Split Screen Architecture: High speed data transmission in industrial machines and vehicles

(1)

UPTEC E 19019

Examensarbete 30 hp Augusti 2019

Split Screen Architecture

High speed data transmission in industrial machines and vehicles

Sebastian From

(2)

Teknisk- naturvetenskaplig fakultet UTH-enheten

Besöksadress:

Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0

Postadress:

Box 536 751 21 Uppsala

Telefon:

018 – 471 30 03

Telefax:

018 – 471 30 00

Hemsida:

http://www.teknat.uu.se/student

Abstract

Split Screen Architecture: High speed data transmission in industrial machines and vehicles

Sebastian From

This thesis examines different interfaces to suggest a concept of a split screen architecture for a company who develops and manufactures computers with integrated displays for industrial machines and

vehicles. Splitting the display from the computer requires a high speed cable link between the display and the computer. This cable link must reach the mechanical and electromagnetic requirements of the intended environment the computers will operate in. Interfaces were found that can send the required video and control data over the same twisted pair or coaxial cables up to 15 m using serializers and deserializers.

To send data in high speed puts tough electromagnetic requirements on the cable link as to not interfere or be interfered by nearby

electrical systems. Electromagnetic properties of different cable solutions were compared to find a suitable cable that is not too expensive for the intended cable lengths.

The study shows that there are interfaces available which can send several data types in a single cable so that a touch display can be positioned several meters away from the computer in demanding environments without losing the functionality it had when integrated in the computer.

Ämnesgranskare: Uwe Zimmermann Handledare: Bo Lööf

(3)

Populärvetenskaplig sammanfattning

CrossControl tillverkar datorer för industriella fordon och maskiner. Deras skärmar är idag integrerade i deras datorer och med det här arbetet undersöks möjligheten att separera skärm från dator så att skärmen blir fristående och kopplas ihop med datorn med en enda kabel.

Att skicka data i höga hastigheter i industriella fordon och maskiner ställer höga krav på kablar för att inte störas av eller störa elektrisk utrustning i närheten samtidigt som signalerna i kabeln inte får påverkas för mycket av kabeln. Detta kräver skärmning av kablaget som hindrar både högfrekventa och lågfrekventa störningar och impedansmatchning av kabel och kontaktdon.

En heltäckande tunn skärm är bra för att stoppa högfrekventa signaler och en tjockare väv kan användas för att stoppa lågfrekventa signaler. Genom att använda impedansmatchade transmissionsledningar minskar reflektion av signalerna vilket gör att signalerna kan ha låga spänningsnivåer utan att förlora data och på så vis minska strömförbrukningen och störningar som kablaget kan skicka ut.

Olika gränssnitt för att skicka video har undersökts, men då de flesta gränss- nitt är utvecklade för kontorsmiljöer är det fördelaktigt att använda gränssnitt som är utvecklade för tuffare miljöer. Att skicka data i en enda kabel kan spara plats och pengar då endast en skärmad kabel krävs. Dett kan göras med serialiserare och deserialiserare för att skicka parallell data seriellt med samma datahastighet men färre ledare. På så vis kan olika sorters data skickas i samma ledare, så att video kan skickas från en dator till skärm och styrdata kan skickas bidirektionellt i samma ledare.

Det finns olika serialiserare och deserialiserare som alla använder liknande teknik för att skicka data. De undersökta kretsarna är gjorda för att serialisera och deserialiser video- och styrdata för att kunna skicka video från en dator till en skärm samt kunna läsa av exempelvis beröring och knapptryck på skärmen.

En sådan implementation i en dator och skärm kan skicka video- och styrdata i ett tvinnat par eller i en koaxialkabel. Strömförsörjningen till skärmen kan injiceras till samma ledare, eller ske över ett separat par ledare. Det finns kablar med två tvinnade par där ena kan användas för data och det andra för strömförsöjning. På så vis krävs det endast en kabel mellan dator och skärm.

(4)

List of Figures

2.1 Transmission line model. . . 7

2.2 Boundary condition. . . 8

2.3 Transmission line ending in another transmission line. . . 9

2.4 Transmission line ending with a load. . . 9

2.5 Electric and magnetic dipoles centered around origin. . . 10

2.6 Braid and foil cable shield on a twisted pair cable. Twisted- Pair S-FTP by Hurzelchen (CC BY-SA 3.0) cropped with labels added [6]. . . 14

2.7 The jitter effect of the eye graph of a signal. . . 17

2.8 Single-ended signaling. . . 18

2.9 Differential signaling. . . 19

2.10 Noise induced in a differential conductor pair removed at the receiver using a subtractor circuit. . . 20

2.11 Typical FPD-Link III interface [35]. . . 26

2.12 Typical APIX3 interface. . . 27

2.13 Video and Ethernet over GMSL. . . 28

2.14 Splitting video stream with GMSL. . . 28

2.15 Power over Ethernet Option B. . . 29

2.16 Power over Ethernet Option A. . . 30

2.17 Power over Coax setup [35]. . . 31

2.18 Impedance of different inductor combinations over frequency [49]. 32 5.1 Proposed system overview of a split screen architecture. . . 48

5.2 Two displays connected to the same computer using two serializers. . . 49

(7)

List of Tables

2.1 Skin depth, speed, wavelength and characteristic impedance for electromagnetic wave propagation with different frequencies in copper [3]. . . 6 4.1 Specifications of two different coaxial cables from Beldin Inc.. 41 4.2 Electrical characteristicts of different multi-pair STP cables. . 42 4.3 Electrical characteristics for X-coded M12 connectors. . . 43 4.4 Electrical characteristics of Rosenberger and Amphenol RF FAKRA

connectors. . . 44 4.5 Electrical characteristics of HSD connectors from Rosenberger

and TE Connectivity. . . 45 4.6 Comparison between the features of the different SerDes inter-

faces. . . 46 5.1 Performance of selected cables. . . 49 5.2 The different cable options for high speed data transmission

using SerDes. . . 50 5.3 Performance of selected connectors. . . 51 5.4 The connectors suitable for high speed data transmission using

SerDes. . . 51 5.5 The cable solutions for high speed data transmission using SerDes. 52 5.6 Interface implementation prices. . . 53

(8)

Abbreviations

4PPoE = 4-Pair Power over Ethernet AC = Alternating Current

APIX = Automotive Pixel link

bps = Bit per second

DC = Direct Current

CML = Current Mode Logic

EMC = Electromagnetic Compatibility EMI = Electromagnetic Interference

EN = European Standards

ESD = Electrostatic Discharge FAKRA = Fachkreis Automobil FEXT = Far end crosstalk

FPD-Link = Flat Panel Display Link Gbps = Gigabits per second

GMSL = Gigabit Multimedia Serial Link GPIO = General Purpose Input Output HDMI = High-Definition Multimedia Interface

