Underwater Optical Communication: An Approach Based on LED

FACULDADE DE ENGENHARIA DA UNIVERSIDADE DO PORTO Underwater Optical Communication an Approach Based on LED Bernardo Miguel Carvalho Silva FINAL VERSION Masters in Electrical and Computers Engineering Supervisor: Nuno Alexandre Lopes Moreira da Cruz Co-Supervisors: José Carlos dos Santos Alves, Luís Manuel de Sousa Pessoa October 30, 2015 © Bernardo Miguel Carvalho Silva, 2015 ii Abstract One of the major obstacles in the use of underwater robots on a large scale is the difficulty in communicating with the platforms during missions, since the most common methods of communication (via radio or other electromagnetic waves) used in commercial systems use signals that are strongly attenuated in water. Alternatively, acoustic-based solutions have long been the default wireless communication method for underwater applications, since they allow reasonable ranges. However, due to the severe limitations in bandwidth and the slow data rates with high latency, these are not the most efficient solutions. Summing up, using acoustics, applications such as monitoring and controlling remote operations are not practical. In this project, it is intended to take advantage of the recent developments of the light emitting diode (LED), (specially with the higher light output and more precise tuning of the wavelength) and the photo-sensor technology. These will be used to develop an improved communication module, in response to the growing demand of robotic solutions for the marine environment that features a high speed communication system at short ranges, where low power, low complexity and small dimensions are pretended. Using high brightness blue, cyan and green LED based transmitters and a blue/green enhanced photo-diode based receivers, the main goal is to achieve, with an abundant number of tests in all sort of conditions and environments, data transmission rates up to 1 Mbps over 5 meters with the capability of transmitting in sea and river waters. The final step consists of the creation of an operational platform, to test multiple combinations of Tx/Rx configurations and relative orientations, and, using the information obtained through the tests, tune up the system, so its maximum efficiency can be reached, matching up all the delineated objectives. iii iv Acknowledgments Começo este capítulo de Agradecimentos por dizer que apesar de ter feito questão de escrever esta dissertação em Inglês, porque achei que seria muito mais interessante um trabalho científico que demarca o final do percurso académico na faculdade, ser legível e ter visibilidade a um público mais abrangente nacional e internacional, escrevo estas palavras de agradecimento e de reconhecimento no bom e velho Português, pois estas palavras são dedicadas a algumas pessoas que me marcaram mais e que me ajudaram ao longo deste percurso. Apesar de aparentemente mais leigas, estas palavras tal como estas pessoas, para mim são as referências mais especiais. Após esta introdução, começa propriamente os ditos agradecimentos, e queria começar por agradecer a quem originou este tema e este trabalho e também quem o apoiou, por isso aos meus orientadores Professor Nuno Cruz e Dr. Luís Pessoa. Obrigado por terem colocado este desafio, e obrigado por toda a disponibilidade, conselhos e troca de experiências. Não teria conseguido assimilar todo o trabalho sem a vossa ajuda. Ao Professor Nuno Cruz que desde o início, foi um dos fatores mais preponderantes na escolha desta dissertação, porque pela minha experiência no meu percurso académico, assumi de imediato que com o Professor iria conseguia trabalhar de forma mais autónoma, mas se precisasse de qualquer tipo de apoio, sabia que podia contar sempre com o Professor. Ao Luís Pessoa, muito obrigado não só por ter muitas vezes “iluminado” as dúvidas que persistiam, mas também por dar uma “luzes” acerca da componente mais científica do trabalho e por todo o apoio dado sempre que foi preciso arranjar algum tipo de material de laboratório, que nunca conseguia arranjar no departamento da faculdade. De seguida, quero agradecer a toda a minha família, em especial aos meus avós, ao meu pai Alfredo, à minha mãe Florbela, e às minha duas irmãs, por ordem hierárquica, a Mariana e a Camila. Toda ela foi capaz de me dar força sempre que precisei, educação porque sempre precisei, motivação, inspiração, carinho,... basicamente um horror de coisas que nunca caberiam numa dissertação só. Agradeço só em especial a ambos os meus pais, por nunca duvidarem de mim, quando eu próprio o duvidei, e de me terem dado toda a liberdade para fazer tudo o que queria e quando queria, pois confiaram (e confiam) cegamente que farei tomarei sempre a escolha correta. É graças a eles que vivi tudo neste cinco anos. De resto, neste curto tempo para a dissertação, foram só boleias incontáveis até a faculdade a todas as horas, ir buscar a faculdade bem cedo (3h, 4h da manhã) para ainda poder dormir umas horitas, fazer diretas a ajudar a fazer testes num tanque de água na faculdade. Depois de cinco anos a viver ao máximo a vida académica, esta dissertação foi um passeio para eles... Às minhas irmãs, à Mariana quero agradecer por me ter mostrado o que é vida académica. Foste um exemplo e uma fasquia para mim, sendo a primeira a entrar na faculdade e a tentar viver tudo a que tinhas direito. Foi o teu amor ao teu curso e tudo que ele representa, que me fez quer ter aquilo que tu tinhas, e amar com tanta força como tu. À Camila obrigada por estares sempre presente quando precisava de descarregar o meu stress em alguém. A partir de agora, já sabes quem é que tens de superar. Boa sorte. Fico à espera dos meus agradecimentos. Ahhh... também deixo uma palavra à minha prima Catarina, por sempre ficar com aquele brilho nos olhos quando falava do primo que era Engenheiro. v vi À minha namorada Tina, também é outra pessoa que poderia dar numa dissertação. Desde de me ligar mais de 20 vezes de manhã para me acordar, a levar a casa quando a estrada estava complicada, a obrigar a descansar quando não queria mas precisava,... não sei como consegues, mas tens uma capacidade de me aturar que não acho normal. Não há ninguém como tu... Obrigado por toda a motivação, amor, carinho, paciência que me deste ao longo destes anos, e obrigado por nunca desistires de mim, mesmo quando nada funcionava. Mesmo NADA! Graças a ti que agora sou capaz de pensar em coisas de uma forma diferente, e de vivê-las de uma forma diferente. Por mais palavras que escreva, estas nunca farão jus a quem és nem ao que representas para mim. Deste-me um novo significado diferente ao número 13. Estrategicamente, coloquei este último ponto nos agradecimentos. Deixo este último para todas as amizades que criei neste curtos cinco anos. Todas elas não fizeram parte da minha vida académica. Todas elas FORAM a minha vida académica. Não é possível escrever uma dissertação ou tese sobre elas, é preciso vivê-las para percebê-las. Foi só graças a elas, que consegui aprender tanto, num tão curto espaço de tempo. Com essas amizades, é que fui capaz de criar ainda mais amizades, e sentir como uma pessoa que já viveu mil e uma coisas e que conheceu tudo que havia para conhecer, para no memento a seguir, voltar a por numa situação totalmente nova e diferente. Foi a melhor experiência de vida, que alguma vez ambicionei ter. Não quero ser injusto para ninguém, mas deixo um abraço especial à F.A.N.F.A.R.R.A. de Electro, por me ter dado algo que nunca pedi, mas nunca irei esquecer, a pessoal mais velho que me moldou e fez pensar, pessoal mais novo com quem me diverti a gozar, e a pessoal do meu ano. A juntar a esta legião, aquele abraço especial ao Chuck, Corrector, Rodriguez, Locomotiva e Pintor. Ao Chuck por ser uma das pessoas que me ensinou mais, e um dos gajos mais filho da mãe que eu conheço. Ao Corrector por ser uma das pessoas que sempre acreditou em mim desde novo e com mais espírito, e um dos gajos mais filho da mãe que eu conheço. Ao Rodriguez por estar sempre aberto a falar comigo sempre que precisava de um ombro amigo, e um dos gajos mais filho da mãe que eu conheço. Qualquer um destes três senhores tem uma das caraterísticas ditas. Ao Locomotiva não agradeço só eu, mas também esta dissertação, pois sempre que tive um problema no percurso académico, muito mais durante a realização desta dissertação, era sempre a ti que dirigia e tirava dúvidas. Este trabalho foi possível também graças a ti. E ao Pintor, por ter sido provavelmente quem sofreu mais comigo em cinco anos, mas que também quem pude sempre confiar para o que der, e vier. E também por ser um filho da mãe. Mereces tudo que ganhaste e conquistaste. Serão sempre estas grandes amizades que nunca esquecerei. Dedico esta dissertação a todos vós. Abraço, Bernardo "Maquetista" Silva “Não chores porque acabou. Ri porque aconteceu.” Chuck Norris vii xiv LIST OF TABLES Abbreviations APD Avalanche Photo-Diode AUV Autonomous Underwater Vehicle BER Bit Error Rate BPF Band-Pass Filter bps bits-per-second BPSK Binary Phase-Shift Keying DEEC Department of Electrical and Computer Engineering EM Electromagnetic ESR Equivalent Series Resistance IC Integrated Circuit IR Infrared LED Light Emitting Diode LD Laser Diode LOS Line of Sight MCPCB Metal-Core Printed Circuit Board MOSFET Metal–Oxide–Semiconductor Field-Effect Transistor OP-AMP Operational Amplifier OTS Optical Telemetry System PC Personal Computer PCB Printed Circuit Board PD Photo-Diode PMT Photomultiplier PSK Phase-Shift Keying QAM Quadrature Amplitude Modulation QPSK Quadrature Phase-Shift Keying RF Radio-Frequency RGB Red-Green-Blue ROV Remotely Operated Vehicle Rx Receiver SMD Surface-Mount Device TIA Transimpedance Amplifier TTL Transistor-Transistor Logic Tx Transmitter VGA Variable Gain Amplifier VSF Volume Scattering Phase Function xv Chapter 