3.3 Development of Cost Reduction Components for Production (Task 2.2)

A primary factor in the deployment of a forward-looking radar (FLR) is the recurring cost of the hardware itself. The hardware configuration that exists today is typically a two piece system that has a millimeter wave radar sensor on the front of the vehicle and large complex processors and computers in the trunk of the vehicle. These systems use many commercial off-the-shelf components and cost in excess of $100,000 to implement. Marketing studies and surveys of automobile manufacturers have shown that the total cost of the system to the consumer must be less than $1,000 in order to achieve significant market penetration. Additionally, the size of the system, including both the millimeter wave sensor and the signal processor, must be reduced to less than 100 cubic inches. Therefore, significant progress must be made in the development of low cost, small size, integrated components.

The FLR sensor consists of three major elements: the antenna, the processor, and the transceiver. Of these, the transceiver is both the highest cost and the highest technical risk. The existing radar sensors typically use individual waveguide components and discrete components that require three dimensional assembly processes that are performed manually.

This task was structured to address this transceiver problem by designing a low cost planar transceiver that is manufacturable in high volume using automated assembly equipment. The task objectives are:

• Development of a low cost 76 GHz transceiver utilizing a single piece planar construction with MMIC chips.
• Development of a high yield fabrication process for 76 GHz MMIC oscillators fabricated in Gallium Arsenide.
• Seek the best compromise between:
  -- Up-integration for reduced parts count, ease of chip placement, and reduced circuit size.
  -- De-integration into multiple (less complex, higher yield) chips, which require more substrate space and manufacturing operations.
• The overall result must satisfy both performance expectations and production cost objectives.

3.3.1 Summary of Progress

The primary design approach featuring planar construction and a MMIC based design was defined, an initial performance specification was generated, and two potential suppliers were placed under contract. A primary factor in supplier selection was that each supplier had an existing MMIC chip set that could be used in the initial concept design. However, this resulted in separate internal block diagrams for each supplier. The "black box" electrical performance specification was identical and the mechanical package was interchangeable. The MMIC based units received from both suppliers did not meet specification and were late in delivery. After multiple meetings, it was decided to proceed with one supplier with a design that used devices from both suppliers. Two transceivers were fabricated and tested with significantly better results than before. It was determined, however, that this design could not meet cost expectations due to the immaturity of 76 GHz MMIC technology at this time in the program.

In order to best meet the overall program objectives, the primary design approach was changed to utilize a Gunn diode transmitter with a design architecture that supports a MMIC receiver. Gunn diode based transceivers meeting the MMIC transceiver mechanical size targets were designed, fabricated, and tested. The ability to characterize multiple transceivers on the bench, in the system, and on the road was a significant benefit of the ACAS program. A total of 50 second iteration transceivers were built and tested over temperature. Analysis of this data showed that fine grain linearity and linearity stability over time and temperature were the major electrical problems. A Design for Manufacturing and Assembly (DFMA) workshop was held with representatives from H E Microwave design and manufacturing, the transceiver supplier, and Delco Electronics manufacturing and process engineering. Several design changes for thermal management and manufacturing costs were identified and incorporated in a third iteration design.

The third iteration design was completed and nine units were fabricated and tested. These units met expectations with regards to fine grain linearity, improved thermal performance, and improved manufacturability.

Throughout the program, attention was given to advancements in MMIC technology and performance, and specifications for MMIC based transmitters, receivers, and fully integrated transceivers were prepared and updated.

3.3.2. MMIC Transceiver Design

The primary design approach was defined, and an initial performance specification was generated for the Forward-Looking Radar (FLR) sensor transceiver. Two potential suppliers were placed under contract. A primary factor in supplier selection was that each supplier had an existing chip set that could be used in the initial concept design. This, however, resulted in separate internal block diagrams for each supplier.

The design shown in Figure 3.13 used an HBT device as the oscillator source. HBT performance at 38 GHz was not adequate, so a 19 GHz source was used followed by amplifier and doubler stages in order to generate the 76 GHz operating frequency. The HBT oscillator design was tried as it offers the lowest phase noise, which is a key performance parameter for the system architecture.