HSD = High Speed Data

IC = Integrated Circuit

IEC = International Electrotechnical Commission ISO = International Organization for Standardization LVDS = Low Voltage Differential Signaling

Mbps = Megabits per second NEXT = Near end crosstalk

OpenLDI = Open LVDS Display Interface PCB = Printed Circuit Board

RGB = Red, Green and Blue

RSDS = Reduced Swing Differential Signaling SerDes = Serializer/Deserializer

STP = Shielded Twisted Pair STQ = Shielded Twisted Quad

TIA = Telecommunications Industry Association TMDS = Transition Minimized Differential Signaling TTL = Transistor-Transistor Logic

USB = Universal Serial Bus

VSWR = Voltage Standing Wave Ratio

(9)

1 Introduction

Data communication between different devices has always been a challenge. A number of different interfaces exist to send specific data between computers and peripherals. More and more companies are working towards single multi- purpose interfaces to standardize data transmission. Some examples of this are the recent HDMI versions also including Ethernet connectivity instead of only transferring uncompressed video data and USB Type-C offering HDMI or DisplayPort pass-through of uncompressed video in its superspeed channels to increase the usability of USB [1],[2].

CrossControl is a company developing and producing automotive displays, computers and control units for e.g. forestry machines, tractors and trains.

Today, their lineup consists of control units and display computers with display and computer integrated into one unit. With an increase of demand for units in different sizes and with different functionality, while maintaining low cost, the question arises if display and computer can be separated to easier meet customer demands. To mix and match different computers and displays could offer tailored solutions at a lower cost for both CrossControl and customers.

One of their older display computers had the option to separate the display from the computer, but the cable used was very expensive and limited to a length of 5.5 m. The computer also featured video out to be able to support a second monitor. With the rise of high speed, multi-purpose interfaces, their goal is to separate display from computer into two separate units with a high speed, low cost interface between them.

CrossControl’s current system architecture with an integrated solution, hous- ing both display and computer in one unit works and is not considered to be phased out any time soon. However, it has its limitations. The different con- nections at the unit can cause the cable harness to be very big, which can be troublesome to route up to the machine operator’s position. Today, CrossCon- trol has a very well-defined lineup of units but development for future products can be limited by this. Changing the computer or display for a new revision can today require a total redesign of the unit.

Separating the display from the computer would ease the development of future products. This makes all displays and computers compatible with each other, making it possible to use the same computer and just upgrading the display, or using the same display, but upgrading the computer. This also

(10)

makes it easier for CrossControl’s customers to upgrade or replace units. If the display malfunctions in one of today’s units, the entire unit needs to be replaced since everything is integrated.

The flexibility of a single interface connecting computer with display would add the option to run two displays from the same computer. This is some- thing that has been requested by customers of CrossControl before but has not been a possibility with their current lineup of products. Instead, a customer would have to use two separate display computers which is very expensive for the customer who tries to keep costs low.

Separating the display from the computer requires a lot of thought and plan- ning. Keeping all features and connection options available at the display creates several problems. Therefore, some features have to be chosen to be placed either on the display or on the computer. The features left on the computer can be integrated the same way as in their current display computers while the features on the display have to communicate with the central processing unit in the computer. This means that a communications link is required between the display and the computer that can handle the required data throughput. Both the computer and display must also be connected to a power source. Power can be limited and power delivery can be uneven in battery powered vehicles and machines which calls for extensive filtering, which current display computers already handle.

This project gives the opportunity to analyze and examine digital interfaces, electromagnetic properties of electrical systems, as well as cost efficiency for possible future interfaces between CrossControl’s computers and displays.

1.1 Project description

CrossControl wants to know if they can achieve a split screen system architecture where they can mix and match different computers with different displays without creating a custom new unit. This could be achieved by splitting the display computer into a display unit and a computer unit. What is the specification of requirements for an interface between the display and the computer?

Is a custom interface required, or are there available solutions that can be adopted? How does the requirements affect cost and complexity of the end products? How is the synergy and flexibility affected by having two separate units instead of one?

(11)

1.2 Methodology

First, information was gathered about CrossControl’s existing lineup to com- pile a list of requirements that the interface must meet. This was done by analyzing circuit layouts of their existing products, analyzing electrical and environmental requirements as well as by looking at the specifications and circuit layouts of hardware intended for use in future products. To get a good overview of electromagnetic compatibility, course material was studied from the Electromagnetic Compatibility course held by Uppsala University.

The second part of this project was a market research to get insight in existing interfaces used in the industry for similar purposes. The information gathered during this part was from articles, journals and informative web pages. A few interfaces were selected to be evaluated in the third part of this project.

The interfaces chosen from the second part of this project were evaluated to try to see how they could be used for this project and what their pros and cons were compared to each other. The interfaces were compared to an older custom interface developed by CrossControl to see if they offered a better solution for this purpose. Different cable solutions were discussed and compared across the different interfaces. Prices of components and cable solutions were obtained by contacting suppliers and manufacturers and will not be disclosed in this thesis.

1.3 Delimitations

This project is limited to study CrossControl’s requirements for data transmission between their computers and displays. This means that some interfaces can exist that reach the physical requirements, but CrossControl has deemed them not useful because of hardware limitations, complexity or cost.

Long distance in the scope of this thesis is a distance of more than a couple of meters, but not further than 15 m. This is the approximate maximum distance between a computer unit and a display unit for the intended placement of CrossControl’s products in the machines.

No testing is to be conducted due to time limitations. Cables and connectors suggested are assumed to be compatible in terms of electrical characteristics.

(12)

2 Theoretical background

2.1 Electromagnetic Compatibility

During the transmission, reception and generation of electromagnetic energy, interference may be caused in other nearby electric systems. The concept of electromagnetic compatibility (EMC) has two main aspects. Firstly, the ability of an electric system to operate without interfering with other systems and secondly, the ability to operate as intended in a specified electromagnetic environment. This means that a system is electromagnetically compatible if it does not interfere with other systems, it isn’t susceptible to electromagnetic emissions from other systems and it doesn’t cause interference with itself.

Electromagnetic problems can be decomposed into three main parts. These are the source of emission, transfer of emission and reception of emission. This means the emitter, the coupling path and the receptor. To eliminate or reduce the effects of electromagnetic interference (EMI), one can suppress the source producing electromagnetic emission, making the coupling path as ineffective as possible and making the receptor less susceptible to emission.

2.1.1 Maxwell’s Equations

To understand the effects of EMI and how to reduce it, one must study Maxwell’s equations. These are four equations that explain the electromagnetic phenomena and describes how one electric system can cause EMI to other systems and how one electric system can be interfered by other system’s electromagnetic emission. Faraday’s law, Ampere’s law and Gauss’ law together are referred to as Maxwell’s equations.

