1.1 Automotive Radar Potential

Each year thousands of people lose their lives in automotive mishaps. Even more people sustain debilitating injury ranging from dislocations to limb loss. On 30 December 1984, Columbia Broadcasting, by way of its popular television program, “Sixty Minutes,” extolled the virtues of automotive air bags in saving lives and preventing serious injury. In support of safety devices (like air bags), statistics are often quoted. In a year’s time, over 45,000 lives will contribute to the American highway fatality statistic [8]. Of course, to believe that a single safety device or even a multitude of safety options could eradicate fatality (and serious injury) statistics entirely is foolhardy. Yet, there exists a class of accidents which does not have to be as violent and as injurious as they are [14]. Further, there is yet a class of accidents which may be entirely avoided. Consider the late-night situation where the unthinking operator is “over-driving” the headlamps: Illumination does not cover a sufficient distance for a safe panic stop. If, let us say, there was a stalled second vehicle on the road without warning lights, a

collision is highly probable. If the road is curved or if the weather is less than optimum, the probability of a mishap is further increased. It is clear in such a case that the accident may be avoided entirely if warning of the obstruction (target) is provided to the driver in sufficient time to react. Even the most conscientious driver would welcome advance warning of an impending hazard. This is the basis for a forward looking collision avoidance radar system. One should not, however, regard such a radar system as a panacea for avoiding all classes of impacts. A situation may be posed, for example, where the physical dynamics dictate that a collision must occur. Yet the severity of numerous collisions could be lessened by the operator responding just an instant quicker. Thus, a forward looking collision avoidance radar system could provide accident mitigation. Besides providing timely operator warning, other possibilities associated with the application of automotive radar become apparent: passive restraint system sensitivity control and “smart” cruise controls. As a safety system, however, automotive radar must be regarded as separate from all others. Other safety measures (like seat belts or air bags) presume the accident has already taken place and attempt to protect the vehicle’s occupants during the impact. In contrast, radar is deployed with the intent of avoiding the impact in advance. Some investigations have suggested the possibility of an

automatic radar controlled brake [50]. It is not difficult to imagine, however, how this over-ambitious application of radar could create more problems than it solves! Product liability is an area that technologists cannot ignore. The experimental work conducted during the initial phase of the presented research involved vehicles that are capable of automatic antiskid braking as prompted by a signal from the forward looking radar if a limit beyond the operator warning limit (light and tone provided) is exceeded. These experimental systems, therefore, have been referred to as: “Collision Avoidance Radar Braking Systems” [105], or “Radar Brakes.” At no time, however, were the vehicles operated on the road with this automatic feature engaged. To take (even a portion of) control away from the operator may result in unexpected and quite deleterious results. It is indeed difficult to imagine that such an automatic brake would ever be practical or even desirable!

1.2 Automotive Radar Practicality

The development of a practical consumer oriented automotive collision avoidance radar system just 20 years ago was unthinkable. This is not at all surprising since two needed techniques were immature at that time: low-cost, reliable microwave components and the necessary signal processing electronics. However, in the past 10 years, this nation has witnessed a phenomenal growth in radio frequency

(RF) communications products and small-scale electronic data processing equipment. To emphasize the profundity of this statement, consider each area separately: First, realize that X-band (10.5 Gigahertz) doppler radar transceivers are replacing door mat switches for activation of automatic doors at supermarkets, airports, and hospitals. It is indeed difficult to imagine that a cost effective and reliable replacement has been made for the simple mat switch! The RF components used in these doppler radar transceivers are not unlike the “front end” of an automotive radar system. Commercially available RF sources are presently available up through the millimeter wavelengths (100 GHz) [35]. Millimeter wavelengths are especially attractive in terms of package miniaturization. Gallium Arsenide and Indium Phosphide Gunn sources are capable of meeting the modest output power (20 milliwatts) requirements for automotive radar. Second, consider that a basic four-function, hand-held calculator with no memory sold for approximately $200 just 15 years ago. The same instrument can be had today for just under five dollars. Microprocessors today are seen everywhere. The 8 and 16 bit personal microcomputers are currently in vogue, with the possibility of a 32 (and higher) bit machine for the home on the horizon. Microprocessors and associated firmware have become so economical that their strengths are being exploited in a spectrum of consumer products including

sewing machines, food blenders, and washing machines. Many persons who have witnessed the impressive performance of the experimental vehicles studied in this research are surprised to learn that the warning decision computation is performed by a basic 8 bit microprocessor (INTEL 8080, NMOS). Since, of course, these experimental vehicles do not have spatial microwave beam scanning capability, the processing being accomplished only embraces kinematic calculations. However, with the aid of high density, inexpensive electronic data storage techniques, microwave beam scanning could provide target signature (pattern) recognition.

State-of-the-art memory technology is represented by random access memories (RAMs) with 256 kilobits per chip. Fujitsu Ltd. has demonstrated a 256 kilobit dynamic RAM with access times below 100 nanoseconds. With gate lengths of only 2.5 microns, the entire chip occupies an area of 34.1 square millimeters. One million bits on a single chip should make its debut this year [35].

It should be recognized that solid state technologies for the production of digital and RF devices are not mutually exclusive. In fact, microstrip and finline techniques are entirely feasible at frequencies beyond 100 Gigahertz [64, 81]. Thus, many digital device production methods (like masking and photolithography) are directly applicable to the manufacture of RF circuitry and components. This single observation explains why production

costs in both areas are plummeting. An additional benefit to be derived for automotive radar capitalizes on the possibility of an integrated system on a single substrate. This may appear quite ambitious at the present stage of development. However, economical and reliable automotive radar sensors would follow from volume requirements. From a technological standpoint, there is presently no reason why such an integrated assembly is not practical. The sensor would incorporate a beam scanned planar antenna, duplexer, microwave generator (source), modulator, control electronics (supervisory and signal coding to eliminate interunit interference), amplification, and detection. The entire unit would be placed directly in back of the rear view mirror of the vehicle. This area allows for an unobstructed high outlook position for the radar beam. This location also poses no ambient stresses beyond those that other sophisticated automotive electronic systems (e.g., electronic fuel injection and spark control) endure. The proposed radar sensor would be no larger than the rear view mirror itself.

1.3 Objectives of this Work

This work is a contributory effort to advance forward looking collision avoidance automotive radar systems from developmental stages closer to a consumer product. Three significant research phases are presented: First, the

technique developed for documenting the performance of an experimental (baseline) automotive collision avoidance radar braking system is described. An 18-minute television production suitable for public broadcasting is the culmination of this exploratory phase. Depicted within the program are actual over-the-road driving situations (panoramic views) with associated real time telemetry (radar processor signals reflecting electronic system response and vehicle dynamics). The television program (interview format) is tutorial by nature. Yet, it adequately demonstrates the impressive performance of an automotive radar system utilizing technology which is almost a decade old. The program also highlights the deficiencies caused by an emitted beam which is not spatially steered with respect to the vehicle. Specifically, “false alarms” caused by highway guard rails and complex traffic situations are demonstrated. These false (or nuisance) alarms would hamper widespread deployment and acceptance of forward looking collision avoidance automotive radar systems simply because of confidence degradation by the vehicle operator. Therefore, the second phase of this research was conceived to explore the possibility of developing an inexpensive (but reliable) electronic means of achieving spatial beam scanning.

If such an effort could be achieved, numerous possibilities for reducing the false alarm rate become

practical. Beam steering, which follows the driver’s steering angle, immediately comes to mind. A more powerful technique entails continuous scanning. This would allow for target signature (pattern) recognition of impending hazards. This would invoke digital signal storage and processing methods which are already available.

