Automobile-Human Technology Edition€¦ · Advanced Technologies for Man/Automobile Harmony by Masaaki Nangaku* T he automobile, while comfortable and convenient for its users, causes

ISSN 1345-3041

VOL. 94/JUN. 2001

MITSUBISHIELECTRIC

Automobile-Human Technology Edition

�Vol. 94/June 2001 Mitsubishi Electric ADVANCE

A Quarterly Survey of New Products, Systems, and Technology

TECHNICAL REPORTSOverview ............................................................................................. 1by Masaaki Nangaku

Application of CFD Analysis to the Development of a Direct-Injection System ................................................................................. 2by Hideaki Katashiba and Kazuhiko Kawajiri

Application of a GMR Element to a Revolution Sensor ..................... 5by Tatsuya Fukami and Masayuki Ikeuti

On-Board Inverter Miniaturization Technology .................................. 8by Hirotoshi Maekawa and Gourab Majumdar

Millimeter-Wave Radar Technology for Automotive Application .... 11by Shinichi Honma and Naohisa Uehara

Technology to Integrate a Vehicle-Mounted Camera and Image-Processing Unit ................................................................................ 14by Yoshiyuki Fujii and Hideki Tsukaoka

Static Steering-Control System for Electric-Power Steering .......... 18by Masahiko Kurishige and Takayuki Kifuku

Development of Car Navigation Software ....................................... 21by Toshiki Kusama and Akio Uekawa

Powertrain Control-System Development Support ......................... 25by Jiro Sumitani and Kiyoharu Anzai

Digital Signal Processing Technology for Car Radios ..................... 28by Masahiro Tsujishita and Eiji Asano

CONTENTS

Mitsubishi Electric Advance is published online quarterly (in March, June, September,and December) by Mitsubishi ElectricCorporation.Copyright © 2001 by Mitsubishi ElectricCorporation; all rights reserved.Printed in Japan.

Cover StoryOur cover expresses a world in whichglobalization is accompanied by environ-mental protection under clean, clear blueskies. Mitsubishi Electric seeks to achieveharmony in the links between humanbeings and the automobile, developing theadvanced technology that addresses theinformational, environmental and safetyconcerns.

Kiyoshi Ide

Haruki NakamuraKoji KuwaharaKeizo HamaKatsuto NakajimaMasao HatayaHiroshi MuramatsuMasaki YasufukuMasatoshi ArakiHiroaki KawachiHiroshi YamakiTakao YoshiharaOsamu MatsumotoKazuharu Nishitani

Shinichi Sato

Keizo HamaCorporate Total Productivity Management& Environmental ProgramsMitsubishi Electric Corporation2-2-3 MarunouchiChiyoda-ku, Tokyo 100-8310, JapanFax 03-3218-2465

Yasuhiko KaseGlobal Strategic Planning Dept.Corporate Marketing GroupMitsubishi Electric Corporation2-2-3 MarunouchiChiyoda-ku, Tokyo 100-8310, JapanFax 03-3218-3455

Editor-in-Chief

Editorial Advisors

Vol. 94 Feature Articles Editor

Editorial Inquiries

Product Inquiries

Automobile-Human Technology Edition

· 1June 2001

TECHNICAL REPORTS

*Masaaki Nangaku is Senior Executive Vice President and a Member of the Board of Mitsubishi Electric Corporation.

OverviewAdvanced Technologies for Man/Automobile Harmony

by Masaaki Nangaku*

The automobile, while comfortable and convenient forits users, causes a number of serious problems, such as pol-lution of the atmosphere and other environmental and en-ergy-related problems. There are also the risks of trafficaccidents and other safety-related problems. The solutionsof these problems are some of the most important issuesfacing us in the 21st century. Only advances in technologyoffer the prospect of solutions, and we need nothing lessthan a complete paradigm shift.

We at Mitsubishi Electric Corporation take “man/auto-mobile harmony” as our watchword, and we are activelyapplying information technology (IT) to advanced develop-ments in environmental and energy-saving technology, andin safety, convenience and comfort. In this way we intendto contribute to progress in our automotive culture.

While these technological developments naturally includethe whole spectrum from miniaturization and weight re-duction to sophisticated electronic control technology, theyalso call for complex and rapid development of commu-nication, information-processing and semiconductortechnologies.We will also see rapid progress in themodularization and systemization of automotive electronicproducts.

Fortunately, the corporation has excellent, in-depth re-sources and achievements in each of these areas, and byutilizing and integrating these different disciplines we ex-pect to be able to take up successfully the challenges ofcreating a new automotive culture appropriate for the needsof the 21st century, at the same time contributing to theeconomic development of society.❑

TECHNICAL REPORTS

Mitsubishi Electric ADVANCE2 ·

Direct-injection gasoline engines have beenwidely introduced in global markets in recentyears as environmentally friendly engines. Toimprove their fuel economy and low-emissionperformance further, it is necessary to have abetter understanding of the fuel spray pattern,mixture formation and combustion state in thecylinder so that fuel supply control and compo-nent characteristics can be optimized. This ar-ticle describes a computational fluid dynamics(CFD) analysis method that has been applied toascertaining these in-cylinder conditions andpresents the results of basic experiments con-ducted during the development process.

Fuel Spray ModelThe behavior of fuel injected directly into thecylinder is a major factor governing mixture for-mation. A fuel spray model capable of calculat-ing fuel spray behavior accurately must bedeveloped in order to calculate the mixture for-mation process. In the CFD analysis, in-cylin-der gas flow was found by applying the finiteelement method to solve equations for the massand momentum of a compressible fluid and theequation for conservation of energy. The stan-dard k-� turbulence model was used. Fuel spraybehavior was calculated based on a discrete drop-let model, taking into consideration dropletbreak-up, coalescence and evaporation. The ini-tial spray condition was given as a function ofthe plunger operation, including the initial injec-tion velocity, injection angle and droplet sizethat were determined from the injector specifi-cations. These values were then corrected basedon a comparison with the results of experimen-tal measurements of fuel spray properties. Thegasoline fuel was assumed to be C8H18 and itsphysical properties were used in the analysis.The distribution function proposed by Nukiyamaand Tanasawa was assumed for the initial drop-let size distribution.

The simulation results obtained with the fuelspray model developed in this study are com-pared in Fig. 1 with the experimental results forfuel spray growth when fuel was injected undera backpressure condition of atmospheric pres-sure. The experimental results are shown as

*Hideaki Katashiba is with Automotive Electronics Development Center and Kazuhiko Kawajiri is with AdvancedTechnology R&D Center.

by Hideaki Katashiba and Kazuhiko Kawajiri*

Application of CFD Analysis to theDevelopment of a Direct-Injection System

images captured with a high-speed camera fit-ted with a photomultiplier. These images arecross-sections of the fuel spray that were visu-alized by means of an Ar laser light sheet. Thesimulation results show overall images of thefuel spray. It is clear from this figure that thenewly developed fuel spray model accuratelyreproduced fuel spray growth behavior. Fig. 2compares the experimental and simulation re-sults for spray penetration and the spray coneangle.

Good agreement is seen between the two setsof results for both spray penetration and thespray cone angle. It was also confirmed that theexperimental and simulation results showedgood agreement with respect to the Sauter MeanDiameter of the spray droplets following injec-tion.

In-Cylinder Mixture FormationVisualization of in-cylinder mixture formationby simulation allows quantitative evaluation ofthe phenomena involved, which is a critical fac-tor in designing the optimum fuel injection sys-tem. In stratified-charge combustion, it isnecessary to make effective use of the cavityprovided at the top of the piston to transport acompact combustible mixture cloud to the vi-cinity of the spark plug for ignition and combus-tion. In this study, an analysis was made of the

Fig. 1 Simulation results for fuel spray growth.

0ms

1ms

2ms

Injector

Elapsed time from end of injection

Simulation Experimental results

Conditions

Fuel pressure : 5MPaBack pressure : Atmospheric

TECHNICAL REPORTS

· 3June 2001

influence of the fuel spray angle produced bythe injector, which represents a key engine com-ponent, on mixture formation in the cylinder.

The configuration of the combustion chamberof the direct-injection gasoline engine that wasanalyzed is shown schematically in Fig. 3. Theinjector was positioned between the two intakeports, and fuel was injected at an oblique angletoward the eccentric piston cavity. The spark plugwas centrally located at the top of the cylinder.Calculations were performed for three types ofinitial spray cone angles of 50° , 60° and 70° andthe spray behavior was compared.

Fig. 2 Spray penetration and spray angle obtained in fuel injection simulation.

Fig. 4 compares the fuel spray behavior in thecombustion chamber. With a spray cone angleof 70° , the broad spray pattern causes the fuelto overflow the cavity, with the result that a

Piston cavity

Spark plug Injector

Fig. 3 Combustion chamber of direct injectiongasoline engine.

Fig. 4 Behavior of fuel spray in combustionchamber.

