3.6 Multi-beam Planar Antenna (Task 3.1)

A major factor in the deployment of an automotive side detection system (SDS) is vehicle styling. Vehicle stylists require the surface area of any sensor to be as small as possible, such that the sensor is ideally not visible on the outside of the vehicle.

The performance parameters defined at the start of the ACAS program required a zone of coverage extending from the side mirror to 10 meters behind the vehicle that is exactly one lane width wide. The system was also expected to have target discrimination capability such that objects stationary with respect to the ground (parked cars, guard rails, bushes, etc.) and objects traveling in the opposite direction were not reported as hazards.

Preliminary work completed prior to the ACAS program indicated that a three-beam antenna system was required. Initial design efforts showed that for a standard planar patch array, the feed structure was 70% of the surface area, and limited the side lobe performance at high squint angles. The antenna board was also one of the highest cost components in the sensor design.

This task was structured to address the surface area and feed radiation problems with a low-cost small-area antenna design. Specific goals and objectives are:

• Development of a 24.125 GHz antenna that has a detection pattern that "looks down the adjacent lane".
• Antenna design must support an adjacent lane target zone 8-10 meters long and 4.5 meters wide.
• Antenna side-lobe performance must be compatible with target discrimination requirements.
• Antenna must feature planar technology in order to be consistent with manufacturing cost and vehicle styling requirements.

3.6.1 Antenna Design Approach

The performance specifications were completed, and the key electrical parameters are summarized in Table 3.10. Each zone requires a separate transmit and receive antenna, and each transmit / receive pair must meet the same performance specification. The surface area was specified as 2.25 x 5.5 inches for the six-antenna array. The combined requirement for high squint angle and low side lobes is the primary technical challenge for this task.

Based on these specifications, a design approach was selected. Theoretical analysis on boresight performance was completed and a concept for construction of the feed circuitry, housing, and radome was determined. The antenna was fabricated on a multi-layered printed substrate material in a planar configuration. This approach is consistent with the vehicle styling objective. Test structures for evaluation of single antenna elements and feed transitions were designed, and work on developing test software to test and characterize these elements was initiated. The original vendor selected to fabricate the test structures backed out after one attempt. A replacement supplier was found and an initial set of test structures were fabricated and tested.

Work continued on developing the planar antenna simulations using a Finite Difference Time Domain (FDTD) analysis program while the first iteration boards were in fabrication. Additional simulations showed that via holes would be required to suppress unwanted stripline modes and a variety of via distributions were simulated and theoretically evaluated.

The first set of boards were received and tested. There were a number of problems discovered with the first set of boards that complicated obtaining accurate and reliable modeling data. First, the boards were slightly over etched in the area of the ground pads for the co-planar waveguide probes, resulting in a ground repeatability problem. Also, there was a problem with the microstrip to stripline transition design that caused a high VSWR. As such, it was not possible to accurately calibrate the measurement system in order to properly de-embed the radiating element impedance data. Finally, the initial circuit design was made before the FDTD analysis program was available, and the test circuits did not have adequate suppression of undesired parallel plate modes. Existence of these modes, as predicted by simulations during the fabrication cycle, prevented adequate calibration of the measurement system.

The microstrip to stripline transition was modified to correct the impedance match and parallel plate mode problems. Design of a partial array test circuit to model mutual aperture coupling coefficients was also completed. These circuit designs were submitted for fabrication.

During the second iteration fabrication cycle work continued on developing the planar antenna simulations. Multiple new simulation programs were reviewed in order to find a simulation program more compatible with the overall analysis problem than program originally selected. A simulation program, named Eminence™, was selected and purchased. Circuit simulation was started with the new program.

The second iteration mutual coupling boards were received in December 1995. These boards were severely over-etched and adequate tests could not be performed. The ground pad spacing was wider than the coplanar probe contacts, resulting in no connection on one or both ground pads. The supplier was notified and replacement boards were ordered. These boards arrived one week after the quarter ended, and were of good quality. The boards were tested, and data showed good correlation with simulation, and good repeatability.

The second iteration transition test circuits were received in January 1996. The quality of the boards was good, and the boards were assembled and tested. Simulation of the circuit, on Eminence™, showed a large unexpected shunt inductance that caused a large mismatch in the transition. Test data correlated very well with the simulation, and a new transition was designed to compensate for the mismatch.

The array architecture was modeled and analyzed. It was found that the addition of vias around the radiating slots, required to suppress parallel plate modes at the radiators, created a very high inductance at the slot. Analysis on the simulation program showed that, although the slot could be matched at a single frequency, the high Q associated with the large inductance made the match very narrow band. The bandwidth was determined to be far too narrow to meet design performance parameters over frequency and over nominal manufacturing tolerances. Additionally, it has been determined that multiple simulation and layout programs must be used in order to fully analyze the entire structure. This discovery forced a change in the physical architecture of the array. Multiple configurations were evaluated before an acceptable solution was found. A new array design approach was completed, and the layout for the array was started.

