Ultra-Wideband Radar Sensors for Short-Range Vehicular Applications

Ultra-Wideband Radar Sensors for Short-RangeVehicular ApplicationsIan Gresham,Senior Member,IEEE,Alan Jenkins,Member,IEEE,Robert Egri,Member,IEEE, Channabasappa Eswarappa,Senior Member,IEEE,Noyan Kinayman,Member,IEEE,Nitin Jain,Member,IEEE, Richard Anderson,Frank Kolak,Ratana Wohlert,Shawn P.Bawell,Member,IEEE,Jacqueline Bennett,Member,IEEE,and Jean-Pierre Lanteri,Member,IEEEInvited PaperAbstract—The recent approval granted by the Federal Com-munications Commission(FCC)for the use of ultra-wideband signals for vehicular radar applications has provided a gateway for the introduction of these sensors in the commercial arena as early as2004.However,the rules governing the allowable spectral occupancy create significant constraints on the sensors’opera-tion.This is further complicated by the variety of applications that these sensors are being required to fulfill.A review of the motivation for the development of these sensors is followed by a discussion of the consequent implications for waveform design and limitations on system architecture.Other practical considerations such as available semiconductor technology,packaging,and assembly techniques are reviewed,and results are presented for conventional surface-mount plastic packages illustrating their usefulness in the greater than20-GHz frequency range.Suitable antenna technology for wide-band transmission is presented that is compliant with the specific restrictions stipulated in the FCC ruling.Finally,all of these considerations are combined with the presentation of a compatible integrated-circuit-based transceiver architecture.Measured results are presented for several critical circuit components including a+12-dBm driver amplifier for the transmitter,an RF pulse generator that can produce sub-1-ns pulses at a carrier frequency of24GHz,and a single-chip homo-dyne in—phase/quadrature down-conversion receiver that has a cascaded noise figure of less than7dB.All circuit components are fabricated in SiGe.Index Terms—Automotive,radar,SiGe,ultra-wideband(UWB).I.I NTRODUCTIONA FTER YEARS of public hearings and rule-making de-bates,the approval granted by the Federal Communica-tions Commission(FCC)through its modified and amended 47CFR Part15regulations allocates over7GHz of usable un-licensed spectrum between3.1–10.6GHz for communications and imaging systems based on ultra-wideband(UWB)devices [1].This ruling has been the subject of much interest and,in-deed,controversy[2]–[6].A large number of both proponentsManuscript received August4,2003;revised May13,2004.I.Gresham,A.Jenkins,R.Egri,C.Eswarappa,N.Kinayman,R.Anderson, R.Wohlert,S.P.Bawell,J.Bennett,and nteri are with Corporate Research and Development,M/A-COM,Lowell,MA01853USA.N.Jain is with Anokiwave Inc.,San Diego,CA92130USA.F.Kolak is with the Mitre Corporation,Bedford,MA01730USA.Digital Object Identifier10.1109/TMTT.2004.834185and opponents of the change in the regulations have been unable to agree upon the level of maximum radiated emission levels that unlicensed users of the spectrum must adhere to in order that existing radio,communication,and other licensed services will be unaffected by harmful interference caused by UWB de-vices.However,the potential ability for UWB devices to occupy spectrum concurrently with existing systems,thereby allowing the scarce spectrum to be used more efficiently,has led to the approval being granted[1].The National Telecommunications and Information Administration(NTIA),who are responsible for the licensing and control of federal and government fre-quency bands in the U.S.,were particularly concerned with the potential cumulative effect sensors could have on the global po-sitioning system(GPS)bands,with the consequence that UWB devices are required to have very low spectral emissions in the 960–1610-MHz band[7].In addition,although maybe less well appreciated,the ruling also approved a second allocation of an unlicensed7-GHz wide spectrum between22–29GHz that is intended exclusively for vehicular radar systems.These systems are intended to detect the location and movement of objects near a vehicle by devel-oping a continuously updated360-radar map of the vehicles’surroundings.Depending upon the sophistication of the system and its intended purpose,this would enable features that can provide,amongst other functions,collision avoidance,enhanced airbag activation,and improved road handling through interac-tion with other vehicular dynamic systems.This allocation is no less controversial than the lower frequency ruling due to its requirement for coexistence with various radio astronomy and other sensitive frequency