Direction of swirl

50o 60o 70o

(BTDC 40o )

Spray angle

1 2 3 4

20

40

60

80

0

1 2 3 4

20

40

60

80

0

SimulationSimulation

Time (ms) Time (ms)

Experimental results


Spr

ay p

enet

ratio

n (m

m)

Spr

ay c

one

angl

e (d

eg)



large quantity of fuel sticks to the walls. It isestimated that a larger quantity of unburned fuelwould be evacuated from the cylinder under thiscondition. With a small spray cone angle of 50° ,many spray droplets overflow the cavity follow-ing collision with the piston top surface, result-ing in a short interval for obtaining the optimumA/F ratio. As described here, simulations allowthe spray behavior and mixture distribution tobe estimated, making it possible to design theoptimum fuel spray properties and other param-eters.

Combustion SimulationThis section describes the combustion simula-tion program that was developed for makingdetailed visualizations and evaluations of com-plex combustion phenomena in the cylinder andpresents an example of the simulation results.This combustion simulation technology is basedon the combustion analysis expertise that wehave accumulated over many years in connec-tion with fan heaters, furnaces and the like.Combustion reactions were assumed to besingle-stage irreversible reactions. A flame areaprogression model was used as the combustionmodel, which assumed that combustion pro-ceeded by means of an advancing of a thin-flamefront in the combustible mixture. The meanpropagation velocity of the flame front � is givenby the following equation, assuming that it isproportional to the spatial gradient of the turbu-lent combustion velocity St and the unburnedmass fraction mfu.

��Cr�uSt|�mfu| ............................ Eq. (1)

St/Su = 1+Csu'/Su .......................... Eq. (2)

where �u is the density of the unburned gas, Suis the laminar burning velocity, u’ is the turbu-lence intensity and Cr and Cs are constants.

TECHNICAL REPORTS


The simulation results for flame propagationbehavior in the cylinder during homogeneous-charge combustion are compared in Fig. 5 withthe experimental data. A single-cylinder opticalengine having the piston head and part of thecylinder liner made of quartz glass was used inthe experiments to visualize in-cylinder combus-tion phenomena by means of a high-speed cam-era fitted with a photomultiplier. The simulationresults are shown only for the visualized regionrelative to the piston diameter. As shown in thefigure, the simulation reproduced flame propa-gation behavior with excellent accuracy as aresult of suitably adjusting the model constantsused in the combustion model.

The newly developed combustion model wasthen applied to an analysis of stratified-chargecombustion. The simulation results for cylinderpressure and burned mass fraction are comparedwith the experimental data in Figs. 6 and 7, re-spectively. Although the peak combustion pres-sure calculated in the simulation was severalpercentage points higher than the experimentalvalue, the simulation results for cylinder pres-sure and burned mass fraction show good over-

0

10

20

30

40

-180 -120 - 60 0 60 120 180

SimulationExperimental results

Crank angle (oATDC)

Cyl

inde

r pr

essu

re (

kg/c

m2 )

Fig.6 Simulation results for cylinder pressure.

Fig.7 Simulation results of burned mass fraction.

0.0

0.2

0.4

0.6

0.8

1.0

-180 -120 -60 0 60 120 180

Bur

ned

mas

s fr

actio

n SimulationExperimental results

Crank angle (oATDC)

all agreement with the experimental data. Asindicated here, the newly developed combustionmodel can be applied to combustion calculationsfor a direct-injection gasoline engine, making itpossible to ascertain combustion phenomenaand to utilize the resulting data in engine de-sign studies.

In future work, this simulation technology willbe applied to the development of next-genera-tion direct-injection engines and their compo-nents with the aim of contributing to thecreation of products that achieve even higherfuel economy and lower exhaust emissions.❑

Engine speed : 1500r/minIntake manifold pressure : -400mmHgSpark timing : 44oBTDC


Simulations

Time

Visualization area

Burned gas

Unburned gas

Fig.5 Behavior of flame propagation.

TECHNICAL REPORTS

· 5June 2001

A giant magnetoresistive (GMR) element, hav-ing a signal amplitude one order of magnitudelarger than a conventional semiconductor Hallelement, has been successfully applied to a revo-lution sensor for automotive use for the first timein the world. Featuring an original signal-pro-cessing IC and a monolithic structure, this sen-sor achieves a high signal-to-noise (S/N) ratio tofacilitate sophisticated sensing applications suchas detection of engine misfire.

Revolution sensors for automotive applicationsare typically used in controlling the engine,brakes and automatic transmission. Sensors em-ployed for engine control in particular must pro-vide high-temperature stability at temperaturesabove 150°C. In just ten years since the discov-ery of the GMR effect, with its large magne-toresistance ratio, GMR elements have beenapplied to magnetic heads, magnetic field sen-sors and other devices. However, their applica-tion to automotive sensors, which must be usableat high temperatures, has lagged behind theirapplication to other devices. This is attributedto a problem with the long-term stability of theGMR film in a high-temperature environment.Our objective was to improve the high-tempera-ture stability of the GMR element by annealingit beforehand under suitable conditions.

The GMR film we fabricated is a multilayerfilm produced by repeatedly depositing a mag-netic layer, consisting mainly of Co, and a non-magnetic Cu layer. The GMR film composition,including the magnetic layer thickness and Culayer thickness, was optimized so as to achievea large magnetoresistance ratio and the small-est possible hysteresis.

Magnetoresistance curves measured for GMRfilms following annealing for ten hours at differ-ent annealing temperatures are shown in Fig. 1.Letting Rmin + DR and Rmin express the maxi-mum and minimum magnetoresistance valuesin the range of the measured magnetic field, themagnetoresistance ratio is DR/Rmin. As shownin Fig. 1, the magnetoresistance ratio increaseduntil an annealing temperature of 250°C. How-ever, at an annealing temperature of 300°C, notonly did the magnetoresistance ratio declinesharply, the saturation magnetic field and mag-

*Tatsuya Fukami is with Advanced Technology R&D Center and Masayuki Ikeuti is with Himeji Works.

by Tatsuya Fukami and Masayuki Ikeuti*

Application of a GMR Element to aRevolution Sensor

Fig. 1 MR curves after 10 hours annealing.

netic hysteresis increased, resulting in charac-teristics that are not suitable for a sensing ele-ment. Judging from the desirable sensingcharacteristics, an annealing temperature rangeof 200°~250°C would be optimum. However, inconsideration of the harsh temperature environ-ment of automobiles, 250°C was selected as theannealing temperature because of the higherthermal stability that can presumably be ob-tained. The magnetoresistance ratio achievedfollowing annealing at this temperature is a high34% and hysteresis is also low, thus exhibitingexcellent characteristics for a sensing element.Fig. 2 shows the long-term high-temperaturestability measured at 170°C for a GMR elementthat was annealed for ten hours at 250°C. Thevertical axis shows the rate of change in themagnetoresistance ratio and from the smallestinitial value. After 2,000 hours, the rate of

-80 -40 0 40 800

5

10

15

20

25

30

35

-80 -40 0 40 800

510

15

20

25

30

35

-80 -40 0 40 8005

10

15

20

25

30

35

Magnetic field (kA/m)

DR

/Rm

in (

%)

DR

/Rm

in (

%)

DR

/Rm

in (

%)

(a) 200oC 10 hours

(c) 300 10 hours

Rmin=1203W

Rmin=1321W

(b) 250oC 10 hours

Rmin=1280W

TECHNICAL REPORTS


change in the magnetoresistance ratio and fromthe smallest value was less than 1%, which issufficiently small. These results show that theGMR element has sufficient thermal stabilityfor use as an automotive sensor.

With the aim of reducing noise, we then ex-amined the possibility of integrating a GMRelement with a signal processing IC on a mono-lithic chip. To achieve a monolithic structure,it is necessary to deposit the GMR element af-ter the IC has been fabricated, taking into ac-count the thermal resistance of the GMRelement. The issue here is how to use the phos-phate-silicate glass (PSG) film, which serves asthe insulator of the signal processing IC, as theGMR substrate in order to obtain excellent GMRproperties. The PSG film has large surface rough-ness, making it unsuitable as the GMR sub-strate. It was found that excellent properties canbe obtained by ion-beam etching (IBE) the PSGfilm surface for at least two minutes before de-positing the GMR film. It is thought that IBEhas the effect of cleaning the insulator surfacethat becomes dirty during the fabrication pro-cess and simultaneously smoothing large sur-face irregularities.

The revolution sensor for automotive applica-tion consists of the newly developed monolithicIC chip, a permanent magnet and componentsfor protection against surge current (Fig. 3). Thissensor measures the rotation of gears made of a

Fig. 2 Long term stability test at 170°C.

soft magnetic material. The gear teeth made ofa soft magnetic material are magnetized whenthey approach the permanent magnet. Becausethe magnetized teeth cause the GMR elementto produce a magnetic field, a magnetic field thatchanges with gear rotation is applied to theGMR element. The changing magnetic fieldapplied to the GMR element causes the electri-cal resistance of the element and also the volt-age between the two terminals to vary. Thevoltage change is amplified and then comparedby the signal-processing circuit, which convertsgear rotation into a digital electrical signal,making it possible to detect the edge of the gear.

The above-mentioned characteristics of thisrevolution sensor were evaluated using a stan-dard-shaped gear. Measurement accuracy wasgiven by the deviation in the output timing ofthe revolution sensor relative to a high-accuracyencoder signal (reference signal) used to moni-tor gear rotation. As the first step, the samemeasurement was performed at different mea-surement temperatures, Ta. The gear was ro-tated at a speed of 7,000rpm, and the distancebetween the sensor and the gear was kept con-

Fig. 3 Structure of the revolution sensor.