The design shown in Figure 3.14 used a Gunn diode mounted in planar configuration. Although the anticipated phase noise will be higher than an HBT oscillator, Gunn devices typically generate significantly higher power than transistor devices operating at the same frequency, thus reducing the number of amplifier stages required.

The "black box" electrical performance specification was identical for both design approaches, and the mechanical package was interchangeable. The units received from both suppliers did not meet specification and were late in delivery.

After multiple meetings it was decided to proceed with one supplier using a design incorporating the best available devices from both suppliers. Two units were fabricated and tested with significantly better results than either of the first two designs. It was determined, however, that the design could not meet cost expectations due to the number of MMIC components required and the state of MMIC development at this time. The block diagram for this MMIC based transceiver is shown in Figure 3.15.

Transceiver results are summarized in Table 3.7. The data is presented in a format such that the recorded value is relative to the specification value. This eliminates referencing proprietary design and performance data in the table.

As seen in the data summary, the transceivers based on existing MMIC designs do not meet FLR requirements, although the 2nd run transceivers were much improved and closer to specification. The circuit data and device data has been analyzed and the reasons for the poor performance are understood. The circuit design changes in the second iteration added MMIC components in order to improve the circuit function. This is not an acceptable solution as it adds cost and power dissipation to the transceiver.

The data presented is for room temperature operation. The units were not tested over temperature. It was expected that the second iteration design changes will introduce potential channel tracking problems as a function of temperature, and it was determined that the thermal rise due to excessive power dissipation could destroy the transceiver if hot temperature tests were performed.

3.3.3 Gunn Transceiver Design

A change request to use a Gunn diode VCO transceiver as baseline in order to meet cost and performance targets and schedules for introductory volumes was presented and approved at the ACAS Third Quarter Review meeting. The Gunn transceiver was designed to be compatible with a MMIC receiver, although the first units used a planar microstrip discrete diode mixer circuit. Concept Gunn diode based transceivers meeting the MMIC transceiver mechanical size and volume targets were designed, fabricated, and tested.

Gunn transceiver data is summarized in Table 3.8. The data summary shows one set of data from both the first and second MMIC transceiver iterations along with the summary from the first two Gunn transceivers. As shown, the Gunn transceivers perform quite well relative to the specification. The only area of non-conformity, modulation sensitivity on one unit, is one that can be accommodated within the overall sensor design if necessary, and is significantly better than the previous units.

The concept Gunn transceivers were tested over temperature and showed good performance relative to the parameters listed in Table 3.8. These transceivers were then integrated into the system, and system performance data was taken. Initial system test results indicated that all transceiver specifications were being met or were at acceptable performance levels at room temperature.

Based on these results, twenty additional Gunn transceivers were assembled and tested over the entire temperature range. Data was evaluated and system performance was predicted based on the measured data. It was determined that the stability of the frequency tuning curve was the most critical parameter, and that small perturbations could cause significant system performance changes. This is a parameter that had not been adequately specified or characterized on the MMIC and initial two Gunn transceivers.

Some Gunn transceiver units were subjected to both burn-in and temperature cycling to determine the long-term stability of the tuning curve. Long term stability is a significant design issue due to the mechanical aspects of the oscillator structure. Initial results of the burn-in test indicate that the tuning curve is prone to small but permanent changes. These changes result in range errors when processing the system radar returns. The sweep waveform applied to the Gunn VCO must compensate for these linearity changes, and the system design had planned for a one-time factory calibration and compensation curve. The data indicated that a room temperature correction was inadequate, and that the compensation needed to be adjusted over temperature in relatively small increments of temperature change. Multiple correction curve look up tables were created in software to accommodate this problem, however, this significantly increased the factory test time to levels that were totally unacceptable for production.