During the sixth quarter, work continued on designing the new array using a radiating patch structure instead of the originally proposed radiating slots. The patches are slot coupled and provide a broader band match to the feed structure. Simulation showed that the design would provide a broader band match, and would also provide more manufacturable tolerances on etched dimensions.

In order to meet the design parameters, the 30-degree antennas require 2 vertical patches and 6 horizontal patches. The -64 degree antennas require 2 vertical patches and 8 horizontal patches. In order to minimize the number of variables, reduce the sources of error, and extract data to verify design models, the antennas were first designed with only 1 vertical patch (1D). The 1D prototype arrays were designed, layouts were completed, and the board was fabricated.

Figure 3.38 shows a sample patch array configuration. The layout is for a 2D-test array that is a 2 vertical patch and 4 horizontal patch configuration. The feed structure is shown as viewed from the bottom side of the multi-layer board, with the topside patches shown as if the board material were transparent.

Tests on the first 1D-array boards fabricated resulted in good performance on the 30-degree squint design. The 50-degree squint performance was moderate, and the 64-degree design was poor.

The assembly was X-rayed and it was determined that some mis-alignment between layers had occurred during the fabrication process. An initial review of the simulation model indicated that this will degrade side lobe levels, however the degradation should not be as severe as the measured data. The supplier was contacted and agreed to fabricate a replacement array of this design. The second 1D array was fabricated to correct for the layer to layer misalignment that was discovered on the first unit. There were no design changes made.

The replacement array was received and tested. Test results for the 30-degree squint angle are summarized in Table 3.11. The design for -30 degrees is the mirror image of the +30 degrees, and the test results are similar.

This data indicates that the buried feed structure antenna gives improved performance relative to both the design target and the existing baseline antenna structure for the 30-degree squint angle. The buried feed structure reduces the required sensor height by approximately 40%, and significantly reduces the sensitivity of system side lobes to the location and spacing of the absorbing material in the radome, thereby opening production tolerances which will both lower cost and increase yields.

Two high squint angle designs were made, one having a target of 50 degrees and the other a target of 64 degrees. Both designs were fabricated in order to help verify the design models, which become much more critical when trying to obtain performance at the high squint angles.

The 50-degree squint data, Table 3.12, shows good performance with regard to gain, beamwidth, and squint angle. Side lobes were significantly higher than specification, and were higher than the 30-degree antenna. The 64-degree squint data, Table 3.13, is similar to the 50-degree data except that the beam width is slightly higher than specification and the gain is slightly lower than specification. The Baseline design did not have a high squint angle antenna due to the poor pattern resulting from top layer feed structure radiation.

A comparison of measured data to design simulations was made for all of the designs. In all cases, the gain, beam width, and squint angle were as predicted, and the side lobe levels were 10-15 dB higher than the mathematical model. The array patterns were improved, side lobes lowered, and VSWR improved on the replacement array. This 2nd array was also X-rayed and although the fabrication alignment was improved, the measured errors exceeded both design tolerances and quoted vendor process tolerances. The design was based on 2 mils maximum alignment error, and errors up to 5 mils were measured on the 2nd fabrication run.

Tests were run with absorber in the radome and with the radome removed to see the effects on side lobe performance. In both cases, the side lobes did not change significantly. At the close of the program, the side lobe issue had not been resolved.

3.6.2 Challenges

The primary task of designing a planar antenna with an electrical boresight squint angle of 64 degrees poses a significant design challenge. The primary problem with this design is maintaining the low side-lobe levels required for the application, and meeting the aperture size requirements dictated by vehicle styling issues. Based on a review of technical journals and discussions with suppliers, this design requirement exceeded all known previous art. Additionally, there was effort placed in trying to select a microwave board supplier that was capable of the high volume delivery requirements expected for automotive requirements.

Vendor Selection

One of the early problems encountered was the selection of the antenna board supplier. The first supplier, after initially indicating the capability to do the task, made one unsuccessful attempt and refused to bid any follow on business. The supplier underestimated the difficulty of the small sizes and tolerances and was not willing to work on his process for the small quantities being ordered at this time. Future production was too far in the future to interest him once he attempted the fabrication. Alternate suppliers needed to be identified, which required two months of calendar time, and resulted in a corresponding two-month delay in schedule.

Simulation Approach

Another early problem was in selecting an appropriate simulation software package. There are multiple microwave analysis programs available with differing degrees of capability. Although multiple programs were evaluated, a final determination of program capability can not be made without attempting to simulate a representative structure.

