development of electronic control system for formula one · 2017-11-28 · – 214 – development...

– 212 –

Kenichiro ISHII* Toshiyuki NISHIDA* Kohei TOSHIMA*

Masaki NEGORO* Masataka YOSHIDA* Yutaka MARUYAMA*

Development of Electronic Control System forFormula One

ABSTRACT

The application of electronic control systems has been rapidly increasing in Formula One cars as well as in othervehicles. System performance is a crucial element in conducting precision control and measurement of cars.

Starting with the 2006 season, the Honda works has been working to apply original Honda systems not only inthe engine control system, as before, but in all vehicle electronic control systems.

In order to pursue higher performance at the same time as enhanced in-vehicle mountability in Formula Oneelectronic control systems, it is necessary to carry out optimization of these systems together with thoroughgoingminiaturization. On-board systems have an electronic control unit (ECU) with integrated functionality linked by a high-speed network with units located in every part of the vehicle. High-speed telemetry is used to coordinate these withthe garage system in order to optimize systems.

The enhancement of unit performance by means of higher speed and greater precision contributes to heightenedcontrollability, and particularly to enhanced precision of driving force control and gearbox control. High-speedcommunication also contributes to greater measurement performance in the pit.

* Automobile R&D Center

1. Introduction

Formula One cars in recent years have incorporatedmore than just the engine control and gearbox controlfound previously. From mid-2001 to the 2007 seasons,traction control and engine brake control have come tobe allowed in the regulations. Clutch control was notprohibited up until 2007.

Traction control and gearbox control, in particular,require precisely coordinated control of the engine andchassis in real time. This has required advancedcomputational capability and measurement performance.Furthermore, Formula One cars demand aerodynamicperformance, so that there are limited on-boardinstallation locations for electronic control units inFormula One cars. This means that systems requiregreater compactness and lighter weight.

The functions of electronic control systems forFormula One are generally for two purposes, for drivingand for analysis. In order to achieve compactness andlight weight, the latter functions are assigned as muchas possible to pit systems. In this way the formerfunctions can be optimized, and this was the basicconceptual approach to the construction of these systems.

This article will provide an overview of Honda

Formula One electronic control systems, introduce theissues involved in development of the systems, anddescribe the contents of the development.

2. System History

Figure 1 shows the history of electronic controlsystems for the third era of Honda Formula One.Development of third-era Formula One systems began in1998 with a view to applying integrated engine andchassis systems to racing.

Under the BAR-Honda system from 2000 to 2005,engine control systems were supplied to the team. From2006, when Honda works joined in the competition, theHonda full system was provided as a chassis integrated

Fig. 1 History of Honda Formula One system

200820072006200520042003200220012000

FIA�system

British American Racing - Honda Honda Racing F1 TeamTeam

System

Honda System�“Athena”�

BAR System

Data�acquisition

Chassis Control Data AcquisitionTelemetry

Engine Control(PGM-FI)Data AcquisitionTelemetry Power Supply

Year

Honda R&D Technical Review 2009 F1 Special (The Third Era Activities)

– 213 –

management system. The system was named Athenaafter the Greek goddess of wisdom and victory.

In 2008, i t was made mandatory to applyInternational Automobile Federation (FIA) standardsystems. The purpose was to standardize engine andchassis control and to reduce costs. As a result, Honda’sin-house developed systems that were used in racingwere limited to the devices for engine use, sensors, anddata logging systems.

3. Configuration of Electronic ControlSystems (Athena)

3.1. Optimization of System Configuration (FunctionalIntegration and Distribution)

A car’s aerodynamic performance, which determinesits body design, makes a very important contribution tolap times. In order not to place constraints on thefreedom of the design, it is necessary to have systemconfigurations that are compact and have outstandingmountability.

The main locations for mounting electronic unit were,as shown in Fig. 2, beneath the radiator ducts on eitherside of the chassis. The only other locations were nearthe throttle pedal toward the front of the car and insidethe cockpit. The systems were therefore mounted in adistributed configuration and were linked by a high-speed network in order to achieve a balance ofmountability and functionality.

Figure 3 shows the system configuration. Thenetwork in the chassis is centered on the ECU and ties

together the various units, sharing sensor data and faildata to operate the systems. The ECU has thefunctionality for integrated control of the engine, chassis,and measurement. It acquires all information from thechassis and conducts the operation of every device aswell as the processing of measurements. Each unit thathas been distributed can manage the measurement andcontrol functions for the different applications, and isconnected to the network as necessary. Of these, theFront Data Acquisition (FDA) unit acquires data fromnumerous sensors at the front of the chassis. Since itprocesses signals used for real time control computations,i t i s connected wi th the ECU by high-speedcommunications at 10 Mbps.

In the pit, the client PC and the ECU are connected

Right layoutLeft layout

Electronic control unit assemblyElectronic power unit assembly

Fig. 2 Electronic unit layout (front view)

Front data acquisition�(FDA)

Steering wheel

AC-DC (PBOX)

Telemetry�(HTX)

Sensors

Sensors

Sensors

EngineGear Box

Differential gear

Sensor terminal (ST)

Arcnet�10 Mbps

SIO�2 Mbps

Ethernet�100 Mbps

C

Model base development

Program�setting data

Electronic control unit�(ECU)

Telemetry data receiver

Telemetry server

Team data logger�(TDL)

2008system -Logger

Video cam

era

Wing driver C

amera

GP

S

Tire tem

p/press

Laser

Ignition (CDI)

Server

CAN�1 Mbps

CAN�1 Mbps

CAN�1 MbpsCAN�

1 Mbps

CAN�1 Mbps

Fig. 3 Athena system

– 214 –

Development of Electronic Control System for Formula One

by Ethernet communications (100 Mbps) so that loggingdata and other high-volume data can be received in ashort time. All vehicle checking applications that handlehigh volumes of data are also placed on the client PCside in order to simplify on-board systems.