1 Introduction 1.1 Context Technology has reached a point where much importance has been given to the creation of remote and autonomous robots, in order to enhance the preservation and comfort of human life, to contribute to the investigation and evolution of technology, among others. With such recent developments, the tasks at hand have been increasing in complexity and precision. In these case scenarios, when it is too dangerous, risky or complicated for a human to perform a task, these are mainly entrusted to a machine. To instruct this machinery and robots, it is necessary to establish communication between the operator and the machine. This communication is also valuable since it can transmit information that has been extracted from sensors. Besides establishing communication via a physical cable, which has some limitations since it is a wire communication, wireless radio waves have been one of the most commonly used solutions in the aerial and terrestrial environments. However, trying to implement communications in underwater conditions raises completely new problems, since radio and other electromagnetic (EM) waves are highly attenuated underwater. Since wireless communications are a big advantage for this kind of environments, alternatives such as the use of acoustic modules became the standard option to communicate underwater. The inherent problems of this solution, despite the low attenuation and long distance range, reveals limitations in bandwidth and presents a high latency and high power consumption. These problems have been the main barrier that restrains the use of underwater robots in a large scale, making underwater explorations an even more challenging and difficult task. Due to such limitations, alternative methods that allow underwater robots (like autonomous underwater vehicles (AUVs) and remotely operated vehicles (ROVs)) to transmit information have been an emerging thematic in the past few years and still represents many challenges, in order to achieve a more reliable solution. 1 2Introduction 1.2 Motivation The necessity of underwater wireless communications results from the mobility requirements of robotic-based activities [12]. This is the main reason why physical connectors are not a viable solution to establish communication underwater [13]. In submarine operations, where a robotic vehicle has to perform complex movements, being dependent on a physical connector could limit the range and the precision of its moves. Besides, it represents an additional risk if the cables, for some reason, end up entangled, which could result in loss of valuable equipment and subjects of study. Even though acoustical-based systems are the default choice and are capable of establishing wireless communications underwater at a long distance range, the slow data-rates and high latency does not allow a real-time control and data harvesting. For these reasons, maintenance and surveillance operations based on acoustics are very limited. There are many areas where an improved performance of the underwater wireless communications may have a big impact, such as in oil, gas and mining explorations, under the sea constructions, the study on the impact of deep-sea raw material resources exploitation, and the measurement and monitoring of physical, biological and bio and geochemical parameters (e.g. climate change) [12,14,15,16,2]. In order to surpass many of the obstacles presented in the underwater wireless communication, an optical light emitting diode (LED) based solution is suggested, since it presents the potential to overcome many limitations shown by other methods. 1.3 Objectives The proposed solution is to create a short range underwater optical wireless communication system based on LED technology, in order to increase the efficiency in underwater communications. Ideally, the solution has to be a viable option to be implemented in an operational platform, such as an AUV. The experiments carried out along the whole period of work will be tested in wide and varied test scenarios, such as in a controlled laboratory scenario, simulating communications inside a test pool, and in real life scenarios, trying to communicate in two distinguished environments: the ocean and the river. To reach this goal the following objectives will be considered: • Assembly of a LED and photo-sensor based test system; • Creation of an underwater wireless link that enables the transmission of information; • Create a system that is able of test different transmission parameters; • Create a real-life solution, that could be implemented in a robotic platform; • Test of the system in the laboratory; • Test of the system in the laboratory pool; and • Test of the system in real life environments (Ocean and River). Chapter 2 State of the Art In this chapter, it is presented all the major scientific work developed about underwater communication. It starts by analyzing the many methods of underwater communication, considering all the pros and cons for each presented approach. Subsequently, it will focus on developing the thematic of underwater optical wireless communication, taking in special consideration the optical properties of the various water types that are used as transmission channels, and it will be studied all the related works and solutions that have been developed until this moment, academic and commercial. In the end, there will be a sum up about all the chapter. 2.1 Underwater Communication Methods As analyzed by [12,17], underwater communications are traditionally made with a physical connector, made by cables or fiber-optics. These represent some advantages, since they are capable of transport electrical power and offer a high speed and reliable communication. However, their use can limit the range and maneuverability of underwater operations, and the presence of a long and heavy cable and the associated hydrodynamic drag, increases the risks of an operation. For these reasons, there is a high interest on the investigation of alternative ways to transmit information underwater, without being dependent on a cable connection. This is why it has been given a big importance to the wireless communication techniques. Next it is presented the most common methods of underwater wireless communication, referring their advantages but also the problems they face [16]. 2.1.1 Acoustics Acoustics has for a long time been the standard solution for submarine missions because of the fact that sound propagates well underwater, it enables the transmission of the information without been dependent on an "umbilical" cable, which gives to the underwater vehicles the necessary mobility to perform complex movements freely. Adding to this, its relative low signal attenuation and long range allows the transmission of information in the order of the kilometers. Despite all these 3 4State of the Art advantages, this solution also carries some problems that begin when the speed of transmission is important[16]. As referenced by [18], this method has its disadvantages, because it presents a low propagation speed (high latency). Moreover, it has a low bandwidth with a maximum data-rate in the order of the tens of thousands bits-per-second (bps), making the acoustics communication a slow option for transmission. This method also presents other problems, such as its susceptibility to multipath propagation, dispersion, frequency fading and ambient noise (specially in shallow waters). The next table 2.1 contain some available acoustical solutions and their main characteristics. Table 2.1: Comparison table of some conventional underwater acoustic solutions [6,7,8,9,10, 11]. Model Distance (m) Data-Rate (kbps) Power (W) Weight in Air (kg) HERMES 120 150 32 — Sonardyne AvTrak6 Type 8220 3000 9.0 50 5.1 AquaSeNT - AM-OFDM-P1 4000 9.0 20 7.6 AquaSeNT - AM-OFDM-S1 4000 1.6 20 2.4 Benthos - ATM-916 6000 15.36 15.75 4.5 Benthos - ATM-926 6000 15.36 15.75 4.1 Benthos - ATM-966 6000 15.36 24.5 6.8 EvoLogics - S2CR 48/78 1000 31.2 60 6.5 EvoLogics - S2CR 18/34 3500 13.9 80 6.5 EvoLogics - S2CR 12/24 6000 9.2 40 7.78 EvoLogics - S2CR 7/17 8000 6.9 80 7.78 LinkQuest Inc. - UWM1000 350 19.2 2 4.2 LinkQuest Inc. - UWM2200 2000 38.4 6 3.0 LinkQuest Inc. - UWM3000 5000 5 12 4.1 LinkQuest Inc. - UWM4000 4000 9.6 7 7.6 LinkQuest Inc. - UWM10000 10000 5 40 21.0 2.1.2 Radio-Frequency On the other hand, there are the EM waves such as radio, a very popular choice to transmit information via wireless in terrestrial communications. As stated by [15,16], EM waves such as radio-frequency (RF) compared to acoustic waves, are much faster and have a higher bandwidth, in this terrestrial transmissions. In addition, radio produces a high throughput compared to sound waves, with speed levels surrounding the hundreds of kbps up to ten meters, and achieving tens of Mbps at the range of one meter. Again, comparing both power consumptions, the radio reveals to be more energetically efficient than acoustic technology, needing less power. Even with these characteristics, there are many limiting factors when using RF in water. The EM waves have different behavior in freshwater and seawater. 