Faraday’s law states that an electric field is generated in a magnetic field with varying magnetic flux with

∇ × ¯E = −∂ ¯B

∂t. (2.1)

Ampere’s law shows that a varying electric field produces a magnetic field and is defined as

∇ × ¯H = ¯J + ∂ ¯D

∂t . (2.2)

(13)

The magnetic field intensity vector is dependent on the current density vector and the partial time derivative of the electric flux density vector.

Gauss’s law is divided into two parts, the first for electric fields and the second for magnetic fields. Gauss’s law for electric fields,

∇ · ¯D = ρ_v, (2.3)

states that the net positive charge enclosed by a surface S is equivalent to the net flux of the electric flux density vector out of the surface S and can be seen in equations 2.3. Gauss’s law for magnetic fields,

∇ · ¯B = 0, (2.4)

shows that all magnetic field lines form closed paths.

2.1.2 Wave propagation

Electric fields, ¯E and magnetic fields, ¯H are uniform and located in the transverse plane to the direction of the propagation. In an xy-plane, this can be formulated as

∂Ex

∂x ,∂Ey

∂y ,∂Hx

∂x ,∂Hy

∂y = 0. (2.5)

This wave structure type is called transverse electromagnetic and wave propagation in transmission lines have this structure. In this xy-plane, the wave propagates in the z-direction and the amplitude of the electric field decreases exponentially in the material. The intrinsic impedance Zm of the medium can be calculated using

Z_m = s

jωµ

σ + jωε (2.6)

where µ is the permeability, ε permittivity and σ is the conductivity of the material. Zm in a free space, or vaccum, is roughly 377 Ω and in good conductors, such as metals, it is much smaller. In a lossless media, σ=0, the intrinsic impedance and the the wave velocity can be calculated using

(14)

Zm =r µ

ε (2.7)

v = 1

√µε = f λ. (2.8)

When a lossy media is used, meaning σ 6=0 and σ/ωε 1, equation 2.8 can still be used. When σ/ωε 1, the following equations are used instead:

Z_m ≈

rjωµ

σ =r ωµ

σ ∠45^◦ (2.9)

v ≈r 2ω σµ.

Zm is also called surface impedance. When a wave propagates in a medium, it travels along the surface of the conductor. An alternating current (AC) will create magnetic fields in a conductor, creating a force on the charge carriers and forcing them to move towards the surface of the conductor. This is called the Skin effect. The skin depth, δ, is the distance into the conductor where the wave amplitude, E0e^−αz, has decreased by 1/e, which is roughly 37%. For a good conductor, such as metals, the skin depth can be calculated using

δ =

r 2

ωµσ (2.10)

Table 2.1 shows different skin depths and characteristic impedance of copper at different frequencies for electromagnetic wave propagation.

Copper (σ_r=1, µ_r=1)

f 50 Hz 1 kHz 1 MHz 1 GHz

δ, m 9.35e-3 2.09e-3 6.61e-5 2.09e-6 v, m/s 2.936 13.13 415.2 1.31e4 λ, m 0.059 0.013 4.15e-4 1.31e-5 Zm, Ω 2.61e-6 1.17e-5 3.69e-4 0.012

Table 2.1: Skin depth, speed, wavelength and characteristic impedance for electromagnetic wave propagation with different frequencies in copper [3].

(15)

2.1.2.1 Transmission lines

When electromagnetic waves are used to transfer energy, it is done over transmission lines, often using two-wire lines. A transmission line will have a series resistance and inductance, as well as shunt capacitance and admittance. Fig- ure 2.1 shows such a transmission line. The R, L, C and G are given in per unit length, where ∆z is the length of the transmission line.

Figure 2.1: Transmission line model.

When a voltage V(z, t) is traveling along the transmission line with the speed v, the propagation speed of the waves as well as the characteristic impedance can be obtained using

v = r 1

LC (2.11)

Z₀ = rL

C. (2.12)

This shows that the propagation speed of a wave in a transmission line with vacuum between the conductors would be the speed of light. Without distor- tion or attenuation the solution to equation 2.11 is V(0, t ± ^z_v). This can be written as

V (z, t) = V (0, t − z

v) + V (0, t + z

v) = V⁺+ V⁻ (2.13) and the current can be written similarly as

I(z, t) = 1

Z₀[V⁺+ V⁻] = I⁺+ I⁻ (2.14)

(16)

and together they can also be used to find the characteristic impedance

V⁺

I⁺ = Z₀ (2.15)

V⁻

I⁻ = −Z0. (2.16)

To obtain the reflection coefficients Γ^V_R and Γ^I_R, a boundary such as seen in figure 2.2 must be considered.

Figure 2.2: Boundary condition.

From the boundary condition, this relationship can be obtained V_R

I_R = ZR = V_R⁺+ V_R⁻

I_R⁺+ I_R⁻ = V_R⁺+ V_R⁻ V_R⁺− V_R⁻

· 1

Z₀ (2.17)

that can be used to find the reflection coefficients

Γ^V_R = Z_R− Z₀

Z_R+ Z₀ (2.18)

Γ^I_R= −Z_R− Z₀

Z_R+ Z₀. (2.19)

This shows that -Γ^VR= Γ^IR. The power reflection coefficient can then be written as

Γ^P_R= Γ^V_RΓ^I_R= Z_R− Z₀ ZR+ Z0

2

. (2.20)

The reflection coefficient, Γ is the square root of the power reflection coefficient, q

Γ^P_R = Γ, (2.21)

(17)

and is used to represent the amount of reflection in a transmission line [4].

When a transmission line with characteristic impedance Z0 ends in another transmission line with characteristic impedance Z1, the line with characteristic impedance Z1 can be seen as a series resistance to the transmission line with characteristic impedance Z0, as seen in figure 2.3.

Figure 2.3: Transmission line ending in another transmission line.

Developing further on this idea, a transmission line of length L and characteristic impedance ZC terminates with a load with impedance ZL as can be seen in figure 2.4.

Figure 2.4: Transmission line ending with a load.

Considering reflection at both ends, the steady-state solution for the voltage and current at a distance x along the transmission line, where l=L-x can be written as Vx and Ix and the impedance at this distance will be Zx.

At a distance x = 0, l = L, Zx can be seen as the input impedance, Zin of the transmission line. The input impedance varies with the frequency and the length of the transmission line. The input impedance is different from the characteristic impedance. However, if the transmission line terminates in a load that is equal to the characteristic impedance, the input impedance will

(18)

be equal to the characteristic impedance regardless of frequency or transmission length. That is, when ZL = ZC = Zin = Zx. The input impedance will always be equal to the characteristic impedance of an infinitely long transmission line and for transmission lines with finite length before the reflected wave reaches the point x = 0. Similarly, if the transmission line is very short compared to the wavelength, the input impedance will be equal to the load impedance.