3.2. Explanation of Each Component(1) ECU

The ECU has three roles, in engine control, chassiscontrol, and measurement control. In engine, it controlsthe ignition timing, volume of fuel injected, and throttleopening, and realizes the requested torque. In chassiscontrol, it uses vehicle behavior information to performwheel drive torque transfer control and seamless shiftcontrol computation for gear shifting. In measurementcontrol, it performs measurement computation of twokinds, in telemetry and data logging.(2) Capacitive Discharge Ignition (CDI)

The CDI is the unit that supplies the energy forignition to the engine ignition coil. The CDI methodprovides greater real time effectiveness than the fulltransistor method and greater precision in drive torquecontrol (traction control and engine brake control) andengine speed control. This unit also simultaneously hasmisfiring detection functionality by means of ion currentdetection.(3) Power Box (PBOX)

The PBOX is the unit that operates as the voltageregulator and power distributor (sequencer function). Thevoltage regulator part uses a highly efficient convertermethod and supplies two power outputs (14 V and 7 V)so that the power components where this power issupplied can be made more compact.(4) Batteries

The system uses NiMH batteries for compactness andlighter weight. Battery temperature monitoring andcharge control are handled by the PBOX.(5) Telemetry transmitter and receiver (HTX, HRX)

The HTX is the data transmitter unit by which dataprocessed by the ECU is sent by radio to the pit. TheHRX is the data receiver unit that receives radio data.(6) Front Data Acquisition (FDA)

The FDA is the unit that acquires data from sensorsat the front of the chassis. It is equipped with input/output (I/O) capable of multi-channel analog input. Thisfunctions to process acquired analog signals by filteringcomputations and then transmits them to the ECU byhigh speed communications (10 Mbps).(7) Sensor Terminal (ST)

The ST is a compact sensor measurement unit usedfor measurement of the environment in the chassis andin the engine. Up to eight ST units can be connected tothe ECU and CAN communication lines.(8) FIA/FOM units

The FIA/FOM units include a data logger, camera,and GPS units. Installation of these units is required bythe Fédération Internationale de l’Automobile (FIA), theumbrella organization for automotive matters, and by theFormula One Management (FOM), the organization thatconducts racing. Car data from the ECU is sent to the

units by CAN communications, and the racing circuitdata for each team is managed and used by the FIA andFOM.(9) Team Data Logger (TDL)

The TDL unit has a logging function and reads datafrom all types of sensors. The use of FIA standardsystems has been a requirement since 2008. This unit istherefore developed so that assistance tool used up to2007 can continue to be used. The TDL is connectedwith FIA standard systems by CAN communications.(10) Garage system

The garage system is system infrastructure composedof the server installed in the pit and the client PCs usedby each engineer. It was designed exclusively for use inFormula One racing. As for the communication betweenthe car and PC, Ethernet was used.

3.3. High-Speed, High-Precision Control (Engine,Chassis, and Measurement Control)

The ECU was required to have high computationalcapability in order to provide satisfactory controlperformance and measurement performance. In the thirdera of Formula One, the main items for controldevelopment have been drive torque control andseamless shift control. The efficiency of this controldevelopment was enhanced by the introduction of modelbase development using floating point operations. Figure4 shows the controls that were applied, together with thechanges in computational capability. The regulationsmade driver assistance control (slip control, typified bytraction control) permissible in mid-2001. From then upto 2007, when it was outlawed, this demanded thegreatest computational capability. This also involved anincrease in the volume of data for control analysis, andthe CPU and data transfer speeds were increased in orderto be able to support the demand for high-speedsampling and logging on multiple channels formeasurement function.

1000

2000

3000

2003 2005 2006 2007

Traction control

Control

Tot

al c

alcu

latio

n in

stru

ctio

n

M/Y

Engine control�Measurement control

Chassis control�Measurement control

Seamless gearbox control

Torque control

(MIPS)

2001

Fig. 4 Processing speed increase


– 215 –

3.4. On-Board LayoutThe main electronic control units were placed beneath

the radiator duct. This location was at the center of thechassis as a whole, as well as at a low position, and itwas the largest area where units could be mountedwithout affecting the vehicle dynamics. As shown in Fig.5, the ECU and CDI are placed on the left side of thechassis between the radiator duct and the under floor.They are affixed to carbon fiber brackets on anti-vibration mounts. As shown in Fig. 6, the PBOX,battery, and HTX are placed on the right side betweenthe radiator duct and the under floor, like the units onthe left side. In order to cool the units, a duct shape

capable of introducing the necessary cooling airstreamwas simulated (Fig. 7), and on that basis a duct shapethat would have minimum effect on the aerodynamicswas created. This meant that the heat radiation structureof the electronic control units was crucial.

4. ECU Architecture and Functions

4.1. ECU Hardware ArchitectureIn order to realize the performance requirements for

the electronic control systems described at the start ofthis article, the hardware performance of the ECU inparticular was important. An architecture with multipleCPUs optimally suited to the various functions carriedout by the ECU was created. The memory architecturethat coordinates those CPUs and their communicationslinks are described below.

The basic architecture of the ECU, as shown in Fig.8, includes three CPUs: the Application CPU (ACPU),the Device CPU (DCPU), and the Gateway CPU(GCPU). The ECU was configured in two blocks, onefor the engine and the other for the chassis, and their I/O were configured for their respective input and outputdevices.

A balance between real time operation of the devicesand high-speed computational processing of applicationswas sought by distributing functions so that processingwould not be concentrated in a single CPU. In the partfor device operation, operations of the injector and theignition device take place at 150 µs intervals when theengine is running at high speed. Consequently, thesewere made into functions handling only device I/Ooperations, and were separated from the processing ofapplications that require high throughput computation.The selection of a dedicated automotive CPU (DCPU)with better I/O functionality therefore satisfied thefunctionality requirement.

