belts to have the belt not bend too tightly. It could be safe to neglect how much torque is required to spin the belt, as our objective is to make the solenoid enclosure and sliding system virtually frictionless. Regardless, a discussion on torque requirements will follow. Power consumption is also not an issue, as typical servo power draw is below 10 1 Watts. It would be unlikely to operate all 12 servos simultaneously. If necessary, a case as such could be programmed out in the exceptions module of code. Taking all of these factors into consideration, Table shows the potential candidate servo motors and their specifications. Every servo with the exception of the Adafruit 154 motor had a maximum rotation of 180 degrees. Vendor PN Voltage (V) Max Angular Velocity Torque (kgcm) Dimensions (mm) Price (ea) Adafruit Adafruit RPM x20.0x Hitec Hitec Hitec 31311S HS S HS S HS-322HD sec/60 Deg x19.8x sec/60 deg N/A 22.6x11.4x sec/60 deg x19.8x

33 Vendor PN Voltage (V) Max Angular Velocity Torque (kgcm) Dimensions (mm) Price (ea) Hitec 33485S HS-485HB sec/60 deg x19.8x SMAKN MG996R 6 Tower Pro (4 pc) MG90S 6 Traxxas sec/60 deg x18.0x sec/60 Deg x32.0x sec/60 deg x20.5x Table : Servo Motor Component Comparison To estimate what kind of torque may be required to move the solenoid assembly, we can use F=M*A. The typical weight of the solenoids we are potentially going to use are between 25 and 150 grams. The casing holding the solenoid could be guessed to be 100 grams, and being liberal would weigh in the belts at 100 grams. The rough estimate rounds in around 300 grams. So, if the radius of our lever arm, i.e. the belt drive is 1 cm, then with servos in the maximum torque range of 1-3 kgcm, F/A = 1-3 kg. Acceleration is then the factor of performance in the servo of choice. It would seem possible for every servo to be able to get the solenoid assembly to move. However, we may prefer one that can perform quicker than others. This parameter, as shown in table 3.3.2, is given in seconds per 60 degrees, or if the servo is continuous in rotation, is given in revolutions per minute. The lower the number, the faster the servo can rotate. As stated before, we wish to be able to travel 43 millimeters in less than or equal to one second. The majority of servo motors we found are 180 degree range of motion, with the slowest one in theory being able to rotate 60 degrees in 1/5 of a second. With a range of 180 degrees, that means that it can travel from stop to stop in 3/5 of a second, within our requirement. To circumvent the size constraints placed on the servo motors due to proximity to each other and the guitar, the orientation that they are fastened to the chassis in can change depending on necessity. Also, the largest dimension the servo motors can be reduced as they are typically molded plastic fastening holes, two on each end. These would be unnecessary to our application and furthermore getting in the way. Hence, paring down our possible servos to use based on size, except maybe for being too small, would be illogical. A more likely factor in paring the possible choices down to a few candidates will most like likely be cost. The two high torque servos we found, the Traxxas 2055 and the Hitec 34485S are both very high in performance, but have unit costs of 29

34 over 17 and 19 dollars respectively. With a total of 12 servos needed to reach our goal, that price takes the cost of servos up to more than one quarter of our total budget. While it may be fiscally feasible, it does not seem fiscally responsible if it were unnecessary. Therefore, we shall, at least to the point of initial testing, narrow the field of servo motor choices to the small to medium, cost-effective options, and upon testing, if necessary, re-examine the more powerful options that we have found. To further reduce the list of possible servo choices, the Hitec 31311S HS-311 and Hitec 33322S HS-322HD both have identical torque, size, and speed, however the HS-322HD is four dollars more expensive. Furthermore, it is not necessary to have continuous rotation. Hence, there would be no reason to use the Adafruit Volt Servo as it is also more expensive, while having the same size specifications as cheaper servo motor choices. The most cost-effective option is the Tower Pro MG90S. It can bought individually for 8.23 on Amazon.com or in packs of four for on Amazon. The dimensions of the MG90S are 23.1 mm long, 12.2 mm wide, and 29.0 mm high, shown in the Figure below. Figure : Tower Pro MG90S Dimensions The pulse width is microseconds, with a pulse cycle of 20 milliseconds. This servo would likely be the most cost-effective, as we would require 3 packs of 4 to fulfill our requirement, which would cost a total of around 51 Dollars. If the maximum torque, of which is 2.2 kg-cm, is sufficient for our task, this servo would be the best candidate. A sample servo has been purchased and will be tested with to see if it will be sufficient in torque. One caveat of the MG90S is that the required pulse width modulation signal voltage level is 4.8 to 6.0 volts. This is greater than 30

35 the voltage of the GPIO pins on our microcontroller. Hence, to use this servo, we will need a driver circuit. If, however, this servo is insufficient in output torque, there is a moderately more powerful option to compare it to. The Hitec 31311S HS-311 has a max torque output of 3.7 kg-cm, which is nearly twice as high as the Tower Pro servo. However, it is 39.9 centimeters long by 19.8 centimeters wide by 36.3 high. This is quite larger than the Tower Pro servo. The unit cost is 7.99 dollars, more than twice as much as the Tower Pro servo. The pulse cycle of the Hitec is 20 milliseconds and the Pulse Width is microseconds. The Pulse Width Neutral is 1.5 milliseconds. There is more data available on this servo than the Tower Pro servo which makes it a much safer candidate to use in our project. Table shows the Hitec HS311 specifications, and Figure shows the Hitec HS311 dimensions. Control System Pulse Width Neutral Required Pulse Operating Voltage Operating Speed(4.8V) Operating Speed(6.0V) Current Drain (4.8V) Current Drain (6.0V) Stall Torque (4.8V) Stall Torque (6.0V) Dead Bandwidth Operating Angle Direction Motor Type Potentiometer Drive Gear Type Weight HS311 Pulse Width Control 1500 microseconds 3-5 Volt Peak to Peak Square Wave Volts 0.19sec/60 degrees at no load 0.15sec/60 degrees at no load 7.4mA/idle, 160mA no load operating 7.7mA/idle, 160mA no load operating 3.0 kg/cm 3.7 kg/cm 5 microseconds 45 degrees one side pulse travelling 450 microseconds Multidirectional Cored metal brush 4 Slider/Direct Drive Nylon 43 grams Table : Hitec HS311 Specifications The operating current of the HS311 is 160 milliamps without a load. This current drawn will likely be higher due to non-ideal forces but would likely not deviate too far. As there is so much data on this servo, it is a truly better choice than the cheaper less powerful servo. 31

36 Figure Hitec HS311 Dimensions In the event that neither of the two servos above are powerful enough to move the solenoid fixture, There is the third option of the Hitec 33485S HS-485HB, which on the expensive side, but would deliver big performance. Table shows the specifications of the Hitec HS-485HB and Figure shows its dimensions. Control System Pulse Width Neutral Required Pulse Operating Voltage Operating Speed(4.8V) Operating Speed(6.0V) Current Drain (4.8V) Current Drain (6.0V) Stall Torque (4.8V) Stall Torque (6.0V) Dead Bandwidth Operating Angle Direction Motor Type Potentiometer Drive Gear Type Weight HS-485HB Pulse Width Control 1500 microseconds 3-5 Volt Peak to peak Square Wave Volts 0.22sec/60 degrees at no load 0.18sec/60 degrees at no load 8mA/idle and 150mA no load operating 8.8mA/idle and 180mA no load operating 5.18 kg/cm 6.41 kg/cm 8 microseconds 45 Degrees one side pulse travelling 400 microseconds Clockwise/Pulse Travelling 1500 to 1900 microseconds 3 Pole Ferrite Motor Indirect Drive Karbonite Gears 45 grams Table : Hitec HS-485HB Specifications 32

37 Figure : Hitec HS-485HB Dimensions Another useful feature is that this servo is continuous rotation modifiable so that if required, it would be possible to implement it. As stated before, upon testing we can verify or reconsider our servo choice but at this stage in the design process it is logical to choose the Tower Pro MG90S due to the size and low cost, with performance being determined in prototype testing. With this in mind, here is a reference schematic use to implement the MG90S. This schematic is a generic reference with typical values shown, and is not a specific servo. This drawing was creating using AutoCAD Electrical Figure shows the reference design of the servo motor driver circuit. Figure : Servo Motor Driver Circuit Schematic 33

