Analog Computers

Reference / Paper · 2022

Analog Computer Prototyping for the Future

Read the PDF (39 pp) ↗

Master's thesis from Malmö University (2022) documenting the design and construction of a low-cost modular single-board analog computer capable of solving second-order differential equations. The project follows design science research methodology (DSRM) and evaluates two hardware revisions (V1 and V2) through demonstration problems including Mathieu's differential equation and a spring-mass damping system. The paper argues that accessible analog and hybrid analog-digital systems are a viable path toward more energy-efficient future computing.

Manufacturer
Malmö University
System
Custom single-board modular analog computer (V1 and V2 prototypes)
Author
Carl Oskar Ahlqvist; Måns Ahlgren
Year
2022
Type
Reference / Paper
Language
English
Learning track
specific applications
Pages
39
  • Custom single-board modular analog computer (V1 and V2 prototypes)
  • Malmö University
  • analog computer design
  • differential equations
  • hybrid computing
  • single-board computer

← Back to the Reference Library

Analog Computer Prototyping for the Future

Analog Computer Prototyping for the Future Carl Oskar Ahlqvist Malmö University, [email protected] Måns Ahlgren Malmö University, [email protected] Master thesis for a degree in computer science Faculty of Technology and Society Malmö University Sweden 2022-05-10 Supervisor: Sven Karlsson Examiner: Dario Salvi ACKNOWLEDGEMENTS Firstly, we would like to thank our supervisor Sven Karlsson for believing in this somewhat unusual project. Your guidance and limitless excitement for this project helped immensely when times were tough. We would also like to express our sincerest gratitude for the excellent work of Bernd Ulmann and the team at Anabrid for keeping analog technology both alive and relevant in a digital society. Without the information you have gathered over the years, this project would not have been possible. Lastly, we would like to thank friends and family for their input, especially Ellen Ek for her contribution of excellent mathematical knowledge that elevated the paper to the next level. The analog computer created during this project. C ONTENTS I Introduction 1 II Background II-A Relevancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 III Purpose and Goal 2 IV Research questions 2 V Research Methodology 3 VI Terminology VI-A Analog computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI-B Boolean algebra and discrete mathematics in computer science . . . . . . . . . . . . . . . . . . . . . . VI-C Differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI-C1 First order linear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI-C2 Second order linear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI-D Simple oscillator circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI-E Mathieu’s differential equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VI-F Spring mass dampening system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3 4 4 4 4 4 4 5 VII Related Work VII-A The computational advantage of analog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-B Analog computers in academia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-C Modern analog computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-D Modern applications of analog computers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-E The different components of the analog computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-F Inverting and non-inverting operational amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-G Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-H Integrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-I Potentiometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII-J Multiplier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 6 7 8 8 10 10 10 11 12 12 VIII Results VIII-A Circuit design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-B Breadboard experimentation - Spring mass dampening . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-C Breadboard experimentation - Mathieu’s equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-D V1 of the analog computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-E Integrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-F Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-G Coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-H Inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VIII-I Multiplier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 13 13 13 14 15 15 15 15 15 IX 15 15 16 16 16 16 16 16 Evaluation of the V1 Analog computer IX-A Integrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX-B Inverter/Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX-C Multiplayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX-D Potentiometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX-E Design choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX-F Usability and function testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX-G Functionality evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X Evaluation of the V2 Analog computer X-A Integrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-B Inverter/Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-C Multiplayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-D Potentiometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-E Design choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . X-F Hybrid system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 17 17 17 18 18 18 XI Evaluation XI-A Feasibility of creating a low cost modular single board analog computer . . . . . . . . . . . . . . . . . XI-B Capabilities of the modular analog computer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI-B1 Physical capabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI-B2 Capabilities for providing academic value . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI-B3 The extended capability of the analog computer . . . . . . . . . . . . . . . . . . . . . . . . . XI-B4 General applications of the analog computer in a digital society . . . . . . . . . . . . . . . . 19 19 20 20 20 21 21 XII Conclusion & Future work 22 References 23 Appendix A: Integrator schematic 25 Appendix B: Power supply schematic 26 Appendix C: Summer schematic 27 Appendix D: Summer schematic 28 Appendix E: Hamilton Analog computer schematic 29 Appendix F: Integrator V2 Schematic 30 Appendix G: Summer V2 Schematic 31 Appendix H: Multiplier V2 Schematic 32 Appendix I: PSU and Potentiometer V2 Schematic 33 Appendix J: BOM-table of V2 34 after the release of Vannevar Bush’s analog computer). The early digital computers relied on the user writing code to solve a problem algorithmically, as the case is still to this day. The early problem was that the digital machines often were slow and expensive, thus hindering the user to solve problems. The digital machines were and still are based on an algorithmic sequence thus not being able to run many calculations simultaneously natively. Analog computers have a long history dating back to around 1000 years ago. The most famous example of this age is the Antikythera mechanism dated to around 100 B.C [5]. One of the more recent famous analog computers is the Vannevar Bush’s differential analyzer released in 1931. The analog computer Bush created was initially meant to model power networks, but its value as a general-purpose analog computer was quickly realized. The computer was a large mechanical construction made of an assortment of gears and shafts driven by electric motors. The disc wheel based integrators could be connected to a number of rotating shafts. The computer could solve sixth-order differential equations, although this was complex to set up. The machine was used in solving problems in physics, seismology, and ballistics [6] [7]. These early examples are all examples of mechanical analog computers. After the release of Bush’s analog computer, the electric computer was on the horizon. The electric analog computer was widely developed by many manufacturers during this time including, Electronic Associates Inc., Applied Dynamics, RCA, and Telefunken among others. The electric analog computer was first used as an aid in missile and airplane design as well as running aerodynamics calculations. This meant that the space and aviation industries were the largest customers during this time. The customer base later extended to nuclearreactor control [8] [9]. Analog computers are not high-accuracy devices. Therefore their usage was not centered around tasks that required high levels of accuracy. The analog computer was used often in engineering and applied sciences as an analog for the problem being studied. One of the most common and least complex calculations is the linear ordinary differential equation (which is described in the terminology section of this paper) [3]. Physics calculations are very well suited for analog computation. In the early days of space flight, analog computers were used to calculate orbit trajectory and other orbit-related equations such as satellite positioning [10]. Calculating particle trajectories was also a well-established field within physics where analog computers were widely used [11] [3]. See fig 1 for an visual example of an commercial electro analog computer. Before the invention of the modern MOS (Metal Oxide Semiconductor) the electric analog computer was based on vacuum tube technology. The vacuum tubes acted as transistors and were later replaced by modern solid-state transistors. The programming of the analog computer was (and still is to some extent) done by interconnecting different computational modules through a patch panel. This way of programming Abstract—This research paper focuses on analog computers and creating a modular low-cost analog computer system in a single board computer form factor. The single-board analog computer will have the capacity to solve second-order differential equations. The capabilities and possibilities of the single board Analog computer will be explored as well as analog computing in general. The paper follows design science research methodology (DSRM) with the goal of creating and evaluating a working artifact. The artifacts’ functionality is evaluated based on a demonstration of its ability to solve Mathieu’s differential equation as well as simulate a spring-mass dampening system. This paper proves that it is possible to create a low-cost analog computer in a modern form factor. The artifact is also placed in a larger contextual setting based on the empirical material provided where its value of it in a digital society is presented. For the world to continue its progression in computational power, but still, limit the already high energy usage, a drastic change is needed. This paper suggests adapting to analog/hybrid technology. To further the progression of analog/hybrid technology it must be made accessible to a wider group of people compared to today. The artifact in this paper offers a solution to this. Index Terms—Analog