A high-speed processing CPU (ACPU) was selectedto handle the control applications used with the engineand the chassis. What had until then been controlledusing three automotive microcomputers was combinedinto one. This reduced the need for data access betweenCPUs for applications and reduced unnecessaryprocessing.

The CPU used for measurement control carries outhigh-volume data communications with PCs andcommunications with telemetry units. A CPU (GCPU)that is high-speed and is equipped with many variedcommunication devices was therefore selected, allowingdata operations that make use of high-speed serial andEthernet communications.

The multiple CPUs with distributed functions have tolook up each other’s data. For example, data that isnecessary for telemetry and logging is sent to the GCPUand the device instruction values that are output bycontrol computation are sent to the DCPU. Thecalibration data and commands that are sent from clientPCs in the garage system are shared by every CPU. Inorder to implement these actions, DPRAM was placedbetween all the CPUs and high-speed synchronized

NACA duct

All outlets open-NACA duct All outlets open-Deep NACA duct

High tem

p

Fig. 7 Cooling duct simulation

ECU

CDI

PBOX

Battery

HTX

Fig. 5 Left side electronic units

Fig. 6 Right side electronic units

– 216 –


access was enabled by means of trigger signals whenthey make connections.

Local memory is assigned to each CPU according tothe functions that are required. Memory functions includeflash memory for program saving, SDRAM for programexecution, NVRAM for data backup, and Compact Flash(CF) memory for use in high-volume logging. CFmemory is installed with two cards in parallel in orderto expand the data bus width from 16 bits to 32 bits.The purpose was to achieve higher-speed data processingand access processing during logging.

The engine and chassis blocks are connected byARCNET communications at 10 Mbps. This is totransfer the parameters required for coordinated controlabove 1 kHz process cycle rate.

Up to 2008, the GCPU performed logging dataretrieval processing and data download processing ina l te rnat ion . How to conduct these processessimultaneously, therefore, became the technical issuewhen a balance between increased logging volume andreduced download time was sought.

The high-speed memory controller shown in Fig. 9was therefore developed. The logging data retrievalprocess and download process could be conducted inparallel by using the PCI bus and the CPU local bus.The increased speed and greater volume in reading andwriting logging data was realized by means of a circuitarchitecture with multiple SD memories connected inparallel. This memory controller was installed in the

TDL developed for the 2009 model, and it achieved a30% increase in data download speed together with anincrease in capacity (to a maximum of 8 GB).

4.2. ECU Software ArchitectureThe ECU software can be generally divided into

these main parts: the engine control part that handlescontrol of engine ignition and throttle, the part thathandles control of the gearbox and vehicle behavior, andthe data measurement part that handles the recording ofcircuit data and its transmission to the garage system.

A real time operating system was implemented oneach CPU. This met the demand for the advanced realtime performance needed to realize high-speed controlcomputations, sensing, and communications. As shownin Fig. 10, the software is divided into three distinct andindependent layers, namely, the OS layer, the middlelayer that performs communications and I/O control, andthe application layer that performs control computations.

Analog IN

Analog IN

Digital IN

CPU (GCPU)

LOGGER and�COMMUNICATION

300 MHz

Flash ROM�Program�8 MB

SDRAM�Work area�64 MB

Flash Disk�LOGGER�2 GB

NVRAM�Back Up�4 MB

ARCNET�10 Mbps�1 ch

CAN�1 Mbps�2 ch

Thermo�Couple(2ch)

DIFFERENTIAL�0-5 V(2 ch)

Ethernet�100 Mbps�2 ch

CPU (ACPU)

Control�APPLICATION

450 MHz

SDRAM�Work area�64 MB

Flash ROM�Program�8 MB

NVRAM�Back Up�4 MB

Single end�0-5 V(6 ch)

CPU (DCPU)

DEVICE I/O

80 MHz

1 Mbyte flash ROM�48 Kbyte RAM

Internal�Monitor(26ch)

DPRAM�Work area�512 KB

ADC�16 bit�6 ch

×64 ×64 ×32 ×32 ×16 ×8

×16

×16

×32

×32

×64

×64

×32

×12

×16 ×16

To�Chassis block

To FIA

To CDI, PBOX�To ST

×16

LAMBDA�Sensors(2 ch)

Single end�0-5 V(4)

DIFFERENTIAL�0-5 V(2)

DAC�12 bit�4 ch


ARCNET�10 Mbps�1ch

CAN�1 Mbps�2 ch


×12 DAC�12 bit�4 ch

Single end�0-5 V(6 ch)

To PC

To�Telemetry

UART�2 Mbps�2 ch

Digital OUT

PVRS solenoid�driver(5 ch)

Lo side�driver(10ch)

IINJECTOR�driver(10 ch)

Analog OUT

MOOG driver�(2 ch)

Analog OUT

MOOG driver�(2 ch)

Fig. 8 Hardware architecture

CPU�Memory�controller�(FPGA)

SD�

SD�

PCI�( 32 bit / 66 MHz )�

20 Mbyte/s

20 Mbyte/s

Fig. 9 SD memory controller


– 217 –

In order to enable independent control development byapplication users in model base development, adevelopment environment capable of automaticallygenerating execute files was set up and controldevelopment efficiency was increased (Fig. 10).

4.3. Circuit Data Measurement Function (Logging)The logging function is used to make settings in the

chassis and power train, to monitor conditions, and toconduct simulations. It therefore needs to provideaccurate data and handle large volumes of data. Thisrequired specifications that include 2000 channels andsampling rates from 10 kHz to 1 Hz.

For the efficiency of circuit tests, it is important forthe engineers to be able to quickly determine conditionsin the vehicle after it has been driven, and make it readyfor the next driving session. It was necessary, therefore,to shorter the time from data download to data display,and to provide techniques for easy analysis.