38 The voltage source value is a range of typical microcontroller power requirements, and the V+ source is a range of typical allowed values for most of the servos we found while shopping Driver Belts To translate a rotational motion of a servo motor into a lateral translation used to move each of the twelve solenoids back and forth, we have decided on using a toothed belt, also known as a timing belt in some applications, most commonly found in automotive mechanical systems. It is important to consider several factors when picking a belt, including tension properties, how tightly it can turn a corner, and load capacity. There are several belt choices, ranging in size and material. To narrow down the possible parts before building a mockup, we used the SketchUp model of the fret board apparatus to give us a good rough estimate as to the required belt length. Shown below in Figure in red is how we would use the belts. Figure : Driver Belt Usage The total measured length on the model was 28 centimeters or roughly 11 inches. The belt width also comes into play. We need to pack the servos and belts relatively tightly together, so if the belt doesn t need to be very wide, it would be more cost efficient to find a belt that is sufficient and not over-engineered. However, we wouldn t want the belts to break from overloading. The width measurement in the simulation allows for a maximum of 8 millimeters, or roughly 1/3 of an inch. It would seem a waste of space to exceed this width. On McMaster.com we found many choices for belts, and have ordered two candidates for testing. Upon prototype testing with these two belts and pending other options we will determine if the lower-budget 1/8 th inch belt will be sufficient in performing the required task. Table shows the driver belt options. 34

39 P/N Description Belt Belt Pitch Price (ea) Length Width 1679K634 Trapezoidal $1.39 Tooth Urethane 1679K24 Trapezoidal Tooth Urethane $2.78 Table : Driver Belt Component Comparison There is, however, an easier alternative to belts, and would be virtually costless and work just as well. This would be to use string in place of a belt. With tension, it could be tied to the windmill or arm attachment of the servo motor and tied onto the solenoid fixture as well. The limitation is on the distance that the solenoid fixture could actually be moved. This is because the windmills attached could be too long to rotate the full 180 degrees if the servos are mounted tightly to the frame. This will be considered in prototyping, however does not require any further research or detailing Solenoids We have determined that in the implementation of depressing strings at frets as a human finger would, linear solenoids would achieve the desired performance. As current is passed through the solenoid, magnetic force is induced that causing a metal shaft to be displaced. The requirements we most considered when comparing solenoids were length, width, and height dimensions, weight, cost, force applied to the shaft, force versus current applied, force versus shaft displacement, coil resistance, and voltage, current, and power ratings. Of primary importance were the length and width of the solenoid. It was desired that the length and width of the solenoid, which are generally equal or close to equal, were both less than 20 mm. This requirement was chosen because of the fret-to-fret width at the 12 th fret, the highest fret location we planned to fit with a solenoid for string depression, was slightly larger than 20 mm. The choice of 20 mm had margin built in, as we project that there may be intermediate structure between the solenoids. As values for these dimensions beyond our requirements would lead to an invalid design, this was the first consideration in solenoid specifications. Of next highest importance was the force applied to the shaft. In initial experimentation it was determined that the force required to fully depress a string on an electric guitar was less than one pound of force (454 gram-force, or gf, units). The minimum required force to depress a string so that characteristic guitar sound could be produced, was estimated to be around 50 gf. Wanting to give plenty of margin on this specification because of the perceived variability of the quality of repeated tones produced from borderline force values, we set the desired force at 200 gf. A small solenoid with length and width below 20 mm each that produced a 200 gf force at the string would fit our first two requirements. 35

40 Coupled with the setting of the force requirement was the understanding that solenoid force output varies inversely with displacement of the shaft, we were compelled to estimate the expected distance the shaft would be required to travel. The combination of the maximum prescribed action, or distance between the neck and a string that is not depressed, for a string (2.38 mm), plus a margin to avoid unwanted damping of the string vibration (1 mm), was 3.38 mm. Removed from this displacement was the projected height of a rubber cap adhered to the end of the solenoid shaft to aid in reliable string depression and characteristic guitar tone, which was estimated to be at least 1 mm. A final expected solenoid shaft displacement was set at 2 mm. That is, the desired force of 200 gf should be achievable at a displacement of 2 mm. Weight was not of great concern for the performance of the solenoid. However, as the solenoid would be a load on the servo motor that was moving it, and considering that the greater load would reduce the maximum speed of the rotation of the servo motor, it was important for the system s overall performance to minimize solenoid weight. The actual effect of the weight of the solenoid on the maximum speed of servo motor rotation, all other weights involved being equal, is difficult to predict accurately. However, we project that the motor speed loss due to solenoid weight would be minimal if the solenoid weighed less than 50 g. As we researched different models of solenoids, we kept that in mind as a loose requirement. Next most significant after the four above specifications was power dissipation. As in many other applications, we desired that this value be kept as low as possible while achieving satisfactory performance. Our desired power range was less than 5 W for each solenoid, and beyond that threshold, the lower the better. Related to this is the coil resistance. As will be seen later, a majority of the solenoids we researched were 12V solenoids. Supposing that two of these 12V solenoids delivered the desired performance, the solenoid with the higher coil resistance would allow less current, and, again considering the voltage values were equal, lower power dissipation. In general, since our design calls for 12 solenoids to be used, we desired to keep the per-unit cost below $10. However, this requirement was not given first consideration in comparing solenoid models. That is, we were generally unwilling to compromise on the solenoid performance for lower cost, since the solenoids reliable behavior is that which characteristic note production, and therefore the entirety of our design, depends. A number of types of solenoids exist, including linear and rotary, push, pull, and push-pull, and tubular, C-frame, and D-frame. Our design requires linear motion and thus a linear solenoid. In addition, we expect the solenoids to push down on the strings and release when current is not applied. We do not need it to perform any pulling actions, so a push solenoid is adequate. On structural criteria, the most advantageous choice is the one that will lend itself best to having a drive belt, or an intermediate structure connecting the belt to the solenoid, fastened to it. Our initial assumption is that the open frame (C- and D- frame) solenoids will be more beneficial for this, as they come with a number of 36

41 mounting holes for fastening, whereas the tubular solenoids have a threaded section just before the shaft for fastening. The use of the mounting holes would likely be easier than the creation or purchase of a component that fits the threads of a tubular solenoid. # Vendor PN Structure L x W, (mm) Body Ht., (mm) Body/Shaft 1 Sparkfun ZHO-0420S-05A4.5 D frame 12 x 11 20/6 2 Sparkfun ZHO-1364S-36A13 D frame 30 x /25 3 Jameco SMT-1632S12A-R Tubular 15.2 x /6.1 4 Jameco SMO-0420S12STD C frame 12.7 x / Jameco SMO-0837S12STD-R D frame 25.4 x /6.1 6 Jameco SMT-1325S12A-R Tubular 12.7 x / LEDEX Tubular 12.7 x / MULTICOMP MCSMO-0630S12STD D frame 19.2 x /12 9 MULTICOMP MCSMT-1325S12STD Tubular 12.7 x / MULTICOMP MCSMO-0630S06STD D frame 19.2 x /12 # Vendor Volt. (V) Res. ( ) Power (W) at 100% Max force (12V); at 2mm (12V) (gf) Wt. (g) Cost ($) 1 Sparkfun ; 140 (25% duty) Sparkfun ; 2200 (36V) N/A Jameco ; Jameco ; Jameco ; Jameco ; LEDEX N/A; 199 (2.54 mm, 27V, 25% duty) MULTICOMP ; MULTICOMP ; MULTICOMP ; 375 (25% duty) Table : Solenoid Component Comparison 37

