MMICs provide the key for lower-cost automotive radar

The forward-looking radar system (FLRS) at 77 GHz allows a detection of objects 1-200 m in front of the car. The angular coverage is around ±5º. The commercially available versions are limited to the detection of moving objects. They are called autonomous cruise control (ACC) radars and are sold as comfort systems. The challenging next development steps consist of adding the detection and classification of fixed objects in order to build real collision warning and avoidance systems. These new generations require better performance including sensitivity and angular resolution, and higher architecture complexity to give multiple channels and fusion with different sensor technologies.

Short-range radar sensors (SRRS), also based on mm-wave radar, are put all around the car. The detection range is 0.1-20 m. The standard frequency for SRRS has yet to be decided due to important requirements on bandwidth (2-4 GHz). The most commonly used frequency is 24 GHz, but 77 GHz is also being considered. These sensors cover a range of applications such as parking and reversing aids (as fulfilled today by ultrasonic sensors), but also side object detection and the "stop and go" function for moving the vehicle forward in slow-moving traffic queues.

There are three main radar principles that can be applied to automotive radar: frequency shift keying (FSK) systems; frequency modulated continuous wave (FMCW) systems; and pulsed systems. Sensors based on these principles are "all-weather" solutions, providing very fast and accurate information on the speed, position and the size of detected objects. Thanks to continuing performance improvements such as radar sensitivity, multi-beam antennas, electronic scanning and sensor synchronization, more accurate and reliable information can be obtained, leading to many new possible functions. GaAs MMIC-based solutionsThe first generations of radar at 77 GHz were based on Gunn diodes. These are now being replaced in many systems by GaAs MMICs thanks to their wide availability, falling cost and high performance. This evolution is well advanced, and some MMIC-based radar versions are already in production for high-end cars. For a more widespread adoption of this concept by car manufacturers, MMIC technology has to show better performance at lower cost.

The main challenge to implementing MMIC-based automotive radar is to find a cost-effective solution fulfilling the following constraints:
• Operation at very high frequency;
• Solutions for high-volume production that do not need post-assembly tuning;
• Ability to withstand the harsh environmental conditions for automotive applications such as vibration, humidity and wide temperature variations;
• High integration level for building very small sensors;
• Relatively high level of performance.

All of these constraints when considered together define a tough set of boundaries for MMIC development, yet GaAs-based MMICs are helping to meet the performance criteria at acceptable costs. The existing processes (PHEMT, HBT and Schottky diode-based MMICs) are mature enough to provide high performance at the mandatory high-quality level for automotive radar. Due to the very high operating frequency of active and nonlinear devices, and the excellent quality factor of passive devices, many functions can be implemented. Silicon heterostructure technology is a candidate for future generations of automotive radar, but much development is needed to prove its feasibility in this application.

Considering the cost and the environmental constraints placed upon MMICs for automotive radar, the required performance level is relatively difficult to obtain. The key requirements can be summarized as follows:
• High sensitivity. This can be defined as the ratio between the output power and the noise figure. Noise is particularly important, as the radar receivers work at very low intermediate frequencies (IFs), and so minimizing the low-frequency noise and transmit-to-receive leakage is critical.
• A fully integrated VCO requires very low phase noise.
• Very-high-frequency stability when no phase locked loop (PLL) is used.
• Relatively high power at 77 GHz (12-17 dBm is required at the component level.)
• Very high switching speed for pulsed SRRS. Cost effective approachesThe cost of GaAs-based technologies is continually decreasing and is now at a few dollars per mm2, which allows competitive solutions for trucks and high-end cars. It is now well recognized that MMICs have much to offer by removing the tuning during front-end fabrication, and also by improving yields when compared to hybrid solutions.

There are two main challenges for MMIC technology in this application. First, the chip-set size must be minimized. This is a permanent objective of circuit designers, which begins with close co-operation with the radar designer in order to find the right architecture. It ends with a clever MMIC design done on the best technologies to optimize performance.

Second, the MMICs must be easily integrated into the front-end modules. This is not a trivial issue, due to the frequency range involved. Much work has already been done on interconnection techniques, allowing the MMICs to be assembled with automatic equipment (pick and place, and bonding). Also, in order to make MMIC solutions more attractive, a lot of emphasis is being placed on MMIC protection, such as passivation against humidity to avoid using expensive hermetically sealed modules. A number of packaging concepts are under test and evaluation. Surface mount technology is now practical at 24 GHz thanks to the latest developments in high-frequency plastic packages. However, at 77 GHz, packaging is still a challenging issue, despite the development at UMS of the multichip package in the W-band (MCPW) concept. This involves a package that integrates one or more MMICs where the W-band interconnects are performed using electromagnetic transitions. All other connections use standard microstrip or coaxial methods. MMICs for frequency generationThe first generations of oscillators were fabricated using 0.25 µm gate length PHEMT processes. They were based on subharmonic generation and often used an external resonator. This was the only way to combine frequency stability, phase noise and good enough repeatability for production. The PHEMT-based VCOs developed at UMS for pulsed and FMCW systems operate at 12.75 GHz, are coupled to an external medium-Q resonator, and have a frequency tripler on the same chip. More than 300 MHz tuning range is obtained at 38 GHz with a phase noise in the range of -68 dBc/Hz at 100 kHz from carrier frequency.

Newer generations of oscillators are fabricated using HBT processes and give large improvements over the PHEMT-based VCOs, including better low-frequency noise characteristics and very high cut-off frequencies. VCOs can now be fully integrated, and provide better performance and the flexibility to include functions such as dual tuning ports and subharmonic output for PLL coupling. Figure 2 shows the results achieved on commercially available components. More than 5 GHz tuning range is obtained with a phase noise of about -85dBc/Hz at 100 kHz from the carrier frequency.