DC-DC Converter Design for Solar Power in Hot Environments

(1)

Bachelor of Science Thesis in Electrical Engineering

Department of Electrical Engineering, Linköping University, 2017

DC-DC converter design for

solar power in hot

environments

(2)

Bachelor of Science Thesis in Electrical Engineering

DC-DC converter design for solar power in hot environments

August Hultman LiTH-ISY-EX-ET--17/0467--SE Supervisor: Tomas Uno Jonsson

isy, Linköpings universitet

Jonas Nilsson Examiner: Dr. J Jacob Wikner

isy_{, Linköpings universitet}

Division of Integrated Circuits and Systems Department of Electrical Engineering

(3)

Abstract

A company in Linköping has a project where a solar power home system is de-signed. The plan is that families that lack grid connection in rural areas of Mali and other countries shall use this system for cooking food and powering every day items. Designing a solar home system for West Africa is more difficult than for other parts of the world, mainly because of the climate, with heat and dust par-ticles in the air, but also because the installation location often is unreachable in short notice. This makes for several specific requirements like high ambient tem-perature, passive cooling, high efficiency and a long mean time between service needed. On top of this, to get the system modular and easy to install, each physi-cal panel should be independent and smart. The system designed is a push pull dc step-up converter that can be assembled to the back of a solar panel. A base platform for the converter is built and a method of power line communication is proposed. Tests show promising results and further development is ongoing.

(4)

(5)

Acknowledgements

By submitting this thesis, I want to give acknowledgements to my project super-visors, Jonas Nilsson at the company and Tomas Jonsson at Linköping university, for giving me inspiration and guidance through all the problems on the way. I want to thank my friends Eric and Johan who always were positive about the outcome and by having practical experience in the area, giving me advice in my work. They have been great sounding boards to exchange ideas with and I want to thank them for their support. Last but not least I want to thank my family and friends for giving me support during my studies.

(6)

Notation

Abbreviations

Abbreviation Meaning

ac _{Alternating current}

avr _{8-bit RISC Atmel micro controller (product line)} cad _{Computer aided design, computer software to design}

hardware dc _{Direct current}

iv _{Current-voltage, a characteristic of solar cells}

mcu _{Micro controlling unit, a unit with a processor and} some peripherals integrated

mosfet Metal oxide semiconductor field effect transistor, a transistor controlled by the voltage over the capaci-tance of the gate

pid Proportional, integral, differential (regulator)

plc _{Power line communication, a method of} communica-tion over power lines

pv _{Photo voltaic, the phenomena in solar cells}

pwm Pulse width modulation, the encoding of a signal into a pulsing signal

pcb Printed circuit board, often a fiberglass core with a copper conductor layer on top

risc Reduced instruction set computer, a processor built on a instruction set with fewer but more powerful instruc-tions relative other types

trms True root mean square, a method of calculating the root mean square that works on different types of wave forms

usart _{Universal synchronous asynchronous receiver} trans-mitter, a communication protocol

uv _{Ultra violet, a specific wave length of light, used in the} process of prototype production

(11)

List of Figures

2.1 Electrical equivalent (a) of a photovoltaic solar cell and its simu-lated, characteristic current-to-voltage and power-to-voltage rela-tionships (b). . . 6 2.2 Method of communication. . . 7 2.3 Example circuit for power line communication between the

de-signed converter and the battery controller using passive compo-nents at the converter side. . . 8 2.4 Block diagram of the system with all the parts separated into blocks

with names that describes their functions. . . 9 2.5 Basic schematic of the converter. . . 10 2.6 Switching voltages. . . 11 3.1 Design method used in this work, an iterative method with clear

steps. Going from research to building the system and testing it, with some iterations of building and testing. . . 18 3.2 The two side layout printed on overhead paper. . . 21 3.3 The back side of the etched board. . . 22 3.4 Bench test with an Agilent oscilloscope measuring different parts

of the circuit, three multimeters measuring voltage out, in and cur-rent out. . . 23 3.5 System test of the designed converter with a capable load. . . 24 4.1 Efficiency simulated for three different suggested converter

topolo-gies. . . 26 4.2 Conceptual design of the converter. . . 27 4.3 Test set up used in simulation, solar panel equivalent current source,

the designed converter and a battery model. . . 28 4.4 Simulation results of output, transformer, transistor drain and gate

voltage. . . 29 4.5 Both prototypes of the converter, first (a) and second (b). . . 30 4.6 Bench test, 450 V (a) and 455 V (b) with different loads and pulse

widths at 31 V input (c). . . 31 4.7 Voltage oscillations at mosfet drain during mosfet switching. . . 32 4.8 Test with maximum power possible (a) and the resulting

trans-former breakdown (b). . . 33

(12)

4.9 Calculations of the power loss in the converter with a visual repre-sentation in (a) and values in (b). . . 34

List of Tables

1.1 Specifications of the converter. . . 2

(13)

1

Introduction

Power efficiency and renewable energy are hot topics today. The focus of this thesis is on solar power and power electronics, being used in Africa.

1.1 Background

A company in Linköping has a project where solar power electronics are designed to be used in Mali. The author of the thesis is hired at the company and will con-tinue the work after that the thesis is finished. In Mali, food is cooked indoors over open fire using a lot of wood and coal. This creates deforestation and pol-lutes both the indoor and outdoor air, killing people every year. As the rural areas of Mali do not have a grid connection, an independent energy system is needed to start using induction stoves instead of open fire for cooking food. Solar power systems are well known in Mali. Multiple solutions for solar power exist, but most of them cannot be used with a long lifetime in the environment of Mali, due to dust and heat. Most of the solar systems used are based on lead acid batter-ies which is really bad for the local environment, as proper recycling options are limited.

(14)

2 1 Introduction

1.2 Purpose

The purpose of this thesis work is to design a step-up converter that will be as-sembled to the back of a solar panel. This converter shall convert 31-50 V up to 360 - 450 V DC to charge a battery, in parallel with other converters. A working prototype of this converter that can easily be modified for production is the goal of this thesis work.