42 Table shows the specifications of the solenoids we considered. All are linear, push solenoids. The items have been numbered for ease of reference. Only one solenoid we researched was of C frame type. Most were D frame or tubular. The length and width of solenoid #2 eliminated it from further consideration; it was simply too large for our design. The remaining solenoids fell within our length and width requirements with the exception of solenoid #5, which is 0.3 mm over specification. The others were either well within specification (#1, 3, 4, 6, 7, 9) or just within margin (#8 and 10). Though height was not of high priority, a very tall solenoid like #3 would require more design work to incorporate in the string selection subsystem. Most of the solenoids we researched had a rated voltage value of 12V. We are cautious to limit our design from incorporating relatively high voltages (> 20V), so if we chose a 12V rated solenoid, would most likely run it at, or close to 12V and 100% duty cycle. However, for solenoids #1 and 10, which are rated at 6V, we would be more willing to run at higher voltages and lower duty cycles. This provides a potential force performance increase for these solenoids with respect to the others. This consideration was expressed in the maximum force column, in which nearly all solenoids were assumed to be excited by a 12V source. Solenoids #5 and 10 had the highest force value at the expected displacement of 2 mm, among solenoids that had not been eliminated. The very low (< 100 gf) force performance at 2 mm displacement for solenoids #4, #6, and #9 made them unlikely alternatives for our design. In addition, though solenoid #7 had a force performance of nearly 200 gf beyond 2 mm, that value was attained by running it at 25% duty cycle, supplying 27V. This force value at this voltage was used since the datasheet had insufficient force data. Since we do not plan to use voltages that high, and since the solenoid s force response to an input of 12V would likely be much lower, solenoid #7 would not likely be considered further. As mentioned earlier in this section, the coil resistance of similarly voltage rated solenoids determined the power dissipation of each. The very high coil resistance of solenoid #4 drove down its power dissipation significantly. The estimated power dissipation of each was calculated from P = V 2 /R and was considered at 100% duty cycle. Those solenoids running at 25% duty cycle would have four times the power dissipation during the operating periods of the duty cycle. That is, solenoids #1 and 10 would be drawing 4.8W and 9.6W, respectively, during that time. Of note is that with the exception of #2 and 3, the power dissipation of all solenoids operating at 100% duty cycle does not exceed 4W. Weight is an important consideration as it directly burdens the performance of the motor for the string selection system. These solenoids are all low-weight, and with the exception of #2 and #5, all meet our loose requirement of 50 g. Solenoid #2 has already been eliminated, but the failure of #5 to meet the weight requirement decreases the likelihood of choosing it. All of the solenoids were priced at less than $20. The midpoint of the cost of the solenoids was around $12, with a high of nearly $20. Three of the solenoids, #1, #5, and #10, met our specification on cost. As so few of these solenoids and of the 38

43 solenoids we browsed through actually met our specification, we were willing to give some of the solenoids that exceeded that value, some additional consideration. The final four solenoids we considered were solenoids #1, 3, 5, and 10. These solenoids had diverse specifications, yet each delivered at least 140 gf force at the desired displacement. Solenoid #1 is a 6V, D frame solenoid. Itwas the lightest component of those we considered, at 13 g, 30 g lighter than the other final four components. At 13 g, this solenoid would put very little load on the string selection motor, easing its design concerns. Additionally, its length and width were the least of any component we researched, at 12 and 11 mm, respectively. If we chose this solenoid, it would satisfy our space concerns and add plenty of margin for potential additional structural components of the fret system. Lastly, was the least expensive of the last four considered at $4.95, a little more than half the cost of the next cheapest component. This would be a great savings if we felt the solenoid could meet the desired performance. Its rated voltage, current, and power dissipation were all low, at 6V, 0.2A, and 1.2W, and would not burden the power or solenoid driver systems. One concern with this and the other 6V solenoid is the application of above-rated voltage. Applying 12V is not extreme, but the maximum recommended amount of time to run these at 25% duty cycle is around 18 seconds, which would seem to suggest that and 18 second use would require a 54 second recovery time, which would not be acceptable for our application. We do not expect to need 18 seconds of continuous performance, as typically long musical notes last between 1-5 seconds, and the from the natural decay of the guitar, the sound produced does not last far beyond that. However, it is not known how the required recovery time varies for different operation periods, whether proportionally or otherwise. Nevertheless, the selection of this or solenoid #10 carries with it the risk of loss of availability at the expected above-rated voltage use. The main drawback of this component was the force performance, the most important specification under consideration. As the force requirement itself is an estimation of the necessary value to adequately depress the string, it is possible that it is an overestimation and that force delivered by solenoid #1 may also be satisfactory. Even so, the risk associated with the loss of availability mentioned above, combined with the risk associated with lower than required force performance do give pause in the selection of this solenoid. Solenoid #10, as mentioned above is also a 6V, D frame solenoid. It has the same above-rated voltage use issue of solenoid #1, but one which raises more concerns. Its coil resistance is half of that of solenoid #1, so its normal rated current and power are 0.4A and 2.4W. When a 12V source is applied, the current and power increase to 0.8A and 9.6W. The power dissipation is above desired values for the system, and the current does not violate a requirement, but requirement of higher current puts higher demands on the solenoid driver circuit. 39

44 Compared to solenoid #1, the length and width (19.2 x 16 mm) and weight (42 g) of solenoid #10 are much closer to the requirement threshold, leaving much less margin, but are adequate and would likely not cause any design problems. Though there is no requirement for height, our perspective is that less is better, and the height of solenoid #10 is 13 mm greater than solenoid #1, but 29 mm less than the longest. The cost, at $11.09, is also in the middle of the extremes. None, with the exception of cost slightly exceeding specification, of these parameters are drawbacks, but none are remarkable. While solenoid #1 has the lowest force performance value, solenoid #10 has the highest, or 375 gf while running at 25% duty cycle, This value is nearly twice our required force, so it gives significant margin and should leave no doubt about the reliability of the string depression. In addition, if this force is be too much, causing undesired noise from striking the string too hard, applying a voltage closer to 6V should alleviate any noise problems while resulting in lower current and power draw. While the force performance of solenoid #10 is more than adequate, the trade-off of non-ideal length, width, height, weight, and cost, and excessive power and current draw may be too much to warrant selecting this solenoid. Solenoid #5 is a 12V, D frame solenoid. Its force performance value was lower than that of solenoid #10, at 300 gf, yet exceeds the desired force as well with plenty of margin. Whereas solenoid #10 required 0.8A of current and 9.6W of power to achieve 375 gf of force, solenoid #5 would only require current of 0.333A and power of 4W. This difference is a significant power system and driver design savings. Also, these values could be achieved while running continuously, which would remove any lack of availability concerns. The height of solenoid #5 was comparable to solenoid #10, at 44.2 mm, compared to 41.5 mm, a negligible difference. A final advantage over solenoid #10 is cost. The satisfactory force performance and lower power cost are available for $8.95, two dollars less per component. While some of these specifications do not compare to the desirable levels of solenoid #1, the reality that solenoid #5 exceeds force requirements, whereas solenoid #1 does not, is significant. The main concerns about solenoid #5 have to do with its physical dimensions. With a length and width of 25.4 x 20.3, the solenoid would have to be positioned so that its width ran parallel to the neck of the guitar; its length would be well beyond the value needed to be feasible in the design. Even when positioning the solenoid as such, the width of 20.3 is 0.3 mm beyond specification and would exceed the width between the last two frets in our design by about 1 mm. A potential alleviation of this problem is allowing positioning the solenoid further down the neck so as to overlap the next fret. This modification would be possible since there would be no other solenoid to interfere with on that side. Taking that modification into consideration would allow us to compare the 20.3 solenoid width with the distance between the next two frets, second and third to the end, the width of which is slightly larger. If the solenoid is similarly too wide, it is possible to attempt to space out the last few solenoids, to ensure they can all fit. This is not an ideal design scenario but would allow solenoid #5 to be used. 40