Logging data is accumulated each time the car isdriven. This is used not only on the circuit, but also foranalysis and simulation at the various developmentcenters. It was necessary, therefore, to associate thecircuit data and the test items.

4.3.1. Measurement data formatIn order to facilitate data management in every

driving situation, data measurement using the loggingsystem was set up to split off the data from each drivingsession and write it to the logging memory (Fig. 11).A single driving data is composed of the Run Headerand the Run Data.

The Run Header area records the Lap Data, which

is updated on every lap; the Straight End Data, whichgives the highest speed point a t the ends ofstraightaways; and the Fail/Warning Data, which givesinformation on malfunctions. These were arranged toenable engineers to readily identify characteristic pointsin each lap. Index data for logging-related data is alsorecorded so that circuit data, calibration data, and the likecan be tied together and accurate data management canbe conducted after driving sessions.

The Run Data area contains records for every channelsubject to logging.

4.3.2. Data recording techniquesAs explained earlier, it is necessary for data logging

to record data from measurements conducted at a highsampling rate. Data logging with a high sampling rateinvolves increased amounts of data, which leads toincreased download times, and this ultimately diminishesthe efficiency of circuit tests. For this reason, a triggerlogging function was developed. Under specificcircumstances, such as when a gear change is performed,when a malfunction occurs, or when some othercondition necessitating acquisition of detailed data at acertain rate occurs, past data is recorded at a highsampling rate for a maximum of one second back froma trigger point. At other times, data is recorded at a lowrate so that the volume of logging data can be reduced.

The trigger conditions are defined on a client PC inthe C language and written into the logging configurationfile using reverse Polish notation. After the ECU hasdecoded the trigger conditions, it implements triggerprocessing when those conditions are met. The systemwas set up to allow the configuration to be changedaccording to test circumstances.

When initially developed, sampled data was simplyput in time series and the data on each channel wasrecorded in logging memory using a discontinuousmethod. Since the aim of this was to reduce theprocessing burden on the ECU. However, when CPUperformance was enhanced, the volume of logging dataincreased and the issue of the time required to displaya graph occurred.

The use of a high-speed CPU (GCPU) enabled thedata recording format to be changed so that the data ineach channel are in continuous sequence and thisupgraded the drawing speed. The data recording formatwas changed so that the measurement data for eachsampling from the same channel is placed in a singleblock with a fixed length of 1 kB, and measurement datais recorded as an aggregation of blocks. (Figure 11shows data from the same channel in blocks of the samecolor in the middle column.)

With this change in format, the block of the samechannel is read by unit and taken united, so an increasein graph display speed was achieved as a result.

This logged data can not only be displayed inanalysis tools, it can also be input to Hardware In theLoop Simulation (HILS) and used to reproduce drivingconditions on a PC. It has also been used to feedbackmeasures to solve circuit-running issues.

RTOS� RTOS� RTOS�

Driving force�Control �

Device control� Measurement�communication�

Ethernet�Arcnet�CAN�Arcnet�

Application�layer

Middleware�layer

OS layer�

Application CPU� Device CPU� Gateway CPU�

Data�exchange�

Data�exchange

Channel IDCannel number

Next channel numberSampling intervalNumber of data

Data 1

Data 2

Data 7

Data 3Data 4Data 5Data 6

Run header 1

Run data 1

Run header 2

Run data 2

Run header 3

Run data 3

Lap data

Fail dataWarning data

Straight end data

Ch1 Ch2 Ch3 Ch4Ch6 Ch7 Ch8 Ch6 Ch8

Ch7 Ch5 Ch1 Ch4 Ch7Ch3 Ch8 Ch5 Ch6 Ch6Ch6 Ch7 Ch4 Ch1Ch2

Ch1 Ch7 Ch8 Ch3 Ch1

Ch4 Ch5 Ch2 Ch6 Ch4Ch6 Ch3 Ch4 Ch8 Ch6Ch8 Ch2 Ch1 Ch5 Ch3

Channel IDCannel number

Next channel numberSampling intervalNumber of data

Data 1

Data 2

Data 7

Data 3Data 4Data 5Data 6

Ch5

Measured data Run header

Index data

Run data

Block data

Virtual connection between same channel data

1 kByte

Fig. 10 Software architecture

Fig. 11 Logging data format

– 218 –


4.4. Garage System FunctionsThe purpose of the garage system was to manage

circuit data, provide race strategy support, and assist inconducting vehicle checks. The functions that areimportant in running the car were built into the ECU.Other additional functions for support purposes were tobe taken on by the garage system. This was theconceptual approach in development, and it was tied-inwith miniaturization of the ECU (Fig. 12).

In the pit, the car is connected to the garage networkwhere such actions as logging data capture by client PCsand transfer of calibration data take place. The loggingdata is stored on the garage system server so that thedata can be looked up by multiple client PCs. It is alsogiven numerous functions for support of racing strategyand test runs, and the system was set up to allowoperational commands to be issued to an ECU that isconnected to the network. Typical support functions andtheir association with the ECU are described below.(1) Data setting support function

At the actual race location, engineers change theengine and chassis settings on the basis of circuit data.Changes to settings are made frequently, and suchchanges need to be made up to the point immediatelybefore a race begins. Therefore, the system was set upso that the data transmitted from client PCs to the ECUwould be the only data changed by engineers on clientPCs, enabling instant setting changes.(2) Auto Warm-up

Performance enhancements in the engine and gearboxhave brought increased complexity in engine starting andwarm-up as well as the engine checking mode. Autowarm-up is a support function that automates theseactivities. This check mode includes numerous checkmodes for determining the engine status, and the engineis controlled using control instruction values sent froma PC. The ECU is not prepared beforehand with a profileof these check modes. Instead, the system specificationcalls for the profile to be stored in the form of files onclient PCs in the garage system. This allows the creationof a variety of check modes adapted to enginespecifications and environment of use without changingthe ECU software.(3) Configuration of Steering Wheel Functions

In addition to driving operation, the steering wheelhas the capability for the driver to check or make

changes to settings and car information while driving, inaccordance with racing strategy and changes in thecircuit environment. Figure 13 shows how the steeringwheel is equipped with multiple switches, buttons, andan LCD display (Fig. 14). The functions given to all ofthese controls and the content shown on the display canbe freely assigned and set up from applications on a PCas requested by the individual driver. By sendinginformation to the ECU through the garage system, thesteering and ECU functions can be linked together.Frequently used and important functions are assigned topush switches and rotary switches so that thoseoperations are carried out with one action. Otheroperations are placed in command hierarchies andaccessed using a Mode Function switch and the “+/-”switch. This realized a balance between operability andmultifunctional switchability.