After much examination of previous antenna beam steering techniques, it was realized that none were suitable for the automotive radar application and the frequencies of interest (10 Gigahertz through 100 Gigahertz). The departure from established techniques was necessitated in order to meet reproducibility, reliability, and cost requirements in the harsh ambient environment imposed by automotive vehicles. A practical solution was sought. To this end, it was necessary to conceptualize antennas in their most basic form. Electromagnetic radiators may consist of only two components: electrical conductors and electrical insulators (dielectrics). If it would be possible to vary the electrical conductivity of the structure at electronic speeds the consequence would be apparent geometry modification (AGM) of the structure with the attendant positioning of the emitted beam. A computer simulation has been performed verifying this theory is correct. Crucial to the success of such a technique is obviously the availability of a suitable material. More generally, a universal device which could be applied to most

any structure would be highly desirable. The control (modulating) signal should be supplied to the variable conductivity material without resorting to an electrical connection. This “action at a distance” would allow for total decoupling (isolation) of the radio frequency (RF) components from the controlling input. A photoconductive compound became the obvious choice for material. In this research, Cadmium Selenide (CdSe) was used with total success at a frequency of 10.5 Gigahertz. Use of this material at the frequencies of interest was heretofore unknown. Even at lower frequencies, substrate capacitive reactance masks photoconductivity variations. Success is achieved through planar (monolithic) minimal capacitance topology with inductive tuning. By careful application of lumped element (LE) techniques, it is entirely feasible to achieve wideband (Q less than 20) electrical resonance. Conductivity modulation of the device at and around resonance mimics direct current performance. The planar configuration of the fabricated microwave photoconductive element yields a device which is not multimodal. A clearly defined (and repeatable) single resonance is exhibited. These devices can also be used to provide variable attenuation along a microstrip signal path. By judicious choice of electrical circuit configuration it is possible to achieve amplitude variation to phase variation conversion. Thus, the microwave photoconductive elements

may be applied to existing conformal phased arrays obviating the need for new structure development. Embedded in phase three of this research is a demonstration of the application of the microwave photoconductive device. Beam steering is accomplished for a rudimentary monopole bihorn radiating structure. Both microstrip fed horn patches are operated against a ground plane. In spite of relatively low forward gain (17.0 decibels), up to 10.0 degrees of beam steering may easily be observed. Only two microwave photoconductive devices are used to feed the structure. The steering command signal is supplied to the chips via Gallium Aluminum Arsenide (GaAlAs) light emitting diodes (LEDs). The spectral peaks of the illumination supplied by these optical emitters (660 nanometers) is extremely close to the peak of the Cadmium Selenide optical spectrum (690 nanometers).

2.1 Research Vehicles’ Description

The Electrical Engineering Department at The Pennsylvania State University has been fortunate to have on loan from the U.S. Department of Transportation, National Highway Traffic Safety Administration, two vehicles incorporating collision avoidance radar braking systems. These vehicles were made available for independent engineering analysis from 1980 through 1985. During this time, many hundreds of hours of over-the-road driving data and impressions were accumulated. Both vehicles were modified by the Bendix Corporation Research Laboratories in Southfield, Michigan, 48076, at a total contract cost of $400,000. The vehicles represent a full-size (1977 Plymouth Fury) sedan and a mid-size (1977 Ford Granada) sedan. Although both vehicles incorporate antiskid braking subsystems, research described herein concerns itself only with the forward looking radar system.

The radar system consists of three major subassemblies: the antenna and associated RF microwave components (the front end), the radar (video) electronics module, and the signal processor console (computer, control, and display).

The radar sensor provides estimates of closure rate and range to a target. These estimates are supplied as inputs to an 8 bit microprocessor-based digital computer. The computer continuously processes these signals along with other inputs representing signal presence, vehicle velocity, and various inhibit functions (driver input commands, such as: steering angle, driver braking, and mode of system operation). In the subsections which follow, the basic warning decision technique will be discussed first, followed by a description of the radar sensor which supplies critical information to the digital processor.

2.1.1 Warning Decision

Numerous algorithms may be formulated to determine when to alert the driver of an impending mishap. However, consider the physics of a typical situation: Kinetic energy (K.E.) increases as the square of the velocity (v) of a vehicle with respect to a reference point (stationary or moving). That is:

where m = vehicle mass. The stopping distance (x) may be defined as that distance which is required to bring the vehicle in close proximity to the target (without actually contacting the target) with a relative velocity of zero at the time of osculation. If one assumes a constant

decelerating force (F) on the vehicle, we note:

but,

where a is a constant deceleration. Therefore:

or

or

where k is a constant based upon deceleration; or

where is the first derivative of x with respect to time.

The implication of this analysis results in a phase plane boundary. Refer to Appendix I. If the range (x) is plotted versus range rate () assuming a constant deceleration (here 0.7 G) we note that a parabolic curve is developed. If 0.7 G is the maximum attainable deceleration, it is recognized that operation of the vehicle to the right of the phase plane boundary will result in an impact. Thus,

to the left of the phase plane boundary are drawn two piecewise linear boundaries: The one on the extreme left represents a warning limit. If the operator crosses this boundary (from left to right), the system renders a visual (light) and aural (beeper) warning. If the second boundary is crossed (left to right), the radar system sends a command to the antiskid automatic braking system. The boundaries are stored as permanent memory in the computer. Where the vehicle is operating with respect to these boundaries is determined by digital processing.

2.1.2 Radar Sensor Description

⚠️ [The original prints this heading as “2.1.1”; renumbered to 2.1.2 to match the Table of Contents.]

The sensor is comprised of a radome, antenna, doppler transceiver, and a separate electronics module. The radome provides protection from the elements. The antenna is an eight inch parabolic dish mounted in a central opening of the vehicle’s grille. Frequency of operation is at 36.0 Gigahertz. The antenna achieves horizontal and vertical half power beamwidths of about 3.1 degrees. Although circular polarization would undoubtedly be a better choice,¹ linear polarization is employed with a 45 degree cant to the horizontal. This provides approximately 20 decibels of suppression of “blinding” from a similarly equipped vehicle. The high gain antenna with low sidelobe levels also aids in

significantly reducing the probability of car-to-car blinding.

The doppler transceiver develops a peak pulse power of approximately 25.0 milliwatts at 36.0 Gigahertz. It is of the Gunn effect variety (Gallium Arsenide) and provides self-detection (homodyning) of the reflected echo signals.

See Appendix II for a block diagram of the radar sensor. The electronics module provides the transceiver with the necessary pulse modulation. ON-OFF pulse modulation results in a positive range cutoff feature. Modulation is accomplished as follows:

1. RF output at 36 GHz + 410 kHz for 730 ns.
2. No RF output for 1420 ns.
3. RF output at 36 GHz − 410 kHz for 730 ns.
4. No RF output for 1420 ns.
5. Cycle repeats.

Within the transceiver, reflected echoes are combined with samples of the transmitted RF energy in the circulator. Homodyne detection occurs in the mixer. The resultant doppler difference signal (video) is amplified in the preamplifier before further processing. The video output, of course, is pulsed and similar to the RF output. At this point, however, it is appropriate to regard these high frequency detected pulses as a “carrier” for the lower frequency doppler modulating component. Only the last 215 nanoseconds of these pulses is gated alternately into the

two amplifier/limiter chains. The sampled data after filtering and further amplification becomes a reconstruction of continuous doppler target information in each of the two channels. Range rate information is recovered as a frequency (doppler) to voltage conversion from one channel only (the range rate detector). Electrical phase comparison between the signals from each channel provides range information (the range detector). The sense of velocity, approach or recede, is accomplished by noting the electrical phase (lead or lag) of the signals between the channels. The threshold channel rejects targets which are too weak (amplitude), or at too low relative velocity. Range rate and range outputs from radar sensor are of an analog (direct current voltage) nature. See Appendix III for typical range calibration data. Ideally, a linear correlation between the data and voltage is desirable. This, in practice of course, is seldom seen. Displacement error () is therefore defined [67]:

where - = actual output voltage - = voltage which would occur for a linear relationship - = full scale (upper limit) voltage.

For this definition, it must be realized that the ideal linear relationship (a line) intersects the end points of

the actual curve. Therefore, represents the largest vertical departure of the actual curve from the ideal relationship. The radar sensor outputs exhibit typical displacement errors on the order of 7.5%. Actual vehicle speed information is derived from magnetic sensor/toothed wheel combinations placed on the front wheels and rear axle differentials of the vehicles. Displacement error is slightly better for the direct measurement devices than for the radar sensor.