45 An additional concern is with the weight of the solenoid, which is 95.3 g, more than twice the weight of the next solenoid, and more than seven times the weight of solenoid #1. It is not yet known exactly how much detriment this difference would be to the performance of the string selection system, but it is reasonable to suspect the effect would be negative. Ultimately, testing would determine whether the trade-off of string selection motor performance would be significant enough to determine whether or not using solenoid #5 would lead to a viable design. Solenoid #3 is a 12V, tubular solenoid. Its weight (45.3 g) and length and width (15.2 x 15.2 mm) are significantly better than that of solenoid #5 and comparable to that of solenoid #10. However, these values are not as much as an asset as they are with solenoid #1. In regard to electrical characteristics, the coil resistance is lower than with solenoid #5, so the rated current (0.57A) and power (7W) are not as desirable. However, the rated voltage is 12V rather than 6V, so those values would not increase further as with solenoid #10, in addition to the avoidance of any lack of availability problems. As with solenoids #5 and 10, solenoid #3 does deliver adequate force (275 gf), with a margin of 75 gf over specification. Whereas the other three solenoids were all D frame solenoids, solenoid #3 is a tubular solenoid, which we project would yield more design problems with connecting to the string selection system than the others. As we intend to move the solenoid by a belt driven by a servo motor, the ends of the belt would need to be fastened to the solenoid or an intermediate structure. All of the D frame solenoids come with mounting holes, which the belt could loop through or which would allow easy installation of another structure that would capture the end of the belt. Since the tubular solenoid can only be fastened by screwing in a part onto threaded structure, the design options are reduced. However, though this situation is less desirable, it does not disqualify solenoid #3 from consideration. Also inherent in the nature of a tubular solenoid is the diameter of the shaft, which is especially narrow. We have not made any requirement on shaft diameter, but we do expect to adhere an end piece made of rubber or similar material, that will serve to reduce noise and ensure a good hold on the string when depressed. Though it is feasible to attach an end cap to a narrower shaft, the process of doing so may not be as simple. Of the last two specifications of solenoid #3 under consideration, one is non-ideal and the other is problematic. The full height of the solenoid is 89.4 mm, which, for better visualization, is nearly 4 inches and is over three times the height of solenoid #1. As before, we maintain that we do not have a solenoid height requirement other than the perspective that we would prefer it to be short, all other things being equal. The cost of the solenoid, though, would certain be well beyond budget, and may be beyond our means or desire to finance. For 12 solenoids, the cost before shipping would be $203.40, nearly a third of our initial total budget estimate. For the cost, we would expect all or most of the other specifications to be exemplary, which is not the case. Table summarizes the comparison between the four solenoids. For each specification, the solenoids are given a qualitative rating on a 4-point scale. 41

46 Additionally, the actual specification values are given on the second row for each specification. The values we have given in our requirements section used to determine the pass/fail status of each, where green denotes a pass and red denotes a fail. As no one solenoid passes all of our requirements, we chose to tally up the 4-point scale values for each solenoid and use those to help make the final decision on a solenoid. Solenoid #1 #3 #5 #10 Force 1 - Poor 3 - Good 3 - Good 3 Good (Spec: > 200 gf) Length x Width 4 - Excellent 3 - Good 1 - Poor 2 Fair (Spec: < 20 mm) 12 x x x x 16 Weight 4 - Excellent 2 - Fair 1 - Poor 2 Fair (Spec: < 50 g) Cost 4 - Excellent 1 - Poor 3 - Good 2 Fair (Spec: < $10) Power (12V) 4 - Excellent 2 - Fair 3 - Good 1 Poor (Spec: < 5W) Current (12V) 3 Good 2 - Fair 3 - Good 1 Poor (Spec: N/A ma) Height 4 - Excellent 1 Poor 2 - Fair 3 Good (Spec: N/A mm) Duty Cycle (12V) 2 Fair 3 Good 3 - Good 2 Fair (Desired: 100%) 25% 100% 100% 25% Type 3 Good 2 Fair 3 - Good 3 Good (Desired: D fr.) D frame Tubular D frame D frame Total Points Table : Final Solenoid Component Comparison As is clear from the last row of Table , solenoid #1 rated the highest on our 4-point system, with solenoid #5 next. This confirms the impression we had, that we should choose solenoid #1. The two areas of concern are the low force output and the low duty cycle required to achieve it. If during our testing we find that these conditions are satisfactory for our design, we will continue to use solenoid #1. If, however, during testing we find that the force performance or delay due to duty cycle is unacceptable, our alternate plan is to use solenoid #5. We select this as our backup component with the understanding that some design rework Microcontroller Having chosen to use microcontroller rather than FPGA technology for our control system, before considering any specific attributes we would desire, we took a survey of the microcontrollers used in Spring 2014 Senior Design projects. We did this to get a sense of what has been chosen previously and what companies and 42

47 models might be good to use as starting points in our search. Table shows our results. From the results in the table, it is clear that TI and Atmel were companies whose products were chosen most often. From this fact, we decided that, barring failure to find any products from either company that met our design and general demands, we would narrow our search to Atmel and TI microcontrollers. Company/Model Number Used Company/Model Number Used Arduino 3 Microchip 2 Arduino Uno 3 PIC18 1 Atmel 9 PIC24HJ256GP206A 1 Atmega168 1 Texas Instruments 11 Atmega Hercules TMS570 1 Atmega325 1 MSP430G Atmega328P 2 MSP430AFE2xxx 1 Atmega32u4 1 MSP430F ATSAM4S16B 1 MSP430F ATXMEGA32A4U 1 MSP430F XMEGA D4 1 MSP430F Freescale Semic. 1 MSP430FG MCIMX6D5EYM10AC 1 Tiva Cortex M4 1 Table : List of Spring 2014 Senior Design Microcontroller Choices Within Atmel and TI alone, the process of choosing one microcontroller out of their entire selection is daunting. TI sells 729 unique microcontrollers and Atmel sells 506. The first criterion we would use to narrow the search was the use of an ARM processor. This criterion was not performance based but chosen solely from the desire of our group to use that specific technology. We decided that since ARM processors are used in an overwhelming majority of mobile devices and embedded systems, we wanted to gain the experience of working with them. Among microcontrollers that use ARM processors, we found that TI sells 205 and Atmel sells 170. In selecting microcontrollers that use ARM technology, we were able to narrow the scope of our search to less than a third of the original numbers. 43

48 One of the most important specifications for our microcontroller was the availability of designated PWM output channels, as the 12 servo motors we planned to use require a PWM input. PWM is common across many applications, but since companies make so many models of microcontrollers, they are able to optimize them for use in specific applications, some of which do not use PWM. As such we narrowed our search further by eliminating products that did not have this feature. Table shows the families of microcontrollers from Atmel and TI that had at least one dedicated PWM channel. Comp. Family CPU Max # PWM Comp. Family CPU Max # PWM Atmel SAM9G ARM926 4 TI AM5K2Ex Cortex-A15 1 Atmel SAM9M ARM926 4 TI 66AK2Ex Cortex-A15 1 Atmel SAM9R ARM926 4 TI 66AK2Hx Cortex-A15 1 Atmel SAM9X ARM926 4 TI AM17 ARM9 3 Atmel SAM3A Cortex-M3 4 TI AM18 ARM9 3 Atmel SAM3M Cortex-M3 4 TI AM335x Cortex-A8 3 Atmel SAM3S Cortex-M3 4 TI AM437x Cortex-A9 6 Atmel SAM3U Cortex-M3 4 TI TM4C129 Cortex-M4 8 Atmel SAM3X Cortex-M3 4 TI TMS570 Cortex-R4 14 Atmel SAM4E Cortex-M4 4 TI RM46L Cortex-R4 14 Atmel SAM4L Cortex-M4 4 TI RM57L Cortex-R5 14 Atmel SAM4N Cortex-M4 4 TI TM4C123 Cortex-M4 16 Atmel SAM4S Cortex-M4 4 TI F28M35 Cortex-M3 24 Atmel SAMD Cortex-M0+ 24 TI F28M36 Cortex-M3 24 Table : Atmel and TI ARM MCUs with Designated PWM Channels The table shows that only one family of microcontrollers features 12 or more PWM channels, whereas six families of TI microcontrollers do. However, the TMS570, RM46L, and RM57L families use Cortex-R processors. These processors are designed for real-time applications, including many safety functions like automotive braking systems. This does not fit the description of our project, and these processors would offer much more and different functionality than our design needs. As such, they would generally be more expensive than we would like, and would include a number of extra features that would go unused. The remaining 44

49 products under consideration all contain Cortex-M processors, which are designed to be low cost and low power, which better describes our desired device. Of the four remaining product families under consideration, we compiled a spreadsheet comparing the specifications from their respective company websites. Since this list still represented 59 unique models, we found it more useful first to consider price and availability. We desired that the microcontroller we chose would cost less than $20, and that it would involve difficulty in acquisition with long lead times. Our vendor search was focused on two of the major electronics vendor websites, DigiKey.com and Mouser.com. Our assumption was that these vendors would be the most likely to have a given device in stock. In Table , we compare the total number of devices under consideration in each family to the total number of in-stock devices that had individual unit sales available. We also averaged the prices of the devices found from each family. Device Family ATSAMD TM4C123 F28M35 F28M36 # of models considered # of models found from DigiKey # of models found from Mouser Average DigiKey Cost $4.34 $10.00 $28.80 N/A Average Mouser Cost $3.14 $9.99 $31.93 N/A Average Overall Cost $3.74 $10.00 $30.37 N/A Table : MCU Cost and Availability Comparison From the table it is evident that the TI F28M35 and F28M36 families were almost entirely unavailable, and otherwise priced beyond our desired range. We also took note of the good and complete availability of the ATSAMD and TM4C123 families, respectively. Of final note was the satisfactory pricing of both families, with extra note of very low cost of the ATSAMD family. Some of the specifications of the remaining 37 models under consideration were compared in Table The values that were compared were cost, memory, processing speed, number of PWM channels, number of general purpose input/output (GPIO) pins, and other common features like various communication interfaces and analog to digital conversion. As memory, processing speed, number of PWM channels, and number of input/output pins are specifications, their relative magnitude for each part was expressed using color coding. The continuum of relative magnitude represented from least to greatest is dark pink, light pink, white, light green, and dark green. Though cost is a specification, it was already determined that the average cost of the TI models was over twice the average cost of the Atmel models, so it was not necessary to mark these values similarly with color. As for layout, the rows were sorted largest to smallest, first by number of PWM channels, by number of GPIO pins, by amount of Flash memory, amount of 45