bands.In fact,it may be more prob-lematic in that it is not just adjacent to these bands,but indeed overlaps them.As we shall see,this has had a profound influ-ence on system and waveform design,and on the perceived way that systems will be introduced,first in the U.S.,and eventually in the global marketplace.In order to understand the rationale for such a controversial move,it is worthwhile reviewing the motivation for the intro-duction of automotive radar systems.In Section II,an overview of the requirements for and of short-range radar(SRR)sensors is used as an introduction to the details of the FCC UWB ruling for vehicular radar applications.Section III considers this ruling0018-9480/04$20.00©2004IEEEin the context of its implications for waveform and system de-sign,and discusses some of the tradeoffs that were considered in architecture selection.Details of sensor level operational performance and consequent system level performance are shown.Section IV highlights the technological choices—me-chanical form,semiconductors,packaging,interconnects,and interfaces—that are imposed by the architecture selection,but more importantly,by both the constraints of the marketplace and the need for the sensors to be networked with other func-tions on the vehicle.In Section V,a specific implementation of a generic integrated-circuit(IC)-based SRR is shown as an example[8]–[10]along with several examples of measured circuit-level performance to demonstrate compliance with the sensor level requirements detailed in Section III.Finally,some comments on the future of UWB radar are presented.A.Case for Vehicular SRRThe subject of automotive radar is not new,and developments have been occurring on a regular basis,as reported in the tech-nical literature for many years[11].One of the main barriers to widespread introduction to date has been the ability to fulfill the extremely demanding technical requirements at a cost level that the average consumer will find palatable.However,signif-icant advancements have been made due to the strong support that the introduction of such sensors has from the automotive industry in general[12].They perceive that the increased safety such vehicles offer to the consumer will provide a differentiating competitive advantage.A far more important,and indeed more persuasive,argument for the introduction of smart vehicles is that any increase in safety in automobiles should have a conse-quent effect in reducing the number of fatalities that occur every year.It is,of course,this second argument that has found favor with the various governmental and regulatory agencies around the world that are required to pass the corresponding legislation. Consider,in2002,the number of occupant fatalities in road crashes in the U.S.increased for the first time since1990to an estimated figure of over42000people annually.The total number of fatalities per year in the U.S.has been relatively con-stant at around40000since1991[13],despite improvements in vehicle safety design and features.This unacceptably high level led to two major initiatives in the U.S.for the Department of Transportation to work in partnership with industry to develop and deploy effective collision avoidance products[14],[15]. In2001,the National Transportation Safety Board(NTSB) concluded that the benefits and effectiveness of collision warning systems(CWSs)and adaptive cruise control(ACC) [16]were of public importance.Independent of this,a similarly high level of fatalities in the European Union—over44000in 1998—led to the European Commission(EC)setting a goal of a50%reduction in the number of annual road deaths by2010 [17],[18].Although this ambitious target can probably not be reached through the introduction of new technology alone,the EC concluded that one of the vital steps in reaching this goal was through the introduction and implementation of Advanced Driver Assistance Systems,and that a key component of such systems was the introduction of UWB radar at24GHz[18].In September2003,the EC committed to work with the European Conference of Postal and TelecommunicationAdministrations Fig.1.Collision probability as a function of increased driver reaction time for varying crash scenarios[22].