1.3 Specifications

The specifications of the system are declared in table 1.1. Table 1.1:Specifications of the converter. Parameter Min Typ Max Unit

Vin 31 39 50 V DC Iin 9.35 - - A Vout 360 - 450 V DC fswitch - 100 - kH z Tambient,air - - 50 ◦C Output power 350 - - W Efficiency η 90% - -Board width - - 100 mm Board length - - 160 mm Height - - 60 mm

The system must be compatible with the solar panels used in the project, which outputs 31 V - 50 V under normal conditions and up to 9.35 A with a maximum power point at about 39 V at 25◦C for the most commonly used one [7].

The battery used is a bipolar Nickel metal hydride battery as it tolerates the high temperature in Mali and does not include toxic metals. Recycling of nickel metal batteries is rewarding with 10 % of the initial cost according to the battery sup-plier. The output voltage must be adjustable from 360 V up to 450 V, as that is the charging voltage range for the battery used. The switching frequency is cho-sen to be high to minimize size and cost of magnetic components, but not to high that the switching losses get too high. Outdoor temperature in the geographic location the system will be used in goes up to about 50◦C and the module must

be able to withstand this. The system must be able to utilize all available power from a 350W solar panel to charge the battery, therefore the power rating must be at least 350 W.

The prototype can be a maximum of 160 by 100 mm2 in size due to prototype manufacturing methods. The height must not exceed 60 mm, as it will be assem-bled to the back of a solar panel. Shipping will be problematic and costly if it protrudes too much out of the back of the panel.

(15)

1.4 Questions 3

To be able to run the system continuously, it must not generate more heat than what can be cooled away with passive heat sinks. As physical space is not an issue other than for the printed circuit board, large heat sinks will be used that have a thermal resistance of about 1 K/W. As many components has a maximum temper-ature rating of 85◦C, at 50◦C ambient temperature, the maximum temperature

increase must be less than 35◦

C. This translates to 35 W power loss maximum

and an efficiency of at least 90 %.

1.3.1 Communication

A central computer should in the future be able to communicate with all the converters connected to the battery to be able to control and troubleshoot them remotely. This is important to be able to shut down the converters, for example, not to overcharge the battery. Communication between converter and battery needs to be thought of in the design.

1.4 Questions

The questions asked are:

• How can a step-up converter be designed to fulfill the specifications in table 1.1?

• What specific problems are there that can arise with such a converter? • What communication methods are applicable to the converter for it to be

able to communicate with connected systems?

As the goal is to have a converter that works in the given conditions, it is of inter-est how it can be designed. Different designs have different problems. To be able to evaluate the converter for future use, these problems should be documented. The goal is to use the converter in a system that will communicate with the con-verter, therefore it is of interest to investigate in methods of communication. This thesis answer these questions.

1.5 Limitations

As the converter is designed as a thesis work, time is a limiting factor. No more than ten weeks of work is put into the design. The converter uses a switching topology with a micro controller to control the switching. Output emissions are not considered, as the system is connected to a large battery together with other systems. Therefore output emissions are of less importance. The company Mi-crochip makes 8-bit AVR micro controllers that includes all the functionalities needed in this project. These will be used for simplicity.

(16)

(17)

2

Theory

The designed converter will charge a battery with power from solar cells. As the system is powered by solar cells and only has battery storage to use during night time, efficiency of electronics after the battery is prioritized. To minimize power loss after the battery, the battery voltage is selected to be over the maximum voltage needed later in the system. This leads to no step-up converter needed after the battery. The battery voltage is therefore several hundred volts, but the solar panels supplies up to a couple of tens of volts. A step-up of voltage is needed to charge the battery.

2.1 Solar cells

The source of power to the converter is a solar panel that has several solar cells in series and parallel circuits. Each solar cell acts like a current source with a diode, resistor and a shunt resistor [3]. The equivalent circuit is seen in figure 2.1a with its characteristics, the current-voltage (iv) and power-voltage curve, in figure 2.1b.

The current source that generates current with the photovoltaic effect is Isolar.

Diode, Rseries and Rshunt are the solar panel equivalent internal diode, series

resistance and shunt resistance. Solar cell characteristics are simulated on the circuit output V + to V −. Over a specific voltage, output current drops faster with increasing voltage. This is the knee on the current to voltage curve. At the point where the derivative of this curve is −1/2, power is at its maximum, as can be seen in the power curve. Voltage, current and power change with temperature and have a specified temperature coefficient. Voltage is most sensitive to temperature changes.

(18)

6 2 Theory

(a)Equivalent circuit.

Voltage

Current Power

(b)Characteristics.

Figure 2.1: Electrical equivalent (a) of a photovoltaic solar cell and its sim-ulated, characteristic current-to-voltage and power-to-voltage relationships (b).

2.1.1 Maximum power point

As seen in figure 2.1b, the power curve has a global maxima at a specific voltage, meaning that a solar cell has a maximum power point at a specific voltage. The voltage generating the maximum power point varies with temperature [3]. This point must be used to get the most power out of the solar cell.

Tracking

Finding the maximum power point can be done in several ways. One way is by sweeping along the iv-curve, finding the maximum point. Another way is by constantly changing the current (load), measuring the power and thereby contin-uing in the direction of positive power change. In reality, some cells in a panel can be shadowed, therefore there may be multiple local maximum power points [9]. This makes power regulation difficult. Periodically searching through all available values gets the voltage at the global maximum power point. The regu-lation is then done on the difference between the maximum power point voltage and the measured voltage at every given time. As temperature does not change abruptly due to thermal mass inertia, this will result in an approximate regula-tion that tangents the highest possible efficiency.

2.2 Communication

Having a distributed system, communication is important to optimize the total ef-ficiency. By only having the power line from the panels to the battery, the options for communication divides into wireless or power line communication (plc).

(19)

2.2 Communication 7

2.2.1 Wireless

Wireless communication is a common solution to minimize the number of phys-ical connections but has several disadvantages over wired communication. Reli-ability is a big factor. Using wireless protocols, the system must compete with other devices in the ever so increasing world of wireless internet and mobile phones. Lower safety is another disadvantage when the system more easily can be disturbed or interfered with. The converter output is high voltage, therefore high safety is important.

2.2.2 Power line communication

As the system is independent of the grid, no noise can be overheard from the public grid onto the local power line communication. Trying to interfere with the power line is hard, as the interferer must electrically connect to the wire. A communication over the power line (plc) is therefore preferred over wireless.