2.1.3 Radar Sensor Traits

Since the transmitter is chirping at two frequencies separated by 820 kilohertz, range determination becomes ambiguous for targets beyond 300 feet. However, restricting the transmitted pulses to 730 nanoseconds duration eliminates responses from targets at ranges beyond 250 feet. This has been confirmed theoretically and experimentally [25].

This radar system possesses no inherent range resolution of separate targets. Because of the limiters in each channel, outputs of the sensor represent the target producing the strongest return signal. In principle, therefore, it would be possible for a large target far away to mask a smaller target which is much closer to the vehicle. This close-in invisible target problem could be

alleviated by spatially steering the emitted beam of RF energy.

2.2 Performance Data Documentation

Numerous drivers were enlisted to evaluate the performance of the experimental vehicles. However, well-meant but ambiguous qualitative reports pointed up the need for a data documentation technique. As initially conceived, the panoramic view of the driving scenario was considered to be of highest importance. With the current availability of reliable, compact video cameras and recorders (VTR) documenting the performance of an automotive collision avoidance radar system reduces to eliminating mechanical vibrations imparted to the TV camera under normal operating conditions and developing a convenient means of simultaneously recording numerical data which represents input and output signals to and from the radar control computer. The first problem is easily solved by a mechanical low pass filter which mounts the video camera inside the vehicle. This “filter” is nothing more than a flat board cushioned with polyurethane foam and secured with compliant rubber straps. The second problem is solved with the aid of a telemetry system which simultaneously records numerical data on the audio channel of the video system as the panaromic view is being transcribed on the video channel. Thus, at the time of playback with a videotape

player (VTP), audio is routed to a telemetry receiver for decoding and display of said data. Refer to Appendix IV for a block diagram of the data documentation technique. At the time of data analysis, actual over-the-road driving situations are replayed as a television display along with the following decoded parameters: vehicle speed, range, closure rate, and warning (light and beeper). With the aid of a studio switcher and a television camera (trained on the data display), a split screen display is generated. On the top of the screen is displayed system operating parameters. Immediately below this display is the panoramic view of the driving situation. These composite driving sequences are then edited for incorporation into the final production: “Automotive Radar, A Technical Review and Analysis,” The Pennsylvania State University, Quad Master #11-1007.

2.2.1 Telemetry System Considerations

In recording data electronically, two decisions are necessary: First, should the format be digital or analog? Second, should the modulation be linear (amplitude) or non-linear (angle)? Angle modulation may be accomplished by frequency or phase variations [10, 102, 107]. In terms of the required resolution (e.g., Speed − 1 km/hr), each channel would require a minimum capacity of 7 bits [57]. Based upon the fact that the data to be recorded is already in analog format and the available audio frequency (AF)

voice channel is inherently designed to handle continuously varying (amplitude) information with low distortion. quantification required for a digital system offers no advantage [57]. Furthermore, bandwidth limitations on the AF channel must be taken into account with regards to the maximum sampling frequency. When system complexity and cost effectiveness are added as primary considerations, analog format is the preferred technique of information transmission. Angle modulation is inherently wideband. While phase locked loops (PLL) may be used in signal demodulation, aliasing or harmonic locking can be a serious problem [10, 41, 107]. Improper “locking” also rules out the possibility of a fail safe system. Our format/modulation choice is therefore Analog/AM. This is a narrowband, fail safe technique. See Chart 1.

Chart 1 — Information Transmission Methods

Method Selected: Analog/AM

To minimize cross talk between telemetry channels, the following system design parameters have been accounted for:

1. The signalling frequencies are not harmonically related. (See Appendix V for telemetry output signals and interface specifications.)
2. The individual channel signals must be low in distortion content.
3. The receiver filters must be very selective frequency wise (high Q with sharp skirts).

In the telemetry system, better than 39 decibels (dB) isolation between channels has been achieved. High Q filters require excellent long-term stability. Because of stability considerations, active filters must be ruled out in the receiver [30]. Digital (switched capacitor) filters would be acceptable within the receiver except for complexity [31]. A novel filtering scheme for decoding the telemetry revolves around an electromechanical filter manufactured by Erie muRata Manufacturing Co. [65]. These filters are miniature tuning forks with piezoelectric transducers attached to the tines. Narrow, reproducible bandpass characteristics are realized with these filters. The high stability and low cost of these so called “Microforks” made them an obvious choice for the telemetry receiver design. The only precaution to observe in the Microfork application has to do with their susceptibility to mechanical vibration. This deficiency was sidestepped by acoustically isolating the circuit boards on which the filters are attached from the chassis with vinyl grommets.

The Microforks proved to be excellent filters for our application. The telemetry system operation is straightforward. The next three subsections describe telemetry system components: the transmitter, the receiver, and the calibrator.

2.2.2 Completed Television Product

The telemetry system described in previous subsections indeed proved to be the key to successful data documentation. Many hours of video tape footage were gathered under actual vehicle operation. Even more time was expended in reviewing the recorded information. As one might expect, much duplication of events was noted. It was possible, however, to extract situations which represented both normal and anomalous events. Nearly all of the documented happenings were spontaneous and unanticipated. A few staged situations were recorded to demonstrate certain system characteristics which might otherwise go unnoticed. Eleven individual segments were selected for incorporation into the final television production. This show (18

minutes) features Dr. Dale M. Grimes as interviewed by Mr. Frank Wilson. “Automotive Radar, A Technical Review and Analysis” is aimed at a general audience. The interview itself serves as a tutorial for the uninitiated. The show is also of interest to technologists since it highlights the problems requiring resolution in quantitative terms. Thus, it is possible to describe the 11 segments and their intended purposes:

Clip 1. This depicts unremarkable rural two-lane driving with oncoming traffic. The road curves to the right. Guard rails abound on-coming traffic. The radar system performs in a normal fashion with no warning provided. Range and closure rates are noted to vary continuously with captured targets.

Clip 2. The research vehicle approaches the rear end of a tractor trailer truck at a moderately high speed (70.0 kilometers/hour). The truck has pulled out into the lane in which the radar equipped car is travelling. The target is acquired at a range of 45.0 meters. A warning is rendered at a closure rate of 9.0 meters/second. A second warning is issued as the driver passes in the left lane. The second warning is a consequence of the high rate of closure on the median guard rail.

Clip 3. A five-lane, in-town road is depicted (two lanes going, two lanes coming, and one center turn lane). The research vehicle is travelling on the outermost lane (adjacent to turn lane). A second vehicle exiting from a bank crosses out over the first lane into the second lane and is now travelling in the same direction as the research vehicle. Of course, the second vehicle is at a much lower speed and blocks the research vehicle’s path. The research vehicle is travelling at 55.0 kilometers/hour. The target is acquired at a range of 18.0 meters. A warning is rendered at a closure rate of 6.0 meters/second.

Clip 4. This segment demonstrates changing lanes at highway speeds (90.0 kilometers/hour). The research vehicle changes from the left lane to the right lane on a four-lane highway (with center divider). A target in the right lane is acquired. The range is 35.0 meters. With a 0.0 meters/second closure rate, no warning is elicited.

Clip 5. The research vehicle is travelling at a moderately high speed (70.0 kilometers/hour). The driver is negotiating a turn in the highway. The radar acquires a metal guard rail as a valid target

(erroneously) at a range of 30.0 meters. The angled closure rate is noted to be 15.0 meters/second. The radar system renders a “nuisance” warning. This clip demonstrates the most important problem hampering large-scale deployment of automotive radar as a consumer product: “the false alarm.”