Additionally, it was found that although most units could be adequately compensated over temperature to pass system acceptance testing, some of the units would experience permanent sets in the tuning curve. This meant that with time and multiple temperature cycles, some units changed to an out of alignment condition that resulted in field returns at the system level for range errors. It was necessary to significantly change the transceiver test station and test procedure such that the linearity could be accurately measured at the transceiver level.

Considerable effort was expended in the area of Design for Manufacturing and Assembly (DFMA). This activity is essential to achieve the overall goals of the program. A formalized DFMA was held that included representatives from H E Microwave design, H E Microwave manufacturing, the transceiver supplier, and Delco Electronics manufacturing and process engineering. The purpose of the workshop was to address the manufacturability of the design, including in process test and screening requirements. Performance and yield issues were discussed as well as potential process flows and new process development requirements. As a result of this activity, some major action items were identified, and some major design approach changes were recommended.

These changes identified in the DFMA did not involve the primary circuit architecture or electrical function which has been developed to date, but were entirely packaging, process, and reliability improvements. Thermal stress at elevated temperatures is a major issue due to the low efficiency inherent in Gunn oscillator designs. A rearrangement of the transceiver housing structure was suggested that greatly improves the heat spreading capability of the design. This layout change also greatly reduces the manufacturing assembly steps and is much more amenable to automation than the original design concept. The primary improvement is in the area of the IF circuit and associated voltage regulator and control circuits. No significant changes were recommended to the millimeter wave circuits previously developed. The new packaging concept also supported a reduced parts count.

A redesign to address the problems encountered to date was started. The circuit redesign was structured to address the issues of thermal dissipation, DFMA recommendations, and the small long-term linearity variations that have been identified. The redesign added an active frequency linearizer circuit as shown in Figure 3.16.

Characterization of multiple transceivers continued even as the last redesign was in progress. This characterization included extensive temperature testing, system level testing, and road testing. The intent of this effort was to fully understand the performance of the transceiver, to identify transceiver test requirements for production, to identify the most critical transceiver parameters relative to system performance, and to assure that the transceiver performance specifications were consistent with system specifications.

Thirty additional transceivers were built and tested over the entire temperature range. Data was evaluated and system performance was predicted based on the measured data. The stability of the frequency tuning as the most critical parameter was reaffirmed, and it was shown that very small perturbations could cause significant system performance changes. Results of the burn-in test indicate that the tuning curve is prone to small but permanent changes, and there has been no way to determine ahead of time which units have this problem and which do not. Additionally, there has been no effective screen identified that indicates the time period required for each transceiver to stabilize. Some have been stable from the start and did not change with screening, and others continue to change after multiple cycles of screening. This is the primary circuit design problem that was addressed in the last redesign.

Transceiver performance in the system was monitored and tracked. It appears that the transceiver, when properly compensated for linearity, meets all system requirements. Primary parameters of power out, frequency stability, receive gain, and noise figure appear to be consistent with system performance requirements. The long-term concerns are fine grain linearity stability, maximum operating temperature, and recurring manufacturing cost.

Nine of the final iteration transceivers were fabricated. Significant effort was spent in testing and characterizing these units both at room temperature and over the temperature extremes. Results of the tests are summarized in Table 3.9. The table also includes a comparison of the second iteration transceiver for reference. The data in the table represents the average performance of the nine final iteration transceivers tested and the average of six earlier transceivers picked at random. Specification values considered proprietary are withheld.

Overall, the performance of the third iteration design was good, and the primary goal of improving thermal management and linearity has been met. The linearity is excellent and is also very stable over temperature and time. This will eliminate the time consuming temperature calibration data collection necessary to generate a temperature dependent correction table that was required with the second iteration design. It will also eliminate the long term time dependent changes in linearity which in earlier designs caused slow degradation in system performance due to changes in tuning characteristics which made the temperature dependent linearity look-up tables invalid. These are very positive results of the new design.