Software implementation

To implement a plc system, one can use existing plc controllers on the market, making the system more physically complex and more expensive. Another option is to use existing hardware in the converter, as high communication speed and bandwidth are not of great importance. A suggested method of communication is shown in figure 2.2. + curr ctrl PLC TX PLC RX Control loop PWM DC/DC Battery Volt div

Figure 2.2:Method of communication.

2.2.3 Power line communication with hardware

There are solutions for hardware implementations of plc with chips that control all of the communication but this can be costly and take up space on the board. By filtering out the direct current (dc) voltage with a series capacitor, plc can

(20)

8 2 Theory

for example be done with the usart protocol. This implies that the data sent is limited to bytes with equal amount of ones and zeros, so that the average voltage is zero and thus not filtered out by the capacitor. With the usart protocol and only one line, communication is only possible in one direction. This could be enough to control the system from the central controller in the battery.

Implementation

The hardware required to create this communication involves capacitors, resis-tors, diodes and inducresis-tors, as shown in the circuit diagram in figure 2.3

Figure 2.3: Example circuit for power line communication between the de-signed converter and the battery controller using passive components at the converter side.

with "converter" being the designed converter with an input for plc via a passive circuit, built by D1, D2, R1, R2 and C_block. The battery controller is connected to a plc transceiver that overlays a communication signal on top of the dc volt-age. Capacitance C_block blocks dc current and lets communication through to the converter communication input, PLC_in. The diodes ensures that the signal will not have voltage spikes over 5 V and under 0 V to not damage the logic in the converter. In practice, the diodes has a forward voltage that must be consid-ered, as it will offset this voltage protection. Resistors limit the maximum current through C_block. If the converter is to send plc communication signals, it is done in software and the battery controller needs to have an inductor in series, before the battery, to let the communication through. Also, the plc transceiver needs to read at a relatively low rate, as this type of sending will be relatively slow.

(21)

2.3 Step-up converters 9

2.3 Step-up converters

To get a voltage of 450 V from a solar panel that has a maximum voltage of about 50 V, a step-up converter is needed. Different topologies use different methods of conversion, including using a transformer to transform the voltage or storing energy in magnetic cores. The basic loss calculation for a transformer converter is

Ploss= Pswitch+ Pconduction+ Ptransf ormer (2.1)

as compared to topologies that stores magnetic energy in inductor cores with a loss calculation

Ploss= Pswitch+ Pconduction+ Pcore (2.2)

as Mohan et al. [6] shows. The difference between these two is the losses in the magnetic components, the transformer on one hand and the power inductor on the other.

2.4 System overview

The circuit schematics is described in a simplified block diagram in figure 2.4, derived from the schematics in appendix A. The layout is in appendix B.

Figure 2.4: Block diagram of the system with all the parts separated into blocks with names that describes their functions.

"MCU" is the micro controller unit (mcu). It is powered from the solar panel via a 5 V converter and controls the transistors, ("Power electronics"), which is powered from the solar panel via a 15 V converter. "Power electronics" feeds the transformer ("Step up transformer") with the solar panel input ("PV panel input 350 W"). "Rectifier" rectifies the output of the transformer and "Output filter" filters it to a stable dc voltage. If power line communication is used, "Optional power line communication" communicates on top of this dc voltage. The output is 360 to 450 V and feeds the battery and other parts not included in this thesis.

(22)

10 2 Theory

2.5 Push Pull converter system overview

The converter consists of the first five blocks in figure 2.4, a micro controller (mcu), switching transistors, a transformer, rectifying diodes and a filter, as usual for this type of converter [6]. As seen in the basic schematic of the converter in figure 2.5 (below), the secondary side is not isolated. As the input is a solar panel and that the output is dependent on the function of the converter, the rectifying diodes protects the solar panel from battery voltage. The battery will have addi-tional protection, therefore isolation is not needed in this application. Thereby the secondary voltage is put on top of the panel voltage, saving some turns on the transformer, getting higher efficiency.

Figure 2.5:Basic schematic of the converter.

When the controller "MCU" switches the transistors "MOS1" and "MOS2" alter-nately, the transformer "T1" will have the current from the solar panel going in alternating directions. Due to the alternating current, alternating flux is gener-ated through the transformer core. This transforms the voltage to the secondary side, where the diodes "D1" and "D2" rectify the transformed voltage and L and C filter it. "OUT" is connected to the load. The load is the battery. As the secondary side center tap is connected to panel voltage, it can be seen as a series circuit where output voltage equals transformed voltage plus panel voltage. Current in secondary side is drawn from panel current and thus making current through transistors to decrease with the output current.

2.5.1 Transformer

To convert ac voltage up or down, a transformer is used. The relation between number of turns, voltage and current, between the two sides of the transformer

(23)

2.5 Push Pull converter system overview 11 is shown below. N1 N2 = U1 U2 = I2 I1 (2.3) where subindices 1 and 2 stands for primary and secondary side, N is number of turns, U is voltage and I is current of each side [6].

All real transformers are non-ideal and have for example a leakage inductance due to the fact that some of the flux escapes the core and goes through the air. This leakage inductance is dependent on core shape, winding method, type of wire and more [6]. the leakage inductance becomes an unwanted parasitic induc-tor that does not interact with the rest of the transformer.

2.5.2 Switching

As the transformer must be fed with ac voltage, some sort of switching mecha-nism must create the ac voltage. This is in this design done by two N-channel metal oxide semiconducting field effect transistors (mosfets) that are switched with a duty cycle of less than 50%, being offset 180◦

from each other. The switch-ing voltages are described in figure 2.6.

t LOW HIGH PWM1 Voltage t LOW HIGH PWM2 Voltage

Figure 2.6:Switching voltages.

The voltage on the y-axis is the voltage over the transistor gate, with ’High’ in-dicating maximum voltage. Time is on the x-axis. As the transistors are placed on each side of the center tap, the transformer gets an alternating flux direction, equivalent of a single primary side winding with an ac input.

The primary side of the transformer consists of two windings that are powered alternately. The winding that is turned off at a given time will have twice the input panel voltage as the primary can be seen as an one to one transformer, leading to twice the input voltage over each transistor.