computing, Hybrid systems, Differential equation, Single board computer, DSRM, Sustainable IT I. I NTRODUCTION Digital computers are the norm today regarding computational power. Since the solid-state transistor became widely available, the digital computer has been dominating computation. Today’s high energy consumption and potential stagnating progression of digital computers have led to a new turning point in the progression of computational speed and sustainable IT. In order to meet the global demand for greater computing power, a drastic change has to be introduced to the world of computation. This is a must if we wish to continue our present computational progression and decrease our ecological footprint. One of the possible solutions to this problem is to change the fundamental view of computation [1]. By switching to solving problems using analog computing a positive impact could be achieved in today’s ecological and computational scene compared to using strictly digital systems. By creating hybrid systems that can utilize an analog coprocessor, that is optimal for solving differential equations, low-energy analog computing can be utilized to a larger extent compared to today. In order to achieve this, a new approach to introducing the world to analog computing is needed [2] [3] [1]. II. BACKGROUND The three major computational paradigms are analog, digital and quantum [4]. The Analog computer has seen a steep reduction in prominence during the last 40-50 years due to the rapid improvement in digital computing. This has led to the fact that very little recent research has been conducted on the analog computer and its potential role in today’s society [3]. The popularisation of digital computers in today’s society stems back to the early days of computing when both analog and digital computers were common (the following decade 1 used 13.2 billion kWh of electricity to power data centers and servers. This was an increase from 12.6 billion Kwh in 2016. The global consumption is estimated to be 350 billion Kwh in 2017. Given the fact that there has been a steady increase in energy consumption between 2010 and 2017, there is evidence that points toward this trend continuing [15]. Most of the high energy usage is due to today’s power-hungry CPUs. A warehouse-scale data center’s total power usage is approximately 33-60% only due to the modern CPU [16]. The digital computer running boolean algebra is a versatile and well-developed system. One of its problems is the aforementioned high energy consumption of modern processors. A hybrid system that utilizes an analog co-processor has been proven to run faster and use less energy than a common digital processor as proven by Köppel et al [2]. The digital computer is also approaching the theorized Moore’s law where the computer chips can not be made faster by physical means [2]. Analog computing has been lost to history due to the digital computer being more versatile and becoming easier to use. The three major computational paradigms are digital, analog, and quantum computing. Quantum computing is still too far away to be implemented in today’s society. Therefore looking to history and investigating analog computation could be a way forward and continue the computational progression and create more energy-efficient systems. Fig. 1. GTE Analog Computer EA22 [12] is very complex and required a skilled programmer to be successful. This is also one of the reasons for their later demise. Another contributing factor to the demise of the analog computer was the introduction of the digital computer. The more straightforward approach to programming, algorithmic computational structure, the ability to easily store information, high precision, and ability to handle any problem given the increase in computational time was a clear advantage over the analog computer. The advancements of MOS and chip technology made digital transistor-based chips faster and cheaper thus quickly rendering analog computers obsolete [8] [5]. This paper will not focus on pre-electric analog computers like the Antikythera mechanism and the Vannevar Bush’s differential analyzer. The focus will lie on electric analog computers, for example, the computer created by the research team at Anabrid [4]. Anabrid has created an open-source analog computer that aims to enable researchers and students to explore the area of analog computing [13]. The device can calculate and simulate several large computational challenges like market economics, the spread of disease, and more [13]. Digital computers are approaching the end of continuous improvement and Moore’s law (where the computer chips can not be made faster by physical means) will take full effect. To keep progress going, something has to be done in order to further the improvements in computational power, since quantum computing is still far off in the future as a viable option. This could be done by creating a hybrid computer that takes all the good aspects to form digital and analog computers and combines them into a hybrid computer like Steiglitz [14] proposes. One of the most challenging parts of analog computing in today’s society is the user experience, availability, and ease of use. In order to have analog computing viable in today’s society, a better way is needed to interact with the computer. III. P URPOSE AND G OAL An analog computer will be created that has the capability to run different first and second order differential equations. The actual component-based computation that is based on the mathematical expressions for the differential