Clip 6. Around town slow-speed driving is demonstrated. Pulling up too quickly (delayed braking) on the rear of a stopped vehicle renders a valid warning response from the radar system. The research vehicle is initially travelling at 20.0 kilometers/hour. At a range of 11.0 meters, the target is noted. The warning alarm is provided at a closure rate of 3.0 meters/second. This clip also demonstrates how the radar system is calibrated such that the zero point for range is slightly in front of the radar antenna. That is, zero range actually occurs approximately 1.5 meters in front of the radar antenna. Of course, any “cushion” is possible. As the vehicle ahead pulls away, a dip in range is observed back to zero. Beyond the 1.5 meter cushion, normal range calculations are resumed.

Clip 7. The research vehicle is deliberately being driven at an abnormally high speed through a parking lot. Numerous alarms are rendered because the radar system is being operated above the low speed cut off. Vehicle speed is observed to vary between 10.0 and 20.0 kilometers/hour. Range varies between 5.0 and 10.0 meters. Typical closure rates are approximately 3.0 meters/second.

Clip 8. A bicyclist is stopped at a four-way intersection. The research vehicle cautiously approaches from the rear at 5.0 kilometers/hour. The final range with the research vehicle stopped is observed to be 2.0 to 3.0 meters. The closure rate is too low to be observed. No warning is rendered. The range dramatically increases as the bicyclist pulls away. The radar clearly has captured the target. The research vehicle then turns into the intersection. The range fluctuates as a building is intercepted by the radar beam.

Clip 9. The research vehicle is stopped at an intersection with the radar operating in a normal fashion. As pedestrians, cars, bicycles, and trucks pass in front of the radar beam system response is noted. The range fluctuates as target interception takes

place. The narrowness of the emitted beam may be clearly observed (~3.1 degrees).

Clip 10. The research vehicle is travelling through a parking lot at a nominal speed of 10.0 kilometers/hour. A lady pushing a baby carriage crosses the path of travel. The range is noted to be 15.0 meters. A warning is rendered by the radar system. The event is of short duration with low (unregistered) closure rate.

Clip 11. This final segment is referred to as a basic complex traffic situation. The vehicle is approaching a red light in the right lane. The road curves to the right before the intersection. Vehicles are stopped in the left lane awaiting a left turn signal. The research vehicle’s initial speed is observed to be 25.0 kilometers/hour. The radar beam intercepts the stopped vehicles. At a range of 15.0 meters target capture occurs. A false alarm warning is rendered at a closure rate of 4.5 meters/second. The research driver is negotiating a normal turn and is braking in an acceptable fashion. Yet, the radar system is “fooled” by the tangential velocity component reflected from the parked vehicles.

In the 11 clips just outlined, an array of most important driving encounters is represented. Other situations could, of course, be envisioned. However, after examining many other encounters, similarities were noted. The 11 clips are by no means all inclusive. They do represent a fair depiction of many like situations. The most important result is contained in video tape footage depicting the false alarm phenomenon. A consumer product must be believable. False alarms reduce operator confidence. The system must not distract the operator unnecessarily. Therefore, system integrity enhancement must embrace a solution to the false alarm. The solution to the problem resides in processing and/or microwave beam steering. The highest dividends reside in the latter area. The fact that this statement is correct resides in the observation that beam steering produces more information to be processed. In other words, the most advanced algorithm renders little performance improvement with sparse system inputs. Therefore, the central goal of the presented research focuses on an inexpensive means of microwave beam steering.

3.1 Beam Steering Techniques

The methods by which an emitted radio frequency (RF) beam may be steered are shown in Chart 3. To select a suitable technique for automotive radar, one must keep in mind cost and reliability. Both concerns are critical. First, the antenna must not drive the cost of the entire system. Second, the antenna is most prominent in determining overall system performance. The antenna characteristics must be reproducible and predictable. No other system component can compensate for an erratic antenna. Thus, a mechanically steered antenna is ruled out of consideration immediately. It would be costly because of tight mechanical tolerance requirements. It would be physically large and inherently unreliable in the harsh automotive environment. Interest is therefore focused only on non-mechanical techniques. A discussion of each non-mechanical method is provided:

Phased arrays have found widespread application for long-range radar systems and commercial communications [16, 44, 48, 49, 52, 84, 92]. The technique employs many individual radiators where the phase/amplitude of the

electrical signal driving each radiator is altered with relation to other radiators. This allows for beam steering at electronic speeds. From a pure geometric analysis of the technique, the effects are easily understood. The obvious drawbacks of the system embrace cost and complexity. Although the system of radiating elements may be fabricated without difficulty, the problem of realizing a suitable phase/amplitude controller presents a major concern. At high power levels ferrite phase shifters [37] are commonly exploited to achieve the required excitation alterations. These devices are quite large and expensive. They certainly do not lend themselves to monolithic fabrication techniques due to the required large magnetic bias. For low power levels, PIN diodes have been used as signal controllers with good success [4, 37]. The PIN diode suffers from two drawbacks: Junction capacitance limits the useful upper frequency limit and the control signal (bias) is electrically connected to the device itself. The latter deficiency presents a serious concern relative to isolating the bias from the RF signal path.

Synthetic prisms have been used for electromagnetic beam steering. A fairly well known method [61] relies on an intricate system of waveguides with apertures so arranged that when the frequency of the excitation is shifted the phase relationship of waves emitted from each aperture is altered. Thus, beam steering is achieved from frequency to

phase (time lag) conversion. This system must immediately be discarded for automotive radar because of its inherent wide band nature. A more modern approach to the synthetic prism is affected from shining monochromatic radiation through a crystal at a non-orthogonal angle. It is well known that the square root of the ratio of dielectric constants of different materials is related to the linear ratio of refractive indices [80]. Thus, beam steering may be achieved by modulating either the frequency [17] of excitation or the dielectric constant of one material (a ferroelectric) [56]. The latter suggestion is the focus of much commercial interest and most information about the subject is of a proprietary nature [15]. Modulation of the dielectric constant of a ferroelectric material may be accomplished by controlling an applied DC electrostatic field along the crystal. The technique is currently hampered by several pitfalls: a) the static field required is quite high (10 kilovolts/cm), b) polycrystalline forms of the material are very lossy at microwave frequencies due to scattering and absorption, c) single crystals are high in cost, and d) the crystal’s environment causes unwanted secondary steering.

Anisotropy modification in various forms may be utilized to achieve electromagnetic beam steering. The possibility of constructing a single radiator (not an array) with the capability of beam steering is indeed an exciting

endeavor. At least one team of investigators [71] has successfully integrated ferrite materials into a biconical radiator. Beam steering is accomplished by varying static magnetic fields. The technique is satisfactory up through L band. For X band and beyond, high static magnetic fields are not only difficult to confine, but also present geometric interference with the electromagnetic radiator itself. This particular area, however, suggests strong promise and may ultimately be attractive up through the millimeter wavelengths. Anisotropy modification may also be accomplished with a plasma [26]. For many years, synthetic dielectrics (dielectrics supporting embedded conductors) have come into being. A plasma may be viewed as a synthetic dielectric in this respect. It should be obvious then that altering the ion population (ensemble or local) could achieve electromagnetic wave steering. For automotive radar it is difficult, however, to envisage this technique in a practical light. Maintenance of the plasma at specified ionization levels for extended periods of time would be most difficult, indeed.

Rectilinear beam positioning to angular conversion is a well established electromagnetic beam steering technique used in ground aviation applications for height finder radars. The basis of operation may be explained with the aid of geometry [92, 95]. Parabolic reflectors are normally illuminated at the focus and emitted rays from the reflector

are parallel. However, if a parabolic torus is illuminated off focus but normal to the directrix, the reflected ray passes through the focus at some angle with respect to the directrix. If the illumination is moved towards or away from the focus, while maintaining a right angle with the directrix, the new ray still passes through the focus but achieves a different angle with respect to the directrix. Clearly, then, to steer the beam requires nothing more than separate radiators with excitation which may be switched on or off. Suitable switches in low power applications make use of solid state diodes [37]. At extremely high frequencies, low junction capacitive reactance must be taken into account. Switching times between 20 and 600 nanoseconds are not uncommon. Variability amongst diodes presents a significant manufacturing (selection and matching) problem.

