MEMS oscillators with improved resilience for harsh automotive environments.
Oscillators are key components in automotive electronics systems. For example, a typical automotive camera module may have three or more oscillators, providing the clocks for microcontrollers, Ethernet controllers, and video chipsets. These oscillators have historically been built around a quartz crystal resonator connected to an analog sustaining circuit driving the crystal to vibrate at its resonant frequency. However, quartz-based devices suffer from poor performance and reliability in harsh automotive environments. SiTime has developed timing solutions based on silicon micro-electromechanical systems (MEMS) technology that exhibit better electromagnetic noise rejection and better performance under shock and vibration. In this paper, we first discuss the design and manufacturing of the MEMS-based device, with emphasis on the specific design aspects that improve reliability and resilience in harsh automotive environments. These aspects include the SOI-based MEMS fabrication process, the oscillator and state-of-the-art temperature compensation architecture, and the manufacturing and packaging process. We then describe the test methods used to evaluate the resilience of the device, including electromagnetic susceptibility (EMS), and performance during shock and vibration. The results show that the MEMS-based oscillator performs better than all quartz oscillators that were tested, with up to 50x better EMS, up to 24x better performance during shock, and up to 100x and 20x better performance during sinusoidal and random vibration, respectively.
CITATION: Arft, C., Lu, Y., and Parvereshi, J., "MEMS Oscillators with Improved Resilience for Harsh Automotive Environments," SAE Int. J. Passeng. Cars--Electron. Electr. Syst. 9(1):2016.
Resonating quartz crystals have been used as time and frequency references for almost a century. Quartz crystals were first used as frequency references for radios in the early 1920s, and the first quartz crystal clock was developed in 1928 at Bell Telephone Laboratories [1,2]. Until the last decade, quartz crystal oscillators were the de facto standard frequency references, and billions of quartz crystal timing products were used each year in a wide range of applications including military, industrial, consumer, and automotive . Examples of these quartz-based products are shown in Figure 1.
In the last decade, micro-electromechanical systems (MEMS) oscillators began replacing quartz oscillators in many applications, including high-reliability automotive applications . A key requirement for oscillators used in automotive applications is resilience and reliability in a variety of harsh environmental conditions. In this paper, we present measured data demonstrating that the resilience of the MEMS oscillators greatly exceeds that of quartz crystal oscillators during shock, vibration, and electromagnetic interference (EMI) that can occur in an automotive application. As an additional benefit, the MEMS-based resonator die does not require special vacuum packaging, which enables smaller size and a wider range of package options.
MEMS OSCILLATOR OVERVIEW
The primary goal of a timing reference is to produce an accurate and low noise clock signal that does not degrade over time and temperature, or due to environmental factors such as stress, shock, vibration, and electromagnetic noise. This section describes the design and implementation of the MEMS-based oscillator to achieve this goal.
The key building block of the MEMS-based oscillator is the MEMS resonator, which is fabricated and encapsulated at the silicon wafer level using the MEMS First [TM] process [6,7]. The MEMS structure is designed to mechanically resonate at a specific frequency that is stable over time and temperature, and is not affected by external forces applied to the MEMS die. The resonator is a passive mechanical structure that requires an active oscillator circuit to induce mechanical vibration. Figure 2 shows a detailed cross-sectional diagram of the MEMS resonator fabrication process. First, photolithography is used to pattern the resonator, and a deep reactive ion etch (DRIE) process is used to define the resonator structure in the device layer of a silicon-on-insulator (SOI) wafer (step 1). A silicon dioxide layer is deposited and electrical contacts are defined (step 2). A thin epitaxial polysilicon layer is then deposited at 1100[degrees]C and small "vent" holes are etched above the resonating structure (step 3). The resonating structure is released using a hydrofluoric (HF) acid vapor to etch the silicon dioxide around the resonator (step 4). The vent holes are sealed using a thick epitaxial polysilicon layer (step 5). This critical step provides a clean, low-pressure enclosure for the MEMS structure and protection from the external environment. The hermetic vacuum encapsulation is critical to achieving long-term frequency stability . The rough polysilicon layer is polished and a thin silicon dioxide passivation layer is deposited. Finally, electrical contacts are created, and metal traces are defined on the surface to connect the contacts to the electrical pads (step 6).