Connector

Chip components

Permanent magnet

Connector

Sensing IC chip

-1-0.8-0.6

00.20.40.60.8

1

0 500 1000 1500 2000

-0.4-0.2

Hold time at 170oC

Cha

nge

from

initi

al v

alue

(%

)

∆R/RminRmin

TECHNICAL REPORTS

· 7June 2001

stant at d = 1 mm. As shown in Fig. 4, the evalu-ated sensor provided excellent angle detectionaccuracy to within ±0.26° over a wide tempera-ture range (−40°C ≤ Ta ≥ 145°C), which is oneoperating requirement of a revolution sensor forautomotive applications.

Fig. 4 Temperature dependence of output timing.

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

-50 0 50 100 150

Dev

iatio

n in

out

put

tim

ing

(dg

eg.

)

Ta (ºC)

An evaluation was then made of sensor re-peatability, which is a key parameter of the mis-fire detection performance of a sensor intendedfor use in engine control. Repeatability is an in-dex that shows the dispersion of the edge posi-tion in the sensor output signal while a gear isrotating. It is defined as the maximum rotationalspeed dispersion when a gear is subjected to re-peated rotation tests. In order to detect enginemisfire, a sensor must have sufficient accuracyto detect minute changes in rotational speed,which requires exceptionally good repeatability.Fig. 5 shows the results of repeatability mea-surements made at different rotational speedsunder harsh conditions of d = 1.5mm and Ta =145°C. Nearly a flat characteristic was obtainedat all rotational speeds. The timing differedslightly depending on the gear teeth, but thiswas caused by timing fluctuation due to sensornoise induced by external interference. This fluc-tuation is sufficiently small, being about one-third that of a semiconductor Hall revolutionsensor. This reduction of noise has resulted fromthe use of a GMR element, which provides a sig-nal amplitude that is an order of magnitude larger,and the integration of the GMR element and sig-nal processing circuit on a monolithic chip.

The measured results show the enormous po-tential of the monolithic GMR sensing elementfor various applications, including its capabilityfor detecting engine misfire, which was previ-ously impossible with conventional devices.❑

Fig. 5 Rotary speed dependence of repeatability.

d=1.5mmTa=145ºC

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0 2000 4000 6000 8000 10000

Rotary speed (rpm)

Rep

eata

bilit

y (

deg.

)

TECHNICAL REPORTS


*Hirotoshi Maekawa is with Automotive Electronics Development Center and Gourab Majumdar is with Power DeviceDivision.

by Hirotoshi Maekawa and Gourab Majumdar*

On-Board Inverter MiniaturizationTechnology

In recent years, hybrid electric vehicles (HEVs)that combine conventional internal combustionengines and electric motors have come into thespotlight as cars that can both improve fuel effi-ciency and reduce exhaust gases. This articlediscusses the development at Mitsubishi Elec-tric Corporation of technologies that are used ininverter equipment—a key part of any HEV—and, in particular, discusses the development oftechnologies for miniaturizing inverters.

Technologies for Miniaturizing On-BoardInvertersTwo major issues to be overcome in miniaturiz-ing inverter equipment for use with high volt-ages and large electric currents are the handlingof both inductive noise and the heat generatedthrough power losses within the elements. If theinverter is to be used in the demanding on-boardenvironment found in vehicles, the equipmentmust also be resistant to physical shocks, ther-mal shocks and power cycling. The equipmentmust provide high reliability under these severeconditions, so such inverters require more ex-acting technical innovations than inverters forgeneral use.

Fig. 1 shows a cross section of an on-boardinverter integrated intelligent power-drive unit(IPU) currently under development. Advancesin miniaturization have required design innova-tions in a variety of areas. Three key technolo-gies contributing to the miniaturization of thepower inverters are discussed below.

POWER CHIPS. Because power chips play a criti-cal role in controlling the performance of the

Fig. 1 Structure cross section.

inverter equipment, improvements to the powerchips have long been a focus of development atMitsubishi Electric.

The power semiconductor element, known asthe insulated gate bipolar transistor (or IGBT) orthe free-wheeling diode (FwDI), is a critical com-ponent of the inverter equipment. While the toppriorities in the development of the power semi-conductor elements at the corporation are thoseof miniaturizing the elements and reducingpower losses, efforts are simultaneously directedat reducing electromagnetic interference andnoise.

For IGBT chips, Mitsubishi Electric has pio-neered the successful advance of microelec-tronic processing from the conventionalthird-generation process rule technologies(three-micron level) to the fifth-generation (sub-micron level) using planar-type IGBTs, takingadvantage of the track record these IGBTs havein on-board applications.

By implementing these advanced microelec-tronic processes, the corporation has succeededin increasing current densities significantly, from130A/cm2 to 200A/cm2 or more, making it pos-sible to reduce the surface area of the chip byabout 30%.

Fig. 2 shows the tradeoff frontier between satu-ration voltage and turn-off time for several gen-erations of IGBTs. Compared with third- generationIGBTs, which are prevalent today, Mitsubishi Elec-tric has succeeded in reducing the saturation volt-age by 0.5V to 0.6V.

The benefits of this process technology are asfollows:

1. The saturation voltage is reduced through theexcellent performance of metal-oxide-semi-conductor structures (MOS) fabricated usingsub-micron fabrication technologies.

2. The saturation voltage is also reduced by op-timizing the distribution of carriers within theIGBT using lifetime control.

3. FwDI soft recovery is enabled through thecombination of heavy-metal distribution andlocal-lifetime control.

Shield plate

Control parts (PCB)

Power chips

Insulatingsubstrate

DC-link smoothing capacitor

CaseBase-plate

Silicon gel

1

2

3

TECHNICAL REPORTS

· 9June 2001

Fig. 3 shows a comparison of the cross sec-tions of a conventional FwDI diode and the newlydeveloped FwDI diode. Because the percentageof time that the energy-recovery mode is usedin HEV systems is relatively high, improvementsin performance of the FwDI are critical. By lo-cally shortening the lifetime of holes in theshaded region of the diagram, Mitsubishi Elec-tric has succeeded in controlling the injectionof holes to levels lower than conventional prod-ucts, and doing so without sacrificing the volt-age drop in the forward direction.

Anode Anode

CathodeCathode

Life-time control area.

Local life-time control area.

(a) Conventional diode (b) New diode

pp

n -

n + n +

n -

Fig. 3 Free-wheeling-diode (FwDi) cross-sectionstructure comparison.

Fig. 2 VCE(sat) - tr Trade-Off

THE CONTROL CIRCUITRY. For the control cir-cuitry, Mitsubishi Electric has succeeded in cre-ating ultra-dense packaging that combines on asingle circuit board the predrive circuits (suchas the power-supply circuit to the power unit),the failure diagnostic circuits, and the high volt-age power-supply sensors, along with the IGBTdrive flash-protection circuits.

Newly-developed sheet-type transformers areused in the power unit power-supply circuitry,enabling a low-profile, compact shape.

Although push-pull circuits have been config-ured in the past using conventional transistorelements in the IGBT drive buffer elements, aswitch to MOSFET elements reduces heatingand allows for smaller IGBT driver circuits over-all.

The IGBT protection circuit is a newly devel-oped custom IC that combines on a single chipa variety of functions such as over-heat protec-tion functions and gate voltage short-circuit func-tions using on-chip temperature sensors, inaddition to providing conventional short-circuitprotection functions. This custom IC integratesthe IGBT gate-drive circuit and all of the its vari-ous peripheral protection circuits, allowing theIGBT peripheral circuitry to shrink substantially.

The CPU also uses a variety of sensors in gath-ering information and making adjustments, notonly making it unnecessary to perform manualadjustments or to sort and match the electroniccomponents but also making it possible to struc-ture control circuits with the bare minimumnumber of electronic components. It does thisby incorporating the functions of the logic cir-cuit in software rather than hard-wiring them.

THE SMOOTHING CAPACITOR. At present, mostvehicle-mounted inverter equipment uses alu-minum electrolytic capacitors as the primary DCpower-supply smoothing capacitors; unfortu-nately, these capacitors account for a relativelylarge percentage of the volume and weight ofthe inverter equipment as a whole, and havebeen a major impediment to miniaturization.

At Mitsubishi Electric, we have focused onthe fact that the primary circuit power supplyfor the vehicle-mounted inverters is always theDC power-supply battery, and have thus engagedin development projects focused on reducing im-pedance in the high-frequency domain ratherthan using capacity as the selection criterionfor the smoothing capacitor. Because of this, wehave been able to set the goal of implementinga solid capacitor with a volume that is aboutone tenth that of a conventional capacitor.

1

1.5

2

2.5

3

3.5

4

4.5

5

0 100 200 300 400 500 600 700 800 900 1000 1100

tf [ns]

(Conditions: Resistive load, IC=100A, Vcc=300V, VGE=+15V, Tj=25oC)

VC

E(s

at)

[V] (

Ic=

100A

, Tj=

25o C

)

600V-Class IGBT

1st Generation

Technology Trend

2nd Generation

3rd Generation

4th Generation

5th Generation (The present

development target)

TECHNICAL REPORTS


The benefits of the solid capacitor currentlyused by the corporation are as follows:

1. It is small, light, and has a three-dimensionalshape that does not take up excess space.

2. Its equivalent series resistance (ESR) and itsimpedance in the high-frequency domain arelow.

3. It has all of the environmental characteris-tics required for on-board use, with a broadrange usable temperature, a long life expect-ancy, and a high breakdown voltage.