Apparent Geometry Modification (AGM) was ultimately considered as the preferred method to achieve electromagnetic beam steering for automotive radar. Sparse information on the subject area is available. Yet, recent developments in materials are about to cause considerable interest in this previously unnoticed area. Although research presented herein concerns itself with photoconductive compounds, other investigators (Japan) are presently studying Apparent Geometry Modification via “conductive plastics.” Certain conductive plastics exhibit

variable local conductivity characteristics with respect to applied mechanical stress. It is understandable that nearly all information on the fabrication and application of these materials is proprietary [1]. The concept of Apparent Geometry Modification is profound but easily understood. The technique is described separately in the next section.

Chart 3 — Steering Techniques

1. Mechanical
2. Phased Arrays — (Ferrite Phase Shifters Commonly Employed)
3. Synthetic Prism —
   • 1. Frequency Variation
   • 2. Dielectric Constant Variation
4. Anisotropy Modification
5. Rectilinear Beam Positioning to Angular Conversion
6. Apparent Geometry Modification

3.2 Apparent Geometry Modification

Apparent Geometry Modification (AGM) is a new and relatively unexplored method for accomplishing electromagnetic beam steering. While geometry modification of antennas has been reported [7], the alterations were of a mechanical nature. Recognizing that a radiating structure consists of electrical conductors, dielectrics, or both, it would appear reasonable that beam position and shape could

be modified by varying conductivity. This may be regarded as changing the conductive material’s geometry. Suitable materials for such application would capitalize on a bulk effect. Photoconductive materials such as Cadmium Sulfide (CdS), Cadmium Selenide (CdSe), Zinc Sulfide (ZnS), and Zinc Selenide (ZnSe) are in this category [21, 56]. Photoconductors have been employed before at microwave frequencies. Sun, Cheng, and Walsh [93] have reported the use of microwaves to bias a photoconductor consisting of mercury doped germanium. No one to date has considered using a photoconductive material as part of an antenna system. This should not be surprising for two reasons: a) For transmitting, resistance introduced into the structure destroys efficiency [44]. b) For receiving, resistance introduced into the structure produces noise [51]. However, in automotive radar applications, efficiencies as low as 10 or 20% mean dissipating powers on the order of only 40 milliwatts. Also, the return signals to the system are so strong that homodyne detection is all that is required. The signal to noise ratio remains good by virtue of the strong return signal. Apparent Geometry Modification should not be regarded as a panacea for beam positioning. A low power situation probably represents the only suitable application of a material which can vary conductivity in a continuous manner. It should be recognized, however, that photoconductors do offer some unique features. Unlike

semiconductor junctions which are small and isolated, a photoconductor can occupy a significant portion of the antenna’s area. Continuous beam steering along with continuous beam shaping is therefore possible. Also, light which would be the controlling signal can easily be confined, focused, and accurately distributed.

3.3 Apparent Geometry Modification (AGM) Simulation

Before actually constructing an antenna which incorporates a photoconductive compound in the structure, it is reasonable to perform a simulation in order to predict the radiator’s performance. On file at The Pennsylvania State University is an Antenna Modelling Program (AMP) [3]. Any surface may be modelled with a finite element grid. This finite element model (FEM) is entirely satisfactory if the grid cells are kept small [68, 92]. Recommendations for segment length [1, 68, 92] place an upper limit of approximately 0.1 wavelength. For grids selected to be analyzed, segments were added and deleted while maintaining a maximum limit on cell size. No alterations were observed in radiated patterns. This technique confirms that the grid used in the investigation is satisfactory and the recommendations regarding segment length are correct. The AMP has a maximum computational limit of 500 segments. The structure examined has 466 segments. In selecting an antenna for analysis, a practical but elementary structure

is desirable. The selected geometry represents a wire fed horn. This is an excellent candidate since in actual practice, microstrip [36, 38, 103] would be used to feed a conformal array [70].

A 3-to-4 relative geometric ratio for the grid cell was established. See Appendix IX. This was necessary in order to secure integral grid connection points. Data is available on wire fed pyramidal horns that use equilateral triangles on all sides [72]. The departure for the model from the equilateral case is not severe (67.38 degrees, 56.31 degrees, 56.31 degrees). The triangle is first developed in the XY plane. The leading edge is 2.0 wavelengths long. The height is 1.5 wavelengths long. In order to establish a perfect imaginary pyramid (with the image reflected in the XY ground plane), it is necessary to rotate the structure about the Y axis by 41.81 degrees:

AMP has this rotational capability. Single point voltage feed is utilized. Ultimately, experiments will be performed at X band. The computer simulation was carried out at 300 Megahertz for the reason that one meter equals one wavelength at this frequency. Thus, all geometric tabulations were easily read in terms of wavelength directly. Since perfect conductivity for the grid and ground plane is specified, no inconsistency in scaling

results [92]. The developed structure is 100% efficient with a maximum gain of 13.33 dB at an azimuthal (phi) angle equal to 180 degrees. As expected, at this angle the wave is linearly polarized in the vertical direction. See Appendix X. In the next simulation, segments were loaded with resistance simulating the photoconductive material. A slot is “cut” in the structure where the material will be inserted. To determine a suitable place to make the cut, the following reasoning is used: From the channel (cut) to the feed point, a smaller horn is established. The difference in forward gains between the total horn and this smaller horn will be called “differential gain.” If the cut is made close to the feed, the slot controls large differential gain but has little effect on the off-center steering angle. On the other hand, if the cut is made close to the leading edge, a wide steering angle is possible with little differential gain control. The mathematical product of geometric steering angle and differential gain is formed. The first derivative of this product with respect to geometric steering angle is calculated. This derivative is set equal to zero. Thus, the optimum geometric steering angle (without regard to current distribution) is found to be 55.81 degrees off center. In terms of the flat surface forming the upper side of the horn, the cut should be made at −0.9331 wavelengths away from and running parallel to the Y axis. The actual loading segments occur between −0.975

and −0.9 wavelengths. See Appendix IX. The affected segments are labelled “J through X.” In series with segment “J” is inserted a 288 kilohm resistance. Segments “K through S” are series loaded with individual resistance values of 240 kilohms. Segments “T through W” have individual series resistance values of 750 ohms inserted. The final segment “X” is loaded with a 900 ohm series element. All other geometric and electrical parameters remain unchanged from the previous case. All resistances selected for this simulation represent typical values which can be achieved in a most reliable manner with Cadmium Selenide material. The radiation pattern for this loaded horn is displayed in Appendix XI. A slight loss in efficiency (91.36%) was noted in the simulation. The beam shape has been radically altered with the appearance of a back lobe. The important observation to be made, however, is that the maximum gain (12.12 dB) is attained off center (185 degrees). Of course, the steering effect would be more pronounced with an array. An important area of research would center on the development of an optimization of individual radiator size versus array size to achieve the desired steering, gain, beamwidth, and sidelobe levels. A conformal array could be developed which possibly would include a dielectric lens or reflector [18, 47, 49, 52, 53, 68, 70, 99]. Parasitic coupling is also a concern which must be accounted for in the inherent design of an array.

Classical analysis techniques coupled with data processing methods may be invoked for these considerations [27, 48, 49, 73, 98]. Travelling wave antennas [101] may be excellent candidates for Apparent Geometry Modification and the application of photoconductive elements.

4.1 Material Characteristics

Photoconductivity is a bulk material phenomenon which is characterized by reduced electrical resistivity across a semiconductor because of impinging optical radiation of the proper frequency. Electrical current may flow in either direction since no PN junction is involved. The effect has been known for many years [21]. To understand its origin it is convenient to examine a form of Ohm’s Law:

where - = current density - = conductivity - = applied electric field

The conductivity in a semiconductor may be represented as follows:

where - = electron concentration - = electron mobility - = hole concentration - = hole mobility - = particle (electron) charge

Conductivity may therefore be increased by elevating particle mobility or making more carriers available. In the case of photoconductivity, this latter mechanism is achieved through broken covalent bonds. It is convenient to refer to an energy diagram for the material. See Appendix XII. Three types of particle transitions may be secured from optical excitation: a) a photon may excite a donor electron with energy into the conduction band, b) a photon may excite a valence electron into an acceptor state with a final excited energy of , and c) electron-hole pairs may be created by high energy photons. The first two transitions are referred to as impurity excitation. The last transition is known as an intrinsic excitation.