Server�

Measured data�

Client PC�

Client PC�

Lap chart�

Logged and telemetry�data chart��

Setting data tuning�

Measured data�

Feedback�

Auto warm-up�

Other statistic tools�

Configuration change�

Measurement data�analysis

Race strategy�support

Driver radio

A/F

Speed limit

Drink pump

Toggle switch

Neutral / Oil check

Traction control

Alarm lamp

Launch/Overtake

Diff

Reverse

Mode function

+/-+/-

Traction control�(tire warm up)

Engine brake control

Shift timing Light

Display�• Gear position�• Mode confirmation�• Message

Clutch paddle�(for right hand)

Shift paddle (up)

Clutch paddle�(for left hand)

Shift paddle�(down)

Fig. 14 Steering wheel display

Fig. 12 Garage system

Fig. 13 Steering wheel functions

5. Hardware Development

The electronic control units were given a three-dimensional structure using multiple circuit boards inorder to increase the package density. Circuit boardsdensely populated with electronic parts have largenumbers of wiring connections, and it was necessary to


– 219 –

consider techniques for making connections betweenboards. Higher-performance CPUs also operate at higherspeeds and generate considerable heat. Techniques forheat dissipation in miniaturized packages were an issue.

5.1. High-Density PackagesElectronic component packaging techniques that had

no track record of use in automotive electroniccomponents at that time were employed in 2003 torealize high-density packages. These were the Fine-pitchBall Grid Array (FBGA) and the Build-up printed circuitboard (PCB), and their use realized miniaturization.Figure 15 shows a circuit board populated with CPUsusing the Build-up PCB technology. For the 2009 modelTDL, a prototype unit was fabricated through theapplication of Device Embedded PCB technology. Thiswas confirmed to achieve 20% greater packaging densitythan Build-up PCB products.

CPU operation at higher speeds was accompanied byfaster bus clock rates and CPU electrical power supplyat a lower voltage (1.8 V). This made the units moresusceptible to being affected by disturbance noise andcross talk. Consequently, there was an even greaterdemand than before for stabilization of pattern shieldsand ground electric potential during pattern design, andcircuit board wiring design techniques based on high-speed transmission path simulation were adopted.

5.2. Inter-Board Connection MethodsIn the case of unit structure from multiple circuit

boards, the method used to connect boards with signalwires becomes an issue.

Larger units involve a great number of inter-boardsignal wire connections. Therefore, not only theconnectors can cause dead space, but it is apparent thateven more space is required when wire routing is takeninto consideration.

In order to resolve this issue, it was decided that the2008 model TDL would reduce the inter-boardconnectors used up to that time by instead using Rigid-Flexible circuit boards to optimize inter-board wiring. Atthe same time, pattern shields could be built in at signal

wire connections for communications and small signals,and the connections could be made electrically stable aswell (Fig. 16).

5.3. Heat Dissipation StructureGenerally speaking, the automotive CPUs used in an

ECU often do not require heat dissipation measures. Thehigh-speed CPUs used in Formula One ECUs, however,have power consumption that rises as high as about 4W, and they exceed the guaranteed operating temperatureof 105°C even at room temperature (25°C).

In order to achieve a balance of high processingperformance and miniaturization in the Formula OneECU, the greatest issue was how to deal with heat. Thiswas a question of whether the unit could be madesuitable for use in the actual car environment.

Heat dissipation for high-speed CPUs and other suchheat generating devices is generally accomplished byforced air-cooling using heat sinks. In Formula OneECUs, for which miniaturization is sought, maximum usewas made of the case and the circuit boards to enableguaranteed operation even in the actual car environmentwhere the ambient temperature rises above 60°C.

The heat dissipation measures will be described here,using the ECU as an example. Figure 17 shows themain electronic parts that generate heat, the heatdissipation structure, and the heat dissipation pathways.It is important to dissipate the heat that is generated andto increase the heat capacity.

Regarding the former, a pattern was formed with acircuit board affixed to an aluminum plate made fromthe same material as the ECU case in order to enhanceheat conduction from the device and the circuit board,thus using the case structure for heat dissipation.Regarding the latter, metal core PCBs were adopted forthe circuit boards that have CPUs mounted. These havehigh heat conductivity in the planar direction, and theheat capacity of the circuit boards was increased as aresult.

Fig. 15 High density PCB surface

Fig. 16 Rigid-Flexible PCB of TDL

A heating element (CPU, Driver)

Thermal flow Copper core PCBPCB

Frame

Al base PCB

Cu core PCB

Fig. 17 High heat conditioning structure of ECU

– 220 –


By means of these various techniques, the unit couldbe made suitable for the actual mounted environment inthe car.

5.4. History of MiniaturizationFigure 18 shows the historical changes in the ECU’s

package size and the technologies employed. There hasbeen demand for higher processing performance sincethe 2004 specifications. This was achieved employing thetechnologies described above, so that higher performancewas realized while maintaining the size of the 2004model.