50 SRAM memory, then by processing speed. This was done to make it clear which models had near the highest values in each category, Digik. Cost ($) Mous. Cost ($) Flas h (kb) SRAM (kb) Max Spd. (MHz) PWM Ch. GPIO SPI I 2 C Part Number TM4C123B H6ZRB TM4C123G H6ZRB TM4C123B H6PGE TM4C123G H6PGE TM4C123B H6PZ TM4C123G H6PZ TM4C123A H6PM TM4C123F H6PM TM4C123B H6PM TM4C123G H6PM TM4C123B E6PZ TM4C123G E6PZ TM4C123A E6PM TM4C123F E6PM TM4C123B E6PM TM4C123G E6PM ATSAMD2 1J18A 5.99 N/A ATSAMD2 0J ATSAMD2 1G18A 6.23 N/A ATSAMD2 0G ATSAMD2 1E18A 6.06 N/A ATSAMD2 0E N/A ATSAMD2 1J17A 4.80 N/A U A R T ADC Res. (Bit) ADC Ch. 46

51 Digik. Cost ($) Mous. Cost ($) Flas h (kb) SRAM (kb) Max Spd. (MHz) PWM Ch. GPIO SPI I 2 C Part Number ATSAMD2 0J ATSAMD2 1G17A 4.56 N/A ATSAMD2 0G ATSAMD2 1E17A 4.35 N/A ATSAMD2 0E ATSAMD2 0J ATSAMD2 0G ATSAMD2 0E ATSAMD2 0J ATSAMD2 0G ATSAMD2 0E ATSAMD2 0J ATSAMD2 0G ATSAMD2 0E Table : Comparison of Final 37 MCU Candidates When the composite of specifications significant to our design is considered, the TM4C123 family of microcontrollers is found to be more satisfactory. More specifically, the TM4C123 s maximum processor speed (80 Mhz to 48 MHz) and maximum number of GPIO pins (120 to 52) were superior. The maximum Flash memory (256 kb) and maximum SRAM memory (32 kb) were the same among the two families, and the rest of the listed specifications were comparable, while the maximum number of PWM channels (24 to 16) was greater with the ATSAMD family. However, as our design only needs 12 different PWM outputs, this was not significant. The major differences between the two families were cost, input/output resources, and processing resources. We felt the performance and resource advantage of the TM4C123 family was worth the cost. Having made the determination to use a TM4C123 family product, the rationale for the choice of specific model was simple; we wanted the highest performing of the group, as the cost difference among them was relatively negligible. Therefore, the microcontroller we chose was the TM4C123GH6ZRB, the part number of which is highlighted in the table. U A R T ADC Res. (Bit) ADC Ch. 47

52 For the purposes of testing we wanted to purchase a TM4C123 evaluation board that would interface more easily with the rest of the system without the permanence of the final PCB assembly. The evaluation board offered by TI is the Tiva C Series TM4C123G LaunchPad Evaluation Kit. The microcontroller incorporated into the evaluation board is the TM4C123GH6PM, the main drawback of which, when compared to the TM4C123GH6ZRB we plan to use, is its number of GPIO pins (43, compared to 120). Our plan with this is to determine during testing whether this amount of pins is satisfactory. If it is, we will likely purchase the TM4C123GH6PM for the final assembly of our design. If not, we will use the TM4C123GH6ZRB as planned Power Supply The power supply was picked by the most convenient way to hook it up to the wall. We wanted a low power high current supply to meet the requirements for the components on the guitar. We also desired a low cost to work with our budget. We found one from TRC electronics that provided what we needed for a cost of $ Below in Table is a table of specifications from the power supply chosen. Specifications Input Voltage 115/230 VAC Output Voltage 24 V Output Current 14.6 A Max Power 350 W Table Specifications table It meets the maximum requirements for our automated guitar. The PN# is SE This product offers protection for short circuit, overload, over voltage, and over temperature. It comes with a built in DC cooling fan with an on and off control. There is a constant current limiting circuit. For these reasons is which why we decided to pick this part for our project. Figure is a picture of the power supply purchased Figure Power Supply PN# SE

53 With the power supply we would need a voltage regulator to be built into the circuit. Voltage regulators are very inexpensive parts. The LM317TG voltage regulator offers a 1.2 V to 37V adjustable output voltage. The price is $1.37 from SparkFun, it offers an output max current of 1.5A which is more than enough for each component. Another voltage regulator that we plan to use is the LM , voltage line regulation; it offers minimum current deliverance and can be purchased from TI. The voltage regulators are both convenient (LM is from Electronics II lab) for our use and an inexpensive component. In Figure is a diagram of how the power will be regulated. Figure : Power Regulation Diagram There are four components that will need power distributed to the device, in order for the device to perform. The components are the servo motors, microcontroller, stepper motors, and the solenoids. Supply 1 will include a circuit and regulator that will provided the proper current and voltage distribution to each of the loads delivered to supply 1 in this case the stepper motors. Supply 2, will have the correct distribution of voltage and current that will lead to the microcontroller. Supply 3 is a different circuit that will require the proper voltage and enough current to deliver to 12 servo motor loads. Supply 4, requires a distribution of voltage and enough current to the 12 solenoids. We desire to have 12 moving solenoids, 1 for 12 frets, this is the reason for the numerous amounts of solenoids. Since we also plan to use servo motors in the design of the pulley system, this leads to the numerous amounts of servo motors. The power regulation circuit, and device, will be dependent on the strategic components that are picked. The regulator will have its own circuit that will lead to each device in order to adjust the voltage and current as needed. The circuit will include capacitors, inductors, and resistors. The design of the circuit will be introduced in section 4, of the design paper. 49

54 4 Project Hardware/Software Design Details 4.1 Initial Design Diagram Electrical Block Diagram Legend Work Status: Build Type: Microcontroller Purchased COTS HW Solenoid Sampling PCB Mounted Servo Research Pre-acquired Stepper Sampling Guitar Assembly Driver Circuits Research Power Regulation Research Signal Flow: Power Supply Purchased Power Guitar Pre-owned Data Computer Pre-owned Mixed Signal Guitar Amplifier Pre-owned 50

55 4.1.2 Mechanical Assembly Block Diagram Legend Work Status: Build Type: Bud Box Research COTS HW Neck Assembly SketchUp Model PCB Body Assembly SketchUp Model Piece-part Drive Belt Sampling Custom Build Wiring Harness Research 51

56 4.1.3 Software/Firmware Block Diagram 52

57 4.2 Stepper Motor Control (Picking System) The stepper motor we chose to use in the implementation of the picking subsystem was the SY20STH A available from Pololu, a 3.9V, two-phase bipolar stepper motor. As previously mentioned, we chose the stepper motor for this application because of its accurate position performance. However, a stepper motor, not being based on continuous current supply but rather on pulses of current, would require a unique driving circuit. As will be explained more below, the behavior of the stepper motor will be controlled by a small number of control lines from output pins on the microcontroller. As our stepper motor is a two-phase bipolar stepper motor, its inputs are arranged as shown below in figure The A phase coils can be energized either positively (current flowing from A+ to A-) or negatively (current flowing from A- to A+). The B phase coils can be energized in these two ways as well. For the motor shaft to rotate one step, a specified angular displacement, a pulse is sent to one of the coils, in one of the configurations mentioned above. If the coil were continuously energized in this manner, the motor shaft would hold its position rather than move. In order for the shaft to continue rotating, more pulses would have to be applied. Not only is this so, the subsequent pulses would have to be at a different coil or in a different configuration than the first. Figure 4.2.1: Stepper Motor Behavior Many pulse sequences exist that result in different motor output characteristics. It is useful for the driver to be able to switch between types of sequences for increased versatility, but this switching requires an extra control line from the microcontroller. For our purposes, as we want to limit the number of microcontroller output pins used, we plan to use a standard full step sequence. The full step sequence, as shown in Figure 4.2.2, takes the alternating configurations and coils concept in the previous paragraph and modifies it to increase torque performance. In this sequence, both coils are energized at all times. The cost of increased torque is increased current. 53