A review of the performance table indicates that performance was degraded significantly in one area: transmit phase noise. Additionally, the Video Noise performance has degraded slightly, but the performance of both the 2nd and 3rd iteration units are out of specification for this parameter, and this parameter is a major driver in overall system sensitivity. All other parameters either meet or are within measurement error of the specification, or had no significant change from the second iteration design, and are not of concern at this time

It is believed that the increase in transmit phase noise is a result of coupling noise from the linearizer circuit onto the VCO control line. When the VCO is disconnected from the linearizer and tuned via an external power supply, the transmit phase noise is reduced by 1-5 dB (3 dB nominal). It is believed that additional filtering of the VCO control line should resolve this problem.

The increase in receive video noise is thought to be caused by two factors. First, because the system is a receive-while-transmit design, transmitter leakage during receive will cause the transmitter phase noise to be down-converted into the receive baseband. This means that any increase in transmitter phase noise will appear directly on the receive video output as an increase in the noise floor. Therefore, the transmit phase noise increase is one of the primary contributors to the increased receive video noise.

The second major contributor appears to be the T/R circulator. The amount of transmitter leakage into the receiver is a direct function of the isolation of the T/R circulator. The second iteration transceiver had a waveguide block circulator design that could be pretested and tuned to verify performance parameters were met. In the current design, the circulator has been integrated into a multiple component structure that is smaller and easier to fabricate, but eliminates the ability to verify and/or tune individual component performance. Tests on the one accessible T/R circulator port indicate that the T/R isolation may be degraded by approximately 5 dB. This then would result in a receive video noise increase of 5 dB independent of the higher transmit phase noise. The sum of the two results in a potential 8-dB nominal degradation of video noise in the system. The result will be a loss in maximum range compared to the second iteration design. The loss in range does not necessarily mean that the system does not meet minimum performance requirements, as these parameters are a part of the overall link budget which include transmit power, receive gain, antenna gain, etc. However, degradation of this magnitude consumes most if not all of the design margin, and needs to be resolved.

The third iteration transceivers were integrated into a full-up FLR system that is currently undergoing system integration and field-testing.

3.3.4 Challenges

The primary objective of this task, the development of a 76 GHz transceiver that meets system performance requirements over the extreme automotive environmental operating conditions and is cost-effective, is in itself the most significant technical challenge. There are a variety of system approaches and architectures that are theoretically feasible, and each system approach has a variety of transceiver design approaches that are also feasible. Unfortunately, there is not enough time and funding to develop all the design approaches and investigate the design trade-offs. Consequently, the design trade-offs were instead based on certain non-verified assumptions regarding technical performance, application requirements, and manufacturability.

For this task, a system design approach was chosen prior to the inception of the ACAS Program. The transceiver architecture was the primary variable. It was decided to try an all MMIC approach based on existing design building blocks. The program was structured to allow design iterations based on facts discovered during the ACAS Program. For example, the MMIC design was fabricated and tested to determine the maturity of GaAs MMIC technology with regard to both performance and cost. A particular architecture had to be chosen early in the program to insure that the overall schedule could be met. As the design was being fabricated, cost estimates were prepared that included, MMIC device processing costs (based on estimates of achievable yields at the foundry), transceiver material costs (molded, cast, machined housings plus other non-MMIC components), transceiver manufacturing costs and yields, and transceiver test costs and yields.

As the design was iterated to meet performance goals, the cost model was updated. It was determined that all MMIC technology had not reached the maturity necessary to achieve all cost and performance goals at this time. It was also determined that transceiver architecture changes could be made that minimize the active component count such that MMIC devices may become cost-effective in the near future, but not within the time frame of the ACAS Program. Emphasis was then turned to transceiver design alternatives that could meet the primary objective of both cost and performance without the exclusive use of MMIC devices, but that were compatible with MMIC insertion in the future.

In anticipation of continued maturity of GaAs MMIC technology at millimeter wave frequencies, the transceiver design was structured to minimize the number of active components and to group circuit functions in a fashion that allows continued development of MMIC devices. New MMIC component specifications based on results obtained from the ACAS Program were prepared and are being pursed with the MMIC industry base.