Growing demand for higher performance and greaterminiaturization of ECUs in mass production vehicles, aswell, will no doubt lead to importance being placed notonly on measures for heat dissipation in electroniccomponents, but also on technology to inhibit thegeneration of heat in devices.

0

500

1000

1500

2000

2000 2001 2002 2003 2004 2005 2006 2007year

Dat

a tr

ansm

issi

on s

peed

[kbp

s]

19.2 kbps

460 kbps

2 Mbps

QPSKFSK

• 1000 channel data�• 100% data coverage�• Trigger function

Standard transmission�speed of 3G wireless phone

Two-way telemetry

38.4 kbps

Fig. 19 Progression in telemetry transmission speed

communication speed was 19.2 kbps.A change in the regulations in 2002 enabled the

transmission of data from the pit to the car, initiating thedevelopment of interactive telemetry systems. Thecommunication speed in one direction was itselfincreased to 38.4 kbps. In 2003, however, the regulationschanged again and interactive telemetry was outlawed.

In 2004, the carrier wave frequency was moved tothe 1.7 GHz band, and the communication speed wasraised as high as 460 kbps.

The communication speed was further raisedsignificantly again in 2006. Of all the data collectedwhile running, the items required for real time analysisamount to approximately 1000 channels. To transmitthese by telemetry would require a transmission speedof approximately 1 Mbps. The demand for high-precisionmeasurement described in section 4.3. required amaximum data rate of 1 kHz. In order to meet thesedemands, the chosen modulation method was Phase ShiftKeying (PSK), which has high-power efficiency andfrequency-use efficiency as well as low error bit ratesat low receiving levels. In addition, the communicationspeed was increased to 2 Mbps, and a trigger functionthat increased the rate of acquisition for certain data atspecified times was also implemented.

In 2006, as a result of the evolution in technologynoted above, practically all the data from running carscould be acquired by the pit in real time. This capabilitybecame an indispensable part of race operations, as thedata from the start of the formation lap could be used,for example, to change the clutch control mode as therace was starting.

6.2. Issues Associated with Increased SpeedTelemetry systems were checked for data acquisition

performance in a simulated desktop environment. Afterthat, circuit tests would be implemented; however, therewere many points where testing on the actual circuit anddesktop simulation indicated different characteristics.This was particularly the case when Quadrature PhaseShift Keying (QPSK) was adopted as the modulationmethod to increase the speed. In desktop simulation, thepacket acquisition rate (the percentage of data packets

’00

100%

Vol

ume

’01�-’02

’03�-’04

50%

’05 ’07’06

• Internal via hole

• Through hole via

FBGA package1608 chip

• Build up

-

QFP, SOP package

• Metal-cored build-up

• Rigid-flexible

’09

• Buried chip�2009 TDL

PerformanceRefer to Fig.4��

Device package technology• Printed circuit board technology�

0402 chip

Year

Fig. 18 ECU package

6. Telemetry

6.1. History of Telemetry DevelopmentTelemetry is the system that takes data from the

various kinds of sensor data and the like while thevehicle is running and uses radio to transmit it to thepit. Telemetry was first developed for use in races byHonda in the mid-1980s, during the second era ofparticipation in Formula One competition. Telemetry wasused during the second era by acquiring important dataon engine speed, engine oil and water pressure, engineoil and water temperature, fuel consumption, and the likewhile the car was passing in front of the pit, and usingthat data to manage conditions in the car. In the thirdera, by contrast, that system evolved to acquire andanalyze all data concerning the engine and the chassisin real time. Figure 19 shows the history of change incommunication speed and communication protocols usedin telemetry.

In 2000 and 2001, the telemetry system developedduring the second era was carried over. Its carrier wavefrequency was in the 400 MHz band, the modulationmethod was Frequency Shift Keying (FSK), and the


– 221 –

received normally) was 90% or higher, but in circuittests, the rate declined significantly. The data was beingretransmitted, however, so the coverage (the percentageof necessary data acquired) maintained a level of about100% (Fig. 20).

Figure 20 shows a conspicuous decline in the packetacquisition rate on the circuit, even in sectors that arerelatively close to the pit. When data acquisitionperformance declines even at short distances, in otherwords, when reception levels are adequate, the cause isconceivably the influence of fading, defined as mutualinterference on the receiving side from multi-pass waves,which are reflections from the circuit road surface, thestands, and the like. The extent of influence from fading

depends on communication speed. Figure 21 shows howthe length of a single frame is reduced when thecommunication speed increases, so that the amount ofinterference is greater proportionate to the frame length.As a result, the influence of fading becomes significantand communication quality declines.

A variety of measures were tried to reduce the abovetype of decline in the packet acquisition rate, including:(1) transmission power was increased to mitigate the effects

of noise;(2) transmission speed was lowered to mitigate the effects

of multi-pass waves;(3) a directional antenna was used on the receiving side in

order to avoid multi-pass waves; and(4) forward error correction (FEC) was adopted to lower the

error bit rate.Ultimately, however, these measures did not achieve

any major reversal in the decline.

6.3. Consideration of a Radio Wave Propagation Modelfor the Circuit

As the development of telemetry depends on desktopsimulation, an accurate grasp of the radio wavepropagation environment on the circuit is of majorimportance. Consequently, telemetry development from2007 on included efforts to raise packet acquisition rateson actual circuits by conducting investigations into radiowave propagation environments that resembled the actualenvironment.

The environment used for Formula One racing differssubs t an t i a l l y f r om tha t o f o rd ina ry mob i l ecommunications (mobile phones and the like). Movementspeeds can exceed 300 km/h at the maximum, there isno clear line of sight from the pit to the farthest point,and adaptive transmission cannot be used because thecommunication is one-way from the car to the pit.