Because the capacitor itself is smaller whenthe solid component is used, it can be positionedcloser to the power element, with the additionalbenefit of allowing the power element itself tobe smaller (optimizing the breakdown voltage)because of the snubber effect.

The smoothing capacitor is used to reduce theripple from the DC power supply. The capacityrequired in the smoothing capacitor is deter-mined by factors such as the type of battery, theinductance in the primary power-supply circuit,and the control software. In this developmentproject, Mitsubishi Electric has worked to mini-mize the required value for the capacitance tocontrol the voltage spike when the IGBT shorts.Fig. 4 shows the waveform when the motor is

Conventional aluminum electrolytic capacitor Newly developed solid capacitor

ConditionsInduction motorPhase current 200ArmsPower mode 25kW Battery voltage 288Vdc @No-load

Motor phase currentDC current ( 50A/div)Vdc Torque reference

(200A/div)DC current Vdc (100V/div)Torque reference (70Nm/div)Time scale (1.0ms/div)

Fig. 4 Experimental Inverter Drive Waveforms Comparison.

driven with the lowest capacitance value. Thisfigure indicates that the new capacitors havecharacteristics in no way inferior to those of con-ventional aluminum dielectric capacitors.

The article describes technologies currentlyunder development by Mitsubishi Electric forminiaturizing on-board vehicle-mounted invert-ers, one step in the corporation’s efforts to de-velop smaller inverters with higher performanceand so to contribute to the spread of environ-mentally friendly vehicles.❑

TECHNICAL REPORTS

· 11June 2001

*Shinichi Honma and Naohisa Uehara are with the Automotive Electronics Development Center.

by Shinichi Honma and Naohisa Uehara*

Millimeter-Wave Radar Technologyfor Automotive Application

Mitsubishi Electric Corporation has developeda millimeter-wave radar for automotive applica-tion. Monolithic microwave integrated circuits(MMICs), developed specifically for this appli-cation by Mitsubishi Electric, are used for theradio-frequency (RF) module to obtain outstand-ing radar performance. This millimeter-waveradar incorporates a signal-processing unit in theradar head, resulting in a more compact, light-weight radar system that is easy to install onvehicles.

The sensors for monitoring vehicle driving envi-ronments embody important technology for theIntelligent Transport System (ITS). Sensing devicesemployed include the ultrasonic sensor, image-pro-cessing sensor, infrared lidar, microwave radar andmillimeter-wave radar. Compared with other typesof sensors, a millimeter-wave radar has the ad-vantage of providing stable detection of targetseven under inclement weather conditions suchas rain or snow. Recent remarkable advances inRF technology have facilitated production of highlyfunctional and efficient radar sensors. Thus, milli-meter-wave radar has been marketed as, a vehicu-lar collision warning sensor.

The RF technology, semiconductor technologyand radar technology that Mitsubishi Electrichas developed over many years for space, de-fense and consumer electronics applications havebeen used to create a millimeter-wave radar foruse on vehicles. One of the features of this ra-dar is that all of its constituent elements, in-cluding the semiconductor devices, have beendeveloped in-house specifically for automotiveradar application.

Measuring Distance and VelocityThe method of measuring the distance from aradar to a target vehicle and the vehicle’s rela-tive velocity depend on the type of radar systememployed. Millimeter-wave radar systems for au-tomotive application include pulse radar, fre-quency-modulated continuous-wave (FM-CW)radar and spread spectrum radar. The newlydeveloped millimeter-wave radar is based on theFM-CW radar system.

The FM-CW radar system allows a simple con-

struction of the RF circuit. Moreover, this sys-tem has often been adopted for automotive mil-limeter-wave radars because it can measureboth the distance to a target and the target’srelative velocity simultaneously. The measure-ment principle of the FM-CW radar system isshown in Fig. 1. The radar transmits a frequency-modulated millimeter-wave signal, and the sig-nal reflected by a target is received by the radar.By mixing the received and transmitted signals,the system obtains a beat signal having a fre-

Fig. 1 Frequency-time relationships in FM-CWradar.

1/

Time

ReceivedsignalFrequency

Transmittedsignal

0

b1 b2

∆ ƒ

ƒ ƒ

ƒ

ƒm

quency of ƒb. This beat signal features a time-delay due to the distance from the radar to thetarget and a Doppler shift due to the relativevelocity of the target. The range to the target,R, and its relative velocity, V, are found withthe following equations using the pair of beatsignal frequencies, ƒb1 and ƒb2. These beat signalsare obtained from the pair of frequency-modulatedwaves having different time variations.

(ƒb1 + ƒb2)cR = __________ ............................ (Eq. 1)8ƒm∆ƒ

(ƒb2 - ƒb1)cV = __________ ............................ (Eq. 2)4ƒ0

where c is the velocity of light, ƒ0 is the centerfrequency, ƒm is the modulation frequency and∆ƒ is the maximum frequency deviation.

TECHNICAL REPORTS


Fig. 2 Monopulse radar signal processing.

sities of α and β. The difference δ betweenthese two intensities, and their sum σ, arethen calculated.

3. For these pairs of field intensities, the ratio ofthe difference to the sum (δ/σ) is comparedwith the angular-error signal voltage of thediscrimination curve. And then the azimuthθt of the target is uniquely identified.

(a) Overlapping two antenna patterns (b) Monopulse antenna patterns

Azimuth θ

Target direction θt

α

β

Field intensity Sum pattern Σ

Difference pattern ∆

σ=α+β

δ=α-β

θ

(c) Discrimination curve and angle detection

Error signal voltage ε

δ/σ

Discrimination curveΣ/∆

θ

Field intensity

θtθt

Item Performance

Frequency 76 ~ 77GHz

Output power 10mW max

Range 0 ~ 120m (Passenger car)

Range resolution 1m

Relative speed -100 ~ +200km/h

Speed resolution 1km/h

Detection angle ±8deg.

Multiple target detection Possible

Data refresh rate 100msec

Table 1 Technical Specification of the Millimeter-wave Radar.

In order to obtain two beams, a sequentiallobing method is adopted in this radar for auto-motive use. For this purpose, a mechanical scan-ning method with an antenna radiating a singlebeam is adopted.

When two or more targets and objects exist,the beat signals obtained from them have mul-tiple frequency components. Consequently, inorder to detect the distance to a target and itsrelative velocity accurately, it is essential toselect the pair of beat signals from the sameobject without error. In order to use a millime-ter-wave radar as the sensor for an adaptivecruise control (ACC) system or an automateddriving system, it is necessary to reduce the tar-get misrecognition rate as much as possible.

Masuring the AzimuthThe monopulse method illustrated in Fig. 2 isoften used to measure target azimuth. The re-flected signals from the same target are receivedas two beams having different directions. Theazimuth is determined according to the follow-ing procedure from the received field intensityof the two beams.

1. The difference and the sum of field intensi-ties for the two beams are obtained for eachangle. A discrimination curve that shows therelationship between the angular-error signalvoltage ε and the azimuth θ is obtained. Theangular-error signal voltage is obtained bynormalizing the difference signal with thesum signal.

2. The reflected signals from the same targetare received as two beams having field inten-

TECHNICAL REPORTS

· 13June 2001

Overview of the Newly Developed Millimeter-Wave RadarAn example of the specifications of the millimeter-wave radar for automotive application is shown inTable 1. Assuming the radar is to be used in ex-pressway driving, it must provide sufficient per-formance for detecting a forward vehicle atdistances greater than 100m from the host vehicle.In addition, a millimeter-wave radar for automo-tive application must also satisfy the followingrequirements:

1. It must be capable of oscillating, amplifying,modulating and demodulating millimeter-wavesignals stably.

2. It must be capable of radiating a millimeter-wave signal toward a target with a suitable levelof power, beam width, number of beams andtiming.

3. It must be capable of maintaining robust per-formance in the environment of the host ve-hicle temperature changes and vibrations.

4. Taking into account ease of installation on thehost vehicle, it must be compact and light-weight.

In order to satisfy all of these requirements,Mitsubishi Electric has developed each of the ele-ments constituting a radar head, including an RFmodule, antenna, antenna scanning mechanism,and radar signal-processing unit.

RF MODULE. The RF module consists of MMICshaving functions such as millimeter-wave bandoscillator, amplifier and mixer. The RF module alsohas an interface function with an antenna and asignal-processing unit. By using MMICs developedand manufactured for this automotive millimeter-wave radar, an RF module has been provided withthe high output and sensitivity to achieve a maxi-mum detection range greater than 100m. The per-formance stability of the RF module has beenconfirmed in the harsh vehicle environment.

ANTENNA. An offset paraboloidal reflector an-tenna is used to obtain high gain and antenna ef-ficiency in order to provide the desired maximum

Fig. 3 Radar head (prototype).