(CEPT)to remove the regulatory barriers that prevented the adoption of UWB radar in Europe to facilitate the rapid deployment of SRRs at24GHz with an eventual migration to 79GHz[19].plete Environment Sensing for Automotive Vehicles Given the multiplicity of various driving and traffic sce-narios,a variety of active(that provide automatic intervention) and passive(that provide driver information)safety systems are required to give a vehicle the capability to not only perceive and understand its environment,but to also act upon it.The more important of these systems in terms of traffic and object aware-ness are largely vision—both video and infrared(IR)—and radar based[20],[21].This emphasis on introducing a variety of sensing technologies to provide a vehicle with a complete environmental awareness as a means of improving safety arises from some interesting statistics.It has been estimated that95% of all road accidents involve some human error and that,in76% of all accidents,a human is solely to blame.Associated with this is the thought-provoking statistic that almost all collisions at intersections,with oncoming traffic,and rear-end impacts, could be avoided if driver reaction time was shifted forward by2s[22],[23],as presented in Fig.1.Confirmation for this comes from a related subject of much research whereby event data recorders(i.e.,black boxes)are added to vehicles to monitor,record,display,or transmit(Tx)pre-crash,crash,and post-crash data element parameters from a vehicle[24].It is postulated that almost all of these accidents could be avoided through the introduction of CWSs and automatic vehicle interventions[18],[25]through the use of a hybrid sensor array to form a safety belt around the vehicle.A typical schematic of one of these sensor arrays is shown in Fig.2.Radar is perceived as a key element in the sensor array due to its ability to offer an immunity to weather condi-tions that is unavailable with other technological solutions.It also offers the vehicle manufacturers a stylistic advantage over ultrasonic or video sensors in that it can be mounted behind aGRESHAM et al.:UWB RADAR SENSORS FOR SHORT-RANGE VEHICULAR APPLICATIONS2107plete environment sensing for automobiles is achieved through a hybrid array of multiple sensor technologies and functions.The functions and typical system range of operation shown are representative only.Short-range sensing requirements for the radar sensors are envisaged to cover a variety of applications,each with differing needs for range,update rate,range resolution,and other key system parameters.fascia that (although dissipative)can be considered transparent to the radar signal without requiring specific cutouts or sim-ilar accommodations.The longer-range ACC systems have been readily available in production vehicles for over three years,and are being increasingly supplemented by video systems to im-prove the quality of information available.In the ACC systems,a narrow beam (2or 3)is typically scanned over the front of the vehicle over an azimuth variationof 5to 8,and the information used to regulate its speed so that a sufficient safety margin of distance to the vehicle ahead is maintained.These systems have a necessarily narrow scanning angle (for adequate azimuth resolution of objects that may be 150m ahead),and a range resolution of 1–2m.This distance requirement also means that the high-gain antennas have a necessarily small beamwidth that inhibits their usefulness in their ability to detect and track objects at close range(20m).Increasing the number of beams to increase the scanning angle is problematic for two reasons.Firstly,the antenna has to become more complex and,therefore,more expensive.Secondly,as the number of beams is increased,the scan rate (and,hence,dwell-time per range gate)is affected.The update rate for tracking algorithms is particularly impor-tant for very short-range operation.A further consideration is the desire to have 360coverage around the car.A single sensor is clearly not a suitable solution,and a system comprised of mul-tiple networked sensors that are distributed at appropriate loca-tions is required [26].The precise number of sensors is a func-tion of the amount of coverage required,and the corresponding coverage overlap of adjacent sensors.However,depending upon the azimuth coverage of each sensor,anywhere from 4to 16in-dividual radar sensors may be required,with placement biasedtoward the front bumper,followed by the rear bumper corre-sponding to the statistically highest crash areas.Proportionally fewer sensors are required for sideways looking operation.C.Short-Range Sensors—The Requirement for UWB Radar at 24GHzFig.2illustrates the variety of applications that the short-range sensors are required to fulfill [27],[28].They range from simple parking aid functions to more elaborate pre-crash detec-tion,and stop-and-go or short-range cruise control functionality.The precise requirements for these applications differ between each other,and also between individual vehicle manufacturers.However,it is possible to define some typical specifications,at least for the more commonplace applications.A