2.1.3 Fields from dipoles

An ideal electric dipole can be seen in 2.5a with length l along the z-axis, carrying a current I0 and an ideal magnetic dipole with area A in the xy-plane can be seen in figure 2.5b, carrying a current I0.

(a) (b)

Figure 2.5: Electric and magnetic dipoles centered around origin.

For the electric dipole, the radiated electric field consists of r- and θ-directed components and the magnetic field consists of φ-directed components. For the magnetic dipole, the magnetic field consists r- and θ-directed components and φ-directed components for the electric field. The characteristic impedance of the medium is Z0 = p^µ_ε. The wave impedance is the ratio of the electric and magnetic field components perpendicular to the propagation direction, meaning that for an electric dipole it is ZW = _H^E^θ_φ and for a magnetic dipole it is ZW = _H^E^φ_θ. In the far field region, the wave impedance will be equal

(19)

to the characteristic impedance of the medium. In the near field region, the electric field will be high impedance, making the wave impedance larger than the medium’s characteristic impedance and the magnetic field will be low impedance, which means the wave impedance is smaller than the medium’s characteristic impedance. The maximum radiated electric field for an electric dipole can be calculated using

|Eθ| = Z₀I₀l

2rλ (2.22)

and the maximum radiated electric field for a magnetic dipole can be calculated using

|Eφ| = πZ0I0A

rλ² . (2.23)

This means that a 10 cm wire used as an electric dipole carrying 170 µA at 30 MHz or 22 µA at 230 MHz or a 10 cm long, 2×3 cm square-shaped magnetic dipole carrying 450 µA at 30 MHz or 8 µA at 230 MHz exceeds the maximum allowed radiation at a 10 m distance from the dipole, according to the CISPR22 standard [3],[5].

2.1.4 High frequency behaviour of electrical components

High frequencies changes the impedance of inductors and capacitors. There- fore it is important to consider the effects of this in conductors and circuit components, as their behaviour will depart from their ideal behavior. This is important when studying EMC as the behavior of each component is important to know for their indented use.

In direct current (DC), and with very low frequency signals, the current will flow uniformly over the entire cross section of the conductor, while with high frequency AC signals, the skin effect makes the signal flow at the surface of the conductor. Due to the smaller area that the current carriers are traveling in, the resistance is frequency dependent and will increase with higher frequencies as the conducting area gets smaller.

Conductors will also have internal inductance due to internal magnetic flux.

Wire inductance for round conductors with the return wire a short distance D from the conductor can be calculated using

(20)

L = µl π

ln 2D d

− D l

. (2.24)

resulting in approximately 1 µH/m for any conductor, mostly from external inductance.

There is no internal capacitance similar to inductance because the surface charges prevents electrical fields to penetrate the conductor, however, there will be capacitance between two conductors. The capacitance between the conductors can be calculated with

C = πε

ln ^2D_d . (2.25)

The capacitance is dependent of the conductor cross section shape, circuit configuration and conductor surroundings. Capacitance between conductors strongly depends on the distance between the conductors up until the distance is around 10 times larger than the diameter. Below this distance, the capacitance of conducting wires is usually between 5-15 pF/m. The external inductance and capacitance can be used to calculate the characteristic impedance Z0 of a transmission line with

Z₀ = rL

C = 120 ·r µ_r εr

2D d

. (2.26)

Electric circuit board components will also behave differently with high frequencies. A capacitor can be seen as a capacitor, resistor and inductor in series. An inductor can be seen as a resistor and inductor in series with a capacitor in parallel. A resistor can be seen as an inductor in series with a capacitor in parallel with a resistor and inductor in series.

2.1.5 Coupling paths

Coupling occurs when two systems interfere with one another. This can happen by crosstalk, where a signal in one circuit creates an undesired effect in another circuit, due to common impedance or electromagnetic coupling.

Crosstalk from common impedance happens when two systems share a conductor, often a reference or conducting plane. With a common reference, the currents in the two systems will affect each other as the reference will have an impedance, especially at high frequency signals. To avoid crosstalk one can

(21)

assign individual conductors for each system to a common reference.

Electromagnetic coupling is divided into magnetic coupling, also called inductive coupling, and electric coupling, also called capacitive coupling.

Capacitive crosstalk occurs between two current loops with a common reference plane. The capacitive crosstalk increases with high frequencies and with increased circuit impedance. Capacitive coupling can be reduced by reducing conduction surface area and increasing the distance between the conductors, keeping the frequencies as low as possible and using metallic screens as shielding.

The magnetic fields from two parallel current loops will cause inductive coupling and crosstalk. The inductive crosstalk increases with frequency and with decreased circuit impedance. Decreasing loop area, decreasing distance between loops, keeping loops perpendicular to one another, cancelling out magnetic fields or keeping frequencies low will decrease inductive coupling.

Crosstalk from both inductive and capacitive coupling is generally present in electric circuits. Crosstalk in transmission lines is measured in near end crosstalk (NEXT) and far end crosstalk (FEXT). NEXT is measured on the same end while FEXT is measured at opposite ends and both are given in dB [3].

2.1.6 Shielding

Shielding is used to protect electric systems from external interference, as well as preventing electric systems from interfering with external systems. For a low frequency approximation, the screen voltage will be zero everywhere if the screen is connected to ground at both ends, preventing capacitive and inductive crosstalk. At higher frequencies, the inductive crosstalk is kept constant after the cutoff frequency of the shield, and as the capacitive crosstalk is kept very low if the shield is grounded at one or both ends, the crosstalk is mostly dependent on the inductive coupling. Most of the external electric fields are reflected on the shield surface, and the magnetic fields are attenuated by the shield, meaning that the thickness of the shield is more important for shielding from magnetic fields.

It is important to take holes and apertures of the shield into consideration.

(22)

They effectively function as half-wave dipole antennas radiating electromagnetic fields. The shield must cover the entire electric system and not only the transmission line to prevent noise from being induced in the conductor.

Shielding transmission lines can be done by covering the cables in grounded metal foil or a grounded metal braid connected to the transmitter and receiver shields. Foil shield is used to prevent high frequency electromagnetic noise due to the lack of holes and can therefore be made very thin. Braid is used to prevent low frequency noise, this means that the braid must be thick accom- modate the skin effect and keep resistance low, but due to not covering the entire wire as it leaves holes in the shield, the braid is not ideal to shield off high frequency noise. This means that to shield both high and low frequency noise from EMI, a combination of foil and braid can be used, although this increases the cost. A cable with braid and foil shield can be seen in figure 2.6.