RADAR SIGNAL-PROCESSING UNIT. The radar sys-tem adopted combines pulse-Doppler radar withFM-CW radar. This combination reduces the rateof target detection error, which can be a problemwith the FM-CW system, and thereby successfullyimproves target detection performance signifi-cantly. This signal-processing unit is housed in theradar head together with the RF module, antenna,reflector-rotating mechanism and power supplycircuit. Fig. 3 shows a prototype radar head. Thecompact, lightweight radar that has been achievedis easy to install on vehicles.

A compact and lightweight millimeter-wave ra-dar for automotive application has been developed.This radar is currently being tested under actualdriving conditions as the sensor for an ACCsystem.❑

detection distance and detection angle range whencombined with the RF module. A reflector-rotat-ing mechanism has been adopted that rotates onlythe lightweight reflector to accomplish beam scan-ning. One feature of this antenna is that beamscanning can be performed while the RF moduleis kept stationary. Another advantage is that itsradiation properties show virtually no changewithin a beam scanning angle of ±10°.

TECHNICAL REPORTS


*Yoshiyuki Fujii is with the Automotive Electronics Development Center and Hideki Tsukaoka is with MitsubishiElectric Engineering Co., Ltd.

by Yoshiyuki Fujii and Hideki Tsukaoka *

Technology to Integrate a Vehicle-MountedCamera and Image-Processing Unit

The imaging camera unit and image-processingunit that were previously separate componentsof a camera system for lane-marking detectionand preceding-vehicle recognition have been in-tegrated by eliminating the interface circuit andincorporating the control circuit in the software.As a result, we have achieved a low-cost, sub-stantially more compact system, 80% smallerin volume than the previous system.

Driving environment-detection and recogni-tion technology is one of the keys to intelligenttransport systems (ITS).[1] Image-processing tech-nology, seen as a fundamental environment-sensing method, has already been implementedon some production vehicles for detecting lanemarkers and recognizing preceding vehicles.However, because the current image-process-ing system for automotive use is still very largein size and expensive, future market expansionwill require a smaller and lower-cost system

Image Processing Using a Small, InexpensiveLane-Marking Detection SensorAn image-processing system for automotive usemust be capable of detecting and recognizingtarget objects in real time from images taken ofthe driving environment. For that reason, thesystem hardware is more expensive than ordi-nary automotive sensors and is also larger. Onthe other hand, the locations where such a sys-tem can be installed on vehicles are generallylimited for various reasons. Consequently, theimaging camera unit and image-processing unitof such systems have generally been separatedin order to reduce the former’s size. The con-figuration of a conventional image-processing sys-tem is shown in Fig. 1.[2] However, this sort ofconfiguration has the following problems withrespect to both performance and cost aspects.● It is disadvantageous in terms of electromag-

netic compatibility (EMC) because the videosignals are transmitted within the vehicle.

● A filter cannot be used to secure EMC be-cause it would attenuate the higher frequencycomponent of the video signals.

● The use of two separate units requires dupli-cation of structural materials, video interface,power source, timing generator and othercomponents.

These problems can be overcome by integrat-ing the camera and the image-processing elec-tronic control unit (ECU). However, simplyintegrating these two elements would causeanother problem in that the increased size ofthe resulting unit would make vehicle installa-tion more difficult. Moreover, integration wouldalso cause problems related to the increasedscale of the circuit, including the issue of heatradiation associated with miniaturization, theneed to develop new debugging tools and thenecessity of providing a separate interface spe-cifically for external inputs.

On the other hand, integrating the camera andimage processing ECU would clearly producelarge performance and cost benefits. Therefore,as the first step, we selected the dimensions ofan ordinary business card as the target size ofthe unit, assuming that it would be installed onthe back of the rearview mirror. Efforts weremade to achieve size and cost reductions fromthree perspectives:● Incorporation of hardware functions into the

software and use of gate arrays● Simplification of the circuitry and inputs/outputs● Elimination of heat factors

As a result, the above-mentioned problemswere resolved by adopting the improvementslisted in Table 1.

Hardware Configuration of IntegratedImage-Processing CameraThe part count was successfully reduced byapproximately 40% by implementing the mea-

Fig. 1 Conventional image-processing systemconfiguration

Camera Image-processing unit

Vehicledynamics

Video I/F

Status I/F

Imageprocessing

Detectionrecognition

OutputI/F

TECHNICAL REPORTS

· 15June 2001

sures noted in Table 1. Fig. 2 shows thehardware configuration of the integrated image-processing camera and its external appearance.

Images photoelectrically converted by theimage sensor undergo various signal-processing

imposition function is extremely effective fortheoretical software debugging. An inexpensivegate array with fewer gates was adopted by hav-ing the software pre-process the images, whichsimultaneously achieved a cost reduction. TheCPU performs lane detection and preceding-ve-hicle recognition based on the images writtento memory and information on the host vehicle’sdynamic status, and the result is output via thecontroller area network (CAN). The CPU alsocontrols the camera exposure and camera char-acteristics. The various operating parameters ofthe camera require no adjustment, which sim-plifies the manufacture of the camera and alsoreduces its cost.

Lane-Detection MethodThe performance of the lane-detection functionis greatly affected by the weather and the ambi-ent environment. However, white-line lane mark-ers on an asphalt road surface are basicallycharacterized as having a higher degree of lumi-nance than that of the surrounding road surface.The top-hat filter (THF) is an image filter thatfocuses on this characteristic. Letting C, L and Rdenote the respective horizontal coordinates andh the width of a white line in an image captured

Table 1 Miniaturization of Hardware

metIlanoitnevnoC

sdohtemnoitargetnI tceffE

UCE

F/IoediV

retlifIME eteleD

noitazirutainiM

rotarapes.cnyS eteleD

LLP eteleD

GSS eteleD

pmaoediV eteleD

sutatselciheV ICS NAC

gnissecorp-erP erawdraH erawtfoS

ylppusrewoP rotalugerseireS retrevnocCD/CD noitcudertaeH

aremaC

DCC hcnidr3/1 hcniht4/1

noitazirutainiM

sirI lacinahceM cinortcelE

lortnocsirI

erawdraH erawtfoSCGA

lortnocaremaC

Fig. 2 Image-processing camera

a) Block diagram

Superimpose

CANCAN

Video

CPULane detectionCamera control

RAM

Gate arrayTHF

Timing generatorMemory control

A/DSignalprocessing

Battery Power supply

Camera control

1/4”CCD

(b) External appearance

operations followed by analog-to-digital conver-sion, after which they are written to memoryvia a gate array. The gate array generates vari-ous timing signals, controls external inputs andsuperimposes the processed results. This super-

TECHNICAL REPORTS


Fig. 5 The effect of a filter for yellow-linedetection

with the image sensor, the THF value at point Cis given by the following equation:

THF (C) = min {Lum (C) – Lum (L),Lum (C) – Lum (R)} ............ Eq. (1)

where L = C – h and R = C + h and Lum (x) is theluminance at point x. Eq. (1) shows that brightpoints having a width smaller than h can beextracted without being influenced by thechanging luminance of the surroundings. Usingthis computational method, white-line lanemarkers can be detected with a simple algo-rithm. This effect of THF is shown in Fig. 3.

Vehicle-Recognition MethodIn general, it is extremely difficult to extract thecomplex features of a detected object, such as apreceding vehicle or an obstacle, under the limi-tations of a simple hardware configuration andreal-time processing. The camera we have de-veloped does not focus on the features of a ve-hicle, but rather it detects “ objects that are notthe road surface” based on changes in the roadsurface luminance. An example of the vehiclerecognition results is shown in Fig. 4.

Fig. 3 The effect of a THF

n line

Shade Sunshine

Video signal(n line)

THF

Result

Yellow-Line Detection MethodOrdinarily, yellow lines tend to be lost very eas-ily because it is difficult to obtain a difference inluminance between them and the road surface.This problem has been resolved by adopting anoptical filter specifically for yellow-line detec-tion, focusing on the differences in spectral char-acteristics between yellow lines and asphalt.The use of this filter makes it possible to detectyellow lines with the same algorithm as thatused for white lines. The effect of this filter isshown in Fig. 5.

Fig. 4 Vehicle detection

(a) Without filter

(b) With filter

TECHNICAL REPORTS

· 17June 2001

Specifications of Integrated Image-ProcessingCameraThe specifications of the newly developed cam-era with an integrated image-processing unit aregiven in Table 2. This camera is not only appli-cable for monitoring the environment ahead, itcan also be used for monitoring all around a ve-hicle, including the rear, lateral and oblique reardirections, by changing the software.

Table 2 Image-Processing Camera Specifications

Image sensor 1/4-inch B&W CCD

Effective pixels (H x W) 510 x 492

Lens 6mm/F4, field of view 33° (H)

Auto-level controlElectronic iris, AGC, selectedarea metering

Signal-to-noise ratio Better than 45dB (AGC off)

Detecting area ~50m at 26° (height 1.3m)

Optical axial aberration Auto calibration and learning

Input/output CAN, SCI

Video outputEIA 1V

p-p (Detection results

can be superimposed)

Power supply DC, 8~32V (less than 3W)

Overall dimensions (W x H x D) 105 x 62 x 44.8mm

Weight Less than 0.3kg

Operating temperature range −30° ~ +90°C

Storage temperature range −30° ~ +80°C

There are strong automotive industry needs forimage-processing sensors capable of recogniz-ing the driving environment, because such asensing device is an essential component in con-figuring various types of in-vehicle systems.Along with the penetration of ITS technologiesin the coming years, these sensors are expectedto find widespread use. In future work, we in-tend to pursue further size and cost reductionsand to improve robust detection performance. ❑

References[1]Shin-ichi Sato, Yuhei Akasu, Kazumichi Tsutsumi and Yoshiaki

Asayama, “Key ITS In-vehicle Technologies,” Mitsubishi ElectricTechnical Review, Vol. 73, No. 10, pp. 17-11 (1999).

