Replacing a well-established technology with a more modern approach can often provide groundbreaking capabilities. For semiconductor devices, the classic justification of smaller, faster, cheaper (and/or higher performance, reliability and more) has been the lure for over 50 years. Today’s digital circuits in automotive applications with critical timing requirements have an even greater need than in the past for microelectromechanical systems (MEMS) oscillators. This article will discuss this emerging need in a variety of automotive applications, explain the differences between MEMS and crystal oscillators and present a new class of automotive-qualified MEMS oscillators that can address the most time-critical applications and provide improved reliability for all applications.
New Demands for Emerging Automotive Applications
Today’s automobiles require precise timing for Advanced Driver Assistance Systems (ADAS) that include on-board cameras, ultrasonic sensing, LiDAR and radar, as well as for infotainment, in-vehicle networks and more. While MEMS oscillators have been produced and used in automotive applications for over ten years, the emerging use of ADAS for autonomous, self-driving cars requires even greater capability and demands more from time-synchronizing devices.
Automotive Requirements
Reliability is a major concern for carmakers and their electronic systems suppliers. Crystal oscillators are mechanically cut and sanded down from quartz to obtain the right frequency and are packaged in a hermetically sealed enclosure. The thin structure of the crystal makes it susceptible to vibration damage and limits it to a fixed frequency. Also, the manufacturing cleanliness level is not high for these devices. Furthermore, the relatively large quartz devices do not hold up well under higher shock and vibration conditions.
In contrast, MEMS oscillators are produced in an integrated circuit (IC) fabrication facility, so they have a much higher cleanliness level like other ICs. In fact, the MEMS oscillator provides 20 times better reliability, 500 times better tolerance to shock and five times better vibration resistance than traditional CRYSTAL OSCILLATOR devices.
Also, MEMS oscillators are inherently very small and rugged. Crystals have a finite size, and the smaller crystals become increasingly more expensive. In the first automotive applications where space constraints were very tight, it was necessary to retrofit some of cameras on vehicles due to the crystal size. In this application, the MEMS approach was a natural solution. Many new automotive applications, such as ADAS, require even smaller packaging, so the MEMS oscillator’s size becomes another driving factor to replace crystal oscillators.
Another aspect of the MEMS oscillators is their ability to maintain their frequency stability at very high temperatures. Quartz devices have a very non-linear characteristic to temperature and have more difficulty in this area. MEMS oscillators available today are rated Grade 1 (-40°C to +125°C ambient operating temperature range per AEC-Q100). The next-generation MEMS oscillators will perform at higher temperatures and address some of the areas in the car where Grade 0 (-40°C to 150°C) is required (see Table 1).
High temperatures can occur in an automotive application because of the ambient temperature of the installed location and/or the oscillator’s required placement on the printed circuit board (PCB). Higher levels of connectivity in vehicles has required higher-power ICs. The heat dissipation from these ICs increases the local ambient temperature for nearby components. For system stability, the crystal oscillator is normally placed close to the IC that it supports, and historically this has allowed a Grade 3 rating. However, this situation is changing.
Microprocessors in infotainment systems dissipate significant heat, and although most car interior components are specified as Grade 2 (up to 105°C), clocks that are physically close to the processors are required to support Grade 1 (to 125°C). A MEMS oscillator is the best solution, because these powerful processors can easily heat up a crystal which experiences sufficient temperature drift and frequency shift, causing the oscillator to be outside of the required frequency range. One solution that allows the continued use of the crystal oscillator is mounting it further away from the processor. This impacts local real estate on the PCB. Another solution is a higher-stability (-50°C to 125°C) crystal oscillator which costs more—perhaps three or more times as much.
In contrast, a MEMS oscillator has active temperature compensation circuitry. The MEMS oscillator’s circuitry can provide real time correction for temperature variations—as much as 30 times per second—by sensing temperature and adjusting to maintain a constant output frequency. This achieves very accurate (as low as ± 20 ppm) temperature stability for high-temperature applications and can provide a cost reduction if it is compared to the cost of a high-stability crystal oscillator.
With increasing performance and processing power for graphics (GPU) and computing (CPU) ICs and their associated power management ICs, the existing crystal oscillators will increasingly be challenged to their limits and beyond.
MEMS Oscillator Technology
The foundation of a MEMS oscillator is a MEMS resonator. This is a structure etched from silicon that produces very precise mechanical vibrations to provide an accurate frequency. The Free-Free beam Short support (FFS) resonator design is shown in Figure 1. Contacting the substrate at four anchor locations, the beam is above it and separated by a narrow gap, so the resonator is free to move.
The electrode below the FFS resonator beam creates an electrostatic transducer. When the beam and the electrode are held at differing voltages, a force is generated between them. Since the transducer gap varies when biased, the structure behaves like a time-varying capacitor that produces an output current at the resonant frequency.
To achieve high quality factors, a capping and sealing process using fusion bonding seals the MEMS resonator inside a vacuum. The resulting wafer-level package can be used in a wide array of injection molded IC packages. Figure 2 shows how the resonator inside the sealed MEMS die is stacked onto a CMOS application-specific integrated circuit (ASIC). Wire bonds connect the MEMS device to the ASIC die.