equations will be converted into circuits. These circuits will then be tested and used as a tool to verify the functionality of the computer. The purpose of this paper is to create an analog singleboard computer to make analog computing accessible for academia and educational institutions. A low-cost, easy-to-use, analog computer can be used as an educational tool to further the innovation within analog technology. Given the fact that most analog computers are either expensive or non-existent on the open market, a low-cost modular analog computer could have a similar impact as the Arduino single-board computer platform had on academia. The created artifact will also be placed in a larger contextual setting where its role in a larger socio-material context will be investigated. The uses for the artifact and general analog computing in a digital society will also be investigated based on the empirical material provided in this paper. A hybrid setup using an Arduino and the analog computer as a co-processor will also be demonstrated. Therefore this project has the potential to be an innovation for change in a digital society for a more computation demanding and greener future. IV. R ESEARCH QUESTIONS A. Relevancy The global energy consumption by data centers, servers, and other digital hardware is staggering. In 2017, Germany alone Based on the information presented in this paper thus far, the research questions for this project are as follows: 2 requirements of the analog computer such as components, size, and complexity. The functionality of the artifact will be tested with a demonstration of its ability to solve specific differential equations. These differential equations are the spring-mass dampening system and Mathieu’s differential equation. A simple oscillator will also be used to evaluate the computational elements. The result of which will be compared to circuitry specifically developed to solve these two equations. The larger role of the artifact and analog computing in a digital society will also be investigated based on the empirical material provided in this paper. The steps this paper will take in order to develop the artifact is: • Investigate the different computational elements of an analog computer. • Translate the computational elements to circuitry. • Develop the PCB that will make out the analog computer. • Evaluate the functionality by demonstrating the predetermined differential equations and compare the result of the analog computer with the custom circuitry. • Present the results found in the evaluation in the context of an ever growing digital society with a focus on education and practical application of analog/hybrid technology. "RQ1 - What is the feasibility of creating a low cost modular single board analog computer that can solve second order differential equations?" "RQ2 - What are the capabilities and usages of a low cost modular single board analog computer?" V. R ESEARCH M ETHODOLOGY The methodology used in this paper is heavily influenced by design science research (DSR) and its methodology as described by Peffers et al. [17]. The focus of the research conducted is to create an artifact that has the potential to further the development of analog computers, therefore being actually useful in its domain. The methodology does not focus on rigid design processes and formal theory. The design science research methodology (DSRM) presented by Peffers et al. [17] is influenced by the DSR forerunners March and Smith [18], Nunamaker [19], and Walls [20]. These authors were focusing on building actual artifacts that would later be used in case studies to examine their impact on a specific institution. A DSRM research project is well suited to start with a loosely specified problem such as a stakeholder request or further development on an existing product. Important to note is that this method of conducting research is more of a suggestion on "how to do it" rather than stating that it is the best way to go about that particular research according to Peffers et al. [21]. The role of theory for DSRM is to create a technical base for the artifact being created to rest upon. Very little contextual theory is included since it is assumed that the artifact contains the context needed in order to justify its creation. This means that more focus can be allocated to the technical base for the artifact and limit contextual inclusions to a DSR paper. Although the artifact created holds context by its very creation, some context will be provided given the lack of prevalence of analog computing in today’s society. To evaluate the artifact, DRSM uses demonstration-based evaluation in order to prove the functionality of the artifact. If the artifact can be demonstrated using a predetermined functionality test, the functionality of the artifact can be proven instantly. If it can be demonstrated over a range of contexts, it can be evaluated [17]. In order to create the artifact, that is the low-cost analog computer, the different computational elements that make up an analog computer need to be investigated. This includes the analog computer programming schematics that is prevalent in a lot of textbooks from the time of the analog computers’ height of popularity. These computational elements will then be translated to circuitry that can be manufactured on a PCB (Printed Circuit Board). This includes choosing ICs (Integrated Circuits), components, and component values based on the technical literature and experimentation. This will form the VI. T ERMINOLOGY A. Analog computer Analog is