2.1 Underwater Communication Methods 5 According to [16], freshwater is considered to the EM as a low-loss medium, and its propagation speed c, and the absorption coefficient αcan be represented as: c≈1 √εµ (2.1) α≈σ 2rµ ε(2.2) In the equations εis considered the dielectric permittivity, µis the magnetic permeability and σis the electric conductivity. By observing this expressions, it can be concluded that the absorption for freshwater is independent of the frequency of the signal, and propagation speeds of EM are smaller than light, but still faster than sound [16]. The only inconvenient for this solution, in freshwater, is the antenna size that is required in order to obtain a large bandwidth. To EM waves, the seawater is considerate as a high-loss medium, and its propagation speed c, and the absorption coefficient αcan be represented as: c≈s4πf µσ (2.3) α≈pπfµσ (2.4) As it can be concluded by the equations 2.3 and 2.4 both are dependent of the frequency of the signal f, meaning that if a high frequency is used, a high speed is obtainable, but the absorption of EM waves is also high. It is possible to conclude that communication through EM waves is limited in freshwater by the size of the antenna needed, and has a critical problem because of the high attenuation in seawater. In figure 2.1 it is shown the different RF waves in different types of water, comparing their performance with the light and sound waves. Although it is known that RF is not the ideal method to communicate, there are some commercial possibilities. WFS Technologies Ltd commercializes a underwater wireless RF system for sub-sea exploration, the Seatooth® , a system series claimed to have a model capable of achieving 156 kbps over 7 meters [19]. Although there are some situations where this system could be used, it still requires a high power consumption, around 15 Watts, and an antenna with large dimensions (1 meter squarial antenna). 2.1.3 Optical An optical solution is another transmission method still in development that shares some advantages of the RF, but not all of its disadvantages. The major advantage comparing to RF communications is the possibility of maintaining the same conditions underwater without considerable losses. As RF, this solution can obtain in a underwater environment, a high throughput in the 6State of the Art 103104105106107108109 103 104 105 106 107 108 109 Frequency (Hz) Propagation speed (m/s) light σ=.001 σ=.05 σ=4 acoustic Light Sea Fresh water Distilled water Acoustic Figure 2.1: RF propagation speeds in different types of water. order of several Mbps to Gbps, a high propagation speed (low latency) and achieve short and midrange link distances in the order of 100-200m [14,1,17]. In [14] also referred that underwater optical communication can benefit meaningfully from the progress made in the terrestrial optical wireless communication, becoming an area in constant evolution. Like any system, optical solutions have some inherent problems. Although optical signal have a low attenuation, it becomes relevant in long distances [14]. For that reason, acoustic vastly outperform the optical channel when comes up to distance range. Other relevant disadvantages, in addition to the absorption, optical scattering is significant and the ambient light can also interfere with communications [16]. Another complex problem associated to this method is the dependence on a clear way for the light to travel, a line of sight (LOS) [14]. In the beginning, without a clear LOS it was impossible to establish communication, since the light cannot reach the destination. For that reason, some alternative methods were created to prevent the loss of the LOS, and to create optical links. Figure 2.2 shows some of these methods. With these variant high bandwidth links options, it is possible to create sophisticated and collaborative paths for planning and observation [14,1]. In figures 2.2a and 2.2b is represented a communication channel based on a clear and simple LOS, where communication exists by emitting light in a narrow beam (similar to laser) and in a diffuse beam, respectively. Figure 2.2c shows a retro-reflector link type, used when limited dual communication is necessary, and the receiver is awfully low powered to perform as a full transceiver (such as underwater sensor nodes). In this configuration the light of the transmitter is 2.2 Underwater Optical Wireless Communication 7 (a) LOS (b) non-direct LOS (c) retro-reflector (d) non-LOS between a transmitter and a receiver Figure 2.2: Different underwater optical wireless link configuration [1]. reflected back from a modulated retro-reflector on the receiver that, in the meantime, is encoded with information. Another link represented in 2.2d is a non-LOS transmission where the transmitter and the receiver position themselves diagonally upwards to transmit and receive the light reflected in the surface of the water. 2.2 Underwater Optical Wireless Communication Seeing the potential of the optic channel in the underwater environment, the attention of this work is focused on study, with some detail, the optical communications. It starts by analyzing the optical properties of the water channel to obtain a detailed overview of the behavior of light in water, and will end in reviewing some major achievements in the field of knowledge. 2.2.1 Optical Properties of Water As concluded by [2], the behavior of light in water is a very complex issue. There are extensive works regarding this thematic, since it has many interests in many fields of science. Since this work is mostly focused in the conception and implementation of a practical solution, this thematic will be briefly referenced, taking into account the work done by [2]. Assuming a differential portion of water (represented by a volume with three dimensions), with a width of ∆rand with an small height and length, it is assumed that the amount of light entering the volume equals the amount of light leaving it, less any loss that occurred on the volume in the form of absorption and/or scattering. This phenomenon is represented in figure 2.3. The incident power is represented by Φi, the absorbed power by Φa, the scattered power by Φsand the transmitted power is represented by Φt. These values are dependent of the wavelength λ. 14 State of the Art external 10 Mbps Ethernet interface. The communication has a data rate of 4Mbps at a distance of up to 50 m in a pool. In a laboratory setting transmitting through air it is 32 Mbps. In both [35,36] is described an optical telemetry system (OTS) with which it downloaded 20 MB of data over a 5 Mbps link at a range of 80 meters. It reports communications at a data rate of 10 Mpps at a distance of 108 meters, 5 Mbps at distances of up to 128 meters, and 1 Mbps at distances of up to 138 meters. In [37] the author tries a different approach by designing a more efficient transmitter and receiver. With an array of 7 LEDs to the transmitter, and 7 PD for the receiver, displayed in the form of a truncated hexagonal pyramid structure with a special lens, the transmitted light gets more focused and the received light can be captured more efficiently, obtaining a quasi-omnidirectional transmitter and receiver. Document [38] gives a continuation work on the AquaOptical project [31,34], which uses the developed hardware, and its upgrades resulting in the AquaOptical II. Also in this paper, by using the AquaOptical II, a base signal strength model to end-to-end communications is projected. It is a great literature that reviews all the recent works done in the area. In [39], using a laboratory set-up, it was investigated the impact of background noise on the receiver in shallow waters. It was used a 2.5 meters test pool to study the effect. The two transmitters used LEDs, but the receiver was constituted by an APD and a PD. For this experiment it was obtained, using Manchester Coding 6.25 Mbps, 12.5 Mbps using NRZ 8b/10b code and 58 Mbps with DMT. It was estimated that maximum distance was around the 60 meters in clear water and 10 meters in murky. Now, it will be shown some of the commercial solution available, for underwater optical wireless communication. As referred before, the Ambalux Corporation [28] manufactures modems that establish communication point-to-point, capable of achieving 10 Mbps and up to 40 meters range. The interface is 10BaseT Ethernet compatible allowing TPC/IP and UDP connections. The transmitter consumes up to 36 W and the receiver 7.2 W. As a product from the investigation in [26,27], the devices created by the author were posteriorly commercialized by the company Penguin Automated Systems, Inc. [40]. Same characteristics of the developed device are considerate. Another commercialized product is the Neptune, produced by SA Photonics [41], that is claimed that it can achieve 250 Mbps over 200 meters, with a power consumption larger than 5 W. From the work of [30], the company Sonardyne International Ltd. adopted the system along with an LED-based model for near-surface communications. The product is the BlueComm [42] and it claims that is capable to achieve up to 20 Mbps of data rate and up to 200 meters of range. 2.3 Summary 15 2.3 Summary This chapter presented the research work that has been done in underwater communications, beginning with a reference on the underwater optical wireless communications. It was possible to conclude that, in all the underwater communication methods, wireless communications are a must-use solution and the optical communications are a new and promising method to make communication links underwater. After all, this is an area that is still in investigation, for its incredible benefits, but also its complex problems. When studying the optical proprieties of water, it was clear that different environments, where water visibility is not the ideal, will obviously limit the propagation of light, not only affecting the speed of the connection, but also influencing the wavelength needed for the attenuation to remain low. In the review of other works, some knowledge was obtained about the architecture of the system, their main components and what are the best materials to obtain the best, expected results, and common errors and difficulties found in this area of work. With this acquired information, a solid and studied design system can now be done. 16 State of the Art Chapter 3 Implementation In this chapter it is presented the implementation and development of the underwater communication module prototype. Initially, it is shown the proposed methodology for the implementation of the system, showing the design and its constraints, the different layers that form the proposed module, specifying the component selection, dimensioning and its design, ending with the display of the produced prototype. 