summary of some of the key requirements for four differing sensing func-tions is given in Table I.These represent the more immediate of the applications that automotive manufacturers want to in-troduce on vehicle platforms as a first step to the eventual in-tegrated sensor solution.Parking aids are already available on many vehicles through the use of ultrasonic sensors.However,radar sensors offer potentially superior range performance,as well as being less susceptible to inclement weather conditions,or requiring custom cutouts in the bumper fascia.The pres-ence of the radar sensor is also considered to then be a gateway to some of the more complex applications,as it can be more easily integrated into the entire sensor network and its opera-tion extended,for example,from being a simple parking aid to a blind-spot detector.One of the more demanding applications is the extension of ACC to a short-range stop-and-go function-ality.This requires not only good range resolution,but also the2108IEEE TRANSACTIONS ON MICROW A VE THEORY AND TECHNIQUES,VOL.52,NO.9,SEPTEMBER 2004TABLE IS UMMARY OF T YPICAL S HORT -R ANGE S ENSOR S YSTEM R EQUIREMENTS FORA V ARIETY OF D IFFERING APPLICATIONSability to detect and track objects such as a bicycle at ranges of over 20m initially,leading to perhaps 30–40m over time as systems become more complex and mutually dependent.The longer-range ACC systems have particular difficulty in moni-toring the corners of vehicles,for example,and are looking to the short-range sensors to provide an increased warning or re-action time for cut-in stly,pre-crash detection net-works are seen as being essential,but demanding,applications that are critical to improving safety.An example of such a func-tion is pre-impact sensing,whereby the closing velocity of an object that is projected to impact a vehicle may be tracked in order that the threshold voltage for airbag deployment may be dynamically adjusted.This is especially important in side-im-pact situations where extra milliseconds for airbag firing could be important.Despite the differing requirements for these appli-cations,the concept of a networked sensor system,plus the cost requirements,oblige a common sensor architecture and design that is adaptable to be used in each worked Sensor OperationA distributed networked sensor structure for the radar re-sults in a system that has a hierarchical detection and tracking process.The totality of information available for compiling and updating the radar map is comprised of information from each of the discrete sensors,plus additional data from the other sen-sors that are included in the hybrid sensor network.The infor-mation available from a single sensor is,therefore,a subset of the total information required by the total radar system.Each sensor consists of an RF/IF front-end that forms the air inter-face —Tx and receive (Rx)circuitry and antennas —plus associ-ated control and signal circuits.Local digital signal processing (DSP)in each sensor is required to perform first tier detection analysis by generating a priority list of detected objects within the sensors ’field of view.The objects and their associated pa-rameters —range,amplitude,and relative-velocity —are sent via a network to a central processor or radar decision unit (RDU).The RDU is also networked via the controller area network (CAN)-bus to accept inputs from other systems within the ve-hicle,as shown in Fig.3.The entirety of the separate sensor in-puts are then analyzed by the RDU,which performs higher level detection and tracking algorithms,and initiates appropriate de-cisions and instructions that are output to the network bus.The requirements for detection ability for each sensor are,therefore,different for the complete radar system.For example,typicalnumbers for the probability ofdetectionand probability of falsealarm of an object may be 0.9and10,respec-tively,at the individual sensor level,but are substantially im-proved at the system level.Here,we will focus on the issues of the discrete UWB SRR sensor that forms a component part of the complete networked radar.For a comprehensive discourse on the much more complex issue of how the networked system functions,see Klotz [26].II.UWB S ENSORS AND THE FCC P ART 15R ULING The key specification common to all of the differing appli-cations that the 24-GHz sensor is required to fulfill is that of range resolution,as indicated in Table I.The accuracy of time information and,therefore,the ability to precisely determine an object ’s location,is inversely proportional to the signal band-width.The precise measurement of object movements is essen-tial for the prediction of trajectories and the