Figure 2.6: Braid and foil cable shield on a twisted pair cable. TwistedPair S-FTP by Hurzelchen (CC BY-SA 3.0) cropped with labels added [6].

A shielded cable can be seen as a transmission line between the conductor and the shield. Another transmission line can be modelled between the shield and ground plane. Because of the skin effect, low frequency fields can more easily penetrate the shield. This means that parasitic voltage between the conductor and shield will decrease with higher frequencies. However, the parasitic voltage induced by magnetic coupling will increase with higher frequencies.

As a shield is equally as effective in both directions, the effectiveness of a shield is the amount of noise that can penetrate the shield from either side.

The shield effectiveness, S, can be seen as the sum of reflection loss, SR, and absorption loss, SA, as

S(dB) = S_R(dB) + S_A(dB). (2.27)

(23)

The noise reflected by the shield is the reflection loss and the noise absorbed by the shield is the absorption loss. The amount of reflected noise is related to the resistance of the shield and the amount of absorption is related to the thickness of the shield. A thick shield with low resistance will therefore be more effective than a thin shield with high resistance. In practice, it is very difficult to calculate shield effectiveness due to imperfections such as joints, corners and holes [3].

2.1.7 Return loss and VSWR

Return loss is when power transmitted from the transmitter is reflected at the receiver. Return loss can be calculated using

RL = 10 · log₁₀(P_l

P_r) = 20 · log₁₀(Γ) (2.28) and is given in dB. Return loss is the relationship between how much power reaches the load, Pl, and how much power is reflected, Pr [7],[8].

As equation 2.20 shows, the power reflection coefficient is dependent of the relationship of the characteristic impedance of the cable and the termination load impedance. To reduce return loss, it is important to match the impedance of these.

Voltage Standing Wave Ratio (VSWR) is a function of the reflection and describes the power reflected with

V SW R = 1 + |Γ|

1 − |Γ| (2.29)

and is unitless [4]. A VSWR of 1 means that 0 % of the power is reflected and a VSWR of 2 means that 10 % of the power is reflected, making the return loss roughly 10 dB. VSWR is a parameter to show how well an antenna, or connector, is impedance matched to a transmission line.

2.1.8 Insertion loss

Insertion loss is a measurement of how attenuated a signal is when it reaches the load. It indicates the amount of signal loss from transmitter, Pin, to receiver, Pout with

(24)

IL = 10 · log₁₀(P_out

P_in) (2.30)

and is given in dB [3].

For all cables used for transmission, there will be signal attenuation due to conduction losses, reflection losses and dielectric losses due to parasitic capacitance. The loss of signal strength is proportional to frequency and length of the cable. The frequency affects the skin depth which increases impedance with increased frequency and the resistance remains almost constant per unit of distance. This means that even at DC, the signal will be attenuated. An- other important aspect that affects signal strength is the insulating materials used between wires and shield. A good dielectric will keep the parasitic capacitance between conductors as well as between conductors and shield low [3],[9].

2.1.9 Jitter and skew

Jittering is short-term variation of the signal that the receiver tries to recover.

This could cause the dislocation of data bits in a data stream due to for example insertion loss, return loss and crosstalk. Jittering can occur at three places of a transmission line: at the transmitter, the transmission line and the receiver. A system is often designed to have a jitter budget, determining how much jitter can be present in a system to determine robustness and bit-rate of the system.

A good way to measure jitter is to look at the signal’s eye diagram. The eye diagram is made by folding the individual bits of the signal on to each other and display them over the same unit interval of time, where one unit interval is the time it takes to transmit one bit of information [10].

The signals are affected by rise and fall times and can also be delayed by electrical components, making events not happen when they are supposed to.

The time difference between the actual and expected event is called skew.

Inter-pair skew comes from transmission lines having different lengths compared to each other in the same cable and intra-pair skew comes from different transmission lengths between conductors transmitting complementary signals within the same cable [11]. The jitter will distort the form of this eye, as seen in figure 2.7, making recovering the signal more difficult. By reducing the

(25)

amount of jitter, the chance to successfully retrieve the data is increased.

Figure 2.7: The jitter effect of the eye graph of a signal.

The total jitter in a system can be broken down to random jitter and deterministic jitter. Random jitter can be minimized by using a good clock signal and deterministic jitter can be minimized with equalization at the receiver [12].

2.1.10 Electrostatic discharge

Electrostatic discharge (ESD) is a very common phenomenon which causing surges in electrical circuits. ESD happens when charge is building up in an insulating material and quickly discharges to an effective ground or electrical component. The human body can for example be charged up to 10 - 20 kV with an energy for several mJ, this results in a peak current of 10 - 100 A for up to a few nanoseconds [13].

(26)

2.1.11 Overvoltage protection

A surge, or electrical transient, denotes a condition in which the circuit is stressed with an overvoltage for a very short period of time. This can happen for example due to ESD. This can cause severe damage in electrical components which can result in system failure, requiring system reboot or component replacement. Protecting from such surges can be done by limiting the current with a large impedance or diverting it from the circuit using a small shunt impedance. Such surge protection circuits interfere as little as possible with the system performance and the surge protection circuit should not be dam- aged by the surge. The most popular surge protectors are gas discharge tubes, metal oxide varistors and avalanche diodes [3].

2.2 Digital signaling topologies

Wire conductors are used to transmit data signals between or within systems.

This is often done using single-ended signaling, but differential signaling is also used for some applications.

2.2.1 Single-ended signaling

In singe-ended signaling, a signal is transmitted with a voltage in reference to a fixed point, usually ground, between or within electric systems. A single conductor carries the signal while another conductor carries the reference, as seen in figure 2.8. This means that each individual signal requires an individual conductor, but several signal carrying conductors can share a common ground cable [14].

Figure 2.8: Single-ended signaling.

(27)

2.2.1.1 High speed interfaces using single-ended signaling

Single-ended signaling is commonly used for very short distance data transmission with frequencies below 200 MHz [15]. Such interfaces are for example complementary metal-oxide semiconductor logic and Transistor-Transistor Logic (TTL) and are often conducted on a Printed Circuit Board (PCB) or in silicon due to the short distances. TTL is a common interface for video data over several parallel transmission lines, and is often converted to other signaling interfaces if transmitted over longer distances and then converted back to TTL at the receiver.

2.2.2 Differential signaling

Differential signaling use two complementary voltage signals in two conductors to transmit signals. One conductor carries the signal, and the other carries the same signal, but inverted. The information of the signal is determined by the difference between the voltage levels in the two conductors. The two conductor’s differential-mode voltage levels are equal but opposite polarity compared to a common-mode voltage, as seen in figure 2.9. The signals are balanced, meaning the voltages cancel each other out and the ground connection will have zero current flowing through it.