Figure 22 shows the radio wave model (fadingmodel) that was used in simulations up to 2007. In thismodel, there are two routes (paths) followed by radiowaves, and the maximum delay time between paths was0.2 µs. There is generally said to be a reflection delaytime of 0.01 µs on land where the line of sight is clear,and a reflection delay time of 0.1 µs (a maximum of 0.2µs) where there are buildings. This model is thought todiverge from the actual situation on the following twopoints:(1) The path delay time is short.

The delay time of 0.2 µs represents no more than 60m when converted into an optical path length.Considering that in the actual environment there wouldbe reflections from distant stands, surrounding

Frame 1 Frame 2 Frame 3

Frame 1 Frame 2 Frame 3

Direct wave

Delay wave

Received signal

d

Frame�1

Frame�2

Frame�3

Frame�1

Frame�2

Frame�3

d

Direct wave

Delay wave

Received signal

Fast data transmission speed

Slow data transmission speed

Many rates of�interference

Usable frame

Interference�(unusable)

Bits/frame is�same as above

d: Delay time

Fig. 21 Difference in influence of fading by datatransmission speed

Max delay time�: 0.2 s1�

2�

Fig. 22 Fading model used for development to 2007

Fig. 20 Packet rate and coverage for circuit oneround (2007 Catalonia)

PIT

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

22212019181716151413121110987654321Sector

Pac

ket r

ate

/ Cov

erag

e

Packet rate

Coverage

– 222 –


mountains, and the like, it can be inferred that the pathdelay time would be longer.(2) There are too few paths.

Since the telemetry transmission antenna is non-directional and transmits radio waves in every directionfrom the car, it can be expected that waves reflectedfrom every structure on the course will be arriving at thereceiving antenna at the pit, and the number of paths cantherefore be conjectured to be greater than two.

Measurements made on site are thought to be theoptimal way of ascertaining the radio wave propagationenvironment . As expla ined ear l ie r , however ,measurement on the circuit presents issues. Therefore, avariety of values were assigned as fading modelparameters and the actual measured packet acquisitionrate and the error pattern in data received with errors oncircuit tests were compared with simulated tests. Thefollowing results were obtained:

• the maximum number of paths is six; and• the delay times of multi-pass waves of dominant

strength were about zero to 1 µs, and weaker multi-pass waves had a maximum delay of 5 µs.

It was found that the above fading model is similarto the actual environment (Fig. 23).

The above fading model is known as the Rayleighfading model. Since the influence of overlapping multi-pass waves in large numbers can cause consecutiveerrors to appear in the data, little effect is generally feltfrom increases in transmission power or forward errorcorrection. Furthermore, the existence of a large numberof paths means that even the use of directional antennas

will not fully avert the effects of multi-pass waves, andsince the delay times are so long, decreasing thetransmission speed of 2 Mbps to just one-half or one-quarter cannot be expected to resolve the issue ofcommunication quality. When the model in Fig. 23 isviewed in this light, it becomes quite understandable thatthe measures described in section 6.2. to address thereduced packet acquisition rate did not yield any majoreffects to better the situation.

In development from 2008 on, the model in Fig. 23was employed to conduct evaluat ions of newcommunication protocols to realize higher-speedcommunications while still maintaining an adequatepacket acquisition rate even under this model.

6.4. Technology for Next-Generation TelemetryThe aim from 2009 had been to do away with

downloading by developing a system that uses real timetelemetry to transmit all the data that has been loggedby the ECU and the logger up to that time. Investigationwas in progress with the goal of achieving furtherincreases in speed. In order to seek still greater speedunder present circumstances, it is important to adopt amulti-carrier system capable of reducing the transmissionrate separately by carrier as a measure to increase fadingresistance. Conceivable specific techniques to realize theaim of increased speed include the use of OrthogonalFrequency Division Multiplexing (OFDM) forenhancement of frequency-use efficiency and fadingresistance.

A prototype for desktop use of OFDM was createdand an evaluation of its characteristics was carried outusing the fading model shown in Fig. 23. Figure 24shows a comparison of the characteristics of thetelemetry transmission equipment from 2007 and theOFDM prototype. The vertical axis in Fig. 24 representsthe Bit Error Rate (BER: the number of bit errorsdivided by the number of bits transmitted) while thehorizontal axis represents the number of fading paths. Itcan be confirmed that the adoption of OFDM hasrealized higher speed and higher reliability.

7. Component Development

As the Formula One engine and chassis have 100 ormore sensors and actuators mounted on them,miniaturization of these devices was required with aview to mountability. Particularly, sensors mounted onthe engine (for measuring pressure, temperature, timing,and throttle position), along with ignition coils and thealternator, were thoroughly miniaturized and made highlyefficient. In addition, the greatest issue was the guaranteeof their sustained functionality under engine vibration.

This article will provide details on the developmentof the following:(1) throttle position sensor; and(2) engine wire harness

7.1. Throttle Position SensorIn the engine of an ordinary automobile, the intake

Max delay time: 5.0 s 1

2

3

45

6

Fig. 23 Fading model used for development from 2008

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

1 2 3 4 5 6 7Number of paths

BE

R (

Bit

Err

or R

ate)

System in 2007 (Data rate 2Mbps)1 carrier data rate 2Mbps

OFDM Prototype (Data rate 4.8Mbps)1 carrier data rate 0.1MbpsGood

Fig. 24 Comparison of conventional system andOFDM


– 223 –

flow is measured by the intake manifold pressure or byusing an air flow meter, and that measurement is usedto control the fuel injection volume. In the highlyresponsive engines used for racing and the like, however,the throttle opening is taken as the basis for setting thefuel injection volume. The throttle position sensor,therefore, is one of the key sensors in a Formula Oneengine, and requires high accuracy (within 1% at FullScale) and reliability.