(24)

12 2 Theory

2.5.3 Micro controlling unit

The transistors are controlled by a 8-bit micro controller (mcu). It creates the two channel pulse width modulated (pwm) signals (see figure 2.6 above) that are phase shifted 180◦ between each others. If communication is used, this is controlled by mcu, along with anything else that may require some logic.

2.5.4 Secondary side

The secondary side of the transformer is rectified with two diodes and filtered with a LC-filter. Because of impedance transformation, an inductor is placed be-fore the output filter capacitor not to have a highly capacitive load on the transis-tors, as that would create high current peaks. The equation for the transformed impedance can be derived from Ohm’s law and equation (2.3) to be

Z1= U1 I1 = U2 N1 N2 I2NN21 = Z2( N1 N2 )2 (2.4)

where Z is the impedance. As capacitive impedance is inversely proportional to frequency, the transformed capacitance from secondary to primary side is the square of the transformer ratio greater than the secondary capacitance. With the inductor, A LC filter is formed. The resulting filter resonance frequency with a 4mH inductor and a 1µF capacitor is

fr = 1 √ LC = 1 √ 4 × 10−₃_{1 × 10}−₆ ≈_{15.8kH z} _(2.5) where fris the resonance frequency, L is the inductance and C is the capacitance.

2.6 Powering the system

The logic and transistor power can be drawn from the solar panel or from the battery. If the solar panel is the source of power to the logic, the system will not be able to take commands when the sun is out, but will not drain the battery without solar power. The transistors can be powered from the battery, but this creates extra expensive hardware, as the voltage must be converted down from up to 450 V to about 20 V or the maximum voltage of the transistor. Driving the system from the battery is important if the system must be installed during night time or when testing the communication without large solar panels, which eases service. There is however no requirement to be able to install the system without sun power at the moment and the solution with driving the logics and transistors from the solar panels is more cost effective.

2.7 Power loss

The power loss is divided into several different parts. The most significant in-cludes transistor switching loss, transistor conduction loss, transformer loss, diode

(25)

2.7 Power loss 13

loss, filter loss and internal converter losses.

2.7.1 Transistor switching loss

Producing an alternating current is done by switching. During the switching time of a metal oxide semiconductor field effect transistor (mosfet), the current and voltage goes from a state to another, which uses energy and dissipates it in form of heat. Shorter rise and fall times produce less heat until the point where the gate resistance dominates the loss. The switching loss of the transistor is divided into two parts, the gate resistance power loss and the actual switching loss. The gate loss equation is

Ploss,gate= fswitch×Qgate×Ugate (2.6)

where Ploss,gateis the power loss due to the charge Qgateat the frequency fswitch

that is needed to charge the gate capacitance to the voltage Ugate. This power is

dissipated in the gate resistance.

The equation for the actual switching loss is

Pswitch= fswitch×(Eon+ Eof f) (2.7)

as Raee et al. [8] and Mohan et al. [6] show. Here, Pswitchis the power loss due

to switching, fswitchis the switching frequency and Eonand Eof f are the energies

needed to perform the switch on and off.

The energies Eonand Eof f are built up of the rise and fall times, voltage over the

switch and current through the switch.

When turning a switch on, diodes on secondary side are conducting the same amount and the transformer is thus not transforming any voltage to the primary side. Current builds up to half through the switch before one of the diodes on the secondary stops conducting and the transformer starts working.

This causes voltage over the transistor to sink to zero after tf all seconds and the

current to rise to maximum after tf all seconds or depending on leakage

induc-tance of the transformer, depending on what is most significant.

When turning off a transistor, one diode is conducting at the secondary side. Cur-rent is high and leakage inductance of the transformer creates an inertia that forces voltage over the transistor to rise quickly to a level where all the energy stored in leakage inductance has been absorbed or transfered to capacitance in the transistor and other components connected.

The energies Eonand Eof f are calculated as

E =

t2

Z

t1

UDSIDdt (2.8)

Where ID is the current through the transistor, UDSis the drain to source voltage.

(26)

14 2 Theory

the on and off switching energy, as the only difference in the equations are t1 and

t2.

2.7.2 Transistor conduction loss

DC resistance of the mosfets also creates power loss in form of a resistive con-duction loss,

Ploss,cond = 2 × ID2 ×RDS,on (2.9)

ID = (IP ×

√

D)2 (2.10)

Where Ploss,condis the conduction loss for two mosfets, D is the duty cycle, ID is

the drain current through the mosfets and RDS,onis the on state resistance of the

mosfet_{s. The current I}_D _{is calculated with primary current, I}_P_{, and duty cycle.}

2.7.3 Transformer loss

Transformer windings are made of copper, which has a resistance, R, of each winding, leading to a power loss

Ptransf ormer = I1,avg2 ×R1+ I2,avg2 ×R2 (2.11)

where R1and R2are the resistance of one winding on primary and secondary side

respectively, and I1,avgis the total average current through both the windings on

the primary side. The current I2,avg is the total average current through both the

windings on the secondary side. This is because only one winding will conduct on each side at a given time.

2.7.4 Diode loss

To produce a dc voltage from the transformed ac voltage, it must be rectified. Using diodes for this purpose is a common practice. With the silicon carbide diode chosen, there is no significant switching loss. Power loss of the diodes is calculated as current times the forward voltage, see equation below.

Pdiode= Uf orward×If orward (2.12)

with Pdiodebeing the power loss, Uf orward the forward voltage and If orward the

forward current. In the converter, only one diode will conduct at a given time, therefore the total power loss is

Pdiodes= 2 × D × Uf orward×If orward (2.13)

where Pdiodes is the total diode power loss and D is the duty cycle, from 0% to

50%.

2.7.5 Internal converter losses

The components in the converter need different voltages. Therefore two smaller step down converters (5 V and 15 V, respectively) are connected to the input

(27)

2.8 Oscillations 15

and creates the biggest power loss after the diodes. As the logic components are powered with a separate converter, all of the power used by this can be consid-ered a loss, as it does not contribute to the output. The loss from the converter that drives the power electronics can only be calculated as 1 − η where η is the efficiency, as the power from this converter is dissipated in the transistor gate resistance and other parts, already accounted for.