[2] Kazumichi Tsutsumi, Yuhei Akasu, Kazuhiro Ikebuchi, ShigekazuMuraoka, Hiromichi Kushizaki and Minoru Nishida, “HeadwayDistance Control Technology,” Mitsubishi Electric TechnicalReview, Vol. 70, No. 9, pp. 13-16 (1996).

TECHNICAL REPORTS


*Masahiko Kurishige is with the Industrial Electronics & Systems Laboratory, and Takayuki Kifuku is withHimeji Works.

by Masahiko Kurishige and Takayuki Kifuku*

Static Steering-Control System forElectric-Power Steering

Electric-power steering (EPS) systems have at-tracted much attention for their advantages withrespect to improved fuel economy and have beenwidely adopted as automotive power-steeringequipment in recent years. The article intro-duces a new EPS control system that reducessteering torque during static steering (i.e., whilea vehicle is at rest) as a means of further im-proving EPS control performance.

Required Control Performance During StaticSteeringAn EPS system controls the current of a motor togenerate assist torque based on the output of asteering-torque sensor. The assist torque is setso that it is nearly proportional to the sensor out-put. Increasing this proportional gain (i.e., mapgain) has the effect of reducing steering torque,but in the vicinity of 30Hz the control systemmay oscillate, causing a driver to feel unpleasantsteering-wheel vibration. In general, drivers tendto prefer a low level of steering torque duringstatic steering. Accordingly, the challenge is tosuppress steering-wheel vibration while at thesame time reducing steering torque.

Damping Control System DesignIncorporating an ObserverFig. 1 shows a model of a steering mechanismequipped with EPS. We linearize this model and

perform frequency analysis[1] to elucidate the os-cillation mechanism during static steering. Asshown in Eq. (1), this model can be representedas a balance between the motor inertia, thesteering torque applied by the driver Thdl, theassist torque produced by the motor TASSIST andthe reaction torque of the steering mechanismTtran. During static steering, the area of contactbetween the front tires and the road surface doesnot change in the region of small steering-wheelangles. Since the torsion of the tires acts as aspring, the steering angle and reaction torquehave a proportional relationship with nearly con-stant gain.

J.d2θs/dt2 = TASSIST + Thdl - Ttran ....................... Eq. 1a

TASSIST = Ggear {KT.Imtr - Tfric.sgn(dθs/dt)} ..... Eq.1b

Thdl = KTSEN (θhdl - θs) + CTSEN (dθhdl/dt - dθs/dt)....... Eq.1c

whereKT motor-torque constantImtr motor currentTfric motor-friction torqueKTSEN spring constant of the torque sensorCTSEN damping coefficient of the torque sensorθs angle of steering-column shaftθhdl steering-wheel angleGgear motor-gear ratioJ motor moment of inertia

The calculated frequency characteristics of thelinearized model are shown by the solid lines inFig. 2. The control system has been designedwith a phase margin in the vicinity of the cross-over frequency where the gain is 0dB. However,when the map gain is increased, the crossoverfrequency shifts to the high frequency side andthe phase margin decreases, which tends to in-duce control-system oscillation.

An effective way of suppressing steering-wheel vibration is to incorporate a control mea-sure that applies damping at the frequencywhere oscillation occurs, but an EPS system doesnot have a motor-speed sensor. Previously, themotor speed has been estimated in the low-fre-Fig. 1 A steering mechanism with EPS.

Supply voltage

Motor

Currentsignal

Voltagesignal

ECU

Detected steering torque Tsens

Torque sensor

Steering torque Thdl

θs

Reaction torqueTtran

Assist torque TASSIST

TECHNICAL REPORTS

· 19June 2001

quency region below 5Hz, but it has been diffi-cult to estimate the motor speed in the vicinityof an oscillation frequency of 30Hz.

To overcome this problem, we have devised amethod of estimating the motor speed by usingan observer[2] and have developed a dampingcompensator that provides compensation basedon the estimated motor-speed signal. This hasresulted in comfortable static steering perfor-mance free of steering-wheel vibration.

The configuration of the observer is describednext. This observer has been constructed by lin-earizing the model in Eq. 1, and has the steer-ing torque and motor current as its inputs. Tovalidate the operation and estimation perfor-mance of the observer, an estimate was madeof the motor speed when static steering oscilla-tion occurred. The results are shown in Fig. 3

where the thicker line indicates the motor speedmeasured with a rotary encoder and the fine lineshows the speed estimated by the observer. Theestimated results agree well with the measureddata in terms of both the amplitude and phaseof the motor speed.

The input signal is filtered with the aim ofapplying damping only at the frequency wheresteering system oscillation occurs, so the esti-mated motor speed fluctuates near a value ofzero. Therefore, when oscillation damping con-trol is performed based on the motor speed esti-mated by this observer, there is no feedback ofthe steering velocity component. Accordingly,since damping control does not manifest itselfas steering resistance, it is possible to achievecomfortable steering behavior.

A block diagram of the new controller thatperforms feedback of the estimated motor speedis shown in Fig. 4, and its frequency character-istics are indicated by the dashed lines in Fig. 2.Although the phase margin is larger than thatof the previous system, the control system isnot affected by high-frequency mechanical reso-nance because the high-frequency gain does notchange.

Vehicle testingTo verify the performance of the proposed con-troller, tests were conducted with an actual ve-hicle using the DSP-CIT controller developmenttool that is capable of generating a control pro-gram seamlessly from the controller model con-structed with MATLAB/SIMULINK. The map gainwas increased 3.3-fold over the conventional level.While oscillation occurred with the conventionalcontroller, see Fig. 5 (a), no oscillation occurredwith the proposed controller, Fig. 5 (b). These

Fig. 4 Block diagram of the contoller withdamping compensation.

+ +

TsensPhasecompensation

Tsens_pcAssistmap

Ibase Iref

Observer Kdamp ω

IdampImtr

Motor

TassistCurrent controller

Fig. 2 Frequency characterics of a steeringmechanism with EPS.

Gai

n (d

B)

806040200

-20-40

10210110010-1

10210110010-1

uency (Hz)

Pha

se (

deg)

0

-100

-200

Propo

Mot

or s

peed

(ra

d/s)

-40

-30

-20

0

10

30

-10

40

20

3.73.4 3.5 3.63.3Time (s)

EstimatedMeasured

Fig. 3 Estimation of motor speed using an observer.

TECHNICAL REPORTS


Fig. 5 Experimental results with conventionaland proposed controllers.

Ste

erin

g an

gle

(rad

)

0

5

10

00 1 2 3 4 5 6 77

000

1

1

2

2

3

3

4 5 6 77

0 1 2 3 4 5 6 70

20

40

Ste

erin

gto

rque

(N

m)

Imtr

(A

)S

teer

ing

angl

e (r

ad)

0

5

10

00 1 2 3 4 5 6 77

000

1

1

2

2

3

3

4 5 6 77

0 1 2 3 4 5 6 70

20

40

Ste

erin

gto

rque

(N

m)

Imtr

(A

)

Time (s)

Time (s)

(a) Conventional controller

(b) Proposed controller

results confirm that the application of an observerand damping control to the EPS control systemhave improved static steering performance.

In future work, efforts will be made to apply thiscontrol system to a high-output motor and toimprove the control performance further. Tech-nologies will be developed for applying EPS widelyto small and midsize cars with the aim of con-tributing to the further development of such ve-hicles. ❑

References1. M. Kurishige, et al., “A Control Strategy to Reduce Steering Torque

for Stationary Vehicles Equipped with EPS,” SAE Paper No. 1999-01-0403 (1999).

2. D. G. Luenberger, “Observing the state of linear system,” IEEETrans. Mil. Electron, MIL-8, 74-80 (1964).

TECHNICAL REPORTS

· 21June 2001

*Toshiki Kusama is with Sanda Works and Akio Uekawa is with the Energy & Industrial Systems Center.

by Toshiki Kusama and Akio Uekawa*

Development of Car NavigationSoftware

Continuing progress in car navigation systemsis rapidly increasing the cost and time neededto develop the associated software. To ease thisproblem, we have developed a car navigation soft-ware development framework called Victoriathat substantially improves software productiv-ity. The effectiveness of this framework has beenconfirmed and is described here together withan overview of Victoria.

The Victoria ApproachThe cost and time required for software devel-opment, including the extension of specifica-tions when a new model is created and the costand time needed to make modifications, are ex-pressed by the following equation.

Development cost and time = cost and timerequired to clarify specifications + theamount of modification work/work done perunit time

The following approach has been devised toreduce software development cost and time.

