The IC Insider looks at a MEMS-based gyroscope in microscopic detail, and finds that the ingenuity in this sophisticated sensor goes far beyond the process technology used to sculpt its mechanical features.
By Randy Torrance, Chipworks -- EDN, 12/1/2008
Gyroscopes have been around for 190 years, ever since Johann Bohnenberger built the first one in 1817. For the first 180 years, the basics of the simple and effective design remained much the same. A gyroscope uses a spinning disk whose axis is mounted such that it can take any orientation. Because the spinning disk has a high angular momentum, it is quite resistant to external torque. By mounting the device in gimbals, a gyroscope can maintain an almost fixed orientation in spite of applied torque. By measuring the angle between the spin axis and the fixed gyroscope frame, one can determine the external torque or rotation that has been applied to the frame (see Figure 1).
In recent years people have realized that a gyroscope would be a great addition to many devices that can benefit from information about their orientation. For instance, I'd like one in my coffee cup so that when I'm walking around the office, it can notify me when I tilt it too far. OK, that may be a bit of a stretch, but gyroscopes are now showing up in many applications. How does your car's ESC (electronic stability control) system know that your car is spinning out of control—and how does it figure out how to adjust the brakes and power distribution to correct the situation? A gyroscope. How do some of today's more advanced digital cameras adjust for a user's shaky hand to improve photo quality? A gyroscope.
But how do they get that spinning disk in there? The answer, of course, is that gyroscopes have moved on from that original design and now come in silicon form. To be exact, MEMS (microelectromechanical systems) are now being used to implement gyroscopes in silicon, and they now come integrated on the same ICs as the standard logic that controls them.
MEMS gyroscopes use a vibrating structure rather than the traditional rotating disk to determine orientation. These gyroscopes measure angular rate by means of the Coriolis acceleration, which can be explained as follows. Think of a rotating disk, such as a 33-1/3-rpm record (younger readers can try to imagine a DVD spinning inside its player). An object near the center of the disk is moving quite slowly, whereas an object near the outer edge of the disk is moving much more quickly. Both objects are also subject to tangential acceleration due to the rotation. This acceleration, which is proportional to the object's distance from the center of the disk, is known as the Coriolis acceleration. Hence the object near the outside experiences a greater Coriolis acceleration than the object near the center. You can measure this acceleration and use it to calculate the externally applied torque.
Chipworks has analyzed the InvenSense IDG-300 two-axis gyroscope—procured by removing it from a camera—which operates according to this principle. The InvenSense MEMS gyroscope uses two resonating proof masses that are linked. The proof-mass system can be thought of as sitting on our rotating disk. As it resonates outward it experiences more Coriolis acceleration, and as it moves inward it experiences less. The mass is micromachined into the silicon and allowed to resonate in only one direction (Figure 2).
Interlocking fingers are micromachined in between and connected to the proof-mass system and the support silicon, such that they can measure capacitance changes (Figure 3).
The Coriolis acceleration will move the proof-mass system proportionally to the value of that acceleration, and the capacitance will in turn change and give a measure of the Coriolis force. InvenSense has integrated two of these MEMS structures orthogonally on the die to allow for two-axis detection.
The design uses a bulk micromachining process, vertically integrated electronics, and wafer-scale packaging. The gyro design includes an integrated dual-mass electrostatically driven actuating mechanism with capacitance sensing. The two linked proof masses are driven into an anti-phase oscillation by electrostatic actuators placed beneath them. This out-of-plane resonating proof-mass system senses the rate of rotation in either the X or Y axis.
The product measures angular velocity with a full-scale range of ± 500º/sec and sensitivity of 2 mV/º/sec. The oscillation circuit precisely controls the amplitude to maintain constant sensitivity over the temperature range of 0 to 70ºC. Two externally connected compensation capacitors are used for the amplitude control loops. An external low-pass filter can attenuate high-frequency noise generated by the vibrating proof masses. The sensor bandwidth is limited to 140 Hz by an internal low-pass filter, and the user can achieve lower bandwidth by choosing an appropriate external filter with a cutoff frequency less than 140 Hz. The MEMS device has an onboard EEPROM for factory calibration of the sensor. Factory trimmed scale factors eliminate the need for external active components and end-user calibration.
Figure 4 shows an annotated die photo of the IDG-300.