58 Figure 4.2.2: Step Sequence For the stepper motor to continue to rotation in one direction, this sequence is repeated, step one through four, then back to one. For rotation in the opposite direction, the sequence is reversed, four through one, then back to four. Since for each phase the current will need to be switched back and forth, a suitable topology to use is an H bridge. In an H bridge configuration, four switches are used, and current flows through the coil in one direction when a pair of switches is closed and flows in the other direction when the other pair of switches is closed. While an H bridge can be constructed out of discrete components including BJTs or FETs, a low-cost, space-saving alternative is to use a ready-made H bridge on an IC. The component we chose was DRV8833PWPR, made by TI. Its output current (1.5A) and voltage (2.7V-10.8V) ratings accounted for our stepper motor s values of 0.6A and 3.9V, respectively. It is designed to receive four input logic level input signals and output four power level output signals to the windings of the motor. It incorporates flyback diode protection and current limiting, if needed, internally. Figure shows the functional block diagram of the device. As shown, a number of external resistors and capacitors will be required to accommodate it. We intend to pull nsl to logic high, and AIS and BIS (current sense outputs) to ground at this time. The device itself will be powered at the same voltage as the supply voltage needed for the motor. Figure 4.2.3: Functional Block Diagram 54

59 There are four inputs to the H bridge device, but we plan to minimize the number of I/O pins used by the microcontroller for the purposes of fitting all functions to the reduced pin count on the evaluation board. To achieve this we plan to use only two output signals from the microcontroller to control the stepper motors. Since the microcontroller will output two signals and the H bridge device we will need an intermediate circuit. Not only do we need to control four lines from two controls, but we need to cause the H bridge to receive a repeated sequence of inputs. To perform these two functions we will use a state machine. If one controls a stepper motor with only two initial control lines, those to controls would include one for step pulse and the other for direction of rotation. In our state machine the step pulse will serve as the clock input to the flip flops. We intend to use D flip flops for their simplicity. Since the direction input will determine the progression of the state machine, it will need to be incorporated at the input of each flip flop. The next table was derived from the figure above in Figure 4.2.2, showing the full step sequence. It is clear that the + side and side of a coil are always inverse of each other. This reduces the number of flip flops from the state machine from four to two, since the state is already mapped by the inverse of the + state. With D flip flops, the next state is the same as the input to the flip flop. This arrangement, as shown in Table 4.2.1, determines the kind of combinational logic needed at the flip flop inputs. Dir A B Da Db A+ B Table 4.2.1: Flip Flop Combinational Logic 55

60 In the above table, all the transitions are assumed to occur at the positive edge of the step signal. The derived input equations for the flip flops are as follows: Da = Dir xor B Db = Dir xor A The implementation of these equations will only require XOR gates, since B is already an output of the flip flop, thus an inverter would be obsolete for Da. Therefore, the full implementation of the state machine for one motor encompasses two XOR gates and two flip flops. We chose SN74HC86N, a 4- channel XOR IC, and SN74HC175N, a 4-input D Flip Flop IC, both from TI. The XOR IC provides for the functionality of 4 XOR gates, and the Flip Flop IC provides for the functionality of 4 D flip flops with differential outputs. Because of this, we will be able to provide the logic for the state machine of two motors with one of each of these chips. Therefore, we will only need three of each for our picking subsystem. Their current and voltage requirements are standard logic level as we expected and their propagation delays were negligible. With the selection of the components of the state machine, the design of the stepper motor driver circuit is complete. In the figure below is the schematic diagram one of the six identical stepper motor driver circuits that will be used in our picking subsystem. Figure 4.2.2: Stepper Motor Driver Circuit Schematic 4.3 Servo Motor Control (Pulley System) The twelve selected servos will be interfaced directly with the microcontroller chip s twelve dedicated individual Pulse Width Modulation GPIO pins, with the microcontroller and servo motors sharing a common ground. The supply voltage 56

61 to the microcontroller will not necessarily be the same as the servo supply voltage source. All twelve servos shall have a common ground and a common supply voltage of 5 volts. There are no required dropdown capacitors or resistors with the servo circuits as they have all the required couplings, stator, and pulse width modulation circuitry is all packaged within the servo. As stated before, the MG90S will require a voltage step up on the PWM input line. There are twelve tri-state buffers that are attached to the input of the twelve servos. The 74VHC244FT buffers Vcc are tied to the 5 Volt line used by the Servo DC power node. The Servos (MG90S) require a pulse width modulation voltage of 5 volts. Thus, when the input of 3.3 volts coming from the microcontroller goes into the 74VHC244FT s lines, they can output the required 5 volts to the servo. The 74VHC244FT comes in a surface mount packaging, and any resistors can be tombstone surface mount components, to conserve space. This circuit is subject to change if the servo picked is insufficient in moving the solenoid assemblies. If the HS311 or HS485 servos are chosen, then the driver circuit is not necessary, as the input voltage for the PWM control line is variable from 3 to 5 volts. This would save space on the PCB and money as well. The unit cost of the 74VHC244FT is 0.49 dollars. With 8 lines inside each chip, this would mean that we only need 2 and would have 4 spare buffer lines. The servo schematic is shown in Figure Figure 4.3.1: Servo Schematic 57

62 4.4 Dynamic Control Dynamic control, in terms of electro-mechanical parts shall be two servo motors, either the HS311 or HS485 servos. They can be tied to the same power, same PWM input control line, and same ground, as they need to be actuated at the same time and travel the same distance. This circuit is easy to implement, as it requires no separate driver circuit, and only needs one PWM line from the microcontroller. As is shown in Figure 4.4.1, one input is used. This should not cause any issues apart from the current drawn from the microcontroller. The current consumed by the PWM line is not a parameter listed on the datasheet for the servos. 4.5 String Depression Figure 4.4.1: PWM line Figure shows the basic design plan of the sub-system for the String Depression. It includes the use of solenoids, with an overall built to hold the solenoids in place. In order to move the solenoids we will use a pulley system that will move accordingly with work from a servo motor. Figure 4.5.1: String Depression Sub-system 58

63 The solenoids will be held by the pulley through a design, with work of a 3D printer. Figure shows the part that is needed to be attached to the driver belt of the pullet system so the solenoid is able to move up and down each fret Solenoids and Control Figure 4.5.2: Solenoid Covering. For our string depression system we chose the ZHO-0420S-05A4.5 solenoid, which had an expected force output of 140 gf when excited with a 400mA current at 12V, at a 25% duty cycle. Its expected continuous current is 200mA, but we do not intend to run it at the level of current, since the force output would be much lower. The solenoid will be controlled from an output pin from the microcontroller. It will be powered not from the microcontroller but from the power supply subsystem, outlined in section 4.5. A solenoid primarily needs to be either on or off. That is, it needs a switch. One of the most common ways to perform this switching function is with a transistor. This is the technology we plan to use. The specific type and model will be discussed later in this section. As a solenoid is inductive, its voltage drop is proportional to the change in current through it. If the solenoid is switched off from an on state, it will resist that change with a proportional voltage spike that is unhealthy for the circuit. A diode will be needed to regulate this occurrence. In addition to the transistor and diode, our design will use a resistor between the output pin of the microcontroller and the base/gate of the transistor to bias it properly. So these three components, a transistor, a diode, and a resistor will complete the solenoid driver circuit. A reference design of a microcontroller-controlled solenoid driver circuit is shown in the figure below. Our design will emulate this layout, which is one of the simpler solenoid driver circuit designs possible. During testing, we will determine 59