REDUCTION OF COST AND TIME REQUIRED TOCLARIFY SPECIFICATIONS. The process of mak-ing ambiguous specification requirements, in-cluding the specifications demanded by thecustomer and others, into rigorously definedspecifications is a major and vitally importantundertaking. A human-machine prototyping toolwas developed and implemented to make it easyto determine the specifications of ill-defined userinterfaces, which are responsible for the mostspecification changes in the entire system.

REDUCTION OF MODIFICATION WORK. The soft-ware was redesigned to localize modifications re-sulting from specification changes and to reducethe amount of modification work required. Thisinvolved separating the elements dependent on thehuman-machine interface specifications, the data-base-dependent elements and the platform-depen-dent elements. Additionally, an object-orientedlanguage was used to develop the framework inorder to localize the places requiring modificationand improve extensibility.

IMPROVEMENT OF WORK DONE PER UNIT TIME.The conventional development environment forembedded systems is referred to as a cross-de-velopment environment in which a host com-puter and the target hardware are connected tocarry out the development work. The newlydeveloped framework has a middleware layerthat absorbs differences between the develop-ment environment on the target hardware andthe development environment on a PC. This fa-cilitates the sharing of higher level car-naviga-tion software and enables the development workto be done on a PC.

An automatic software generator based on ahuman-machine builder was also developed toboost the development efficiency of the user in-terface element, which requires the largest num-ber of modifications on each model.

Object-Oriented DevelopmentC++ was adopted as the object-oriented language,and the software architecture was restructuredto minimize the amount of software modifica-tions needed to effect specification changes.This was done to reduce linkage between mod-ules and to localize modifications.

Development in a PC EnvironmentA middleware layer is provided to absorb devel-opment environment differences between thetarget hardware and a PC, allowing higher levelcar-navigation software to be shared.

The higher level software does not issue a sys-tem call directly to the operating system (OS)but rather through the middleware. Themiddleware for the target hardware environmentand that for the PC environment are interchange-able, facilitating software development on a PC.

Developing the software on a PC makes itpossible to utilize the outstanding developmentenvironment available for PCs. Because the en-tire process from software development to sys-tem testing and improvement can be done onthe same PC, the debugging cycle can be short-ened.

Fig. 1 shows an example of the developmentenvironment on a PC. Operations can be ex-ecuted via input from keyboard or mouse. A dia-

TECHNICAL REPORTS


Fig. 1 The development environment on a PC.

Fig. 2 The development environment on target hardware.

log box is provided to reproduce sensor data, suchas the signals of the global positioning system(GPS), gyro and vehicle-speed pulses, and vehicleinformation and communication system (VICS)data. It is also possible to simulate driving op-erations and reception of VICS data. At the finalstage of development, the middleware can besubstituted for that of the target hardware envi-ronment, as shown in Fig. 2, and a cross-com-piler can be developed in the target hardware

environment. Microsoft’s Windows NT4.0 hasbeen adopted as the OS of the PC environmentand Microsoft’s Windows CE as the OS of thetarget hardware environment.

Separation of Map Database-Dependent Partby Using a Map APIMap databases for car navigation systems haveyet to be unified, and each company currentlyuses its own map format. Previously, we devel-

TECHNICAL REPORTS

· 23June 2001

oped car-navigation software separately for eachcompany’s format, as shown in Fig. 3. To avoidduplication of development work for each for-mat, we separated the map database-dependentpart as a map-access library and specified a mapapplication program interface (API) as the pro-gramming interface for map access.

As shown in Fig. 4, the higher level softwareaccesses maps in each format via the map API,

making the software independent of the mapformat and allowing common usage.

Software Development Using a Human-Machine BuilderThe elements dependent on the human-machinespecifications were separated and a prototypingtool was adopted to support specification deci-sions and a human-machine builder was applied

Fig. 3 Conventional access to map database

Fig. 4 Access to map database through map API

rI

Navigation software (common) Navigation software (common)

Displaymap

Locator Routecalculator

Displaymap

Locator Routecalculator

Map access libraryfor format A

Map access libraryfor format B

Map DBfor

display

Map DBfor

locator

Map DBfor

routecalculation

Map DBfor

display

Map DBfor

locator

Map DBfor

routecalculation

Map DB (format A) Map DB (format B)

Inde

pend

ence

of

map

DB

Dep

ende

nce

on m

ap D

B

Navigation software (format A) Navigation software (format B)

Displaymap Locator

Route calculator

Displaymap Locator

Route calculator

Data access

Data access

Data access

Data access

Data access

Data access

Map DBfor

display

Map DBfor

locator

Map DBfor

routecalculation

MapDBfor

display

MapDBfor

locator

MapDBfor

routecalculation

Map DB (format A) Map DB (format B)

TECHNICAL REPORTS


to the automatic software generator for the userinterface element. The human-machine buildermakes it easy to modify the areas of the soft-ware that have to be revised. It also minimizesduplicated effort because consistent support isprovided from the upstream design process wherethe human-machine specifications are exam-ined.

As shown in Fig. 5, the map screens of a carnavigation system are created with a screeneditor and the operational transitions and actionsare described in state transition diagrams.Thanks to the use of an automatic code genera-tor, the screen parts and actions thus createdbecome the source code that can be run as it ison the PC environment and the target hardwareenvironment. The entire cycle from confirma-tion of an action to its modification can be ex-ecuted on the same PC. The screen editor is atool that we have newly developed for creatingcar navigation system screens and can be usedin the same way as an ordinary drawing tool. Acommercially available tool is used to describethe state transition diagrams.

The new software support framework promisesto increase the speed and flexibility with whichMitsubishi Electric can accommodate the rap-idly increasing complexity and sophistication ofcar navigation systems. ❑

Edit screen elements Edit state chart

Code generation

Car navigation software

User interface element

Screen elementsource code

Relation source code

CompileCompile

Target environment PC environment

Rev

ise

Fig. 5 Development user-interface element using the human-machine builder

TECHNICAL REPORTS

· 25June 2001

*Jiro Sumitani is with Himeji Works and Kiyoharu Anzai is with Mitsubishi Electric Control Software Corp.

by Jiro Sumitani and Kiyoharu Anzai*

Powertrain Control-SystemDevelopment Support

Today’s automotive powertrain control systemsmust meet a wide variety of performance require-ments, from improving the necessary inherentvehicle performance to satisfying demands forlower fuel consumption, cleaner exhaust emissionsand improved safety. Control-system specificationshave thus become extremely complex. The articlepresents an overview of a powertrain control-system development support environment createdto facilitate more efficient development of opti-mum control systems.

Powertrain Control-System DevelopmentCharacteristics and IssuesAn automotive powertrain control system is typi-cally developed through a repeated process ofidentifying and resolving problems based on ac-tual vehicle tests until the desired performanceis attained. The flow of this development pro-cess is shown in Fig. 1.

Global environmental concerns and other fac-tors today mean that the specifications of apowertrain control system must meet a widen-

ing range of performance requirements. As aresult, the control specifications have becomemore complex and require much greater devel-opment time and expense. However, because ofintensifying global competition, the develop-ment lead time of vehicles must be shortened.Accordingly, more efficient development ofpowertrain control systems is a high priority.

Development Measures Incorporated in theSupport Environment (see Fig. 2)

IMPROVING THE DEVELOPMENT PROCESS. Ac-celerating the study of control-system specifi-cations is one such important improvement.Simulations are conducted based on control-model diagrams created with commerciallyavailable control-system design tools and pro-grams can now be generated automatically, aswill be described below. These measures enabletheoretical studies of control-system specifica-tions and actual vehicle tests to be conductedmore efficiently.

AUTOMATING HARDWARE CALIBRATION. Apowertrain control system contains numerouscontrol parameters generally set according tocalibrations based on tests conducted with theactual vehicle. An automatic calibration envi-

Start of development

Study of control system specifications

Development of control program

Hardware calibration

Evaluation conducted with vehicle

Desired performance satisfied?

Completion of development

No

Yes

Fig. 1. Development flow for a powertrain controlsystem.

Objective

Accurate control-specification description

Facilitate early validation of control-system specifications

Reduction of program development time

Reduction of calibration time and cost

Development measure

Specification description with control model diagrams

Simulation based on control-model diagrams

Automation of each development stage

Automation of calibration

Fig. 2. Development measures incorporated indevelopment support environment.

TECHNICAL REPORTS


Fig. 3. ALICE development environment.

Creation of specifications/validationby simulation

Code generation

Calibration

Document generation

Specificationdocuments

ALICE workbench

ALICEdesigner

ALICEauto coder

MATLAB/Simulink (The MathWorks, Inc.)Targetlink (dSPACE GmbH)

ALICE data server

ALICElibrary

Compiler/Linker

Controlmodel

Objectcode

ALICEcaliber

ALICEdoctor

ronment has been developed to improve devel-opment efficiency.

Improved Software Development.

AUTOMATIC SOFTWARE GENERATION. If thecontrol program could be generated automati-cally from the control-model diagrams that rep-resent the control-system specifications,development efficiency would be substantiallyimproved. Automatic program generation itselfis possible today through the use of commer-cially available tools. However, with just thestandard functions of the commercial tools cur-rently available, there are some problems in thedescription of data processing. The developmentsupport environment was therefore created tosupplement the capabilities of commerciallyavailable tools.