Electron-hole pairs are produced at a rate of . The carriers will exhibit a statistically distributed lifetime of (mean). This lifetime is not unlike current injected carriers (as in a transistor). Steady state carrier density may now be defined:

This is a special case of the solution to the more general continuity equation for excited carriers. Thus, conductivity increases due to the increased number of mobile holes and electrons:

Commercial photoconductors exhibit much larger conductivity excursions (for a given optical energy deviation) beyond what this equation predicts. This is a consequence of “trapping” [21]. Trapping essentially hinders normal recombination. If one type of carrier is easily trapped, the other type remains in its conduction band. Recognizing that this phenomenon contributes significantly to conductivity variations, it is possible to formulate a new equation accounting for excess electrons:

where = concentration of excess electrons

Trapped carriers exhibit a mean lifetime which is much longer than . The effects of traps may be expressed in terms of an effective quantum yield Y [21]. This is a measure of the number of electrons passing through external connections to the device per photon absorbed in the sensitive region. If Y is less than unity, the photocurrents are known as primary. If Y is greater than unity, the photocurrents are referred to as secondary. For commercial Cadmium Selenide (CdSe) and Cadmium Sulfide (CdS) photoconductors, the quantum yield is between 100 and 10,000. This indicates the important role of trapping in such devices. During the experimental phases of this research Cadmium Selenide was selected as the most promising candidate for investigation. This choice was a consequence

of low ON (saturated) resistance, wide dynamic resistance range, low hysterisis, and rapid speed of response.

4.2 Device Configuration

Commercial photoconductive cells are normally fabricated in three layers: a) the substrate (Alumina), b) the photoconductive compound, and c) the ohmic contacts (Indium). The channel formed between the ohmic contacts in the plane of these contacts exposes the photoconductive compound to optical radiation. The channel may be fashioned in a straight line or as a meandering path. The latter configuration is achieved by interdigitating “fingers” of the ohmic contacts. It provides for long channel linear lengths and establishes a low device resistance. Low device resistance may be attained by long channels and/or narrow width channels. Upon examination of a commercial photoconductive cell its appearance is not unlike a lumped element (LE) microwave capacitor [2]. See Fig. 1 of

Appendix XIX. This observation prompted the following goal: to fabricate a “universal” microwave photoconductive device. If such a universal device could be realized, its application would not be limited to the Apparent Geometry Modification antenna technique. In fact, beam steering could be accomplished with existing commercial planar antenna arrays by incorporating microwave photoconductors in the structure. Small (0.15 inch diameter) commercial photoconductive cells were considered as candidates for application at X band (8.0 to 12.0 Gigahertz). Several difficulties arise, however. First, even for the small chip (0.15 inch diameter) residual capacitance across the channel is quite high (~ 1.0 picofarad). At 10.0 Gigahertz this corresponds to a reactance of only around 16 ohms. This low value would clearly mask resistance variations of the photoconductor. Notwithstanding this difficulty, one might consider resonating the device with a shunt inductive element. However, at X-band it is necessary to evaluate the true equivalent circuit. For the capacitor itself, the interdigitated fingers exhibit inductive reactance. Capacitance is then distributed between adjacent fingers. See Fig. 2 of Appendix XIX. It is noted that this “capacitor” is in fact a network of distributed inductance and capacitance. For convenience, refer to this combination as capacitor “.” In Fig. 3 of Appendix XIX, an inductive shunt is added in parallel with . This inductive element

passes directly through the substrate from the backside and makes electrical contact with both capacitor “plates.” Leads exiting the device exhibit inductance. The vertical components of the shunt through the substrate also exhibit inductance. The horizontal portion of the shunt exhibits inductive and capacitive characteristics. Capacitive coupling is achieved through the dielectric (substrate) to the interdigitated fingers of . Since the fingers of are interdigitated, crossed distributed capacitive coupling is achieved in addition to the progressive distributed capacitance. An accurate equivalent circuit model for this configuration is given in Fig. 4 of Appendix XIX. This arrangement exhibits a strong multimodal nature. Under swept frequency laboratory studies, numerous interlaced resonances and antiresonances were observed. Since the resonances were quite shallow (~ 2 dB in a 50 ohm transmission configuration), the three-dimensional device was abandoned from further study.

To avoid multiple modes in a microwave device, a symmetric, monolithic structure must be realized. Refer to Appendix XX. A cross section of this element reveals only three planar layers: the substrate (), the photoconductive compound (CdSe), and the top conducting sheet (Indium). The straight channel which communicates with two end holes provides for the lowest possible capacitance of all configurations. With the exception of

this channel, conducting Indium extends over the entire top surface. Thus, around the outer edges of the holes are formed two identical flat inductive loops which shunt the coplanar capacitor. With this geometry, four electrical components may be identified: two end inductors (L), a low inductance coplanar capacitor (C), and a resistor (R) which represents the photoconductor itself. All four components appear in a parallel electrical configuration. This is depicted as network “A” in Appendix XXI. The two end inductors electrically combine to form an equivalent inductance value of L/2. This is shown as network “B” in Appendix XXI. This system is a parallel RLC circuit with a clearly defined resonant frequency. If the device is operated at this point of resonance, the magnitudes of capacitive and inductive reactances (X) are equal. However, since these reactances are imaginary with opposite signs ( negative, positive), the total impedance of the device reverts to pure resistance as provided by the photoconductor. See model “C” in Appendix XXI. In essence, the inductive reactance of the end inductive loops has effectively cancelled the capacitive reactance of the coplanar capacitor.

Electrical connections are made perpendicular to the channel and directly on to the conducting Indium. For the experimental devices which were fabricated, these connections consisted of silver epoxy bonds to flat copper

conductors leading to center pins of type SMA connectors. This facilitated rapid testing and tuning. This technique would, of course, not be suitable for production units. Production units would incorporate integral microstrip connections with no bond.

The physical dimensions for the experimental units are as given in Appendix XX. A special test fixture was fashioned for testing device performance. The test fixture and its performance are described in a later section (4.4). The prediction of electrical parameters for the microwave photoconductive element are described in subsections 4.3.1 and 4.3.2.

4.3 Device Characterization

The fabricated devices behave as true lumped element RLC networks. This was confirmed by swept frequency analysis. For a device with dimensions as given in Appendix XX, a natural resonance around 16.85 Gigahertz is observed. For optical saturation of the photoconductor, a typical bandwidth (BW) of 555 Megahertz is noted. Experimental results are in good agreement with calculated expectations. The next two subsections demonstrate capacitance (4.3.1) and inductance (4.3.2) calculations.

4.4 Test Fixture Characteristics

Successful testing of the microwave photoconductive chips requires a standard test fixture. This is necessary to secure reproducible results at the frequencies of interest. In early experiments, type N bulkhead connectors were used in a back to back configuration. These connectors have quite large dimensions and therefore afford little connector to connector isolation. Shielding hoods were used to achieve isolation at the expense of introducing

undesirable modes to the fixture. A superior fixture was devised utilizing the smaller SMA bulkhead connectors.

The test fixture used for all measurements was ultimately incorporated into the demonstration antenna system. This fixture accommodates two microwave photoconductive chips simultaneously. The test fixture is constructed from two aluminum blocks which are each 0.186 inches thick. Each block surface (1.6875 in. x 0.575 in.) mounts two SMA bulkhead connectors (for a total of four connectors) with 4-40 hardware. The blocks are maintained apart by 0.25 inches using 0.19 inch diameter aluminum spacers over the same machine screws which mount the bulkhead connectors. Thus, a 0.25 inch channel is formed between the aluminum blocks. This dimension accepts the microwave photoconductive chips with 0.0135 inch clearance on each side. Two chips reside isolated in the fixture on centers appearing 1.125 inches apart.