derived from the Greek words ana-logon. The translation is interpreted as "according to a ratio". This could be explained as a connection between different relationships in regards to material and constructs. A sail catching the wind, an airplane wing creating lifting force, and the submarine deep in the Atlantic ocean are all different constructs but analogous in the way they deal with physical forces. The electric analog computer’s output from operational amplifiers is a response to its input signals. The output varies depending on the input. Utilizing this, a system that can analogously respond to different conditions can be modeled to simulate different real-world phenomena [22]. Analog computers were and still are based on variable voltages to calculate and represent the solution to a given problem. Analog computers handle computation by utilizing logic built by different operational amplifier-based circuits for different mathematical calculations based on variable voltages in the circuit [5]. The analog computer is highly suited for calculating second order differential equations. The problem the user wants to solve (by the equation) is first translated into an equation. The equation is then rearranged, in this example using the Kelvin feedback technique [5], where the highest derivative is isolated on the left side of the equation. By then feeding this derivative into a specific chain of computational elements (integrators, summers, etc.) the right side of the equation will be obtained. By creating an analog system, a physical system is configured and translated to mathematical abstraction. By introducing an initial condition that corresponds with the problem’s 3 initial value the analog computer will simulate the problem over time starting at T0 to Tf (meaning the start and stop of the simulation) and present the results continuously [8] [5]. Digital computers rely on relays representing ON (1) and OFF (0) to calculate and display the solution to a problem. This is called boolean algebra and is the norm for digital computers today. dN = λdt N 2) Second order linear differential equations: A second order differential equation is defined by the equation below: B. Boolean algebra and discrete mathematics in computer science This is essentially the same as a first order differential equation, except that y second derivative ÿ also is included. Boolean algebra is the logical language where you are able to express any statement through a binary form [23]. In computer science this often refers to machine language, ones and zeros or on and off pulses. Due to the vast versatility of the different logic gates that can be created by combining transistors boolean algebra and discrete mathematics quickly rose as the leading language of computer science [24]. There are many advantages with discrete mathematics and boolean algebra when it comes to computer science, mainly giving in this case the programmer the ability to be very precise when running calculations. It takes a lot for a one to become a zero in comparison with a 1,00001 becoming a 1,00002 in the field of continuous mathematics [25]. Despite being the computer language of today, discrete mathematics and boolean algebra has a big bottleneck when it comes to its use in computer science. This disadvantage stems from the lack of variability in the throughput which forces digital computers to work sequentially. A digital computer can only process either a one or a zero at a time which limits the computational speed to the rate in which you can switch between a one and a zero [25]. D. Simple oscillator circuit • Radioactive decay: − F (t, y, ẏ, ÿ) = 0 The oscillator circuit is described in Ullmann’s book “Analog and Hybrid Computer Programming” [5] the equation for the oscillator goes as follows: ÿ = −ω 2 y where ω 2 is some arbitrary weight. The logic schematic is: Fig. 2. Simple oscillator [5] The oscillator can be used as an effective tool to test the functionality of analog computers (integrators, invertors and summers). It was used in this project as the first functionality test and the output (solution) should be displayed as a steady waveform. C. Differential equations As Braun and Golubitsky state in there book "Differential equations and their applications" differential equations occur in many areas of science [26]. In this section we will briefly explain what a first and second-order linear differential equation is and give some examples of what they are used for. Important to understand is that a derivative can be interpreted as a rate of change seen over time. The first derivative is the rate of change and the second derivative is the acceleration of change. 1) First order linear differential equations: A first order differential equation is defined by the equation below: Fig. 3. Simple oscillator output [5] F (t, y, ẏ) = 0 E. Mathieu’s differential equation In other words, an equation consisting of the unknown function y and its first derivative ẏ with respect to time t. First order differential equations occur in a lot of different fields of science, a few examples of those are: • Newton’s law of cooling: Mathieu’s differential equation is a well known, linear homogeneous, second order differential equation. The equation has many different applications ranging from the vibrations of an elliptic drum to balancing an inverted pendulum [27]. As described by Ulmann [27], the general expression of Mathieu’s differential equation is: ẏ = k(M − y) • ÿ + (a − 2qcos(2t))y = 0 Eikonal equation: H(x, ∇u(x)) = 0 With the initial conditions: 4 y0 = 