prevention of false alarms.For a requirement of sub-10-cm resolution,therefore,the spectral occupancy of the main lobe of the incident wave has to be at least 3-GHz wide and more likely needs to be of the order of 4-GHz wide to allow for errors,variations,and de-sign margin.The availability of such a broad spectral frequency band has to be balanced with other practical considerations.The gain,or focusing ability,of an antenna is proportional to the antenna aperture.The antenna of each sensor,therefore,has to be small enough that an aperture of suitable electrical size can be physically mounted in the restrictive space behind a vehi-cles ’bumper fascia.In addition,as a percentile of center fre-quency,it becomes easier to produce antennas that support sig-nals of at least 4-GHz wide by increasing the frequency of op-eration.However,there are also several practical considerations that conversely argue for as low an operating frequency as pos-sible [29].The requirement for multiple sensors in a networked configuration forces the average cost per sensor to be low.De-spite the advances made in ACC radar at 77GHz,technology has not yet progressed enough such that a 77-GHz multisensor system can be cost competitive.In addition,fascia and atmo-spheric loss is far higher at 77GHz than at 24GHz —an impor-tant issue in any power-limited stly,it is believed by Short-range Automotive Radar frequency Allocation (SARA)1that the short-range sensor market is more likely to grow with a frequency allocation that is not dependent upon rapid technolog-ical advances.Weighing all of these factors together,the (nearly)1SARAis a consortium of automotive manufacturers and suppliers that wasformed in Spring 2001to coordinate frequency allocation issues for SRR sensors with the FCC,European Telecommunications Standards Institute (ETSI),and the International Telecommunication Union (ITU).GRESHAM et al.:UWB RADAR SENSORS FOR SHORT-RANGE VEHICULAR APPLICATIONS2109Fig.3.Schematic representation of multiple short-range sensors connected to a local network.Each sensor passes a detected object list onto the network for analysis in the RDU.global industrial,scientific,and medical (ISM)frequency allo-cation around 24.125GHz is ideal.Well,almost.Unfortunately,there are several licensed frequency bands that surround the 24.125-GHz ISM band that are used for predominantly low-power and sensitive appli-cations.In particular,a global spectrum allocation between 23.6–24GHz is used for radio astronomy and remote sensing.It is this band,in particular,that has caused consternation amongst other spectrum users,particularly the European Space Agency (ESA),European Meteorological Satellite Service (EUMETSAT),Committee on Radio Astronomy Frequencies (CRAF),and others [30].Passive satellite-based sensors are used by the Earth Exploration Satellite Service (EESS)for monitoring water vapor and trace gas concentration in the Earth ’s atmosphere,and the lifting of the background noise level through the introduction of millions of albeit,low-level,radiators is seen as a cause for concern.There are also concerns about the effect of increases in wide-band spectral emissions on fixed service communications (also in the 23–26-GHz band in Europe).This has led to the specific regulations imposed by the FCC,which calls for reduced emissions at certain elevation angles above the horizon over time.The specification is written in terms of an emitted power spectral density,or effective isotropic radiated power (EIRP),which allows for either a reduction in the peak transmitted power or,as is more likely,a reduction in the elevation sidelobes of the Tx antenna.Similar proposals in Europe are at the time of writing still under consideration by the ETSI.One mitigating solution that is the subject of much research —notably by the Radarnetconsortium 2—is the proposal by SARA in conjunction with the European Radiocommunications Office (ERO)to impose a sunset date on the introduction of 24-GHz radar systems of 2014or ten years after their introduction,whichever is later.In conjunction,SRR sensors would slowly migrate in frequency to be based at 79GHz.This would set an upper limit on the maximum number of sensors produced at 24GHz and,thus,limit the increase in background noise that the EESS sensors would be exposed to.It should be noted that the corresponding spectrum at 79GHz has also yet to be allocated.Table II summarizes the key elements of the Part 153ruling governing UWB radars for vehicular applications.The asso-ciated spectral mask is illustrated in Fig.4.To be considered an UWB device,the fractional bandwidth of the spectrum measured at the 10-dB point from the peak must be at least 20%or 500MHz,regardless of the fractional bandwidth.The vehicular radars must operate between 22–29GHz in such a way that the center frequency and the frequency at which the highest level emission occurs must be greater than 24.075GHz.Thus,to be considered UWB,the vehicular radar must have at least 500-MHz bandwidth to satisfy the regulations.Normally,the spectral density of the average emission in this band should not exceed 41.3dBm/MHz.To reduce potential interference2Radarnet —AEuropean-based consortium of car manufacturers,electronicssuppliers,and researchers who are developing SRR applications for 77GHz.