No attempt was made to fabricate a test fixture which would maintain a constant impedance. Dispensing with a constant impedance requirement was necessary to achieve flexibility in chip tuning (see subsection 4.3.3). More important considerations deal with the maintenance of uniform amplitude transmittance over the frequencies of interest (conducting chip present) and high electrical isolation (conducting chip absent). These two parameters were tested for each side of the test fixture (NO DOT END

and DOT END) over the frequency range of 8.0 Gigahertz through 12.4 Gigahertz. See Appendix XXV. Two blank chips were fashioned with Indium covering one side of the substrate entirely. These chips bonded into the test fixture permitted measurement of amplitude transmission characteristics. The resistances caused by the silver epoxy bonds were noted to be 0.4 ohms and 1.2 ohms (NO DOT END and DOT END, respectively). Up to 11.0 Gigahertz, less than 2.0 decibel variation is noted in transmission for both sides of the fixture. Out to 12.4 Gigahertz maximum amplitude variation is approximately 3.0 decibels (total) for either side. Isolation (chip removed) for either fixture position is better than 50 decibels. For all measurements, the amplitude/frequency characteristic of the crystal detector was taken into account.

For the frequencies of interest, the chip and test fixture must be considered as a unitary system. Amplitude transmission characteristics may be adversely influenced by geometric alteration of the center line conductor (the chip) or the test fixture. For the sake of completeness, this phenomenon is also demonstrated in Appendix XXV. By replacing the blank chips with cylindrical conductors (0.0159 inch diameter), it is noted that the transmission characteristics for either position of the test fixture are greatly affected. Wide amplitude excursions (5.5 decibels) are noted for small frequency shifts (10.7 to 11.0

Gigahertz). The effect is even more pronounced (~ 8.0 decibels) at higher frequencies (11.0 to 12.3 Gigahertz).

The foregoing analysis is necessary for proper interpretation of tests performed on the microwave photoconductive chips. Without measurements of the test fixture alone, ambiguity would result. Thus, based upon the predictable performance of the test fixture, any observed resonance may only be attributed to the chip under test.

5.1 Bihorn Array

Development of the microwave photoconductive elements proved to be entirely successful. It is therefore logical to demonstrate the applicability of the devices in their initially intended function: electromagnetic beam steering. In order to accomplish such a demonstration, a radio frequency (RF) radiating structure must be established. Apparent Geometry Modification (AGM) of a single radiating element would be most impressive. AGM would, however, require the fabrication of an integrated electromagnetic radiator/photoconductor. Local facilities for such fabrication are unavailable. Yet, on-hand fabricated microwave photoconductive chips could be incorporated into an easily constructed array. This latter suggestion demonstrates wide versatility. Existing antenna array technology could be expanded to accept microwave photoconductive elements. Selection of a suitable radiating structure was based on two criteria: simplicity and reasonably high gain (directivity). Simplicity facilitates fabrication. On the other hand, high gain permits ease of observation with regard to beam steering. That is, the

primary lobe of a low directivity antenna is quite bulbous. Because of the low amplitude to angle gradient of such a radiator, main beam steering may be difficult to observe. This is especially true at the angle of highest gain (head on). Thus, high gain structures are desirable for demonstration purposes because of dramatic steering effects. No attempt was made to secure the gain requirements of automotive radar. This would require unnecessary expenditures in time and cost. A modest two element phased array ultimately was selected to be the preferred demonstration candidate. Referring back to section 3.2, it is noted that the described wire fed horn (2.0 wavelengths long leading edge) achieves a very respectable forward gain of 13.33 dB above an isotropic radiator. However, a beam steerable phased array requires individual radiators to be placed at odd multiples of half wavelengths. Therefore, if the leading edge of such a horn could be shortened slightly to 1.4 wavelengths (while maintaining the same aspect ratio), two identical radiators could be positioned 1.5 wavelengths apart, side by side in the same plane, establishing a Bihorn structure.

To assist comprehension of individual array elements, refer back to Appendix IX. Recall that the depicted horn is first developed in the XY plane. The leading edge is 2.0 wavelengths long. The distance from the leading edge to the feed point is 1.5 wavelengths. If the leading edge is

shortened to 1.4 wavelengths, the distance from the leading edge to the feed point becomes 1.05 wavelengths. Thus, the original aspect ratio is maintained. This shortened horn (as before in section 3.3) is rotated about the Y axis by 41.81 degrees:

A perfect imaginary pyramid is thus established with the image reflected in the XY ground plane. In fact, the open sides which exist between the edges of the radiator and its image are of the same dimensions as the radiator or its image. That is, the leading edge of the radiator taken with the leading edge of the image form a square mouth. The geometric similarity between these radiators and the Great Egyptian Pyramid of Gizeh prompted my colleagues to refer to the large horn (2.0 wavelengths leading edge) as the Pharaoh. The smaller horn (1.4 wavelengths leading edge) naturally was nicknamed Mini Pharaoh.

The radiation characteristics of the Mini Pharaoh were determined with the aid of the antenna modelling program (AMP). Refer to Appendix XXVI. The main lobe exhibits a forward gain of 12.64 dB above an isotropic radiator. Feed point impedance was noted to be 84.26 + j73.99 ohms.

5.2 Demonstration Equipment

The Bihorn radiating structure (copper) is fashioned on glass epoxy substrate of 0.060 inch thickness. The nominal relative dielectric constant of this material is 2.5. The Mini Pharaoh radiators are fed with a common phasing line. The phasing line is 0.086 inch wide microstrip. The effective surge impedance and effective relative dielectric constant for this line are obtained from tabulations [36]:

Refer to Appendix XXIX. The phasing line is provided excitation at two points spaced 0.3 inches apart. The microstrip feed lines are 0.174 inches wide and achieve surge impedance values of 50.0 ohms. All other RF system connections are accomplished with rigid 50.0 ohm coaxial lines. It may be seen that an X-band Gunn source⁴ feeds RF energy to a Wilkinson type power divider.⁵ The

commercial Gunn source combines varactor tuning facilities along with an integrated circuit voltage regulator (MC 1569) and series pass transistor (2N3997). In order to permit the source to accept square wave (1000.0 Hertz) pulse modulation, three modifications were performed: The integrated circuit regulator speed was increased by removal of a 0.1 microfarad noise suppression capacitor, the series pass transistor response was improved by removal of a 15.0 microfarad emitter capacitor, and additional heat sinking (aluminum) was provided for the GaAs Gunn element module. The subject microwave source is capable of supplying continuous (CW) output power of approximately 60.0 milliwatts at a frequency of 10.5 Gigahertz.

Two in-phase microwave signals are provided from the power divider output ports. These signals are applied to small ferrite isolators⁶ (circulators with matched 50.0 ohm terminations at port 3). These isolators prevent reverse power feed for high voltage standing wave ratios (VSWR). Thus, as feed forward power is varied by the microwave photoconductor, interaction between power divider output ports is avoided.

The microwave photoconductive elements reside between the outputs of the isolators and the 50.0 ohm microstrip

feed points which connect to the phasing line of the Bihorn array.

If the two in-phase excitations at the phasing line are of equal amplitude, a single virtual feed exists at the phasing line midpoint. Thus, the main beam is radiated straight ahead (180.0 degrees) as depicted in Appendix XXVIII. However, if the photoconductive elements are illuminated with unequal levels of light, the devices provide unequal amounts of attenuation. Thus, the two excitation signals appearing at the phasing line of the Bihorn array remain in-phase but are of unequal amplitude. By phasor addition it is seen that the single virtual feed point has been shifted from the midpoint of the phasing line. This shift establishes an electrical phase displacement of signals which arrive at individual radiators. Thus, beam steering is achieved in accordance with the equation specified for the array factor (AF).