3.1 System Design In the primary system design, shown in figure 3.1, it was intended that a pack of data would be present in a computer and, by modulating this data into a light signal, it would be sent by a transmitter (Tx), then this light signal would be captured by the receiver (Rx), and the end terminal would demodulate and show the data. Figure 3.1: Primary system design overview. After some development, the more detailed redesign solution in figure 3.2 was achieved. It is similar to the initial one, where a computer terminal tries to communicate with a peer, using the optical channel, but it is intended that the computer communicates with a micro-controller that connects to the computer (PC) and the Tx/Rx via serial communication. Also, after some considerations, the design would be physically able of establish bidirectional communications with two distinct colors. 17 18 Implementation Figure 3.2: Detailed system design. 3.1.1 Design Constraints The reviews that were made show that underwater optical wireless communication is a viable solution to be implemented in an operational platform, such as an AUV, for communication purposes. So, if it is pretended to implement a module in an AUV, some standard necessities must be taken into account. Because an AUV is an autonomous vehicle that operates in an underwater environment, the device that will be attached to it, in addition to the full protection of the electronics from water, which has to be assured, must be as efficient as possible, mainly in terms of energy, and also have a small physical size. For that reason, these initial simple requirements are noted: 1. Maximum range - up to 5 meters; 2. High data-rate - at 1Mbps; 3. Low power - below the 5W; 4. Small size - relatively smaller than 15 centimeter long; 5. Low weight - made with light materials; and 6. Low complexity - simple and modular. There are other requirements, such as a low-cost solution, that could be considered, but for the purpose of this dissertation, these will be somewhat overlooked, making the choice based on common sense. 3.2 Optical Transmitter - Tx With the optical transmitter, it is pretended to design a circuit that is capable to output as much light as possible. This operation will be done through a driver that feeds the chosen light source to obtain the maximum performance of it, to be able to achieve a wide range. The data rate will also be limited by the time response of the components that the transmitter is made of, so a studied 3.2 Optical Transmitter - Tx 19 selection for the module components is advised. Figure 3.3 gives a simple view of the optimal solution. Figure 3.3: Simple Tx system design. In the following subsections, a study and some functional tests will be performed to better choose the best light source and adequate driver for it. 3.2.1 Light Source Since one of the objectives of this dissertation was the creation of a LED-based system, primarily it was necessary to decide a light source for the transmitter. Two possibilities for a general LEDbased light engine were considered: a diffuse light source, available as a LED or a concentrated beam, as a laser, in this case a laser diode (LD). Between the two choices, the diffuse LED light source was considered as a more appealing solution. Although using a LD would be efficient since it has a narrower EM spectrum, higher light intensity at long distances and faster switching speed, the LD presents a complex problem of misalignment at long distances, and a diffuse light engine is able to adjust its light beam using an optic, making it a much more versatile solution. Other issues regarding the LD were also taken into account, such as a higher power consumption and a much greater cost. The light source for the Tx ended up to be high-power LEDs, where characteristics such as small size, high-brightness, narrow EM spectrum and other optical features were desirable. Also, the existence of the three colors (blue at 470 nm, cyan at 505 nm and green at 530 nm) originally pretended would be an advantage. Figure 3.4: A MCPCB with green LUXEON Rebel LEDs. For these reasons, for the initial testing and understanding of the system, it was only used LUXEON Rebel green LEDs, shown in figure 3.4, since it was available an example of these LEDs (provided by INESC TEC) and these met up the principal necessities described before. Table 3.1 shows some of the most relevant information about the LUXEON Rebel LEDs, distributed according to their light color. Since the seven LEDs are connected in series, to obtain the maximum output of light, but still respecting the typical datasheet values, it is necessary to supply about 20.3V at 700mA. But since one of the objectives is to be implemented in an operational 20 Implementation base or an AUV, the LEDs supply will be set to more typical values of 24V at 700mA (which corresponds to two batteries of 12V), with the remaining voltage been dissipated by a dimensioned resistor, indicated for this operation. Because the system is supposed to be controlled by a micro-controller, the switching signal respects the transistor-transistor logic (TTL) levels. So it was necessary to dimension a LED driver that was able to switch the LEDs at the necessary frequencies and, at the same time, supplying the necessary current and voltage they require. Table 3.1: Electrical and spectral characteristics of the LUXEON Rebel LEDs. Component Typical Voltage (V) Typical Drive Current (mA) Typical Wavelength (nm) Green: LXML - PM01 2.9 700 530 Cyan: LXML - PE01 2.9 700 505 Blue: LXML - PB01 2.95 700 470 As referred, although with the sample there was no optical lens available for testing, for the final prototype it is intended to use some optical device to increase and optimize the light output of the 7 LEDs. 3.2.2 Light Source Driver As explained previously, it is expected that the signal respects the TTL, meaning that it will vary between 0V and 5V (depending if the bit value is ’0’ or ’1’, respectively). Figure 3.5: Primary driver design. There are some driver solutions available in the market that can drive these kind of high-brightness LEDs, just like the previously used, but for this application in this dissertation, the LEDs must be switched in the highest frequency rate as possible, maintaining a maximum throughput power at the same time. Although commercially there are some solutions to drive this sort of LEDs with high power outputs pretended or with relative high frequency switches, finding a solution that can perform both operations at the same time is another issue. The solutions presented in the market are too complex, large in size or inefficient. For those reasons, it was considered more beneficial to design an appropriate LED driver for this application, which is capable of meeting the requisites, while maintaining a simple complexity. One of the initial ideas was to control the high current supply of the LEDs through a transistor. There are many types of transistors that could perform the pretended function (BJT, JFET, IGBT and more), but for this kind of application, precisely because of the high current and relatively high voltage to control, a MOSFET was considered a more logical choice, not only because of its low cost, with a small size and a fast response time but also because a micro-controller, by default, 3.2 Optical Transmitter - Tx 21 is not able to produce a high current, and a MOSFET is able to operate using a signal with low current, needing just to be controlled mainly by the voltage applied to the gate. The idea is that a command signal is applied to the gate of the MOSFET. If the signal is at +5V, the LEDs turn on and when the signal is at 0V the LEDs turn off. The proposed LED driver, as shown in the circuit 3.5, is consisted of a MOSFET, where the drive signal is applied to the gate, the 24V and the LEDs are connected to the drain and the ground to the source. After some primary considerations, an additional resistor was placed between the 24V and the LEDs to protect the LEDs and to dissipate the residual power that typically remain from the LEDs. The value of the LEDs load resistance RLis obtained: VL=RLIf⇔RL=VL If⇔RL=Vsupply −Vf If⇒RL=24−(2.9×7) 700×10−3≈5.3Ω(3.1) And RLmust be able to dissipate a power of: PL=If2RL⇒PL= (700×10−3)2×5.3≈2.6W(3.2) The value of the current If, used in the previous calculation, was established by the datasheet in order to ensure that the device could sustain the worst case scenario: a continuous transmission at full power. 3.2.3 Transmitter Tests For these first tests, to analyze the behavior of the LEDs and check if the concept would work, it was used some power and switch MOSFETs available in the DEEC. For this first stage tests, the input signal is created by a function generator that outputs a square wave that operates between the +5V and 0V. The objective was to see the delays generated by the transistors and to study their behavior at different signal frequencies. To achieve this goal, a test setup, as shown in figure 3.6, was assembled. Figure 3.6: The test setup for the MOSFET testing. Initially, it was assumed that MOSFETs for power applications, although able to switch greater power loads, they would only be capable of producing slower response times. By using a sample founded in the laboratory, in this case the P30NF10, it is visible in the print 3.7, the influence of the slow response time in a transmission. 