5 V converter loss

The logic components (mcu and its peripheral components) are driven by the 5 V converter. The power used by the converter is

Pconv = Uconv×Iconv×(1 + (1 − η)) (2.14)

where Pconv is the total power consumed by the 5 V converter, Uconv is the

volt-age out of the converter and Iconv is the current. Here, η is the efficiency of the

converter. The power is multiplied with (1 + (1 − η)) as the total power used is the power to the components plus the converter power loss.

15 V converter loss

The power electronics are powered with the 15 V converter. The power loss is

Ploss = Uconv×Iconv×(1 − η) (2.15)

Where Plossis the power loss in the converter, Uconv and Iconv are the voltage and

current out from the converter. The efficiency of the converter is η. The power is multiplied with (1 − η) as the power loss is the power loss of the converter. The rest is lost in other components such as the transistors, which is already accounted for.

2.8 Oscillations

Parasitic inductive and capacitive elements create resonance circuits. This can create oscillations, for example when switching off the current to an inductor, as the voltage will spike due to the inductance equation

v(t) = Ldi(t)

dt (2.16)

where v(t) is the voltage over the inductor, L is the inductance, and di(t)/dt is the derivative of the current. Because of this a high voltage will be built up over the switch when switched off due to the current inertia of the inductor, possibly leading to failure.

2.8.1 Snubber circuit

To address the issue with oscillations, a so called snubber circuit is built. This is a circuit that absorbs the energy stored in the inductor, by a capacitor to charge with the energy and a resistor to dissipate the energy. The maximum allowable voltage over the switch together with the inductance value give the value of the

(28)

16 2 Theory

capacitor, as it should take all energy stored in the inductor. The energy is

EL=

LI2

2 (2.17)

which then is equal to the energy in the capacitor,

EC= CU

2

2 (2.18)

with ELbeing the energy stored in the inductor, L being the inductance, I being

the current through the inductor, EC being the energy stored in the capacitor, C

being the capacitance and U being the maximum allowable voltage. The resistor dissipates the energy as heat and thereby provides damping. The snubber circuit is connected from the oscillation node down to ground with the resistor in series with the capacitor.

The damping ratio ζ, which defines how much oscillation that will occur, is given by ζ = 2 R₁ ωrC = 2qR L C =√1 2 (2.19)

Note that the resonance angular frequency ωris

ωr =√1

LC (2.20)

and that the damping is

ζ =√1

2 (2.21)

(29)

3

Method

The converter design is divided into different steps. The first step is to read about DC-DC converters, building up a specification and learning the basic theory of switching converters. A topology is selected for a starting point, to be able to start simulating. Simulations are done and new topologies are selected if needed. A prototype is constructed, tested and eventual problems are solved. With a work-ing prototype, more tests are done, with more realistic and capable loads as a system test. If the system test shows that the prototype will fulfill the specifica-tion, a final design might be done if there is time left. This is described in figure 3.1, derived from a NASA Engineering method [1].

Conclusions can however be drawn from the bench tests and specification.

(30)

18 3 Method

Figure 3.1:Design method used in this work, an iterative method with clear steps. Going from research to building the system and testing it, with some iterations of building and testing.

(31)

3.1 Literature survey 19

3.1 Literature survey

To learn what others have done and not to make their mistakes again, a literature survey is made. In the survey, different topologies are studied as well.

3.2 Choosing the topology

To choose a suitable topology for the converter, different topologies are evaluated by following parameters:

• Efficiency, as the power loss generates heat, which is bad. • Cost, as the system is being designed for a commercial purpose • Complexity, as the project is limited by time

• Construction time, as its being designed under limited time

The parameters described are derived from the specification and the commer-cial application that the converter is being designed for. The evaluation is docu-mented.

3.2.1 Efficiency

The converter is going to operate in a high ambient temperature without a fan, therefore it should generate as little heat as possible, and thus have as high effi-ciency as possible. Simulations with powerEsim are made on effieffi-ciency for rec-ommended topologies, results are documented.

3.3 Conceptual design

From the chosen topology, a conceptual design is made to be able to evaluate if the chosen topology is able to fulfill the specifications in table 1.1. This is a simplified basic design to evaluate basic functionality.

3.4 Test system

To simulate the topology and design, a test system is set up. To test a model, one way is to set up a test system and do experiments on it, as Min et al. [5] do.

3.5 Simulation

All parameters are evaluated in simulations to see if there are fundamental prob-lems with the design and to optimize it at good as possible before testing with a physical construction. Simulations are done with SPICE in MultiSim. Results from the simulation is documented.

(32)

20 3 Method

3.6 Construction

When a theoretically working design is finished, a prototype is constructed to be tested in practice.

3.6.1 Implementation of design

The design for simulation is implemented with real components.

Logic

Switching logic is implemented in software in an ATmega328 [4], an 8-bit micro controller (mcu) from Microchip. The mcu runs at up to 5.5 V and has a risc based processor that can be run at 16 MHz system clock. Important included features in the mcu are:

• 10-bit analog to digital converter with several multiplexed inputs

• Three timers that can be configured for hardware pwm and interrupt gen-eration.

• Hardware serial interfaces

Power electronics

The transistors used are metal oxide semiconductor field effect transistor (mosfet). As the voltage over each transistor will be twice the input voltage, as seen in chapter 2.7.2, and that spikes will appear, the maximum tolerable voltage of the transistor must be greater than 100 V. Transistors selected are IRF2415 from In-ternational rectifier, which has a max voltage of 150 V. The internal resistance of the transistors are 42mΩ. The transistors can tolerate a continuous current of up to 43 A and a junction temperature of up to 175◦C. High temperature

rat-ing of the transistors is important, as they are passively cooled in a high ambient temperature environment.

To achieve high speed switching, mosfet drivers that can charge the gate capaci-tance faster than the MCU was added. The chosen drivers are NCP81074ADR2G from ON Semiconductor.

The rectifier is composed of two diodes, GP2D005A170B from Global Power, sili-cone carbide schottky diodes with zero switching losses and zero reverse recovery current. The diodes tolerates 1700V , 5A and 175◦

C.

A transformer is wound by hand, as it is expensive to order a single special trans-former and so that the system design is not limited to which transtrans-former can be found.

3.6.2 Hardware and manufacturing

The prototype is built in house with simple methods. A circuit board is designed, etched and components are soldered on to it.