Appendix XXX describes the electrical circuitry necessary to augment the Bihorn radiating structure as a complete microwave transmitter. The schematics indicate three regulated power supplies, current steering control for the light emitting diodes which excite the microwave photoconductors, and the square wave pulse modulator which supplies the Gunn microwave source. An avalanche diode (1N2999A) establishes regulated voltage for the tuning varactor within the Gunn microwave source. Frequency tuning

is accomplished via a ten turn 10.0 kilohm linear taper potentiometer (0.0 − 48.0 volts). Two integrated circuit voltage regulators (LM317) provide 15.0 volts D.C. power to the modulator and the light emitting diode (LED) circuitry. LED currents are steered via series/shunt resistance arms. Variation is affected by a 10 turn 1.0 kilohm linear taper potentiometer. When this control is physically centered, equal currents are established through the diodes. If the potentiometer setting is changed, one diode receives higher current while the other diode current is diminished. Full range current control for each diode spans 0.25 to 19.75 milliamperes as monitored by two individual analog meters. Provision is made for turning off the light emitting diodes completely by a DPDT switch. In the LED off position, the meters double as voltage monitors for the regulated power supplies (self diagnostic).

The modulator section of the microwave transmitter uses a CMOS integrated circuit timer (ICM7555) connected to achieve astable operation. Frequency of modulation is adjustable via a ten turn 100.0 kilohm linear taper rheostat. A square wave with a 50.0 percent duty cycle (approximately) may be achieved over the frequency range of 635 to 1555 Hertz. This frequency range was established so that the modulation frequency could be adjusted to the receiver’s tuned passband. (The receiver consists of a horn antenna with 17.0 dBi gain, a crystal rectifier, and a tuned

AC coupled vacuum tube voltmeter (VTVM). The rectifier is a detector exhibiting square law transfer characteristics. The VTVM (Hewlett Packard, Model 415B VSWR indicator) is tuned to a nominal center frequency of 1.0 kilohertz.) The modulator is completed by a pulse amplifier consisting of complementary symmetry output transistors (2N6049 and 2N3054A) fed by three small signal drive/predrive transistors (all 2N2222).

5.3 System Performance

In order to predict the beam steering capability of the demonstration antenna, measurements were first performed on the microwave photoconductors in a matched 50.0 ohm system. The relative range of attenuation secured for total darkness to optical saturation (> 150.0 foot candles) for either of the two devices was noted to be approximately 10.0 decibels. Maximum LED optical output is considerably below saturation levels for the photoconductors. The reference RF signal level (0.0 dB) was established at an LED current of 9.75 milliamperes. If the LED current is reduced to 0.25 milliamperes, the relative RF signal falls to −3.5 decibels. If the LED current is raised to 19.75 milliamperes, the relative RF signal rises to +1.0 decibel. The observed asymmetry is indicative of the extreme non-linear characteristics of the LED and photoconductor. Accompanying relative RF voltage levels for the three test conditions are

calculated:

6.1 Summary Remarks

At the onset of the presented research, the utility of an effective consumer oriented collision avoidance automotive radar system appeared promising but for the high false alarm rate. This has been adequately documented via a complete television production. The subject radar system incorporated no provision for steering the emitted electromagnetic beam with respect to the vehicle. The beam steering provision yields the highest gain in system performance enhancement. Yet, this area has heretofore received little to no attention. This is understandable since cost effectiveness must be considered for successful implementation. Thus, the high challenge level provided impetus for study area selection. The developed method of antenna beam steering has the potential for making the previously most expensive system component the least expensive. The breakthrough was made possible by merging existing and established technologies. Apparent Geometry Modification (AGM) and microwave photoconductors are, however, of limited application. Yet, within these applications it is indeed difficult to imagine any other

methods competing successfully on all salient considerations. Low power short range radar systems represent the most appropriate class of application for the microwave photoconductor. This research area proved most rewarding. Exhaustive literature searches⁷ proved that no previous investigator has considered using photoconductors as parts of an antenna system. Even more surprising is the fact that no other previous investigator has considered development of a universal microwave photoconductive element which could be applied as a light controlled microwave attenuator. Extensive literature searches by Dialog Information Services, Inc. revealed nothing comparable or even similar to the presented technological contribution. The microwave photoconductor should prove even more versatile in a host of related high technology applications. Signal processing at ever increasing frequencies should inspire continued research in this exciting area. Secondary features of photoconductive compounds may provide solutions to otherwise overwhelming requirements. For example, in terms of nuclear hardness, the junctionless photoconductor is comparable to a carbon resistor. At least one team of researchers [91] has studied the effects of 3-10 MeV accelerated electrons on CdS and CdSe. Radiation doses of

rads. have been reported to enhance material crystal structure with attendant improvements in electrical properties. Thus, based upon inherent nuclear hardness and insensitivity to electromagnetic pulse (EMP) environments, the microwave photoconductor could find application in a battlefield tactical radar system.

6.2 Recommendations for Further Research

This research effort has been by nature interdisciplinary. Consequently, limited time was expended on individual constituent areas. The decision not to dwell on any supporting area was made during early research phases. This decision was necessary because the ultimate project goal was to culminate in the construction of a realizable product. The developed beam steerable antenna obviously is not suitable for immediate automotive radar application. It is a feasibility demonstration model only. A practical automotive radar system would address the development of a totally integrated RF subsection [35]. Thus, the antenna, the photoconductors, the RF source, the detector, and controlling electronics would exist on a single substrate. Reliability and cost effectiveness dictate this approach. The presented research relied on commercially available subassemblies and devices which could be fabricated with resources on hand. Although good reliability was realized, cost effectiveness was given no

consideration. Areas meriting further research are not difficult to identify.

Apparent Geometry Modification (AGM) is an exciting area which needs much attention. So little has been accomplished in this area. A single radiator (not an array) which incorporates an electronically adjustable radiation pattern will offer many new application possibilities. Analysis and synthesis methods will have to complement material developments in order for this technique to realize full potential. AGM should be of interest to antenna designers from academic and practical considerations alike.

Microwave photoconductors are applicable to presently existing antenna array technology. Optimization of an array suitable for automotive radar applications would provide an immediate cost effective beam steerable antenna. For minimal commercial effort a high performance radiating system could be had. Dialogue between antenna fabricators and photoconductor manufacturers would have to be established, however.

Cadmium Selenide (CdSe) proved to provide good performance at X-band microwave frequencies. More research should be conducted on this material at even higher frequencies (up to and beyond 100 Gigahertz). Improved fabrication methods for this and other photoconductive compounds should be explored specifically for microwave applications. The geometry used for the presented research

is very close to optimum because of low device capacitance. However, a substrate with a lower relative dielectric constant could be used to good advantage at millimeter wavelengths. For example, quartz [11] with a relative dielectric constant of 4.0 would be a better substrate candidate at higher frequencies. Consider the fabricated devices presented in this research. By scaling down the physical dimensions, resonant frequencies corresponding to millimeter wavelengths may be achieved. Even higher frequencies may be realized with a quartz substrate.

The inductive end loops utilized in the fabricated devices provided a high degree of tuning flexibility. The loop configuration also establishes repeatable inductance values, even for rudimentary fabrication methods. However, straight line channel shunting inductors can provide lower values of total device inductance. Thus, higher resonant frequencies are possible with straight line inductors which are physically short (the channel width). The success of a device utilizing such inductors resides in adherence to narrow margins of manufacturing tolerances. Producing devices with straight line inductors may lead to wide variations in device resonant frequencies. In fact, the resonant frequency of the device could be used as an indicator of manufacturing tolerances. This obviously requires further study.

Finally, many advanced solid state device fabrication techniques are available. However, nearly all of these techniques are dedicated to transistor and integrated circuit production. The successful commercial manufacture of microwave photoconductors must consider the full gamut of existing production methods. Obviously, an assessment of volume requirements is dictated. Widespread application of collision avoidance automotive radar could warrant dedicated production facilities.

Total Device Capacitance Calculations

where and are complete elliptic integrals of the first kind

and farads/inch

and = length = 0.195 − 0.026 = 0.169 inches