22 Implementation Figure 3.7: P30NF10 response time test: F = 1MHz, T = 250ns. So, since the pretended operation frequencies demand a fast switching speed, a more appropriate MOSFET for these operations is needed. The challenge behind this type of switching MOSFET, is to find a sample that can bear the supply voltage and drive current. In order to test the concept the BS170, which is a more appropriate MOSFET for this application, was used. However in the future, a more considered choice will be made, since this MOSFET is not capable of supporting the necessary 24V of voltage supply and 1A of drive current. In figure 3.8, it is possible to see that the BS170 is capable of following the transmission signal, with a considerable small delay along the broadcast. Figure 3.8: BS170 response time test: F = 1MHz, T = 250ns. Based in the work done by [15], it was also the possibility of using a MOSFET driver was also explored, in order to increase the current of the transmission signal inserted in the MOSFET’s gate. This is a desirable amplification because if a higher current is injected in the MOSFET’s gate, the intrinsic capacities of the MOSFET would be saturated, which could decrease the commutation delay produced by it. In figure 3.9, it is shown a possible circuit, that would work based on a MOSFET driver. The complexity of this layout is higher, comparing with the previous scheme, but still relatively simple. 3.3 Optical Receiver - Rx 23 Figure 3.9: Setup scheme for Tx with MOSFET driver. Despite of its advantages this option was overlooked, since the delay produced by this driver, in the output, was superior in comparison to time gain by charging the MOSFET capacitance. 3.3 Optical Receiver - Rx Just like the optical transmitter, it is also necessary to design another important component of the system, the optical receiver. Basically, as shown in 3.10, this part has to be capable of capturing the emitted light signal, transmitted by the Tx, through a photo-receiver. Then the light signal sent, passes through some processing to be able to retrieve the original message. This block is one of the most complex and important parts of the system, because not only the response time of the layer is important, again, not to limit the data rate, but also the receiver must be able to receive the light of the signal ignoring the ambiance light that adds noise to the transmitted signal. Figure 3.10: Simple Rx system design. In this section it will be described the selection of the photo-receiver, the many stages of the acquisition and processing of the received signal, and testing each element of it, finalizing with a test to the all receiver layer. 30 Implementation 3.3.3.2 Transimpedance Amplifier So a more simple and direct setup was created to be able to acquire the transmission as shown in figure 3.17, by using the already tested Tx. Although this setup is more limited, since it cannot test the range with precision, it permits to better testing the TIA time response. Figure 3.17: The final test setup for the TIA testing. This simple test is performed by placing the Tx and the "Rx" in a fixed position, where a plausible transmission might occur, and test the output of the TIA with some advisable values for the resistors and capacitors, at different transmission frequencies of 100kHz, 500kHz and 1MHz. The values of the resistors used were 10kΩand 100kΩ, and the capacitors values were 1.8pF, 2.2pF, 2.7pF and 3.3pF. In a first approach, the output gain was studied by testing the TIA using only the feedback resistor. As it can be seen in the prints 3.18, the gain of the TIA, is ten times greater with the 100kΩresistor, as it could be predicted. (a) Rf=10kΩ. (b) Rf=100kΩ. Figure 3.18: Feedback resistor comparison: F=100kHz, without CpF. Although a higher gain can be seen as an asset, with a bigger feedback resistor, the greater is the delay generated by it, since the RC time constant τ(expressed in seconds) is related to the resistance of a circuit: τ=RC (3.10) 3.3 Optical Receiver - Rx 31 In addition, as it can be seen in the prints 3.19a, at a frequency of 1MHz, the output signal, even without a capacitor that could extend the delay, the signal is highly affected. Although that at a first sight at 3.18b, seems that it does not need to be filtered, if a capacitor is placed, as figure 3.19b shows, the signal becomes almost impossible to be reconstructed, making it impracticable. (a) Without Cf. (b) Cf=1.8pF. Figure 3.19: Time response of the TIA with high gain: F=1MHz,Rf=100kΩ. Because of this visible delay, it was assumed that using a feedback resistor of 10kΩwould end up being the best option. Even thought, the output signal of the TIA at 1MHz requires some filtering, as it can be concluded with figure 3.20. Figure 3.20: Output signal of TIA: F=1MHz,Rf=10kΩ, without Cf. The next step for the TIA testing only the feedback resistor of value 10kΩwas used, varying only the values of the feedback capacitors. Never forgetting that the TIA delay is always dependent of equation 3.10, which means the larger the capacitor, the higher the delay. In the lower frequencies tested, the delay generated by the capacitors is not entirely perceptible, the only significant difference is in the output noise of the signal. The larger the capacitance, the less noise remains in the output signal, as it can be assumed by the prints 3.21. 32 Implementation (a) Cf=1.8p f . (b) Cf=2.2pF. (c) Cf=2.7pF. (d) Cf=3.3pF. Figure 3.21: Time response of the TIA: F=100kHz,Rf=10kΩ. By critically analyze and compare with 3.18a, the results in 3.21, at 100kHz, it was considered the best result were the 2.2pF and 2.7pF capacitances, the figures 3.21b and 3.21c respectively. Figure 3.22: Output signal of TIA: F=500kHz,Rf=10kΩ, without Cf. At the transmission frequency of 500kHz in figure 3.22, it is noticeable the signal slowing down and the noise still remaining, and with the capacitors the delay begins to increase, although 3.3 Optical Receiver - Rx 33 there are not very significant variations between capacitances, being the noise remove factor more preponderant, as shown in figure 3.23. (a) Cf=1.8p f . (b) Cf=2.2pF. (c) Cf=2.7pF. (d) Cf=3.3pF. Figure 3.23: Time response of the TIA: F=500kHz,Rf=10kΩ. Again by analyzing the figure 3.23, the responses of 3.23b and 3.23c seem to be the better results. Finally, the TIA is tested at 1MHz, where an improvement must be done to the output signal represented in figure 3.20. In the prints presented in 3.24 are displayed the results from the various capacitors. 34 Implementation (a) Cf=1.8p f . (b) Cf=2.2pF. (c) Cf=2.7pF. (d) Cf=3.3pF. Figure 3.24: Time response of the TIA: F=1MHz,Rf=10kΩ. At this frequency, the response times are at their lowest obtaining the highest delay. In the print 3.24d, although it might show a more filtered output, the delay displayed in the fall time is too high to be considerate a viable choice. Taking into account all the previous tests and this last one, it seems that the print 3.24b, with a capacitor value of 2.2pF is the most viable choice for the feedback capacitor of the TIA. With these tests, it is possible now to present the recommended features for the TIA, as shown in table 3.2. Table 3.2: TIA block specifications. Transimpedance Amplifier Component THS4631 Power Supply (VS+,VS−)+12V, -12V Feedback Resistor (Rf)10kΩ Feedback Capacitor (Cf)2.2pF 3.3.3.3 Bandpass Filter The next block of the processing is the BPF, and as explained previously the objective of this test is to analyze the behavior of the output signal by changing the frequency of the input transmission 3.3 Optical Receiver - Rx 35 and detect functional results, as when the magnitude of the signal declines and if the phase shifts. Just as argued in the past subsection 3.3.2.2, it was believed that using a single supplied amplifier for the filter, would be more beneficial, since it simplifies the supply necessities and limit the output for the desired levels, but in very first tests, the output signal was not remotely the expected, since the single supply distorted the filtered output, where an input sinusoidal wave did not have the desired magnitude output, and phase shifts were not visible. Luckily, the chosen amplifier was able to withstand single a dual supplies of +5V and -5V, which would allow to obtain the desire output. For that reason, a setup similar to the figure 3.25 was arranged. Figure 3.25: Test setup for the BPF testing. Due to the absence of components with the proposed value to obtain the precise cutoff frequencies, for these tests it was used components with similar value to approximate as close as possible of the ideal values. The values of the components for these tests and respective cutoff frequencies are:                R1=100Ω R2=1kΩ C1=1µF C2=18pF ⇒   fc1≈1.6kHz fc2≈8.8MHz (3.11) The first frequency tested was at 1MHz, because it is one of the pretended transmission frequency rates and also has been used as a base test frequency all along the work. The resulting output wave is shown in figure 3.26 36 Implementation Figure 3.26: BPF output at 1MHz As it may be assumed, at a first sight the output signal has the pretended magnitude and right phase, since it is an inverting filter, but is difficult to see the delay caused by it. Because of that problem, the prints of the oscilloscope were inverted to enable a better time response perception. Figure 3.27: BPF at 1MHz with inverted output. Now in figure 3.27 it is possible to clearly see that the output is suffering from a natural delay, since this frequency is closer to the region of high cutoff frequency. It is confirmed that the closer it gets from this region the greater the delay from the filter, as shown in figure 3.28, where the input is a 5MHz signal. As also can be seen the magnitude of the signal is affected as well. 3.3 Optical Receiver - Rx 37 Figure 3.28: BPF inverted output at 5MHz But eventually, if the signal reaches or surpasses the cutoff limit region, the magnitude of the signal is highly attenuated and the phase is completely shifted, as it is possible to see in figure 3.29, where the input signal is filtrated. Figure 3.29: BPF inverted output at 10MHz Then, if a signal approximates the high frequency cutoff region suffers from a delay, contrary, if a signal dislocates to the low frequency cutoff region the output is in advanced. The figure 3.30 shows exactly this phenomenon, where the output trace is ahead of the input trace. 