(33)

3.6 Construction 21

Layout

A board layout is made in Eagle CAD. Special consideration is taken to place the power electronics on the edge of the Printed Circuit Board (pcb), as they dissipate heat and need to be connected to a heat sink so to not overheat.

Prototyping

As the layout needs double layers, the prototype needs to be constructed with precision to get the bottom and top layer aligned. Starting with the top layer, the bottom layer is protected with tape during the work. The layout is printed on overhead paper, see figure 3.2. It is then put over a copper clad with photo re-sistive coating and is exposed to uv radiation to transfer the layout to the board. The exposed photoresist copper clad is developed in caustic soda to remove

(34)

22 3 Method

cess photoresist. The tape on the other layer is protecting that layer from this process.

The developed board is etched in sodium persulfate to remove excess copper. Again, the tape on the other layer protects it from the process. Selected holes are drilled with reference to the top layer for alignment with the bottom layer. The bottom layer layout overhead paper is aligned with the holes and fastened with tape.

With the top layer finished, it is protected with tape and the process starts over again from exposure to development and etching.

Finished PCB

When the bottom layer is finished, see figure 3.3, components are soldered to the board and the prototype is ready for programming and testing.

Figure 3.3:The back side of the etched board.

3.6.3 Software

Software development is done in C programming language in Atmel studio. At-mel studio features debugging, compiling and uploading of code to an avr micro controller (mcu). As the mcu features hardware pulse width modulation (pwm) and several timers that can be configured to control these pwm channels and generate interrupts, the software will be built around this. A programmer from Atmel is used to write the code written in C to the prototype. Programming is done via the isp interface on the micro controller.

(35)

3.7 Testing 23

3.7 Testing

To be able to evaluate the converter, testing is done to eliminate simulation errors such as not finding all the parasitic components or any other type of unrealistic behavior.

3.7.1 Bench test

The prototype constructed is tested with simple methods to eliminate measure-ment errors. Iterations of the design are made until satisfactory results occur, or if the design does not work with the tests, a new design or topology is selected. The prototype is connected to a bench power supply and a suitable load to see how it performs. The output is monitored with an oscilloscope and voltage mea-surements are taken with a True Root Mean Square (trms) multimeter. Oscil-loscope measurements are saved and imported to Matlab for further processing with filtering and composing graphs.

A bench test set up can be seen in figure 3.4.

Figure 3.4:Bench test with an Agilent oscilloscope measuring different parts of the circuit, three multimeters measuring voltage out, in and current out.

(36)

24 3 Method

3.7.2 System test

The prototype is tested in a realistic scenario with actual or mimicking load and source. Results are evaluated. Iterations will be made to optimize the prototype. If there are fundamental problems, a new design or topology is selected. If there are uncertainties about the performance, new simulations are done. A typical test scenario could be a battery or an induction hob heating up water as a load. The input could be solar cells that powers the converter. A system test set up is shown in figure 3.5 with two variable resistors in series capable of dissipating 6000W each seen in the foreground.

(37)

4

Result

4.1 Literature survey

Mohan et al. [6] gives an extensive theory about power electronics and switching converters. Inspiration for simulation test systems was found in a report by Min et al. [5].

4.2 Choosing the topology

A push pull converter was suggested by both a supervisor and by the book power electronics by Mohan et al. [6], due to the high ratio of input to output voltage and the power range. A push pull converter, and other converter topologies sug-gested, were studied and simulated with a general setup in the web application PowerEsim. Results of efficiency are seen in figure 4.1. As seen, the push pull converter achieved the highest efficiency over all.

4.3 Conceptual design

To get started with converter simulations, a conceptual design of a push pull con-verter was made, as explained in 2.4. This design was made as a base for further development to find out optimal values of for example transformer ratio, filter frequency, snubber circuits and more. It resembles all the critical functionali-ties that is expected from the finished system. This specific design was never intended to be physically produced and tested, only simulated. To be able to sim-ulate it, the micro controller was replaced by a signal generator. The conceptual design is shown in figure 4.2.

(38)

26 4 Result (a) F orw ard con v erter . (b) A civ e clam p con v erter . (c) Push pull con v erter . F igure 4.1: Effi ciency sim ula ted for three di ff eren t sug g ested con v erter topol ogies.

(39)

4.3 Conceptual design 27

(40)

28 4 Result

The transformer, T1 has its windings wound in the same direction. S3 and S4 are rectifying diodes, L2 is a filter inductor, C1 is a filter capacitor, C2 is the input capacitor. Sources V4 and V5 generates the pulse width modulation signal (pwm). Transistors U1 and U2 are switching transistors and R10, C9 and R8, C8 builds up the snubber circuits.

4.4 Test system

A test system for the simulation was set up in Multisim with models of a solar panel and a battery as shown in figure 4.3.

Figure 4.3: Test set up used in simulation, solar panel equivalent current source, the designed converter and a battery model.

The current source, I2, has a function that models the solar panel, equivalent to the solar cell model seen in 2.1 for the solar panel supplied. The converter is shown as a sub block, converter_box, to get an overview. The battery is modeled as a voltage source, V2, and a series resistor, battery_esr.

(41)

4.5 Simulation 29

4.5 Simulation

The results from the simulation are shown in figure 4.4

Figure 4.4: Simulation results of output, transformer, transistor drain and gate voltage.

The figure shows transformer voltage (solid line) and output voltage (dotted line) in the top box. The middle box shows transistor drain voltage and the lower box shows transistor gate voltage. All of the curves are time synchronized. As seen, voltage drops soon after starting the converter when the inductance is charged up and the converter starts using energy. As capacitors charge, voltage rise to a steady state level. In the simulation, the leakage inductance of the transformer was estimated to be 15 nH on the primary side. That is the inductance that creates a voltage spike over the switch. Knowing this and that the current through the primary will be approximately 10 A, the energy is given by equation (2.17) as

WL =15 ∗ 10

−₉

102

2 = 750nJ = WC (4.1) The transistors used are tolerant of up to 150 V and the allowable voltage of the oscillation is set with a margin to 100 V. This gives the capacitance, from equation

(42)

30 4 Result

(2.18), 150pF. The resonance angular frequency, ωr, is then

ωr = 1 √ LC = 1 √ 15 × 10−₉_×_{150 × 10}−₁₂ ≈_667Mrad/s. _(4.2) With dampening factor, ζ = 1/

√

2, the resistance R is calculated with equation (2.19) to be R = ζ 2× 1 ωrC = 1 2 × √ 2 × 667 × 106×_{150 × 10}−₁₂ ≈3.5Ω (4.3)

The snubber circuit at the time of measurement was C = 150pF and R = 3.3Ω, but the real leakage inductance was not known when ordering components. Ac-tual leakage inductance was measured on the wound transformer to be 0.4µH.