When Honda returned to Formula One racing in2000, the brush type of contact sensor (potentiometer)was initially being used. There were various issues withit, however, including wear of the brush contact surfacefrom vibration and foreign matter (such as oil or dust).As vibration from the engine reaches approximately 500G, there was an urgent necessity to isolate the sensorfrom contact for the purposes of accuracy and reliability.From late 2003, therefore, this sensor was changed to anon-contact type using a magneto-resistance (MR)element to detect the magnetic vector. As this is thebasic sensor for fuel injection control, it is necessary toprovide high accuracy across the entire temperaturerange. An IC using two MR elements was adopted, andaccuracy was assured by a technique that cancels outtheir respective temperature characteristics.

A split type sensor was adopted that has the sensingelement mounted on the throttle body, and a magnet(polarized to generate the desired magnetic field)attached to the shaft of the throttle butterfly, which is arotating body (Fig. 25).

This sensor was mounted toward the back end of theengine, and heat damage occurred frequently.

The main cause of heat damage was exhaust heatfrom the exhaust manifold. Due to the aerodynamicrequirements of the chassis, the engine cowl was beingsqueezed to a smaller size every year. This had animpact on the flow of air inside the engine cowl, inaddition to which the layout of the exhaust manifold wasalso changed accordingly so that it was closer to theengine where the sensor was mounted. The result wasan increasingly harsh thermal environment. Even whenthe MR element and other IC parts are guaranteed athigh temperatures, such guarantees ordinarily cover upto 130 °C. In order to guarantee up to 150 °C for racing,high-temperature solder and other such materials were

used, while the range of guaranteed accuracy wasrestricted and adjustment made by means of software.Through these and other such measures, sensor accuracywas able to satisfy the condition of staying within 1%FS.

7.2. Wire HarnessThe Formula One wire harness was designed with an

emphasis on reduced weight, durability, and on-sitemaintainability. Up until engine homologation wasprescribed in 2007, there were no engine weightregulations. Consequently, measures were specialized inweight reduction, and the wire harness was fixed in placeusing simple stays that were mainly placed around theairbox and cylinder head. When engine weightregulations were instituted, there was a shift inorientation to maintainability and reliability rather thanweight reduction. A junction box was placed above thecylinder head and the circuits for all the sensors werebrought together in that box (Fig. 26).

In terms of reliability, it was a struggle against breaksin the wires. The top speed of a Formula One engine isdouble or more that of a mass production vehicle, andnormal engine speed extends across the entire range, sothat areas around the engine are subject to constant high-frequency vibration. In the case of a mass productionvehicle, the wire harness is bunched together with tapeor other such material and then encased in outersheathing material (plastic tube) to counter wire breakageand wear at points of contact due to vibration. InFormula One cars, since the use of finer wires andreduction in weight are also important objectives, heat-shrink tubing made of woven polyester fiber is used toprotect against external contact as well as to realize theuse of finer wiring and lighter weight. Wire breakageresulting from tension load on the harness generated byvibration and acceleration has further been addressed bythe use of silver-plated high-strength copper wire forracing use. The harness wires are then twisted togetherand bundled inside shrink tubes, enhancing harness

Magnet

Hall IC

Sensor case

Throttle butterfly fitting

Throttle body fitting

Wire harness

Fig. 25 Throttle sensor

Junction Box (J-Box)

Fig. 26 Junction box on engine

– 224 –


Kenichiro ISHII Toshiyuki NISHIDA

Masaki NEGORO

Author

Kohei TOSHIMA

Masataka YOSHIDA Yutaka MARUYAMA

strength against breaks (Fig. 27). This twisted structurenot only enhances strength against wire breakage, it alsoheightens the flexibility of the harness itself andenhances engine mountability.

Even when these techniques are used, however, thereare still places on the engine where vibrations at thelevel of several hundred Gs are generated. There havebeen cases, therefore, where just the copper wire hasbroken inside its cladding or inside the shrink tubingeven when no external damage is visible. This is thoughtto be generated by sympathetic vibration of the copperwire or of the harness itself due to the high-frequencyvibration, and the vibration is conjectured to result indisplacement and tensile loads that cannot be absorbed.The method was therefore adopted of fixing the entireharness in place between the attachment points, includingthe oscillating parts, by encasing it in rigid retainingparts made of carbon material. This has succeeded inpreventing wire breakage.

Vibration has been an unavoidable obstacle in thecourse of Formula One development. In the initial phasesof development, it was important to grasp the levels andpatterns of vibration in advance and to skillfully combinethe measures taken to counter them. Another point is thatthe short Formula One development period allowslimited opportunities to conduct the durability checksunder engine operating conditions that are necessary toguarantee reliability. Consequently, it is necessary toascertain the symptoms of faults and the effectivenessof countermeasures early on, and methods of analysisusing X-ray and other such non-destructive inspectiondevices in durability testing have been introduced to theharness development process.

It has been necessary to manage the design effort soas not to interfere with weight reduction measures bypursuing vibration-resistant design that goes beyond whatis required in actual vehicles.

Fig. 27 Harness

regarding new technology has been obtained. Theelectronic control systems developed for Formula Oneare for unique vehicles and are further specialized forracing functionality. However, the conceptual approachto system construction and the basic electrical systemtechnologies that were employed are also likely to beheld in common with the advancing automotivetechnology of the future.

It has been decided that the development of electricaland electronic technology for Formula One racingunderway since the adoption of FIA standard systemswill be superceded by the development of kinetic energyrecovery systems (KERS), which is employed from2009. It is to be hoped that Formula One will continueto stand as the highest achievement in automobile racing,as well as the highest achievement in automotivetechnology in times to come.

8. Conclusion

In retrospect, Honda’s Formula One systemdevelopment proceeded without break from the time ofthe test runs at Pembrey Circuit in the United Kingdomin 1999, one year after their development started, up tothe Brazil Grand Prix in 2007, and a wealth of findings

development of electronic control system for formula one · 2017-11-28 · – 214 – development...

Documents