38 Implementation Figure 3.30: BPF inverted output at 10kHz And just as at the high frequency cutoff region, at 1.6kHz the output signal suffers from high attenuation in the magnitude, and gets in advance of the input signal, as can be concluded in figure 3.31. Figure 3.31: BPF inverted output at 1.6kHz With these tests it was possible to confirm that it works like it should and it was able to analyze and predict the behavior of the filter in the range of frequencies. The possible specifications of the BPF are shown in table 3.3. Table 3.3: BPF block specifications. Bandpass Filter Component AD8041 Power Supply (VS+,VS−)+5V, -5V High Pass Components (R1,C1)100Ω,1.2µF Low Pass Components (R2,C2)1kΩ,2.2pF 3.3.3.4 Comparator For this final block, to test the comparator, a very simple setup, as shown in figure 3.32, was assembled to test its response. 3.4 Physical Casing 39 Figure 3.32: Test setup for the comparator testing. Because the threshold was fixed as half of the supplied voltage, the resistors Rd1and Rd2must be equal. Also the resistor Rpworks as a pull-down resistor that is responsible to speed up the comparator time response. (a) T = 250ns. (b) T = 100ns. Figure 3.33: Time response of the comparator at 1MHz. As shown in the images of 3.33, the time response of the comparator is more than satisfactory, the only detail to tune in the prototype is the threshold. Table 3.4: Comparator block specifications. Comparator Component AD8561 Power Supply (VS+,VS−)+5V, 0V Pull-Down Resistor (Rp)10kΩ Voltage Divider (Rd1,Rd2)1kΩ,1kΩ 3.4 Physical Casing As referred in the beginning of this dissertation, the main goal is to create a communication module that is able to communicate in an underwater environment. In order to achieve this it was necessary to create a waterproof casing to place all the electronic components inside. Additionally, it must 46 Implementation Also some additional decoupling capacitor were included in all the integrated circuits (ICs), such as in the regulators, according to their datasheet, in order to reduce their noise. For each it was used a pair of ceramic and tantalum capacitors with a lower equivalent series resistance (ESR) than the typical electrolytic capacitor, to boost the decoupling performance, and to ease the overall PCB layout. Finally the last change made to the circuitry was to include a MOSFET driver. Despite previously this was a discarded option, it was thought that with this driver a better isolation of the ground from the MOSFET, and a smaller distortion of the Tx signal was possible. With a new market search it was found the UCC27517 that looked like a viable option, since it is suitable for the use with micro-controllers and is compatible with the TTL, working with a logic level signal of 3.3V or 5V, and as a fast response time (rounding about the 20ns) that does not induce a significant delay to the transmitted signal. All the values of the gate resistor (represented by Rgate) where the recommended from the datasheet. In figure 3.41 it is shown the evolution and the differences of the first prototype and the final version. Figure 3.41: The primary Tx prototype on the left and the final Tx PCB on the right. As it will be refereed later on this work, it was possible to achieve the possibility of have only the circuitry working with a supply of -5V and +5V, which resulted in removing the +12V regulator from the Tx layer, improving the system efficiency and simplicity. The negative voltage converter was relocated to the Tx layer. In the appendix Ait is possible to examine more precisely the PCB schematic and its layout. 3.5.4 Rx Layer Finally, in the Rx layer is where the most complex operation of receiving the light message, is processed. Because of that, this layer has more components in it, to be able to perform the necessary operations. Just like the Tx layer, the Tx required more than one version to achieve an acceptable prototype result. In the first approach, beginning with the TIA, the choice of components was maintained the same (the THS4631 with the Rf=10kΩand the Cf=2.2pF) as the previous tests, taking into 3.5 Proposed Prototype 47 account this same tests. In the BPF the amplifier used was changed, in order to avoid the necessity of having an additional regulator producing -5V that would only be used by this component. So the LM7171 was chosen to perform the filtration, since is more than able to handle the frequencies needed and as a similar voltage supply like the TIA, the rest of the components stayed the same as tested. For the comparator, the only change is the threshold value Vre f will be regulated with a potentiometer. Also because of the necessity, a voltage regulator was placed in the PCB to output the necessary -12V for the components. The only problem in the assembly of this first version was only the fact that the pretended electrolytic capacitors to be used in the negative converter were not delivered until the end of June, making it impossible to be used according to the initial planning, so some default ceramic capacitors were used. Figure 3.42 shows the final product of the first version of the Rx PCB. Figure 3.42: First assembly of the Rx layer. Still short after the first tests on the PCB, some defects were found in the design. In addition to the inability of acquiring the signal while transmitting due the noise generated by the MOSFET, the normal acquisition was not possible by motives unknown. As well an audible sound noise was generated by the PCB from this Rx layer. It was presumed that a component was damaged or malfunctioning. The first debug test was to cut and separate the function blocks from each other and analyze were the problem was. Meanwhile, it was identified the necessity of a buffer between the TIA and the BPF in order to separate the operations in current of the TIA, and the voltage amplification operation of the BPF, making it a good practice in this circuit. In one of the Rx layers there was also an issue regarding the negative voltage converter, because one of then was working and the other was not. Later this complication was identified due to an unpredictable problem with this device, where in one of them a latch occurred because of residual charges in the capacitors. This problem can be fixed by following the datasheet steps, and use a default diode (LS4148) for this random issue, that can occur. 48 Implementation It was checked that the TIA was working properly, but after cutting the blocks connections, it was diagnosed a problem in the BPF were no signal was been filtered. So a narrower debug test was applied to the BPF, by firstly using only the resistors as a simple Inverting Operational Amplifier and then adding the capacitors to check if the output is the expected. By redoing and testing step-by-step the BPF, it was possible to make it work like it was initially pretended. Subsequently an improvised buffer was included between the TIA and the BPF (using a BUF634, acquired in INESC TEC) and it was possible to see the receive signal with the correct filtration and amplification. But even with this corrections when attempting to connect the BPF output to the comparator the audible noisy sound reappeared, so it was assumed that the problem resided in the comparator or even in its connection. Initially between the BPF and the comparator it was placed a default diode (again the LS4148) in order to avoid a negative input in the comparator, making it basically a safety measure and nothing more. But it was presumed that this diode could represent a complication, and that the potentiometer that control the threshold Vre f of the comparator was improper for this precision application necessity. So even more changes were made to this first version of the PCB, by eliminating the connection diode, connecting a proper trimmer, to get a more precise Vre f and replacing the comparator to exclude any error regarding damaged material or other possible causes. After applying all this changes, for the first time, a successful output signal was possible to be extracted. In the end of the preliminary testing of the PCB, in figure 3.43, shows the necessary modifications made. Figure 3.43: Visible improvised modification in the first version of the Rx layer. On account of the same problem with the PCB traces such as the Tx initial layer, and all the other mistake mentioned above and some more considerations, it was deliberated that an enhanced version of this PCB would be an advantage. 