The two MEMS structures including the vibrating proof masses can be seen in the center of the die. Each MEMS device has its own dedicated drive circuitry (located below the MEMS sensors), and signal-detection and measurement circuitry (above the MEMS sensors). At the top of the die is a large amount of trimming and calibration circuitry, which includes nonvolatile memory arrays. The bottom of this die photo is mainly occupied by voltage- and bias-generation circuitry.
InvenSense has designed a lot of interesting and novel circuits onto this chip. The Coriolis sense circuitry needs to be able to detect very small capacitance changes with high accuracy. High-voltage circuits are required both to drive the MEMS proof masses and for the nonvolatile memory. The proof mass drive circuits and oscillator must be both accurate and powerful. Finally, the voltage- and bias-generation circuitry must be very accurate, and must account for variations in process, voltage, and temperature. Let's consider this circuitry, and how it can maintain accurate voltages, currents, and oscillation frequency to the MEMS devices.
One of the key components in a MEMS gyroscope is the circuitry used to drive the proof mass into resonance. The proof mass will only vibrate effectively at its resonant frequency. The oscillator subsystem must therefore drive the proof mass at a precise oscillation frequency to maintain the system in resonance. This driver system must include methods to compensate for process, voltage, and temperature variations. The IDG-300 does this using multiple methods. A large percentage of the die is dedicated to on-chip trim and calibration circuits that can be set at the factory to compensate for process variations. An interesting CMOS-compatible OTP (one-time programmable) memory is used to store these factory settings. Figure 5 shows a row of four of these cells.
However this still leaves voltage and temperature variations to handle. InvenSense starts with a bandgap reference to generate a supply- and temperature-independent voltage reference to the oscillator. However, the resonant frequency of the proof mass is a function of temperature. Hence the oscillator actually needs a temperature-dependent reference in order to maintain the proof-mass resonance. We believe InvenSense has modified the bandgap reference circuit to add in this characteristic.
Referring to Figure 6, we see the heart of the bandgap reference.
PNP transistors X1619 and X1167 are used in conjunction with the resistors and op-amp to compensate for temperature variations. It is interesting to note the lengths to which InvenSense went to match these bipolar devices. Referring to the image of the polysilicon layer of the chip shown in Figure 7, we can see these two PNP transistors highlighted with the white rectangles.
X1167 can be clearly seen to be eight unity PNPs, compared with the two of X1619. What is of interest is just how much larger these devices are than a standard MOS device in this process, one of which is highlighted by the small white rectangle in the upper left of the bipolar devices. This is likely done for matching reasons, to ensure the best temperature response possible. It is also interesting to note that the feedback path to the op-amp actually travels through two more PNPs that are also scaled to allow for a large Vbe difference.
Another interesting feature of this circuit is the method by which it creates the bias currents through these two feedback PNP transistors of the bandgap core. The circuit shown in Figure 8 is used to create these currents.
After analysis, one can see that the current generated by this circuit is proportional to absolute temperature (PTAT) voltage divided by R1257. Assuming that the temperature dependence of the polysilicon resistor is much smaller than the bandgap voltage temperature dependence, the output current is also PTAT. This may be used to compensate for the temperature dependence of the feedback PNPs in the bandgap core of Figure 6. A more likely scenario, however, is that this PTAT current, along with appropriate selection of component values in the bandgap core, creates a temperature-dependent characteristic that will compensate for the temperature dependence of the proof mass resonant frequency. Taken all together, this is one of the most complex bandgap circuits we have ever seen.
The overall MEMS market is a bright spot in the semiconductor industry, and inertial sensors represent one of the fastest-growing subsegments. Driven by accelerometer applications like the Apple iPhone and the Nintendo Wii, and by the coming legislation requiring stability-control systems in all vehicles, these devices have moved out of industrial segments and into consumer ubiquity. While most designers might think that the innovation in the MEMS industry lies in the hands of the technology teams, this device shows that these ICs deliver sophisticated designs of their own. And as the competition heats up, we can only expect that it will get even better.
Author informationRandy Torrance leads the Circuit Analysis team for the Technical
Intelligence group at Chipworks. During 22 years in the technology industry he has held senior technical and management positions in the IC design and electronic systems areas. He holds bachelor's and master's degrees in Electrical Engineering from the University of Waterloo.
¡Obtén la mejor experiencia en la web!
Descarga gratis el nuevo Internet Explorer 8