4.6 Construction

To test the converter physically, prototypes were made of the conceptual design with some modifications for reality with all peripheral components needed to control it. The steps of design and construction are explained in 3.6. Two pro-totypes were made. The first prototype had some design problems, why a new prototype was made. The finished prototypes can be seen in figure 4.5a and 4.5b.

(a)First prototype. (b)Second prototype.

Figure 4.5:Both prototypes of the converter, first (a) and second (b). In addition to different layouts and designs, the first prototype has a larger trans-former and a smaller filter inductor (the inductor is placed out of sight in 4.5a) than the other prototype. The other prototype lacks input relays and has differ-ent internal voltage regulators compared to the first prototype (regulators are out of sight in 4.5a). Prototypes were mounted on a wooden board with an electrical outlet to connect the load, as seen in 4.5b.

(43)

4.7 Testing 31

4.7 Testing

Testing was done in two different labs for bench tests and system tests that had different possibilities of testing the system.

4.7.1 Output

In the bench test, the system was tested with a voltage controlled power supply as a source for the converter and a secondary load of 600Ω and 1679Ω. The input voltage was set to about 31V , pulse width was set high to get high power out and output was measured with a voltmeter and ampere meter. The output can be read from figure 4.6 where the yellow, upper instrument measures current and the red measures voltage. The output reaches above 450 V and the system should be able to charge the battery.

(a) 450 V output, 1679Ω load.

(b) 455 V output, max pulse width, 600Ω load.

(c)Input

Figure 4.6:Bench test, 450 V (a) and 455 V (b) with different loads and pulse widths at 31 V input (c).

(44)

32 4 Result

4.7.2 Switching and inductor current

The diode and transistor drain voltage are seen in figure 4.7. The top graph shows voltage of the node where both diode cathodes connect. Transistor drain voltage can be seen in the lower graph. Both nodes experienced oscillations due to reso-nance circuits. 1 1.5 2 2.5 3 3.5 Time [s] 10-5 -200 0 200 400

Diode output node

1 1.5 2 2.5 3 3.5 Time [s] 10-5 -50 0 50 100

Transistor drain voltage

Figure 4.7:Voltage oscillations at mosfet drain during mosfet switching.

At the moment that a transistor switched on, oscillations occurred at 649kH z on the voltage over the rectifying diodes, as seen in figure 4.7. This could be caused by the parasitic capacitance of the rectifying diodes of ca 20pF at high voltage [2] together with the inductance of the inductor of ca 4mH, which theoretically creates a resonance at fr = 1 2 ∗ π √ 20 × 10−₁₂_×_{4 × 10}−₃ ≈_{563kH z} _(4.4) As there are other parasitic components, this number should be considered close to the actual resonance.

As also seen in figure 4.7, the oscillation amplitude varies with the switches, go-ing over the limit of the oscilloscope to the left, indicatgo-ing that the balance be-tween the two is not perfect.

(45)

4.7 Testing 33

4.7.3 Snubber

The snubber circuit used was calculated with theoretical, estimated, values, which was wrong. This is the cause of the inductive spike over the transistor at turn off, as seen in figure 4.7.

4.7.4 Power

The highest possible power that could be tested was limited to where the oscil-lations caused breakdown in the insulation between the windings in the trans-former. At the moment of breakdown, the input power was, as seen in figure 4.8a, Pin= 33.53V × (4.23A + 4.06A) ≈ 278W , at an output power Pout = 245W .

That gives an efficiency η = Pout/Pin≈88%

(a)Maximum tested power. (b) Broken transformer insulation mea-sured after testing.

Figure 4.8:Test with maximum power possible (a) and the resulting trans-former breakdown (b).

The transformer experienced insulation break down in the test. The insulation between the secondary windings was damaged, as seen in figure 4.8b.

Nearly half of the power loss was in the transistors and the next biggest part was transformer conduction losses, see figure 4.9.

Other/sens is all the logic components and the sense resistor. The chart does not include transformer core loss.

(46)

34 4 Result (a) Chart of losses. (b) T able of losses. F igure 4.9: C al cula tions of the pow er loss in the con v erter with a visual represen ta tion in (a) and v al ues in (b).

(47)

5

Discussion

The design has been proven to work in some conditions. Several problems were found and no full system test with solar panel and battery was made.

5.1 Result

Since the transformer winding insulation was damaged during testing and time was short, a full test against the specification could not be performed. The simu-lated system showed similar results as in the tests, but problems with oscillations were not experienced in simulation. This is because actual data on transformer and other components were not available during design and simulation, leading to a design that performed well with somewhat different components. As simula-tions showed a working design, the only problem seems to be that there were not enough iterations of the prototype design. This is done as the project continues.

5.2 Power and oscillations

Although there were oscillations on the secondary side, the converter was able to output 455V at 245W which makes it possible to top charge the battery. The transformer experienced over voltage due to the oscillations and the insulation broke down. It should theoretically tolerate more than the measured 1000V peak ripple, but as it was hand wound by the author who is not an experienced trans-former winder, best practices may not have been followed. Also as the bench test was iterated, the transformer was soldered and desoldered several times, possibly weakening the insulating material on the winding.

The system was equipped with a hardware power line communication converter

(48)

36 5 Discussion

to be able to communicate using the usart protocol. There were not enough time to test this part, it was postponed to further development.

The converter had an efficiency of approximately 88 % at 245 W power output, close to the specified 90 %. Other transistors may be used that has lower switch-ing losses and lower on state resistance, as the transistors had a significant power loss. The transformer bobbin did not allow for the thick wire of the primary, which was wounded last onto the transformer. Therefore thinner wires (half the diameter) were used and the transformer became a big contributer to heat dissi-pation in the system. This was possibly making it fall below 90%. In the future, copper foil may be used for the primary, as it only needs a couple of turns, or another core and bobbin. The resistive loss in the transformer will not be a prob-lem.