For automotive applications, there are several advantages of the MEMS oscillator in comparison to a traditional quartz resonator or oscillator. Due to its small size, the mass of the MEMS resonator is much less than that of a quartz crystal, making it less susceptible to the effects of shock and vibration. The MEMS resonator is centrally anchored, making it more immune to externally applied stress. In addition, the MEMS resonator is vacuum encapsulated inside the die, such that it does not require external vacuum packaging that increases the size of the product and may leak over time.
Figure 3 shows a comparison between a MEMS resonator and a quartz resonator. The quartz resonator is sealed in a 3.2 mm x 2.5 mm ceramic cavity vacuum package. Note the sealed cover was removed for the image. The quartz crystal is approximately 1.5 mm x 2.5 mm and is soldered to the package. The MEMS resonator is encapsulated inside a tiny 0.4 mm x 0.4 mm silicon die that does not require an additional sealed vacuum package. The resonator itself is much smaller than the quartz crystal and is centrally anchored, making it robust against stress, shock, and vibration.
Programmable Oscillator Architecture
The MEMS resonator is combined with a programmable oscillator IC as illustrated in Figure 4. The MEMS oscillator's patented architecture has been described previously  and for clarity is summarized here. The IC includes the active sustaining amplifier which drives the MEMS resonator and forms the oscillator circuit. The resonator uses differential drive and sense circuits that inherently reject any electromagnetically-coupled common-mode noise. The architecture also includes a temperature sensor and temperature compensation logic that controls a fractional-N phase-locked loop (PLL). The PLL serves two purposes. First, it is used to synthesize a large range of unique frequencies from the MEMS reference frequency. Second, it is used to compensate for small frequency offsets of the resonator due to fabrication tolerances and temperature. The two die are stacked, assembled, and molded in a plastic DFN package, a plastic SOT-23 package, or assembled into a chip scale package (CSP). An optical image of an assembled DFN package with the outer plastic compound removed is shown in Figure 5.
RESILIENCE TEST METHODS AND RESULTS
In this section, we describe the experimental methods used to compare the resilience of the MEMS oscillator to several commercially available quartz oscillators. We performed the following tests: electromagnetic susceptibility (EMS), performance during shock, and performance during sinusoidal and random vibration. The tests and performance metrics are summarized in Table 1.
Electromagnetic Susceptibility (EMS)
For oscillators, phase noise is defined as the frequency domain representation of rapid fluctuations in the phase of the clock signal. Increased phase noise can lead to performance issues such as bit errors or corruption of video frames. It is known that radiated EMI adversely impacts the phase noise performance of quartz oscillators [10,11]. In this study, we measured the increase in oscillator phase noise due to exposure to radiated EMI. We followed electromagnetic compatibility standard IEC EN610004.3  which specifies test conditions for evaluating the response of electronic components to EMI. The test was conducted in an anechoic chamber using the setup shown in Figure 6. The setup included a phase noise analyzer (PNA) to capture the phase noise of the device under test (DUT). The DUT was positioned so that it was aligned with the axis of the antenna as shown in Figure 7. The signal generator created an electromagnetic field with strength of 3V/m and scanned the frequency range from 80 MHz to 1 GHz in steps of 1%.
Exposure to radiated EMI at a single frequency causes a sharp peak in the phase noise spectrum referred to as a phase noise "spur." An example phase noise spectrum including a spur is shown in Figure 8.
The frequency of the noise spur induced by the EMI occurs at offset frequencies that are aliases of the oscillator frequency. This aliased frequency can be calculated as
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
where [f.sub.EMI] is the frequency of the radiated EMI, [f.sub.osc] is the output frequency of the oscillator, and n is an integer. As an example, Figure 8 shows the phase noise spectrum of a 26 MHz oscillator in the presence of EMI at 80 MHz. In this case, the noise spur is aliased to 2 MHz offset, since 80-(3x26) = 2 MHz. In this study, noise spurs caused by the radiated EMI were found to be the largest contributors to the oscillator phase noise degradation due to their large amplitude.