64 whether we need to add greater complexity to modify such performance aspects as solenoid current draw and turn-on and turn-off transient response. Figure : Microcontroller Solenoid Driver Circuit The transistor in the figure is a bipolar junction transistor (BJT). The output from the microcontroller is either high (3.3V) or low (0V). When it is low, the biasing voltage is below turn-on voltage and the transistor is in cutoff mode. The collectorto-emitter terminals act as an open circuit and the power supply is not connected to ground. The solenoid remains in the off state. When the output of the microcontroller pin is high, the transistor is biased beyond the turn on voltage. The collector-to-emitter terminals nearly act as a short circuit; the transistor is in saturation mode. The power supply is connected to ground and current is passed through the solenoid but not through the diode. The current in the solenoid does not change instantaneously but exhibits a transient response that at this phase of our design we are considering to be of negligible length of time and current variation. By the end of the transient response, the solenoid should be passing the expected amount of current, resulting in the desired force applied to the shaft and, by extension, the guitar string. When the microcontroller output transitions from high to low voltage, the solenoid is expected to return immediately to its off state. This is not the case as the sudden change in voltage supply induces opposing current through the solenoid, causing it to continue operating after the desired time and with the spike in voltage potentially damaging the circuit components. To avoid this occurrence, the diode, commonly referred to as a freewheeling or flyback diode, is placed in parallel with the solenoid to dissipate the current quickly. 60

65 The type of transistor we will choose for our design will be a BJT, rather than a field-effect transistor (FET), as a BJT does not require as high of a voltage applied at the base to operate as a FET requires at its gate. There are tens of thousands of models of BJTs and FETs available to choose from, with many FETs that do not have such a high gate voltage, yet these are often more expensive. Though the drawbacks of using a BJT are current flow at the base terminal, voltage drop across the collector-to-emitter terminals, and often higher power dissipation, these are not expected to present design problems. We plan to use the TIP102 Darlington NPN BJT that is referred to in the reference design. Its absolute maximum current ratings are listed in Table It can handle up to 8 A of collector-to-emitter current, whereas we expect to need around 500 ma. We do not expect our base current to approach the same order of magnitude as the maximum of 1 A. Symbol Parameter Rating Unit IC Collector Current (DC) 8 A ICP Collector Current (Pulse) 15 A IB Base Current (DC) 1 A Table : NPN BJT current ratings The saturation current gain for the TIP102 is 500. From our desired Ic of 500 ma, our base current IB will need to be 1 ma to keep the BJT in saturation. With a microcontroller output voltage of 3.3V, and a base-emitter voltage of approximately1v, the resistor in our driver circuit will need a resistance of at most 2.3 k. To provide margin, we chose a value of 2 k. We note that the corresponding collector-to-emitter voltage drop is 0.8V, which would require a higher supplied voltage if we expect to keep a 12V drop across the solenoid. The power dissipated through the BJT would be VCE*IC = 0.4W, much less than the maximum rating of 2W. We plan to use the 1N4004 diode that was recommended in the reference design. It is a part of a family of 1N400X diodes (1N4001 1N4007) whose overall behavior is similar except for maximum reverse voltage, which ranges from 50V for the 1N4001 to 1kV for the 1N4007. All of these values are beyond the safety factor of 2 recommended above the nominal coil voltage of the solenoid. However, to comfortably handle all voltage spikes, we chose the 1N4004, which has a maximum reverse voltage of 400V. The average output current rating for each model is 1 A, which is greater than our operating current and is satisfactory for our circuit. Our circuit is shown in Figure

66 4.6 Power Supply Figure : Solenoid Circuit The power supply that was chosen and purchased for the use of this project is from TRC electronics PN SE It has a rated output voltage of 24VDC and an output max current of 14.6 A. The power it used for distribution and needs to be distributed to the six servo motors, six stepper motors, and 12 solenoids. Along with distributing to the electro-mechanical parts the power supply will need to be distributed to the microcontroller. In order to distribute power it is needed to know what components we are using, and what the rated voltage and current is required for these parts. In doing research, I have noticed that it may be better to even distribute power, that is a couple volts over the rated voltage in order to insure and moving device. In Table below it shows the rated voltage and current of the known components. The power supply design is then created from these known components. The first is the power design to the stepper motor. Component Manufacturer Part Number Rated Rated Current Voltage Power Supply TRC Electronincs SE VDC 14.6 A Stepper Motor Pololu SY20STH A 3.9 VDC 0.6 A Servo Motor (Pulley System) Tower Pro MG90S VDC mA/idle ma no load operating Servo Motor (Dynamic Control) Hiltec HS VDC Solenoid Average rate Average Rate 6-9 VDC 0.5 A mA/idle ma no load operating MCU TIVA TM4C123GH6PZ 3.3 VDC 19.7 ma Table 4.6.1: Rated Voltage and Current 62

67 In Figure shows, how a DC power supply, will be feed through a supply and converted for the rated voltage and current required to each stepper motor. The stepper motor will be using a driver circuit, and the power circuit will deliver current and voltage to the load. Since there are 6 stepper motors required, the current output of the supply will be that of the 6 stepper motors added together. So the voltage output required is 3.9 VDC, and the output current is 3.6 Amps. The LM3150, is ideal for this circuit because the voltage allows adjustment through the use of changing resistors values. Figure 4.6.1: DC Conversion to Stepper Motors In order to regulate the power going into each load component requires the use of power regulators. The regulators can be designed where once the power has been controlled it can be feed into the multiple components. For instance, for the stepper motor to receive the substantial power it requires the Texas Instrument voltage regulator LM3150 will be useful. The LM3150 is known as the simple switcher, and is a simplified step down power controller. It offers features such as thermal shutdown, under-voltage lockout, over-voltage protection, short-circuit protection, current limit, and output voltage pre-bias startup. This allows for a reliable product. For instance during testing or operation of the automated guitar, instead of frying the circuit or a motor we will, the regulator will allow for shut down. The LM3150 has an operating voltage input range is 6VDC-42VDC. The LM3150, will be designed in conjunction with the use of other components such as capacitors, inductors, and resistors. In Figure shows the schematic for the distribution of the power supply to the stepper motors. The output, can be connected to each load, so only one circuit is needed. The schematic includes the use of capacitors, resistors and MOSFET transistors. The output is the rated voltage, and the current required through each load added together. 63

68 Figure 4.6.2: Schematic for Stepper Motor In figure 4.6.3, this shows how a DC power supply will connect to the microcontroller load. Supply_1, signifies the power regulator converter that will be used a long with a circuit to provide the rated voltage of 3.3 VDC for the microcontroller, along with the logic components. The driver circuit of the stepper Figure 4.6.3: DC Conversion to Microcontroller and Logic Components motors, in the string picking subsystem will rewuire the use of 3 flip flop ics and 3 xor gates. We are able to set the voltage equal to that of the microcontroller to save design room on a PCB board. Since the TIVA microcontroller is high performance and varies, the rated current was choosen from the data sheet ma, is rated at room temperature at 16 MHz, while the microcontrolller is running. This is a very small current running to the microcontroller, after analysis of the data sheet, it was noticable that this specific controller is able to handle current more 64

69 than the rated and as little as 1.6µA. Since the current is so small, in the microcontroller using two regulators is recommended for better effiency. The first regualtor would be used to step-down the voltage and meet the recommend rated current, and the second regulator supply would be used to step-down the voltage to that of 3.3 VDC. The recommended regualators, for supply_1 is TPS The TPS62177 is a high efficiency synchronous step-down convereter. It has a voltage input rand of 4.75 V to 28 V DC. The device is used to proive up to 500 ma output current. The features of this device, is best suited for out project because it automatically enters power save mode at light loads, to maintain high efficiency across the whole range. It offers a feature a sleep mode to supply applications with advanced power save modes with ulter low power microcontrollers. This step- down converter is used to bring down the voltage as well as the output current. Figure is a schematic, for the power regulation of the microcontroller load and of the logic components. Figure 4.6.4: Schematic for Microcontroller and Logic Components Above in Figure shows how our DC power supply, will deliver DC voltage to the 12 servo motors. The servo motors are used in the design of the pulley system to move the solenoids, forward and backwards. The rated votlage ranges from 4.6-6V. The design of the power system is for a 6V servo motor. The greater power Figure 4.6.5: Servo Motor Load Distribution 65