Figure 2.9: Differential signaling.

As every signal requires two conductors, differential signaling requires almost double the conductors as single-ended signaling. However, because differential signaling has very little return current or none at all, the ground conductor becomes less important [14].

(28)

Another important thing to take into consideration when choosing between single-ended and differential signaling is that EMI and crosstalk will have equal effect on all conductors. Because the conductor voltages are equal but opposite polarity in differential signaling, the induced noise due to EMI will be cancelled out at the receiving end, as can be seen in figure 2.10. This means that EMI and crosstalk has less impact on differential signals than single-ended signals.

Figure 2.10: Noise induced in a differential conductor pair removed at the receiver using a subtractor circuit.

Outgoing EMI will also be affected by the choice of signaling. Both single- ended and differential signaling will generate EMI to their surroundings. How- ever, due to the nature of the differential signaling, using twisted pair conductors will cause the generated electromagnetic fields to cancel each other out if the wires are twisted in such a way to minimize the distance between the wires.

It is much more difficult to suppress electromagnetic emissions for single-ended signaling [14].

2.2.2.1 High speed interfaces using differential signaling

Differential signaling is used by a lot of high speed data interfaces. Using low voltage levels, a lower power consumption and low electric emissions can be achieved. Further improved performance is achieved by the termination re- sistors, which reduces reflections during transmission. Such interfaces are for example Low Voltage Differential Signaling (LVDS), Reduced Swing Differen- tial Signaling (RSDS), Transition Minimized Differential Signaling (TMDS) and Current Mode Logic (CML) [15],[16],[17] .

LVDS is a standardized interface covered by TIA-644 and a common variant of this is Open LVDS Display Interface (OpenLDI) which is commonly used

(29)

to connect video data to a display where it is converted to TTL. LVDS use voltage swings of up to 500 mV between the logic states [16]. RSDS is a further development of LVDS, using lower voltage levels and lower power consumption compared to LVDS [17].

CML has faster rise and fall times than LVDS and RSD. This means the switching rate can be higher, offering higher data rates [18]. TMDS works very similar to CML, but use a specific form of 8-bit/10-bit coding to reduce EMI [19].

2.2.3 Bit rate and channel bandwidth

Sending digital bits implies sending either 1’s or 0’s. In digital data transmission, these are often represented by different voltage levels of a signal. The bit rate of a signal is how many bits that can be transmitted over a specified time, often given in bits per second (bps). The bandwidth is then the frequency spectrum that the transmission is done across. The bit rate of a transmission can be linked to the transmission bandwidth using

R = B · log₂M (2.31)

where R is the channel capacity bit rate in bits per second, B is the bandwidth in Hz and M is the level of modulation [20]. If there is no modulation, i.e. two signal levels, the bit rate equals twice the bandwidth.

2.3 Cable and connector solutions for electromagnetic compatibility

2.3.1 Twisted pair cable

A lot of the interfaces using differential signaling use twisted pair cables. Un- shielded twisted pair cables are readily available and cheap. A differential twisted pair consists of two conductors. Each conductor has insulating sheaths of the same length and are twisted together inside the cable sheath. Twisted pair cables often come in single, dual and quad pairs. Twisted conductors offer good EMC and shielding them improves this further. Shielded Twisted Pair (STP) cables are more expensive, but have better electrical characteristics. Twisted pair cables can be shielded in two ways. The cable can have an

(30)

overall shield that prevents external EMI. The overall shield can either be a braid to suppress low frequency electromagnetic noise, a foil to suppress high frequency electromagnetic noise or both to suppress a wider frequency band.

Each twisted pair can also be individually shielded to prevent crosstalk within the cable. To suppress a large frequency band of electromagnetic noise, an overall shield can be combined with shield over each individual differential pair. This is very expensive, but offers very good immunity to EMI.

As a twisted pair cable can consist of several twisted pairs, causing the twisted pairs to not be an equal distance from the noise sources. A solution to this when using two twisted pairs is to use a twisted quad cable, also called star- quad cable, where the four wires of the two pairs are twisted around the same core. This causes noise to affect both pairs equally. This topology also decreases crosstalk between the differential pairs in the cable. Shielded twisted quad (STQ) cables are common in two pair differential cables.

Twisted pair cables usually have a 100 Ω characteristic impedance. A transmission line using twisted pair cables should pair the termination impedance with the characteristic impedance of the cable to reduce crosstalk [21].

2.3.2 Coaxial cable

Coaxial cables are used when EMC requirements are high, for example for high speed data transmission over long cables. A coaxial cable consists of a conductor surrounded by insulating material, a shield covering the insulating material and an outer insulating material. The shield is grounded, allowing the conductor to use single-ended signaling. Coaxial cables usually have a 50 Ω or 75 Ω characteristic impedance and are often quite cheap and readily available [21]. Due to the good impedance control and noise reduction of a coaxial cable, a CML signal can be passed through the coaxial cable without the need for a differential pair.

2.3.3 Connectors

The connector is an important part of data transmission. A large connector can be seen as electrically long, essentially acting as a transmission line segment.

This means that it is important to control the impedance of the connector. It is also important to consider that twisted pair cables will have a short distance

(31)

of being untwisted near the connector and this will affect the impedance. For coaxial cables, the shielding will be connected to ground within the connector which can also affect impedance of the connection. Because of this, a larger impedance tolerance is usually allocated to the connector regions while the cable itself have a tighter impedance tolerance [21].

2.3.4 PCB layout

When sending signals to the connector over the PCB, it is important to keep the PCB traces at equal length for differential signals. Different lengths could affect high frequency signals in several ways. The most important aspect of high frequency signals is timing. Different propagation lengths could mean that signals arrive at the receiver at different times. A differential signal would lose its advantages if timing was off as the common mode noise rejection would not work and electromagnetic emissions would increase. Different trace lengths can also affect the impedance and cause unbalance in the transmission line.

It can also be beneficial to add an ESD protection circuit at the connector to prevent discharges to reach sensitive circuitry.

When designing PCB traces for high speed data signals, it is also important to not use right-angle bends and try to avoid vertical interconnect access points.

The traces should be at least two times the width of the traces apart for single-ended signaling and two times the spacing of the differential pair, and differential pairs should be placed as close as possible to each other to minimize noise [22].

2.4 Interfaces for high speed video and data transmission over long distances

2.4.1 Commercial interfaces for video transmission

HDMI is used to transfer video and audio signals over a standardized interface created by HDMI Licensing, LLC. The video and audio is transmitted over three TMDS data channels and one TMDS clock channel, making four differential pairs in total. HDMI can carry many different pixel sizes and resolutions, and as of HDMI 1.4, an extra differential pair can be used to pass through 100 Mbps Ethernet [23]. The HDMI specification does not include a maximum length, but a maximum distance of 3 m could be achieved with passive cables when running data rates of up to 48 Gbps using HMDI 2.1. Slower

(32)

speeds such as using HDMI 1.3a or HDMI 1.4 can be used on passive High Speed cables of lengths up to 5 m. However, running HDMI over coaxial or twisted pair cables can increase range to above 50 m, but requires converters at transmitter and receiver [24].