For each DUT, the calculated average power, P, of the noise spurs over the 80 MHz to 1 GHz range was used as a measure of the EMS of the oscillator. The average spur power was calculated as
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)
where [S.sub.p] is the magnitude of EMI-induced spur for each EMI frequency and N is the number of EMI frequencies in the scan.
The results of the EMS testing are shown in Figure 9. The test revealed that the average phase noise spur power due to EMI was lower for the MEMS oscillator than for any of the quartz oscillators in this study. The data shows that the MEMS oscillator outperforms other quartz based oscillators by up to 34 dB, which is equivalent to approximately 50 times greater immunity to a radiated electromagnetic field.
This data is particularly interesting since the MEMS device in this study utilizes a standard plastic IC package, not the typical ceramic package of most quartz oscillators. Without this EMS data, one may assume that the metallic lid used on the ceramic quartz crystal package would shield it from EMI; however, our data shows that ceramic packages do not guarantee superior EMS performance. In fact, the metal lid of the ceramic package is not grounded in most quartz-based oscillators and therefore does not provide additional shielding against EMI. Other factors, such as the resonator and the oscillator circuit, have a much greater effect on the EMS. Quartz crystals are piezoelectric materials that can accumulate electrical charge. Their operating frequency can therefore be affected by a radiating electromagnetic field. The MEMS resonator, on the other hand, is electrostatically driven and sensed, an actuation scheme that is more immune to EMI. In addition, quartz crystal oscillators use a single-ended oscillator circuit, whereas the MEMS device uses a fully differential circuit, which helps to mitigate the effects of external EMI.
Performance during Shock
Typically, shock and vibration testing are carried out as reliability tests only, and device performance is not measured during the shock or vibration event. However, both shock and vibration can induce an electrical signal in quartz and MEMS resonators that negatively affects frequency stability and jitter. To evaluate device performance during shock, the transient frequency deviation was measured during a mechanical shock event following the specifcations of MIL-STD-883H Method 2002  for a 1 ms half sine wave shock pulse with an acceleration of 500 g.
The shock test setup is shown in Figure 10. For each DUT, frequency measurements were taken every 100 [mu]s continuously for 10 seconds to monitor the frequency before, during, and after the shock event. Because of the asymmetric nature of the mechanical resonator and package, the oscillator response to shock depends on the direction of the applied force. The test was therefore repeated in the x-, y-, and zdirections, and we report the maximum frequency deviation for the worst-case orientation.
Figure 11 shows an example of the frequency data collected during the shock test, with the x-, y-, and z-direction data overlaid on the same plot. Because the start of data collection was not synchronized to the shock event, the shock occurs at slightly different times in the x-, y-, and z-direction data. For this DUT, a maximum frequency deviation of 13 ppm occurred when the shock was applied along the z-axis.
A summary of the maximum transient frequency deviation observed during shock is shown in Figure 12. All of the quartz oscillators exhibited between 2.5 and 14.3 ppm frequency deviation. The SAW devices (Quartz1 and Quartz4) are especially sensitive to shock. In contrast, the MEMS oscillator exhibited a frequency deviation of less than 1 ppm. As discussed previously, this is due to the stiffness and small mass of the MEMS resonator structure, resulting in minimal defection during acceleration.
Performance during Vibration
Next, we measured the oscillator performance during sinusoidal and random vibration that may occur in a harsh automotive environment. The vibration test setup consisted of a controller, power amplifier, and shaker, as shown in Figure 13. The phase noise of each oscillator was recorded before and during each test, and the test was repeated in the x-, y-, and z-directions. The same test setup was used for the sinusoidal and random vibration testing.
In this test we measured the response to sinusoidal vibration at frequencies ranging from 15 Hz to 2 kHz. The periodic nature of sinusoidal vibration creates frequency modulation that induces spurs in the phase noise spectrum at frequencies offset by the vibration frequency as shown in Figure 14. To characterize the oscillator sensitivity to vibration, the vibration-induced phase noise spur in dBc is converted into an equivalent frequency shift in parts per billion (ppb), then normalized by the peak acceleration of the sinusoidal vibration and expressed in ppb/g. Mathematically, the sensitivity, [GAMMA], can be expressed as
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
where [f.sub.vib] is the frequency of vibration, [S.sub.p] is the spur level in dBc, A is the peak acceleration, and [f.sub.osc] is the oscillator output frequency.