3.3.5 Completeness of Task and Major Benefits

This task was completed on schedule and on budget. The task was modified to use the best compromise between design approach and deployment constraints in order to meet the critical objective of performance expectations and production cost. In order to accomplish this, the all MMIC and all planar approach was changed in favor of a wave guide Gunn oscillator, with a discrete planar receiver and frequency linearizer. Given these changes in design approach, there were significant benefits derived from the ACAS program in the areas of cost reduction, performance improvement, and reliability. These benefits would not have been achieved in the same time frame without the ACAS program.

Primary performance improvement was achieved in the area of waveform linearity and thermal management. Waveform linearity is a critical parameter for FMCW radar systems. At the start of the contract, the transceiver typically had a linearity 5 times the specification limit. This required extensive temperature characterization and multiple look-up tables for correction over temperature. Even with these tables, intermediate temperatures often had linearity errors greater than the specification resulting in range errors during system operation. Addition of the active linearizer resulted in a transceiver with linearity up to 1/5 the specification limit at all temperatures. The need for temperature characterization and look-up tables was eliminated, and performance was significantly improved at all temperatures.

Additionally, the transceivers at the start of the program had a thermal dissipation 1.5 times the specification limit. Due to internal temperature rise, this resulted in an inability to operate at temperatures greater than 50^oC baseplate. With the improvements made during the ACAS, the present power dissipation is at 50% of specification value, a factor of 3 improvement. This allows transceiver operation at baseplate temperatures of 85^oC, which is consistent with automotive requirements.

Significant progress in cost reduction was also made during the ACAS contract. Development volume (<50 units) transceiver cost was reduced by a factor of 3. Additionally, estimated volume production costs were reduced by a factor of 5. The costs associated with temperature characterization for linearity were eliminated, resulting in an estimated 40% reduction in assembly and test labor costs for production quantities due to this parameter alone.

Reliability was also dramatically improved. Prior to the reduction in DC power consumption, all reliability predictions indicated that units would incur non-recoverable failures at 85^oC baseplate temperatures, and concept-engineering units were never operated at the high temperature extreme. With the 3 fold improvement in power dissipation, and with additional thermal resistance path design improvements, the units are now tested and operated at the high temperatures without damage. Additionally, a field failure rate in excess of 20% was observed for linearity drift alone. Since linearity was achieved by a look-up table, any change in linearity due to either electrical or mechanical aging created errors in the correction tables. There was no mechanism available to automatically recalibrate in the field, and after drift occurred, the corrections applied often resulted in greater non-linearity than the un-compensated oscillator.

Addition of the active linearizer has eliminated field returns for linearity failures. At this time, cumulated field time for the active linearizer is approximately 25% of the total for the passive look-up table units. The absence of failures coupled with the fact that drift failures tended to occur soon after deployment is a strong indication that the problem is solved, and not just an artifact of less field time.

3.3.6 Future Directions

There are certain aspects of the current task that can use further work. The purpose of the task was development of cost-effective millimeter wave components for production. Cost targets and product performance requirements including physical size were estimated based on the facts known at the start of the program. In addition to this task, other tasks of the program were structured to measure system performance and to establish requirements for collision warning products. Independent of the ACAS program, automobile manufactures were also working to establish performance parameters, and to derive consumer acceptance criteria. As a result of these activities, target performance and target cost were continually changing. Over the two-year program cycle, size and cost targets changed by greater than 50%.

Additionally, the state-of-the-art for MMIC devices was not as well established as anticipated at the start of the program. Cost-effective devices that met performance criteria did not exist as commercial off the shelf components, and the design cycle for MMIC devices was longer than the two-year program could tolerate. As a result, the transceiver developed during this program met the original size and cost targets and provides acceptable performance, but it does not meet the end of program changes desired by the automotive customers.

During the course of the program, significant progress has been made by some MMIC suppliers in both cost and generic designs that are compatible with FLR block diagrams. Use of these building blocks in the next generation transceiver will further improve performance, size, and cost, but are beyond the scope of the current program due to schedule and budget allocations.