5.2.1 Method

When measuring the efficiency, measurements were taken with multimeters at the output, but not at the input. The input power was calculated from the power supply display readings, which had its outputs parallel connected, possibly lead-ing to higher current readlead-ings than what actually was used by the converter. This is because of the nature of parallel connections which in theory could have cur-rent flowing between the outputs. It is possible that the actual efficiency was higher than 90 % in that test.

Because of the limited time, every possible analysis and precaution could not be done. A system test with an actual solar panel and battery was only done in simulations with the characteristics of the solar panel and battery. In reality, characteristics may differ and the converter design may need to be modified. The tests with battery and solar panel will be conducted during the summer with a maximum power point tracking routine. This fall, the finished solar power solution will be tested in Africa, probably Kenya or Mali, by me and the other thesis workers.

5.2.2 Topology

When converting with such a high ratio (over ten times) of input to output volt-age, the switching duty cycle goes to extreme values, leading to high current peaks and high flux in the inductor. As there is a limitation of the size of the con-verter, the inductor core size is limited and Pcoreis easily getting out of hand due

to the high flux density in the core. Because of the high input to output voltage ratio, a converter with a transformer was selected. The cost, complexity and con-struction time for the different transformer converters was assumed not to differ much. Only efficiency was considered.

5.3 Solar converters in a wider context

As solar power emerges in Africa, less wood and coal is burned, leading to lower deforestation and lower emissions. Cooking food on an induction cooker is much

(49)

5.3 Solar converters in a wider context 37

faster than over fire, freeing up time, especially for women in Africa. The modu-larity of the system creates possibilities for the local population to install it with-out technical expertise, possibly creating jobs in rural areas. When people stop burning wood and coal indoors, indoor air pollution decreases drastically, lead-ing to fewer deaths. If the project is successful, people in Mali can get more self-sufficient on electricity, wood and coal. The expected increase in prosperity for the rural population of Mali might lead to an increased possibility of finding a peaceful solution to the civil war.

(50)

(51)

6

Conclusions

The designed converter was able to step up voltage from 31 V to 450 V. Output power up to 245 W was tested with an efficiency of 88 %. Oscillations occurred at the secondary side, leading to breakdown of insulation in the transformer. This limited testing and implies that more care must be taken when designing the sec-ondary side, as well as there is a need of a snubber circuit here. Communication methods were presented, with minimal changes to the current prototype.

From what can be seen in the result, the converter should work in the given con-ditions if the problems with oscillations are solved. Further development is un-dergoing. The converter fitted into the given dimensions and should be able to reach above 90 % efficiency, given that better transistors are used and that the transformer windings are redesigned.

All the specifications should be fulfilled if the specific problems with the trans-former and oscillations are fixed. Suggestions for communication methods were presented, not needing extra hardware on the converter, other than maybe a ca-pacitor and protection circuits.

The maximum output voltage is enough to fully charge the battery. Although output power over 245W was not tested, no other problem than the oscillation that could limit this was seen.

(52)

(53)

Bibliography

[1] Engineering design. URL https://mynasadata.larc.nasa.gov/ engineering-design/. Accessed: 2017-03-01. Cited on page 17.

[2] 1700V SiC Schottky Diode. GP tech group, Aug 2015. Available at http:// www.gptechgroup.com/pdf/GP2D005A170B.pdf, Rev. 2. Cited on page 32.

[3] Shengyi Liu and R. A. Dougal. Dynamic multiphysics model for solar ar-ray. IEEE Transactions on Energy Conversion, 17(2):285–294, Jun 2002. ISSN 0885-8969. doi: 10.1109/TEC.2002.1009482. Cited on pages 5 and 6. [4] 8-bit AVR Microcontrollers ATmega328/P DATASHEET

COMPLETE. Microchip/Atmel, Nov 2016. Available at http://ww1.microchip.com/downloads/en/DeviceDoc/

Atmel-42735-8-bit-AVR-Microcontroller-ATmega328-328P_ Datasheet.pdf. Cited on page 20.

[5] Chen Min, Member Student, Rincón-mora Gabriel A., and Member Senior. Accurate electrical battery model capable of predicting runtime and i-v performance. 2012. URL https://login.e.bibl.liu.se/login? url=https://search-ebscohost-com.e.bibl.liu.se/login. aspx?direct=true&AuthType=ip,uid&db=edsbas&AN=edsbas. ftciteseerx.oai.CiteSeerX.psu.10.1.1.227.4482&lang=sv& site=eds-live&scope=site. Cited on pages 19 and 25.

[6] N. Mohan, T.M. Undeland, and W.P. Robbins. Power electronics: convert-ers, applications, and design. Number v. 1 in Power Electronics: Convertconvert-ers, Applications, and Design. John Wiley & Sons, 2003. ISBN 9780471226932. URL https://books.google.se/books?id=ToYoAQAAMAAJ. Cited on pages 9, 10, 11, 13, and 25.

[7] PPAM Onyxium 350W. PPAM solkraft. Available at http://ppam.se/ wp-content/uploads/FolderOnyxium350.pdf, Accessed: 2017-05-28. Cited on page 2.

[8] H. Raee, A. Rabiei, and T. Thirnger. Analytical prediction of switching losses

(54)

42 Bibliography

in mosfets for variable drain-source voltage and current applications. In 2013 IEEE 8th Conference on Industrial Electronics and Applications (ICIEA), pages 705–709, June 2013. doi: 10.1109/ICIEA.2013.6566458. Cited on page 13.

[9] B. Raja, M. R. S. Kumar, S. Vikash, and K. Hariharan. Maximum power point tracking in solar panels under partial shading condition using equilibration algorithm. In 2016 International Conference on Communication and Signal Processing (ICCSP), pages 2073–2077, April 2016. doi: 10.1109/ICCSP.2016. 7754543. Cited on page 6.

(55)

(56)

(57)

A

Schematics

Schematics are not published because of company policy.

(58)

(59)

B

Layout

Layout is not published because of company policy.

(60)

DC-DC Converter Design for Solar Power in Hot Environments

Bachelor of Science Thesis in Electrical Engineering

Department of Electrical Engineering, Linköping University, 2017