DisplayPort is another modern interface used to transfer video and audio and is developed by VESA. DisplayPort use differential signaling over four lanes featuring embedded clocks on each lane [25]. DisplayPort can achieve up to 8.1 Gbps per lane, with a total data rate of 32.4 Gbps [26]. In excess of the high speed data lanes, an auxilliary channel is available for half-duplex bi-directional data such as I²C or USB 2.0. DisplayPort can be converted to and from HDMI using passive converters. Similar to HDMI, DisplayPort does not state a maximum cable length, but instead state a minimum bit rate for different cable lengths to be expected. For example, DisplayPort cables over 2 m must support at least 5.4 Gbps per lane, although it does not state if the cables must do so passively [27].

USB is used by many peripherals and connects many devices and computers together. The current USB version is USB 3.2 Gen 2×2 and features extra dual simplex channels for SuperSpeed mode allowing 20 Gbps transfer rates over these channels using shielded differential pairs or coaxial cables, as well as offering full backwards compatibility over a single differential pair offering 480 Mbps bi-directional communication. The USB Type-C connector can be connected either way, while type A and type B can only be plugged in one way. The extra SuperSpeed channels can be used to pass through other signals, such as HDMI or DisplayPort by reassigning pin functions. USB 3.2 Gen 2×2 with 20 Gbps data rates can use Type-C cables up to 1 m, USB 3.2 Gen 2×1 with 10 Gbps data rate can use Type-C cables up to 2 m [2],[28]. The USB specifications does not allow the use of extensions cables by connecting an additional cable in series [29].

These are not the only common interfaces found for commercial products for video and high speed data transfer. However, due to these interfaces being primarily used for commercial and office products, they are not designed for automotive or industrial requirements. Special automotive cables and cables less susceptible to EMI are available, but for an increased cost. The limited cable lengths and loose requirements for EMC make these solutions non-ideal for use in industrial vehicles and machines. CrossControl has earlier conducted standardized tests on their units with USB devices connected using different shielded cable lengths with varying results.

(33)

2.4.2 SerDes

Serializer/Deserializer (SerDes) is a way to group several signals together to increase throughput and reduce the number of conductor wires required and then separating the signals after transmission. The transmitter consists of a multiplexer to combine the signals into one and the receiver consists of a demultiplexer that splits the signal back to its original parallel signals. The data clock can be parallel to the data or embedded in the transmission using start-stop bits, encoded to the signal or using interleaving bits [30].

Parallel clock requires an additional signal pair that the receiver will use to deserialize the data from the serialized data streams in the data signals. This is used for example in the LVDS interface. Parallel clock SerDes is an inexpen- sive SerDes method, but requires more wiring compared to using embedded clock.

Serializing the clock and data to a single bus lowers the required signaling wires down to a single pair. Embedding the clock using start-stop bits allows the receiver to synchronize with the embedded clock without external inter- vention. The start-stop bit SerDes method has very relaxed requirements for transmitter and clock sources. Using 8-bit/10-bit encoding, every byte of data is encoded to a 10-bit code. A special symbol is sent to mark symbol boundary.

This makes it possible for the receiver to realign the data to the original byte.

8-bit/10-bit interleaving is well suited for byte oriented data packages but has very tight requirements for the clock. Bit interleaving takes the data from several serial bit streams and multiplexes them into one faster serial stream by interleaving the bits. This serialized data stream is then demultiplexed by the receiver into their original serial data streams. Bit interleaving requires very precise external clocks due to the high data speeds. Bit interleaving SerDes has a higher cost than the other methods, but can be used to increase speed of slower serial data streams such as multiplexing several LVDS channels to a single CML channel.

When serializing data, long runs of only ones and zeros can occur, ruining the DC balance of the data stream. These patterns are unwanted as it can be difficult to find when one bit ends and another bit starts. Encoding the 10-bit data words to obtain a spread spectrum of ones and zeros keeps the DC balance and keeps the occurrence of transitions dense [31].

(34)

2.4.2.1 FPD-Link

Flat Panel Display Link (FPD-Link) is an interface utilizing SerDes technol- ogy. FPD-Link is developed by Texas Instruments and was first released in 1996 to create a high bandwidth interface between graphics controllers and displays. It converts parallel single-ended TTL to a narrow LVDS interface, lowering cable costs, power consumption and EMI [32]. FPD-Link is very similar to OpenLDI and they are often interchangeable with each other.

The current version is called FPD-Link III and uses CML outputs to achieve a higher bit rate. FPD-Link III offers improved performance and additional features such as embedded clock signal using start-stop bits, bi-directional communication with an I²C control data channel and I²S audio transmission and a video data stream. When using dual output for higher data rate on some FPD-Link III integrated circuits (IC), an additional SPI or high speed General Purpose Input Output (GPIO) communications channel can be used as well. FPD-Link offers backwards and forward compatibility between the versions, as well as support for different video interfaces at transmitter and receiver, such as LVDS, parallel Red, Green and Blue (RGB) and HDMI. A typical FPD-Link transmission setup can be seen in figure 2.11. FPD-Link can transmit up to 3 Gbps over a single cable, and some serializers can output two signals using two cables or both pairs of a STQ cable to transmit up to 6 Gbps [33]. FPD-Link III can be used with coaxial cables up to 15 m and twisted pair cables up to 10 m [21]. Some FPD-Link III ICs supports single-ended transmission using coaxial cables as well, meaning that some ICs can be used with either twisted pair cables or coaxial cables [34].

Figure 2.11: Typical FPD-Link III interface [35].

(35)

2.4.2.2 APIX

Automotive Pixel Link (APIX) is also an interface that uses SerDes to serialize video data streams. APIX is an automotive high speed data interface developed by Inova Semiconductors and was first released in 2007 [36]. The current version, APIX3, had its first products released in 2017 and offers backwards compatibility with older APIX versions. APIX transfers serialized video, audio using I²S, control data using SPI or I²C as well as an Ethernet channel using CML. The control data use the SPI interface on the transmitter for APIX3, but the receiver can handle both SPI and I²C.

APIX ICs can all use twisted pair cables for transmission and some can also use coaxial cables instead. A typical APIX interface can be seen in figure 2.12.