The original antenna architecture that was pursued was an array of radiating slots etched in a ground plane, and fed by stripline transmission circuitry. The simulation package required for this type of structure must support a multi-layer design, including plated through holes, radiating apertures, and stripline source excitation. The initial analysis package available was a "Method of Moments" package which did not support apertures in ground planes. A non-commercial simulation package under development that used Finite Difference Time Domain (FDTD) analysis and could handle arbitrary 3D structures, including radiating apertures, was available. Time was spent developing this tool for the 24 GHz antenna application, but it was discovered that the program was deficient in a number of ways. First, moderately sized problems (even single radiating elements) took many hours to simulate, and the results of any given simulation were questionable. For some problems, the results matched measured data well, for others there was a discrepancy. Second, the implementation of plated through holes appeared inadequate, and the accuracy of the effects of through-holes was doubtful. Third, the software lacked a user-friendly interface. Obtaining the S parameters from a simulation required significant external spreadsheet calculations, which further added to the overall cycle time and complexity. Commercial FDTD programs are available, but the high costs of these packages made making an immediate purchase without careful evaluation of the adequacy of these programs relative to this particular application less attractive than the decision to use the available non-commercial FDTD program.

Because of the problems encountered with the simulation programs available to us, and using the knowledge gained attempting to simulate the required antenna structure, an investigation into alternate existing commercial programs was conducted. After completing the investigation, a commercial program that utilizes a Finite Element Method (FEM) that can simulate arbitrary 3D structures, that can analyze radiating slots, and that can simulate stripline input excitation, was purchased. Additionally, the program had been on the market for a number of years and appeared to be a mature user-friendly program. The major drawback to the package was that it required a very large memory even for moderate sized structures, and again required a considerable length of time to execute a simulation. The program did give very good correlation between simulated and measured results for individual stripline fed radiating elements.

The array was simulated using the new FEM software. This iteration showed that the input impedance of the radiating slot was highly inductive due to the presence of plated through holes. These vias must be located very near the radiating slot in order to suppress the excitation of unwanted stripline modes. It was found that it was possible to tune out the inductive effect with a capacitive stub, but the resulting element impedance had a very narrow theoretical bandwidth and would not be adequate for an array design due to manufacturing tolerances.

When designing the final array, three different simulation programs were required. The radiating patches were simulated in one program that was optimized for microstrip design. A second simulation program was used for the stripline junctions in the feed structure. The resulting S-parameter matrices were imported into a third simulation package that was optimized for strip line design.

In summary, it is important to define the proposed architecture well enough prior to selection of a simulation program to ensure that the program is capable of accurately analyzing the proposed structure and that the time and equipment required for the analysis is allotted. When evaluating a new simulation program, simple representative elements of the design should be fabricated and test data available such that these test structures can be analyzed by the software under evaluation and the simulation results can be compared to measured data. Doing this gives the evaluator a method for quantitative evaluation of the simulation program's accuracy, speed, user interface, and host equipment compatibility in a timely fashion before the purchase of the program.

Fabrication Processes

Finally, the current suppliers experienced similar problems as the first supplier with their initial fabrication runs, resulting in circuits that were not suitable for accurate testing. These suppliers have been willing to work the processing issue, and have now delivered parts that are closer to required tolerances. It did, however, require two fabrication cycles to understand the process. In the future, delays can be avoided by having the supplier fabricate a representative sample board during the actual circuit design phase. This will give the supplier adequate time to work process development in parallel with the detailed circuit design such that the fabrication process will be correct when the first circuit iteration is ready for fabrication. It should be noted that the most significant fabrication problems encountered were with structures required to design and test the antenna elements. These structures are required only during the design phase and are not a part of the final design. The final design tolerances are compatible with quoted standard fabrication processes.

3.6.3 Completeness of Task and Future Directions

The overall progress on this task did not meet the goals and objectives planned. Time delays associated with simulation software, vendor selection, and fabrication process problems prevented completion of the final 2D full array. Additionally, interest in SDS systems at the automotive platforms waned, and the interest that remained was directed at lower performance, lower cost systems that did not have target discrimination. The system architecture required to meet system performance specifications established at the start of the ACAS program was more sophisticated and more expensive than needed to meet evolving specifications. The combination of schedule delays and business decisions based on customer desires resulted in a decision to terminate efforts on this task prior to completion of a full 2D array. The task was stopped with $300,000 left in the original budget.

During the course of this task, we have identified certain items that need development in order to better achieve low cost commercial systems. Improved simulation software packages are required. Additionally, a process to quickly determine the optimum simulation programs for the particular design task needs to be developed. Fabrication processes for thin multi-layer microwave printed circuit boards need to be improved before high volume cost objectives can be met. There are "specialty houses" that can fabricate boards to the accuracy required in small quantity at "development level" prices, but the ability to achieve the required accuracy and repeatability at high volume, cost-effective facilities has not yet been demonstrated. A solution to this fabrication issue will be required for future side, rear, and forward systems.

A materials and fabrication process development program for multi-layer microwave and millimeter wave circuit boards should be considered in the future.