1 ẏ0 = 0 a and q are parameters that has been assigned a value based on the given problem. To simplify the calculations one can set a = 2q. By doing this one gets the calculation: ÿ + (a − acos(2t))y = 0 One can then factor out a out of the parentheses which results in: ÿ + a(1 − cos(2t))y = 0 Fig. 5. X for Mathieu’s equation [27] The resulting equation is now dependent on y (the sought after value) and t (the solution to the equation when taking time into account a is still a parameter one can assign any value to. The parts that is dependent on t can be seen as an "input signal" that can be modeled. Therefore one can call it x(t) or just x. To ease the calculations we introduce the x variable to the aforementioned equation by the following statements: x(t) = 1 − cos(2t) ÿ + axy = 0 Based on the aforementioned explanation, these statements holds true. The general expression for y can be exemplified by analog computer schematics, see fig 4. Fig. 6. Typical solutions for Mathieu’s differential equation [27] F. Spring mass dampening system The spring mass dampening system is a good example of an oscillator with dampening. This simulation is an analog of a weight suspended on a spring with the mass m. The oscillation is the weight bouncing up and down with the position y, thus the vertical position. According to ´ Ulmann [5], to simplify the calculations, the gravitational forces acting upon the weight will be neglected. The force due to the moving mass (weight) can be expressed as: Fm = ma = mÿ The spring will introduce a force depending on the strain applied to it (pulling the string) and the force of the linear velocity dampening can be expressed as: Fig. 4. Y for Mathieu’s equation [27] The general expression for x in the Mathieu’s differential equation (5) can be expressed as: Fs = sy ẍ + 4x = 4 Analog computers are ideally suited for solving differential equations. To specify what differential equation to use, the Mathieu’s differential equation was chosen because of its prevalence in academic resources. The analog computers schematic for solving Mathieu’s differential equation can bee seen in fig 4 and 5. Typical solutions for Mathieu’s differential equation can be seen in fig 6 Fd = dv = dẏ The spring mass dampening analog is a closed physical system. This means that all the forces in the system needs to add up to 0. The second order differential equation based on this information can be expressed as: mÿ + dẏ + sy = 0 5 The m is the mass, d is the dampening constant and s is the spring constant. Important to note, with out Fd no dampening will occur. The actual equation the analog computer can solve, and has the possibility to be easily converted to a schematic, is: dÿ + sy m The schematic that will be used in this paper to create a circuit is a simplified version in order to save one summer. The disadvantage of this approach is that constants s and d can not be set independently. The simplified approach has s and d as a fraction of m. The simplified version introduces two new parameters called v and µ. The mathematical expression for these two are: s v= m ÿ = − Fig. 8. Typical solutions for spring mass dampening. Top graph v = 0.6 µ = 0.8 Bottom graph v = 0.8 µ = 0.6 [5] The comparison between analog and digital is something academia has covered in regards to time, the complexity of problem, and energy usage. A paper written by Köppel et al [2] uses a common differential equation. Their experiment showed that the analog computer did not take longer to compute given the complexity of the problem but the digital computer did. The energy consumption did also increase for the digital computer (which was significantly higher initially) when the complexity increased, whereas the analog computer remained at low levels of energy consumption throughout the experiment. This is due to the fact that the analog computer does not run on a sequential boolean algorithm. Instead, the Analog computer runs continuously and directly using the op-amp (operational amplifier) based computational elements. This is the main advantage of using an analog computer according to the writer. The common digital computer is a so-called a "stored program computer" and has a fixed internal structure. This fixed internal structure is then controlled by an algorithm which is stored as a program in memory inside the machine. This structure has the tendency to create data dependency and memory bottlenecks that impact the overall performance negatively. The analog computer is in comparison not reliant on memory or having a fixed inner structure. The different computational components are arranged in a way to create an analog of the mathematical problem that is being simulated. This speeds up the process of computing and simulation of a given task due to all calculations running in parallelism to each other [28] [2]. In a paper written by Holzer and Ulmann [28] they discussed the integration of analog computing when working with machine learning (ML). They created a hybrid system where the analog computer was closely connected to a digital machine. The analog computer was continuously simulating the balancing of the inverse pendulum and the ML algorithm used this information as data for reinforced learning. The analog computer was utilized as a powerful coprocessing unit to solve the differential equation-based prob- d m. The two potentiometers −ẏ0 and y0 are used to set the initial condition of this analog. y0 sets the initial deflection and −ẏ0 sets the initial velocity of the mass which is set to 1. Different values for s and d