APIX3 can reach speeds of up to 12 Gbps upstream using two twisted pairs in a star-quad cable or two coaxial cables and a single coaxial or STP cable can reach speeds up to 6 Gbps. The downstream data rate can reach up to 187.5 Mbps. It also has the ability to serialize two independent video streams using HDMI input, transmitting to two independent displays simultaneously using up to 10 m cables [37]. Older APIX variants can reach up to 3 Gbps transmission speed with Ethernet, I²C or SPI, I²S and video with OpenLDI or parallel RGB video interfaces. This requires two cables, one for video and one for full duplex data communication and can be used with up to 12 m cables [38].

Figure 2.12: Typical APIX3 interface.

The cable link suggested for the first APIX version should have electrical characteristics of a STP category 6 cable with -17.7 dB attenuation at 0.5 GHz or better with a connector with 3 dB bandwidth of minimum 1.5 GHz for 1 Gbps transmission [39].

(36)

2.4.2.3 GMSL

Maxim Integrated developed their automotive video link Gigabit Multimedia Serial Link (GMSL) based on SerDes for infotainment systems and camera driver assistance systems. The link is capable of transmitting uncompressed video data as well as I²S audio, peripheral control communication using UART or I²C up to 15 m with up to 3.12 Gbps data rates using CML. All GMSL ICs support differential signaling using twisted pairs and some can also use coaxial cables for transmission. The next generation, GMSL 2, which is not yet commercially available, will also feature a gigabit Ethernet channel over the same cable and reach data rates of up to 6 Gbps [40],[41]. An illustration of this can be seen in figure 2.13. GMSL embeds the clock to the same cable as the video and control data. The inputs of the serializers and outputs of the deserializers varies between different ICs but can for example be LVDS or HDMI for display compatibility [42]. This means that the serializer input can be HDMI while the deserializer output can be LVDS [43].

Figure 2.13: Video and Ethernet over GMSL.

GMSL 2 variants will also feature the ability to split video streams which makes it possible for a video data stream to be split in the serializer and sent to two separate units for deserializing, as seen in figure 2.14 [44].

Figure 2.14: Splitting video stream with GMSL.

(37)

2.4.3 Power delivery

Power delivery between systems is done using transmission lines. For automotive and industrial battery powered applications, the main power source becomes the battery offering DC voltage levels. The voltage can be stepped up or stepped down to achieve application specific voltage levels.

The most common way for power delivery in interfaces is a separate transmission line for DC voltage levels in the interface cable. A typical example is a USB cable. The USB cable has separate wires for power delivery and keeps them separated from the data cables. The USB Type-C connector and cable has for example four dedicated pins for VBU S, which is the voltage bus used for Power Delivery which can supply up to 100 W with up to 20 V [45].

Another way of delivering power over a cable is to use Power over Differential Pairs. This means that one or more pairs in a twisted pair cable are dedicated to power delivery. This is used for example by Power over Ethernet. There are three ways to use Power over Ethernet: Option B, Option A and 4PPoE specified in IEEE 802.3af, IEEE 802.3at and IEEE 802.3bt [46],[47]. Option B is where two of the four conducting pairs are used for only power delivery and two pairs for only data, as can be seen in figure 2.15.

Figure 2.15: Power over Ethernet Option B.

Injecting the DC bus to the signal wire will not have any effect on the data

(38)

signals. Power over Ethernet offers two solutions called Option A and 4PPoE where the DC bus is injected to the data wires in such a way. Option A uses two of the conducting pairs, as can be seen in figure 2.16 and 4PPoE use all four conducting pairs for data and power [46].

Figure 2.16: Power over Ethernet Option A.

Power delivery and data over the same conductor can be done by dedicating a frequency band for power, for example by using 0 Hz frequency DC power.

Using this principle, it would be possible to transfer data and power over the same transmission line using the DC offset for power and filter out the high frequency signals data signals by AC-coupling.

FPD-Link III and GMSL have solved the power delivery in this way [35],[43].

The power can be transmitted over the same conductors used for data transmission. This is called Power over Coax when using coaxial cables, but the same principle applies when sending power over the same differential pairs as the data, similar to Power over Ethernet Option A and 4PPoE [48]. A typical Power over Coax setup can be seen in figure 2.17. Power over Coax splits the signal into two branches, the DC power and the data.

By AC-coupling the data signals with a capacitor at the transmitter and receiver, the data signals are not affected by the DC voltage. However, as the transmission line has a controlled impedance, the filter at the power injection

(39)

Figure 2.17: Power over Coax setup [35].

and separation must have high impedance to not interfere with the data signals. It should be at least 20 times higher than the characteristic impedance of the transmission line [35]. Because of the wide frequency range of the data signals, a single inductor will not suffice due to parasitic capacitance at high frequencies. A coaxial cable with characteristic impedance of 50 Ω requires a filter impedance of at least 1 kΩ. A single 100 µH inductor will handle frequencies up to 70 MHz before parasitic capacitance will decrease the filter impedance below 1 kΩ. This inductor will therefore only block the control data signals, so adding more inductors in series with lower inductance will block the higher frequency video data. Figure 2.18 shows how two or three cascaded inductors impedance varies with frequency. To keep the physical footprint of a Power over Coax circuit small, increasing voltage will decrease current over the cable. This allows for smaller components to be used [35].

(40)

Figure 2.18: Impedance of different inductor combinations over frequency [49].

(41)

3 Interface and cable link requirements

3.1 Hardware limitations

The hardware used by CrossControl offers a few limitations. Their goal is to keep costs as low as possible, while still delivering the features that their customers want. This means that some of their units will have different features such as different peripheral interfaces, different processing units and different display sizes and resolutions.

The computer unit and display unit would be placed in different locations of the vehicle. The display would be accessible by the vehicle operator, while the computer could be hidden away. Apart from video data over the display interface, control data and audio also has to be transferred both ways over the same cable. The control data includes sensor data gathered from the display unit, backlight control and button data. The display should have a button to boot the computer from a low-power mode where power consumption of the display and computer has to be very low.

3.1.1 Processors

CrossControl use different processors for different units. Both x86-based processors and ARM-based processors are used as central processing units for different computers. The display interface used should be common between the processors so that one display can be used with different processors using the same interface. The processor should be able to output two separate video feeds to two different displays, both displays connected using the same interface but to separate video outputs from the processor. A computer unit with a slower processor is allowed to only support one display in case two separate video outputs is too demanding. If the processor physically lacks the ability to output two separate video feeds with the preferred output signals, one video output is accepted, unless converting the video signals or using components compatible with the specific video interface does not add significant extra cost.

Video output interfaces supported by the current processors used by Cross- Control include, but are not limited to, OpenLDI, parallel RGB and HMDI.

It is therefore necessary that the interface should support at least one of these interfaces as input.

Split Screen Architecture: High speed data transmission in industrial machines and vehicles