[Online].Available:.3The Electronic Communications Committee,CEPT issued an approval no-tice for the allocation of the 77–81-GHz spectrum for Automotive Short-Range Radars on March 19,2004.2110IEEE TRANSACTIONS ON MICROW A VE THEORY AND TECHNIQUES,VOL.52,NO.9,SEPTEMBER 2004TABLE IIS UMMARY OF THE K EY E LEMENTS OF THE FCC R ULING G OVERNINGUWB R ADARS FOR V EHICULAR A PPLICATIONS[1]Fig.4.Spectral mask of the UWB ruling for vehicular radar.Note that the peak radiated emission limit overlaps the 23.6–24-GHz astronomical band and,thus,has specific additional limitations imposed in this area.with radio astronomy observations and passive earth sensing satellites,the FCC further limits the radiated emissions by requiring that,in the 23.6–24.0-GHz band,the EIRP of the antenna sidelobes beyond 30above the horizontal plane not exceed 66.3dBm/MHz until 2010,and dropping to 76.3dBm/MHz beyond 2014.In addition to the average limit of 41.3dBm/MHz,the FCC also effectively limits the peak EIRP density emission to 17dBm/MHz in a 50-MHz band around the frequency of the highest power emission.These two constraints effectively dictate a maximum duty cycle for a sensor in pulsed operation to around 0.4%to take full advantage of the average power specifications.The FCC is clearly con-cerned about the interference generated by potentially millions of radars operating simultaneously.Of course,there are also limits placed on spurious emissions;see the details in [1].One proposed solution to the restriction on elevation sidelobes is to center the spectrum of the transmitted signal above 25GHz sothat the first null ofthespectrum falls at the restricted band,thus reducing some of the demands upon antenna design.This may well become the preferred interim solution,although early sensor production is still planned for an ISM-band center frequency.TABLE IIIS UMMARY OF THE P RINCIPAL S HORT -R ANGE S ENSOR L EVEL R EQUIREMENTSFOR O BJECT DETECTIONIII.S YSTEM A RCHITECTURE AND W A VEFORM D ESIGN Table III summarizes some of the more important operational and practical requirements for a generic short-range sensor.By combining an understanding of the EIRP spectral limitations with our knowledge of the operational performance require-ments of the networked radar system,it becomes possible to compare the relative merits of different sensor architectures,and select the most appropriate.A.Sensor Architecture ConsiderationsOne can broadly characterize radar architectures of interest in vehicular applications into three generic types,which are:1)pseudorandom noise (PN)coded continuous-wave (CW)radars;2)classical frequency-chirped systems;and 3)pulsed radars [31].Simple CW Doppler radar systems are not consid-ered here,as almost all of the vehicular applications require the detection of zero relative velocity mon to all of these techniques is the ability to employ a process known as pulse compression.This is a traditional technique for enhancing radar performance and may be used to improve sensitivity for detecting targets at long range.It is also possible to combine techniques to form hybrid radar systems.Frequency chirp radars that use pulse compression are ideal for long-range higher power applications.Resolution is related to how fast and over how wide a bandwidth it is possible to generate a well-defined chirp.Close-range and closely spaced targets also put stringent requirements on local oscillator (LO)phase noise.Both of these requirements can be challenging in the UWB context in any cost-effective manner.PN-coded radar —essentially a spread-spectrum-type system —also provides the possibility for pulse compression,often referred to as coding or processing gain in the context of PN codes.Here,flexibility exists with the choice of the code,allowing designers to trade various performance parameters,with the limitation that dynamic range is often restricted by code auto-correlation properties [28].The ability to assign different codes to systems that share the same spectrum also provides an inherent resilience to interferers making it a good。