3.5 Proposed Prototype 49 So for new final version between the TIA and the BPF a Buffer was included, the only aspect to watch for is if this buffer is able of operate in the required frequencies and that would delay the signal as less as possible, and the previously used buffer BUF634 was more than able to operate in the required frequencies without delaying the signal significantly. Its simple connections and circuit layout were also attractive traits for its choice. Like was mentioned earlier the trimmer was also an important change, so for this version a SMD 10kΩtrimmer was used to control the threshold with more precision, in order to make transmissions more effective. Since it was given a new opportunity to create a new version of the Rx, a new hypothesis was explored to be able to operate the Rx layer without resorting to the +12V and -12V. To make this hypothesis to be true, it was searched an AMP-OP that would manage the pretended frequencies of operation, with a low noise input voltage and that can work at only +/-5V supply. After an extend market search, it was found an good option, the LMH6624, which has a remarkable performance at a low supply voltage of +5V and -5V, with a low noise input voltage and is able to operate in the intended frequencies. With this modification, a new improvement was made to all the circuitry, making unnecessary the existence of the +12V regulator, simplifying the circuit. With this, as mentioned previously the negative voltage converter was placed in the Tx layer, changing slightly the module pin-out. The remaining modification were similar to the Tx layer, by creating a recommended groundplane in the PCBs and reinforcing the decoupling capacitor in each function block of the layer. For this layer, due to the complex layout it was additionally necessary to make this layer a dual plane layer. This made actually easier to test sequentially the layer blocks one-by-one. In (a) Initial Prototype (b) Final Prototype Figure 3.44: The change from the initial prototype to the final. 50 Implementation To finalize this layer, in figure 3.44 it is shown the evident changes of the initial model and the end prototype. 3.5.5 Assemble and Connection To assemble all the electronic components of the Tx and Rx it was necessary to create different PCBs for each layer. To be able to do such a thing, each PCB would have to match up the dimension requirements of the physical casing, which means that the ideal solution is to create round PCBs for each layer with a maximum diameter of 40mm. Figure 3.45: Preview of the order of the layers. The module as shown in figure 3.35 begins with a physical cable where the necessary 24V and ground are supplied alongside with the transmitter and receiver cables. They connect to the first PCB, the Tx driver, by soldering the cables directly to the PCB. This one is fixed by screws to the casing, and then this layer connects to the Rx PCB through pin head (resembling the connection between an Arduino and a shield). Some of this pin heads are just some supply voltages for the Rx PCB, but also as some pin heads that have to connect to the LED PCB and other output pin heads to connect to the PD PCB. Finally, by placing the PD layer in ring shape between the optical lens and the LED layer and attach the lens to the LED PCB, the module is complete. The optical lens focus each LED of the PCB and the PDs have some clear space to detect light through the hole present in the lens. The figure 3.46 shows the final aspect of the module, showing the connections between all the layers of the system that are inside the casing. 3.6 Summary In this chapter it was presented all the main components of the communication system separately, showing the preliminary tests made to validate the concept and to take some prior conclusions on how it will perform in different test conditions. It was obvious that this real implementation of the prototype was way more elaborated than the expected, and a substantial amount of time was invested in it, in order to achieve a working 3.6 Summary 51 Figure 3.46: The nucleus of the communication module, assembled all together. and testable prototype. With all this work is accurate to say that the physical conception of this prototype, is the solid core of all this work. It goes without saying that a lot of improvements can be done to all the prototype, some which will have reference in the chapter of conclusions and future work. Now that the implementation of the prototype is concluded some tests on bench and underwater must be done to validate the concept, and to see if there are any adjusts to be made to the system. 52 Implementation Chapter 4 Test Results In this chapter, it is presented all the tests that have been done to validate the implemented solution. The test methodology started with the simplest tests in a workbench with different physical factors and when the prototype is all tested in air-borne, its transmission is subjected underwater in the laboratory pool presented in the department. First the test procedure is presented, explaining what important data is collected, how the test is done and how a result is registered. Then it is presented the results of the tests performed is different environments. 4.1 Test Procedure For every test initially, the parameters which are recorded and varied, in each different test, are the following: •Date and Hour of the test; • The type of Test Scenario; • Type of Casing to be tested; • The Light Color of the transmission; • Which Frequency is transmitting; • The Deviation of the modules between each other; • The ambiance Light Level; • The Time of transmission; • And the value of the comparator Threshold; There are two test scenarios, firstly testing the prototype out of water in the workbench an in the test pool. In the casing parameter it is possible to see the prototype performing without 53 54 Test Results casing nor the optical lens, only with the focusing crystal and with no casing, or with two different types of casing. One with a long end, able to better reject the influence of the ambiance light, but obviously more sensitive to deviations, and a short end without the deviation problem but more affected by ambiance light. The testing end parts of the casing can be seen in figure 4.1. Figure 4.1: The two tested casings. On the left, the long one, and on the right, the short one. Also, the two studied high brightness LEDs light colors were the green (figure 4.2c) and blue (figure 4.2a), testing at two different frequencies of 1MHz and the 100kHz. In terms of the deviation, the holder had some marks marking the deviations of 0, 15, 30 and 45 degrees, being this deviations subject of testing (figure 4.2b). Since the tests were performed in a laboratory, the ambiance light level was also somewhat controllable, so three cases were studied, with medium lights, maximum ambiance light and with no light at all. The remaining parameters were only monitoring variables, so no tests were performed by variating them. (a) (b) (c) Figure 4.2: The blue Tx, the module holder and green Tx. 4.1 Test Procedure 55 In addition, to connect the terminal of the modules cables to the respective source or output, it was necessary to implement a "derivation" board, in order to perform this task. So a very simple circuit board was created to fulfill this necessity. It features a connector to each cable of the module, some test points, a command switch and a fuse in order to protect the battery from a short-circuit if needed. The board is shown in figure 4.3. Figure 4.3: The derivation circuit board. In every test it is followed a certain procedure, that can be interpreted like a protocol, that is roughly the same in each test scenario. The pair Tx and Rx are placed in a certain range to be tested, fixed and then they are both supplied with their voltage suppliers. At this time, both of the modules are in active Rx mode, which means that both are able of receiving a light signal. Then a signal, generated by a function generator or a microcontroller, is placed in the pretended transmission cable and the Tx turns on. By now, the receivers of the Tx are ineffective because they get absolutely dazzled by the light of the Tx. Then the voltage is regulated until a plausible communication is established. When a transmission is settled, the initial consumed current is written down and this operation is maintained during a predefined time period. When this period ends the final current is recorded. The testing case ends when the maximum supply, around 24V, is reached. At this point, if the device does not transmit at a certain range, this distance is known as its range limit. 4.1.1 Determining a Result and Related Problems Right after the implementation started to show some promising performances, a question occurred in terms of testing and retrieving results: How can a transmission be classified as successful or even plausible, and how to measure error rates or derivatives? 62 Test Results (a) Green Tx (b) Blue Tx Figure 4.9: Test results using different transmission frequencies of 100kHz and 1MHz. Summing up, it is concluded that in environment out of water, in terms of the wavelength of the emitting light, the relation of the PD and the blue transmission LEDs seems to be the strongest combination, rather than the green light, which seems to be way less bright than the blue Tx. It was confirmed that the acrylic lens does enhance the deice range significantly, by performing its function of focusing the output light of the Tx, and still enabling the reception of light by the Rx. In the other experimental cases, the light level interference tests were pretty much inconclusive, since the high influence of the deviation due to the using of the acrylic lens, produced 4.3 Workbench Tests 63 inconsistent results that did not allow a more critical verdict. Meanwhile, the impact of the threshold that controls the comparator output, seemed to be another preponderant factor to achieve a better transmission. Still this is a delicate matter since if a too low threshold is defined, higher the probability of error captions and erratic outputs due to the ambiance light interference. Finally, in the two different transmission frequencies, no major influence was detected, which means that in both frequencies the transmission is much alike. 