70 delivered may help in resulting a greater torque for the servo. The figure shows, how only one voltage regulator will be needed and can be distributed to the multiple servo motors. TPS54336 is the best use voltage regulator for the servo motors. This step down converter is adjustable to the correct circuit to support the ideal current needed in this design. Figure shows the schematic for the servo motor. The schematic includes design requirements for the rated voltage and current, of the servo motors. Since each load requires a 0.18 ma of current, the ouput of the supply will have to be 2.34 A. The TPS54226 regulator is ment for output currents of 3 amps or less. It has up too a 28 voltage input, which is just enough for our DC voltage supply. Figure 4.6.6: Schematic for Servo Motor The schematic for the servo motor was desing in WEBENCH, a tool for power archeticture that Texas instrument offers. This supply will be regulating voltage and current to the twelve servo motors in the load. The servo motors are being used in the design to control the pulley system, under the fret. Since our design requires the use of 12 solenoids, one for each fret, we will also require 12 servo motors. The power distribution of the servo motor, may be redesigned in the future. The trick with the motor systems, is that there will have to be enough power delivered to each motor, so there is enough torque produced. Figure shows the DC conversion to solenoids. Since power can be distributed through multiple loads of the same voltage, only one power regulation circuit will be required for the power supply to the solenoids. The example above has twelve loads in this supply, because it signifies the use of twelve solenoids. The twelve solenoids will be used in the String Depression system. The current running through each solenoid is added in series to the supply, and this result in the output current at the regulation circuit. The solenoids require a LM25117 synchronous buck controller; this is intended to 66

71 Figure 4.6.7: DC Conversion to Solenoids step down regulator applications from a higher voltage source, such as our 24VDC power supply. The use of the emulated control ramp that this regulator carries is a feature that reduces noise sensitivity of the pulse-width modulation circuit. Figure shows how the power will be distributed throughout the load. Supply_1 includes the LM25117 regulator. The figure below shows how one supply regulating circuit can be used to distribute over multiple loads. For instance, to power the solenoids uses 0.5 A of current; therefore the output of the supply is 6 amps. The schematic to make this plausible can be found in Figure Figure 4.6.8: Schematic for the Solenoid 67

72 Figure shows the schematic for the regulation of power to the twelve solenoids we are using. Since testing is still required for the solenoids, the schematic is for the average rated voltage we plan to use and the average rated current. These values were chosen, because in the research section, it would be ideal to have a solenoid between 5-12 VDC. During testing periods we may conclude that a use of a greater rated voltage may be required, and then the design of the solenoid schematic will need to be re-worked. However for twelve solenoids rated at 6 volts with an input of 0.5 amps of current per each load. Figure 4.6.9: Dynamic Control Servo Motors Figure is the schematic that will be used for the dynamic control power regulation using the servo motors. The max rated voltage was at 6V, with a 180 ma rated current. This design was created for the max current. It reguires the us of the TPS62175 step-down converter. This converter is ideal as it can handle an input voltage from 4.75V-28V. Since there is now driver circuit required for these servo motors used our output current will be under 500 ma. This specific stepdown converter provides up to 500mA output current. Another great feature of the regulator is that it offers both adjustable and a fixed output voltage. This is great for future use if the schematic will need to change, we can just adjust the value of the capacitors. Component Regulator Price Stepper Motors LM3150 $1.42 Microcontroller TPS62177 $1.42 Servo Motors TPS54336 $2.26 Servo Motors TPS61725 $1.44 Solenoids LM25117 $4.30 Table 4.6.2: Regulators and Price In Table shows the results for the given regulators that we will require. Included in the table is the part number of the regulators and the recorded price for each component. The components will be used for testing purposes. We plan to use Webench, software offered from Texas Instrument, to create our power architecture. Webench allows the design of a multi-load power distribution, which may be integrated into the design of our PCB layout for the project. 68

73 5 Design Summary 5.1 Electrical Design Summary Power Regulation Power Regulation Schematic for stepper motors: Power Regulation Schematic for Microcontroller and Logic Components: Power Regulation Schematic for Servo Motors (Pulley System): 69

74 Power Regulation Schematic for Solenoids: Power Regulation Schematic for Stepper Motors (Dynamic Control): Servo Drivers Tri-state buffer driver array: 70

75 Servo Array: PWM Line: Stepper Motor Drivers Stepper Motor Driver Circuit: 71

76 5.1.4 Solenoid Drivers Solenoid Driver Circuit: 5.2 Mechanical Design Summary Our Hardware will be comprised of three main assemblies. There shall be a framework that has the 12 servos attached, and the belt-driven solenoid mechanical assembly. A separate assembly shall rest flush with the surface of the guitar body, and consist of the dynamic control system, which contains the six stepper motors. These two assemblies are connected to an enclosure which houses our PCB board and Power supply via wiring harnesses. 72

77 73

78 The prior figure shows the two subsystems attached to the guitar that will be used for this project. The dimensions of the static items have been fitted specifically for this instrument and would likely not be compatible with another instrument, apart from one of similar make and model. Each subsystem will be detailed in later portions of the paper PCB Enclosure Our PCB and power supply shall be encased inside of a metal casing, with chassis grounding for the power supply and potentially for the PCB as well. The power supply shall be mounted to the enclosure directly, while the PCB board will be mounted to the enclosure using standoffs, giving reasonable clearance for all surface mounted components on the PCB as well as any heat sinks String Selection Assembly The String Selection assembly, shown in Figure , would fit and or be assembled around the neck of the guitar. It would comprise of 2 main bulkheads, of which the neck of the guitar would pass through the bulkheads. Also fixed to the two bulkheads are two parallel metal dowels, on which the 12 belts would be rolled over, with some form of friction-removing bit, similar to a paper-towel roller or spool. Attached to the bottoms of each bulkhead shall be a floor on which the servo motors are fixed to, in staggered order. Twelve solenoid assemblies, separated by grooved wedges, shall be suspended above the fret board and strings of the guitar. The belts on each servo motor shall be passed around the dowels and be fixed to both sides of the solenoid assembly. Upon rotation of a servo, it would pull on one side of the solenoid assembly, which has free motion within the grooves in the wedges, would be pulled in the desired direction. Figure : String Selection Assembly (Top View) 74

79 As is shown in Figure , the dimensions between frets are shown in decrementing fashion from the left to right. Shown in Orange are the two metal dowels. The bulkheads are shown in magenta as well as the grooved wedges. As shown in Figure , servo motors are represented by red block items. Figure : String Selection Assembly (Side View) Solenoid assemblies are represented by blue block items. Spacing and dimensions are shown in the diagram. The grooved wedges are the purple items separating the solenoids, and their required spacing dimensions are specified. As guitar fret boards and string placement is slightly beveled, the grooves need to be slightly beveled as well, so that the solenoid action is the same distance for each string. This is better detailed in figure , shown below. Figure : Solenoid Enclosure 75

80 Materials for the bulkheads and wedges is not specified at this point in time but a likely candidate is balsa wood, or some other light wood. Solenoid enclosure material candidates include 3D printer-made or similar material. A mold would likely be even more expensive than 3D printing Picking Assembly The picking assembly comprises of a base piece, which would rest flush on the surface of the guitar, conforming to several key contour features, which include the 2 humbucking pickups of the guitar. Four corners of the base shall each have a vertical groove. Matched with each groove is a corner of the dynamic control assembly. The dynamic control assembly consists of a boxlike framework, containing six servo motors, with 3 on each opposite side, staggered and 20 mm apart center to center. Fixed to each stepper motor is a guitar pick, using some method yet to be physically actualized but rather conceptualized at this stage. Two armatures protruding from the box structure would include a tooth that rests on a worm gear. The worm gear is fixed a servo motor, of which is mounted on the base fixture. This is the dynamic control system. When the servo motors are rotated, the worm gear rotates, and would in theory lift up the entire rack of stepper motors, which in turn changes the depth of the picks in reference to each string. These characteristics are visualized in Figures , , and Figure : Picking Assembly (Top View) As is shown in Figure , the six stepper motors are represented in lime green. These items are directly modeled on the chosen stepper motor, purchased from Pololu. Its dimensions fit very conveniently to our use. The worm gears are shown in light blue. Beneath them are misleadingly modeled servo motors. The servos currently chosen are Hitec HS311 s which are rectangular and tall. They will be more accurately represented in the concept sketch below, Figure The base is shown in dark pink, while the dynamic control frame is shown in light pink. 76

81 Figure : Picking Assembly As seen in Figure , guitar picks are shown as orange items. The attachment feature used to connect them to the stepper motors is not shown. The picks could be cut accurately and form fitted around the rotating arm of the stepper motors, or securely fixed to the flat side, to simplify the design. Figure : Dynamic Control Assembly 77