will result in different results as the spring and dampening will be seen in the oscillation of the mass. The analog computer schematic can be seen in fig 7 and typical solutions can be seen in fig 8. µ= Fig. 7. Simplified spring mass dampening [5] VII. R ELATED W ORK A. The computational advantage of analog The analog computer does not run on a sequential basis like its digital counterpart. The added flexibility of the digital computer is due to the ability to run it on a joined clock system (enabling time sync between two digital computers enabling precise simultaneous operation) and being able to solve a variety of tasks due to the aforementioned sequential algorithm structure. The main problem with the current development of the digital computer is that the digital computer is rapidly approaching Moore’s law, thus hindering further progress [2] [5]. 6 lem of balancing an inverted pendulum. The reinforcement learning was done episodically where one episode is from the start to the pendulum tipping over or the cart moving outside the designated area. The digital computer asked the analog for the real-time simulated values, the communication went through a hybrid controller (by serial communication) that converted the information between the two systems. The digital computer, running the ML algorithm, then decided if the episode was to be terminated and reset or continued. If the pendulum was mounted on a fixed cart the equation would have been: g φ̈ − sin(φ) = 0 l Since the cart was not fixed, the mathematical expression was greatly more complex than if the cart would have been stationary. The authors goes great lengths proving and simplifying the differential equation actually being calculated by the analog computer. The differential equation that was being solved was: The usage of electronic analog computers can also be applied to academic endeavors in other ways than simulating physics or engineering problems, such as a pedagogical instrument (see fig 9). Students can be enrolled in mathematics classes without ever visualizing a differential equation. Students are thought to instead create an expression that satisfies the equation through a systematical (perhaps even mechanical) approach that gives little deep understanding of the process. The implementation of analog computers in this context is both cost-effective and shows a positive impact on the learning outcomes due to an increase in intuitive calculations by the students. The analog computer allowed the students to work on nonlinear and linear systems while emphasizing the relationship between the mathematical and physical models. This meant that there was a knowledge combining element between the math and physics-focused classes [30] [31] [32]. φ̈ = ẍcos(φ) + gsin(φ) And the acceleration of the cart can me modeled as: M ẍ = F The length of the pendulum was assumed to have the length 1 (l = 1) and is therefore removed in the above equation. The φ is the angle between the the pendulum and the vertical axis (y). g is the gravitational acceleration, x is the position along the horizontal axis (x), M is the mass of the cart and F is the force applied to the cart to stabilise the pendulum. The results of the study by Holzer and Ulmann [28] showed that a hybrid system is possible when dealing with ML and that having an analog computer simulation the problem was more stable and more energy efficient than a purely digital system. The paper also suggest that great computational improvements could be achieved if there was an integrated analog coprocessor embedded in modern digital systems, thus increasing computational and simulation capabilities. Fig. 9. A demonstration of the solution to a differential equation based problem using an analog computer [30] A low-cost, desk-sized, analog computer for use in academia has been of interest to technical faculties in the past. Hamilton [33] proposes a design for a machine (see appendix E for schematic) in a paper called "An Analog Computer for Educational Laboratories". The machine is a general-purpose analog computer that features several computational elements such as integrators, summers, dividers, and coefficient potentiometers. The design also features an overload detection, meaning when an overload of the operational amplifiers occurs. An example of a desk-sized analog computer that saw a long life in academia is the Comdyna GP-6 released in 1966 [34]. The GP-6 longevity in academic institutions can be explained by its use in control unit education. The example given in Spiess [35] paper describes the students acquiring more knowledge of electronics and the connection to mathematics while using an analog computer. The computer used in the paper is a GP-6 and is a staple in the control systems laboratory at the University of Illinois. Programming an analog computer requires an understanding of the fundamental building block B. Analog computers in academia Since the invention of the general-purpose electronic analog computer the usage in engineering and physics laboratories has been prevalent. Given the nature of the analog computer and its ability to solve differential equations efficiently, the analytical solution to problems was not the default approach. The analog computer made the numerical approach possible as the approximation needed was easily introduced in programming. In the early days of analog computing, the experimentation process was greatly sped up by having the possibility to easily program (compared to its contemporaries) a problem and see the results almost immediately. An example