AUTOMATIC DOCUMENT GENERATION. The pro-cedure usually followed in developing apowertrain control system is to make improve-ments on the basis of repeated tests conductedwith the actual vehicle. Such tests require de-tailed control-system specification documentscovering control parameters and control vari-ables. These documents must be created con-currently with the control program every timetests are conducted. An environment has beenconstructed that generates control-system speci-

fication documents automatically from control-model diagrams and other information so as toimprove development efficiency.

ALICE Development Support EnvironmentThe advanced logic integrator for car electron-ics (ALICE) environment (see Fig. 3) has beendeveloped for powertrain control-system devel-opment. ALICE is built around MATLAB/Simulink (from The MathWorks, Inc.), a com-mercially available control-system design sup-port tool, and the TargetLink code generator(from dSPACE GmbH) that runs on MATLAB andgenerates fixed-point program code. To these corecapabilities we have added the above-mentionedunique functions developed by Mitsubishi Elec-tric Corporation.

ALICE WorkbenchThis is an integration tool built to enable vari-ous development support tools created in-houseto be used with consistent operating procedures.These tools include ALICE Designer, which in-corporates additional functions for specificationdescriptions based on control-model diagrams,for comparing control-model diagrams and forconducting information searches. It improves thedevelopment efficiency and reliability of control-model diagrams using the ALICE Auto Coderthat automatically generates optimized objectcode based on control-model diagrams; the

TECHNICAL REPORTS

· 27June 2001

Fig. 4. Conventional calibration procedure

Parameter changes and measurement of control results

ECUControl

Set

ting

of te

st c

ondi

tions

KEYECU Powertrain control computer unit

Automatic calibration tool

Load control device

Fig. 5. System configuration of automaticcalibration environment

Start of calibration

Setting of test conditions

Control system evaluation

Completion of development

No

Yes

Control performance OK?

Everythingcompleted?

Sco

pe o

f pos

sibl

e au

tom

atio

n

Parameterchanges

ALICE Doctor (document generator) that auto-matically generates detailed specification docu-ments from control-model diagrams and otherinformation; and the ALICE Caliber that auto-matically generates the information for calibrat-ing the actual powertrain hardware.

ALICE Data ServerA control model generally describes the process-ing operation in terms of actual physical quanti-ties. However, functional limitations of themicrocomputer that is used require conversionto a fixed-point processing description. Conver-sion to a fixed-point description can be accom-plished with TargetLink, but the various typesof information required must be entered manu-ally in each mathematical function block.Therefore, an environment has been created thatfacilitates automatic entry all at one timethrough centralized management of the infor-mation needed for conversion to a fixed-pointdescription.

ALICE LibraryIt is difficult to describe interrupts and the likewith just the standard control-model descriptionblocks of MATLAB. To make such descriptionspossible and reduce the size of the program thatis generated, we created an independent blockfor powertrain-control descriptions and incorpo-rated it into the ALICE Library.

Development of an Automatic CalibrationEnvironmentCalibration has traditionally been performed asshown in Fig. 4. Control parameters werechanged manually and measured for each oper-ating state of an actual vehicle, and the processof changing the parameters was repeated untilthe measured results coincided with the targetcontrol values. One problem with this approachhas been the enormous time and cost involvedin processing the data.

To resolve this problem, input/output func-tions for controlling the operating state of theactual vehicle were added to the previously usedcalibration tool along with an automatic process-ing capability, as shown in Fig. 5. This resultedin the creation of an automatic-calibration en-vironment configured as shown in the diagram.

The article presents the overview of an environ-ment created to support more efficient develop-ment of complicated powertrain control systems.The use of this environment and associated toolsmakes it possible to improve overall efficiency,from the initial study of the control-system speci-fications through to vehicle calibration. Throughour work of developing car electronics, we in-tend to contribute to the future development ofthe automotive industry. ❑

TECHNICAL REPORTS


*Masahiro Tsujishita is with the Imaging Systems Laboratory and Eiji Asano is with Sanda Works.

by Masahiro Tsujishita and Eiji Asano*

Digital Signal ProcessingTechnology for Car Radios

In order to improve performance and reduce thecomponent count of an AM/FM car radio re-ceiver, we have converted it to digital operation.The article describes the system configurationand the signal-processing technology.

Development of a Digital FM radioA block diagram of the newly developed digitalAM/FM radio receiver is shown in Fig. 1. Thesystem mainly consists of an AM/FM front-end,limiter, mixer and digital signal processor. Asshown in Fig. 2, the digital signal processor per-forms the tasks of FM demodulation, multipathnoise reduction, pilot synchronization, stereodemodulation and pulse-noise reduction. Anexplanation follows of the method adopted toreduce FM demodulation distortion and the pro-cessing procedure to reduce external noise, bothcharacteristic features of the newly developeddigital signal processing technology.

Improving Digital FM DemodulationDistortion

CORRECTION OF IF SIGNAL-AMPLITUDE VARIA-TION. The bandwidth limit of the low-pass filterplaced before the A/D signal converter causesthe carrier amplitude of an FM signal to vary.FM demodulation of the signal under these con-ditions generates distortion. To eliminate thisdistortion, the amplitude compensator detectsthe carrier amplitude X and multiplies the out-put signal of the FM demodulator by 1/X, asshown in Fig. 3. With the newly developed digi-tal signal processing technology, this amplitudecorrection achieves a distortion of less than 1%(for 30% modulation at 1kHz) while allowingthe cut-off frequency of the forward low-pass fil-ter to deviate by ±20%.

Fig. 1 Block diagram of an AM/FM receiver

Fig.2 Block diagram of digital signal processing

Demodulated AM audio signal

AM/FM front end

BPF 10.7MHz

FM IF signal

Limiter Mixer Digital signal

processor

Poweramplifier

IMPROVEMENT OF DIGITAL FM DEMODULATIONDISTORTION. In the arc-sine compensator in Fig.3, an approximation of the arc-sine curve is usedto correct the distortion caused by the staticcharacteristic of the sine curve generated by thequadrature FM demodulator. This has the effectof reducing the distortion by about 10dB for 100%modulation.

Improved Noise ReductionFor good radio reception in the automotive envi-ronment, it is necessary to reduce pulse noisecaused by electromagnetic wave noise andmultipath noise caused by signal reflection frommountains and man-made structures.

REDUCING FM RECEIVER PULSE NOISE. Thisnewly developed signal processing technologysuitably switches between interpolation meth-ods depending on the frequency components ofthe FM stereo demodulated signal. Moreover,

Fig. 3 Block diagram of FM demodulation

IF signal

Quadrature FM demodulator

Delay X XLPF

Amplitudecompensator

ArcSinecompensator

Compositesignal output

Near 1/X3rd-order

polynomial approximation

Carrieramplitudedetector

Lch

Rch

Lch

Rch

FMIF signal Audio

output level

Field strengthAudio output (FM)

AM demodulatorsignal (audio signal)

Audio output (AM)Linear

interpolation

High-speednoise detection

Stereo demodulation

Pilot synchronization

FM demodulator

Pulse noise

reduction

Noisedetection

Multipathnoise

reduction

Noise detection

TECHNICAL REPORTS

· 29June 2001

as shown in Fig. 2, the L-channel and R-chan-nel signals are corrected separately following ste-reo demodulation. This approach is especiallyeffective in reducing high pulse noise in stereo-signal reception. The signal waveforms beforeand after interpolation are compared in Fig. 4.

REDUCING MULTIPATH NOISE. All of the spikynoise lasting several µs that arise in the FM de-modulated signal are processed one by one withthis new technology to maintain the originalsignal values. This procedure is highly effectivein reducing multipath noise. The signal wave-forms before and after interpolation are com-pared in Fig. 5.

Fig. 4 The effect of pulse-noise reduction (FM)

Fig. 5 The effect of multipath noise reduction

REDUCING AM RECEIVER PULSE NOISE. Asshown in Fig. 2, reduction of AM receiver pulsenoise involves detection and interpolation ofnoise components in the AM demodulated sig-nal. Noise correction is performed by linear in-terpolation, whereby the correction interval isadapted to match the noise occurrence intervaland pulse height. This approach is highly effec-tive in reducing pulse noise. The signal wave-forms before and after interpolation arecompared in Fig. 6.

Fig. 6 The effect of pulse noise reduction (AM)

Time(b) After interpolation

(a) Before interpolation

0

-20000

-10000

10000

20000

Time

-20000

-10000

10000

20000

Leve

lLe

vel

0

Approx. 28ms

Level 0

Level 0


0 3Time (ms)

0 3Time (ms)

(b) After interpolation

00 100100

Level 0

Level 0

Time (ms)Time (ms)

0 100Time (ms)


(b) After interpolation

For current FM broadcasts, the newly developeddigital receiver can handle the signal from theIF stage right to the audio signal. An adaptivesignal processing technology using a digital sig-nal processor has been adopted that provides agreater noise reduction effect for both AM andFM broadcasts than the analog system it willreplace. ❑

MITSUBISHI ELECTRIC CORPORATIONHEAD OFFICE: MITSUBISHI DENKI BLDG., MARUNOUCHI, TOKYO 100-8310, FAX 03-3218-3455

Documents

Automobile-Human Technology Edition€¦ · Advanced Technologies for Man/Automobile Harmony by Masaaki Nangaku* T he automobile, while comfortable and convenient for its users, causes