In the ASIC, on-chip one-time programmable (OTP) memory and a crossbar switch provide product flexibility. The PLL and divider values that set the output frequency are stored in this memory, along with temperature calibration settings, choice of output protocol, rise/fall time control, enable pin pull-up/down, and other values.
In fact, many functionalities can be added to the MEMS ASIC. There can be multiple outputs, which help to reduce the required space and functionality that cannot be added to a quartz crystal. Another example is a spread spectrum function to reduce or avoid electromagnetic interference (EMI) issues. EMI can also be affected by output rise and fall times of the clock. With the programmability of the ASIC in the MEMS oscillator, changing the rise and fall time of the clock solved the problem in a very timely manner and resulted in completion of the design.
Automotive-Qualified MEMS Timing Solutions
The recently introduced DSA11x1 and DSA11x5 are automotive-grade MEMS oscillators and clock generators. The AEC-Q100 qualified units have a frequency stability as low as ±20 ppm over a -40 to +125°C temperature range and are designated for AEC Grade 1, Grade 2 and Grade 3 applications.
With phase jitter below 1 ps (typical), these MEMS oscillators operate over a frequency range from 2.3 MHz to 170 MHz. The AEC-qualified units are available in small industry-standard footprints of 2.5 mm x 2.0 mm, 3.2 mm x 2.5 mm and 5.0 mm x 3.2 mm and all are 0.85-mm thick. Functionally equivalent to the DSA1101/21, the DSA1105/25 have longer rise and fall times for EMI reduction. Figure 3 shows the integration in these MEMS oscillators.
With other circuitry, the temperature sensor in the MEMS oscillator produces a digital representation of the die temperature that is passed to the PLL to correct for natural spreads in the absolute frequency of the resonator, as well as its temperature coefficient. Figure 4 shows an example of the temperature stability produced by this technique.
shows dramatic improvement, especially at higher temperatures up to 125°C.
Multiple Output MEMS Oscillators
One of these new AEC-Q100 qualified and Grade 1 rated MEMS oscillators, the DSA2311, is the industry’s first dual-output MEMS oscillator. Available in a 2.5 x 2.0 mm package (Figure 5), it can replace two crystals or oscillators on a board (see Figure 6). The unit’s two simultaneous CMOS outputs each range from 2.3 MHz to 170 MHz. This saves PCB space, reduces purchasing, inventory and installation costs and ultimately leads to even further integration.
With the dual-output MEMS oscillator, two crystals can be replaced with a single device, reducing the bill of material (BOM) cost. Infotainment essentially has a motherboard and many processors, and each requires a reference frequency. In this case, the dual-output MEMS oscillator can replace multiple clocks. Since PCB space can be quite valuable and hard to come by, this MEMS oscillator provides options and can solve a few problems. Figure 7 shows how the DSA2311 and other Microchip devices come together in an automotive circuit.
Microchip has a history of product longevity, and its client-driven obsolescence practice means that Microchip will make products available if possible when there is a demand. Carmakers and their suppliers can count on a consistent supply of MEMS oscillators much longer than they can from other semiconductor suppliers.
Time is of the Essence
For any design change, design-in support is essential. With Microchip’s online ClockWorks Configurator tool, designers can easily select and customize the right MEMS oscillator for their application based on frequency, package size and temperature range and order free samples as well. Even the two output frequencies of the DSA2311 dual-output clock generator can be customized by using Clockworks Configurator.
While customers can receive samples within two to five days from the configurator, with the TimeFlash 2 Field Programming Kit, designers can program blank field-programmable oscillators to a custom frequency and perform design verification within seconds. Plugging the kit into the PC’s USB port allows Flash programming right on the user’s desktop. The kit also provides the ability to measure frequency accuracy and power consumption of standard oscillators, as well as measuring current and stability.
Change is Good
Over the past 20 years, reliability has become a much greater concern for carmakers than it once was. On a PCB, the ICs have the highest reliability. Other components, including crystal oscillators, fall below this benchmark level. In contrast, MEMS oscillators raise the oscillator reliability to the ICs’ level, which is a great benefit to automotive customers. With applications such as autonomous driving, the highest level of reliability is required, so the MEMS oscillator solution becomes the compelling choice for automotive suppliers.
If there is any hesitation in making the change from crystal oscillators to MEMS oscillators, the increased frequency stability, space saving, temperature and shock and vibration capabilities tips the decision towards MEMS oscillators. Because of these advantages, more and more automotive manufacturers are adopting the new MEMS oscillator technology.
About the author
Song Li is the Product Marketing Manager for the Timing and Communication Group at Microchip Technology. He has nearly 20 years of experience in the semiconductor business, 15 of them are associated with clock and timing products. He is currently responsible for product strategy and marketing of MEMS oscillator and clock generator product lines in the Timing and Communication Group of Microchip. He was product marketing manager at the Clock and Timing division of Silicon Labs. Prior to Silabs, he led business development in the Asia market at SpectraLinear and served as an application engineer at Cypress Semiconductor. Song has a MSEE in Electronics and Computer Science from the University of Alabama and a BSEE from Tsinghua University in China.