A peak acceleration of 4 g was applied for each sinusoidal vibration frequency of 15, 30, 60, 100, 300, 600, 1000 and 2000 Hz. Each sweep of the vibration frequency took about 15 to 20 minutes, and the dwell time at each frequency was about 1 minute. The test was repeated in the x-, y-, and z-directions, and we report the result of the worst-case orientation.
The normalized sensitivity of the oscillators to sinusoidal vibration across vibration frequency is shown in Figure 15. These results demonstrate that the MEMS oscillator outperforms the other devices during sinusoidal vibration by a factor of 10 to 100, depending upon the frequency of vibration. Again, these results illustrate the advantage of the small and stiff MEMS structure, compared to a quartz crystal resonator.
Oscillators may experience random vibration during operation in an automotive environment with vibration frequencies ranging from a few Hz to several kHz. These random vibrations increase broadband phase noise. Several standards specify test conditions for random vibration profiles that vary with the expected operating environment or type of electronic equipment tested. We conducted tests according to MIL-STD-883H, Method 2026 , as it is the most applicable to electronic components. This standard specifies a vibration profile and allows for various intensity levels as shown in Figure 16. Condition B, with a composite power level of 7.3 g rms, is suitable for a high vibration automotive environment. For this test, the controller in the test setup of Figure 11 used digital signal processing to synthesize random vibration in the specified frequency range, based on the power density level defined in the vibration profile.
Random vibration causes an increase in phase noise at offsets corresponding to the vibration frequency. For each oscillator, we measured the phase noise before and during random vibration, and then calculated the integrated phase jitter (IPJ) from 10 Hz to 10 kHz offset for each case. The induced jitter due to random vibration was then calculated as the root-mean-square difference between the IPJ values measured before and during vibration. Figure 17 shows an example of the increase in phase noise observed when an oscillator is exposed to random vibration. In this example, the IPJ over the 10 Hz to 10 kHz integration band increased from 2.18 to 90.1 ps rms when the oscillator was exposed to the random vibration profile shown in Figure 16.
Results of calculated induced jitter during random vibration are summarized in Figure 18. The data shows that all of the quartz oscillators exhibit significant increases in jitter during random vibration, from nearly 40 to over 100 ps rms. In contrast, the MEMS oscillator is 20x less sensitive to random vibration, exhibiting only 5 ps rms increased jitter.
Quartz crystal-based oscillators have become the de facto standard timing reference for many applications including automotive electronics. But, our experiments have shown that quartz-based devices can suffer from poor performance in automotive environments. We have developed an alternative based on MEMS technology that exhibits better electromagnetic noise rejection and better performance during shock and vibration that may occur in a harsh automotive environment. In this paper, we presented test methods and results showing that the SiTime MEMS-based oscillators perform better than all competitor quartz oscillators that were tested, with:
* Up to 50x better EMS performance
* Up to 24x better performance during shock
* Up to 100x better performance during sinusoidal vibration
* Up to 20x better performance during random vibration
[1.] Marrison, W.A., "The Evolution of the Quartz Crystal Clock," Bell System Technical Journal, 27:510-588, 1948, doi: 10.1002/j.15387305.1948.tb01343.x.
[2.] Bottom, V. E., "A History of the Quartz Crystal Industry in the USA," in Thirty Fifth Annual Frequency Control Symposium. 1981, pp.3-12, May 27-29, 1981, doi: 10.1109/FREQ.1981.200452.
[3.] Vig, J.R, "Quartz Crystal Resonators and Oscillators; For Frequency Control and Timing Applications-A Tutorial; Rev. Nov 2008," http://tf.nist.gov/sim/2010_Seminar/vig3.ppt, accessed Sept. 2015.