4.3 Workbench Tests Since the first tests are concluded, it is necessary to test the module with the waterproof casing, to check if it can transmit even with the closed casing. This casing capability test will be confirmed first out of water and after in the underwater environment. Now that the casing is going to be tested, all the structure that will be used in the underwater tests will also be added. So for this simple setup, as it is shown in 4.10, the casings of the Tx and Rx were fixed to a holder with marked positions for the deviation angle to be tested. This holder it is settled in a sort of runway with marks on it, in order to mark the various range positions. Figure 4.10: Workbench test setup. The testing parameters are the same as the previous tests, but now the deviation angle is considered in four different positions: no deviation, 15 degrees, 30 degrees and 45 degrees. Also it was possible to test this casing with more than one end terminal, and with a long head and a short head. In a more important note, since the control of the Vre f is not accessible with the casing closed, it was needed to calibrate the threshold in a fixed value. Since, for the Rx, a lower value would enhance its performance, but at the same time, makes it more susceptible to the noise, a pondered middle term needed to be reached. So both thresholds were calibrated to 740mV, in order to improve the communication, without inducing so much noise. So for the first test, it will be tested the performance of the green and the blue light at 1MHz, by applying no deviation, and with an average interference of light. 64 Test Results Figure 4.11: Results of the green and blue transmissions, using the two casing types. As it is shown in the graph 4.11, the use of the casing, independently if it is the long or short type, the range of the transmission, visually comparing to the control subject, is more limited. This is probably because of the design of the casing and the positioning of the PD ring. Even though it is able of captioning the signal, at a satisfactory range, it always limits the capture of the signal. In terms of which casing in this test was the effective, their difference is almost irrelevant. In the green light performance, the traces of the long and short casing are almost coincident, so the different casings did not influence the transmission in terms of the range or the power consumed. Only in the blue light case, when no deviation is presented in the transmission, the long casing seems to perform a little better than the short one. In a first impression, the long casing comparing to a short one, in a transmission, such as experimented in the test 4.11, indeed does not represent a very relevant difference, since the entry and exit point are supposedly well aligned. The most relevant difference between this two casings might occur in two different scenarios: when the interference of the ambiance light is higher, where in this situation, the long casing potentially has an advantage, since it properly isolates the Rx from this effect. And when a deviation between the two modules exist, where now in this situation the long casing might have a disadvantage comparing to the short, since it can limit the Tx light output. So for the following experiments, firstly, the ambiance light influence will be tested by applying the three different light intensities. Then after some conclusions are made, the deviation behavior will be tested, in order to confirm or not the presumed conclusions. 4.3 Workbench Tests 65 (a) Long Casing (b) Short Casing Figure 4.12: Results of the interference of different levels of ambient light. The results displayed in the graphs 4.12, shows the predicted behavior. By analyzing the results on the green light, although it is a very small difference, in the long casing end, shown in graph 4.12a, the reception for the three different interference maintained practically the same, and in the short end, shown in 4.12b a small difference between the receptions can be detected. As for the blue light, it has thinner trace all along its domain, in the long end, meaning that has a smaller variation, and in the short end shows a more unstable output. Despite all this conclusions, even with the different light interference all the outputs are satisfactory. 66 Test Results Since the impact of a deviation, on the previous control tests, was very visible, it is possible to assume, that by deviating the modules between each other, the output will respond to the stimulation. It is only necessary to analyze how big the casing will limit the reception and transmission. (a) Long Casing (b) Short Casing Figure 4.13: Results of the different module deviations. As expected, in the graph 4.13 the deviation add a huge impact on the light output, both in the green and the blue Tx. The deviation that achieved the maximum range was the value of 15 degrees. The other deviation values greater than this had a worse performance than the average signal. 4.3 Workbench Tests 67 It is important to state that by comparing the long and short casing outputs, respectively graph 4.13a and 4.13b, is clearly visible the limited range of the Tx when the deviation is superior to 15 degrees, which means that the long casing limits the transmission of the signal to deviations superior to 15 degrees. The only experiment that is left to do is the transmission at 100kHz, in order to stipulate the behavior of the module working at a lower frequency, capable of being serial communication supported. For this test, it was only performed using the long casing, since no advantage was predicted in testing in the short casing. Figure 4.14: Test results using different transmission frequencies of 100kHz and 1MHz. In this transmission at 4.14 the only particular occurrence, only applicable to the blue light, is the more limited range, which started to be insignificant but in the end it got considerably large. Also at 100kHz the power consumption is higher. Now that the behavior out of water is known, the concept will be tested underwater. 68 Test Results 4.4 Laboratory Pool Tests Figure 4.15: Laboratory pool test setup. Before beginning the transmission underwater, some attention like avoiding the cables being wet were also necessary. Figure 4.16: Results of the green and blue transmissions, using the two casing types. The results of graph 4.16 show a result other than the predicted. Due the communication underwater, the signal output is slightly attenuated, being the maximum range of 1.80 meters, not as good as other cases. The test of the green light with the long casing, was not very successful, inducing some error to the measurements. 4.4 Laboratory Pool Tests 69 Now the system will be tested for the different ambient light level interference. Since this happened in the previous test, some attenuation is predictable for this case. (a) Long Casing (b) Short Casing Figure 4.17: Results of the interference of different levels of ambient light As it can be seen in the graphs 4.17, it maintained the similar performance, being the long casing trace thinner than the trace of the short case, due to the interference of light. The error is the green measurements at the default light level. For next tests the performance of the modules will be tested by applying the deviations marked in the holder. Since the previous workbench tests, the best range results were obtained when the 70 Test Results deviation was applied to the modules, and although the general performance of the system declines underwater, it is predicted that the best range results underwater will be obtained also by applying the deviation. (a) Long Tx (b) Casing Tx Figure 4.18: Results of the different module deviations. In the graphs 4.18, more precisely in the 4.18a, the signal of the green light with no deviation, that was assumed like an error measurement, looks like an output response of the already deviated light. In 4.18b, also the green does not represent a big deviation like it should show, which may indicate that the green light is more influenced by the water than the blue light, which is a bit attenuated, but not in a such significant difference. 4.5 Summary 71 Figure 4.19: Test results using different transmission frequencies of 100kHz and 1MHz. In this last case, as shown in the graph 4.19 the transmission in 100kHz using blue light, against the normal tendency shown by the past experiments, is reaching higher ranges with lower consumption. The green light presents a very predictable behavior, but does not shown deviation flaws. 4.5 Summary In this chapter, it was possible to test and analyze, even if briefly, the characteristics of the communication module. Although far from perfection some basic concepts were also validated, which contributed to the value of the work. As some performance specifications, the prototype was able of transmitting up to 2.80 meters with a deviation of 15 degrees out of water, and was able of transmitting up to 2.00 meters with the same deviation of 15 degrees in underwater. This tests enable the possibility of validating and fulfill the primary objective that is creating a prototype able of transmitting information underwater. The recommended improvements and new ideas to increase the potential of all this project will be described in the future work section. 78 PCB Footprints and Schematics A.1 Initial Version 79 A.1 Initial Version A.1.1 Tx Layer Figure A.1: Initial PCB Schematic - Tx layer. 80 PCB Footprints and Schematics Figure A.2: Initial PCB Footprint - Tx layer (only top layer). A.1 Initial Version 81 82 PCB Footprints and Schematics A.1.2 Rx Layer Figure A.3: Initial PCB Schematic - Rx layer. A.1 Initial Version 83 Figure A.4: Initial PCB Footprint - Rx layer (only top layer). 84 PCB Footprints and Schematics A.2 Final Version 85 A.2 Final Version A.2.1 Tx Layer Figure A.5: Final PCB Schematic - Tx layer. 86 PCB Footprints and Schematics Figure A.6: Final PCB Footprint - Tx layer (only top layer). A.2 Final Version 87 94 REFERENCES [42] Sonardyne. BlueComm Underwater Optical Modem. http:// www.sonardyne.com/products/all-products/instruments/ 1148-bluecomm-underwater-optical-modem.html. [Online; accessed 13/February/2015].