[4.] Ng, E., Yang, Y., Hong, V.S., Ahn, C.H. et al., "The Long Path from MEMS Resonators to Timing Products," in Micro Electro Mechanical Systems (MEMS), 2015 28th IEEE International Conference on, pp.1-2, Jan. 18-22, 2015, doi: 10.1109/MEMSYS.2015.7050869.
[5.] HM International, http://www.hminternational.be, accessed Sept. 2015.
[6.] Partridge, A., Lutz, M., Kim, B., Hopcroft, M. et al. "MEMS resonators: getting the packaging right," in Proceedings of the 9th SEMI Microsystem/MEMS Seminar (SEMICON), Makuhari, Japan, pp. 55-58, 2005.
[7.] SiTime Corporation, "SiTime's MEMS First [TM] Process," http://www.sitime.com/support/application-notes, accessed Sept 2015.
[8.] Candler, R.N., Hopcroft, M., Kim, B., Park, W.-T. et al., "Long-term and accelerated life testing of a novel single-wafer vacuum encapsulation for MEMS resonators." Journal of Microelectromechanical Systems 15(6):1446-1456, 2006, doi: 10.1109/JMEMS.2006.883586.
[9.] Partridge, A., Lutz, M.,"Frequency and/or Phase Compensated Microelectromechanical Oscillator," U.S. Patent 6,995,622, Feb. 7, 2006.
[10.] Rohde, U.L. and Poddar, A.K., "Electromagnetic Interference and Start-up Dynamics in High Frequency Crystal Oscillator Circuits", in Sarnoff Symposium, 2010 IEEE, pp. 109, 2010.
[11.] Laurin, J.J., Zaky, S.G., and Balmain, K.G., "EMI-induced failures in crystal oscillators," Electromagnetic Compatibility, IEEE Transactions on, 33(4):334-342, 1991.
[12.] IEC Test Method Standard, "Electromagnetic compatibility (EMC) - Part 4-3: Testing and measurement techniques -Radiated, radio-frequency, electromagnetic field immunity test," "IEC 61000-4-3 ed3.2, Rev. March 2010.
[13.] Department of Defense Test Method Standard, "Microcircuits," MIL-STD-883H, U.S. Defense Logistics Agency, Rev. 2010.
SiTime Corporation, 990 Almanor Ave., Sunnyvale, CA 94085
DRIE - Deep reactive ion etch
DUT - Device under test
EMI - Electromagnetic interference
EMS - Electromagnetic susceptibility
IPJ - Integrated phase jitter
MEMS - Micro-electromechanical systems
PLL - Phase-locked loop
PNA - Phase noise analyzer
SOI - Silicon-on-insulator
Carl Arft, Yin-Chen Lu, and Jehangir Parvereshi
Table 1. Summary of resilience tests and performance metrics. Reliability or resilience test Performance metric Electromagnetic Susceptibility Average phase noise spur (dB) Performance during shock Peak frequency deviation (ppm) Performance during Frequency deviation per g (ppb/g) sinusoidal vibration Performance during random Induced integrated phase vibration jitter (ps rms) Table 2. Summary of the MEMS and quartz oscillators tested in this study. Label Technology Output MEMS MEMS + PLL 156.25 MHz (LVPECL) Quartz1 Quartz, SAW 156.25 MHz (LVPECL) Quartz2 Quartz, 3rd Overtone 156.25 MHz (LVPECL) Quartz3 Quartz, 3rd Overtone 156.25 MHz (LVPECL) Quartz4 Quartz, SAW 156.25 MHz (LVPECL) Quartz5 Quartz, 3rd Overtone + PLL 156.25 MHz (LVPECL)
|Printer friendly Cite/link Email Feedback|
|Author:||Arft, Carl; Lu, Yin-Chen; Parvereshi, Jehangir|
|Publication:||SAE International Journal of Passenger Cars - Electronic and Electrical Systems|
|Date:||May 1, 2016|
|Previous Article:||GaAs Optical Field Effect Transistor (OPFET): A high performance photodetector for automotive applications.|
|Next Article:||Introduction of a MR rotor-position sensor for off-axis application and comparison to competitive technologies.|