Method of Making an Electrostatic Silicon Accelerometer

The present invention pertains to accelerometers and particularly to solid state accelerometers. More particularly, the invention pertains to electrostatic solid state accelerometers.

The related art includes accelerometers of the piezo electric, silicon piezo resistive and capacitive kinds. However, there has been a need for an advanced-technology, near-chip-size accelerometer with high reliability and long-term accuracy. The present invention is a response to that need. The invention, the electrostatic silicon accelerometer, is a solid state accelerometer having the capabilities of complementing future inertial navigational systems based on either ring laser or fiber optic gyroscopes. The greatest problem plaguing present-day accelerometers is long term drift that is virtually eliminated by the present invention. The invention has exceptional stability in terms of accuracy over extended periods of time, thereby eschewing the typical periodic calibrations necessary for related-art accelerometers. The major factor contributing to such stability is the single-material construction of the accelerometer.

The present invention is a capacitive pickoff accelerometer which uses doped (conductive) silicon for the mechanical parts such as the diaphragm or pendulum, and other parts. These parts have a dual purpose which is performing the mechanical and electrical functions of the accelerometer. The use of doped silicon provides for mechanically stable bonding (such as fusion) and for electrically effective conduction.

The accelerometer is an all-silicon assembly having virtually no creep which is a material trait that is crucial for low bias drift and hysteresis. Mismatched coefficients of thermal expansion among the acceleration sensor capsule materials constitute a dominant factor in bias drift. Related-art silicon accelerometer designs contain material mismatches in two areas involving the metallization of the pendulum and stops to provide capacitor-plate pickoff and balance, and the use of Pyrex for the stops to fascilitate thermal-electric (TE) bonding. The invention has eliminated such material mismatches by exclusively incorporating doped (conductive) silicon throughout its structure. The accelerometer is a monolithic, fully integrated silicon structure created at the wafer level to assure mechanical stability over time and temperature, thereby eliminating thermal and mechanical drift, and material creep. The accelerometer design is symmetrical and flextures of the pendulum or diaphragm are oriented through the pendulum or diaphragm center of gravity to reduce cross-axis sensitivity. The conductive silicon pendulum or diaphragm, and flextures in lieu of metallized silicon eliminate any bimetal effect in these structures. The solid state accelerometer utilizes capacitive pickoff and electrostatic rebalance for low cost and high stability. The pickoff and electrostatic rebalance are effected through common electrodes for enhanced performance and low cost. The associated electronics are placed on a separate integrated circuit chip to maintain high performance and low cost.

FIG. 1 is an exploded, perspective view of the invention.

FIG. 2 is a cutaway view of the invention.

FIGS. 3a-f reveal the steps of the process for making the all silicon accelerometer.

FIG. 4 shows a diaphragm which may be used in lieu of the pendulum as another embodiment of the invention.

FIG. 5 is a schematic of the electrostatic rebalance loop electronics.

FIG. 6 is a diagram of the digitizer electronics.

FIG. 7 is an illustration of the solid state accelerator sensor capsule, housing and electronics.

Sensor capsule 10 of FIG. 1 has three major components which are plate 12, pendulum 14 and plate 16. Pendulum 14 is supported by flexures 18. Wafer 20 forms a rigid frame which holds flexures 18 in place and is a support structure to which plates 12 and 16 are adhered thereby forming capsule 10. Flexures 18 allow for movement of pendulum 14 relative to plates 12 and 16 such that pendulum has the shape of a plate that is approximately parallel to the surfaces of plates 12 and 16. Without any inertial or gravitational force affecting pendulum 14, the neutral position of pendulum is equidistant from plates 12 and 16. An inertial or gravitational force perpendicular to the broadest surface of pendulum 14 causes pendulum 14 to move in a direction toward plate 12 or 16. For instance, if pendulum 14 is closer to plate 12, then it is obviously farther from plate 16; if there is an indication of a force in one direction and if pendulum 14 is closer to plate 16 than plate 12, then, of course, the force is in the other direction. Voltage is applied to pendulum 14 and plates 12 and 16, to electrostatically drive pendulum 14 back to its equidistant position with respect to plates 12 and 16.

Pendulum 14 and plates 12 and 16 are made of a conductive material which is a conductive silicon for the present embodiments. Plates 12 and 16 are composed of doped silicon (p+) and pendulum 14 is composed of a combination of layers of doped silicon (p+ and p- as separate layers), along with coatings of SiO.sub.2 and SiN.sub.3. Separate electrical connections are made to plates 12 and 16, and to pendulum 14, respectively. The varying distances between pendulum 14 and plates 12 and 16 are electrically determined as respective quantities of capacitance. For a force acting on the pendulum 14 in one direction of one plate or the other (12 or 16) increases the electrical capacitance between pendulum 14 and the respective plate, which electronically is converted, through high gain rebalance loop 50 of FIG. 5, into an indication of the amount of force in terms of gravitational units (g's) acting on pendulum 14. Rebalance loop 50 drives pendulum 14 back to its zero or null position tending to keep pendulum 14 stationary, with electrostatic force through voltage being applied across pendulum 14 and plates 12 and 16, respectively. The electrostatic force counteracting the accelerational or gravitational force on pendulum 14, is proportional to the square of the applied voltage. However, electronic linearization of the applied voltage results in a voltage proportional to the electrostatic force and indicative of the accelerational or gravitational force.

The inventive aspect of sensor capsule 10 is that all the elements are constructed with silicon thereby providing the advantages of accuracy, sensitivity, range, stability, miniaturization, power dissipation, ruggedness, and life span.

Typical specifications of solid state accelerometer 10, utilizing pendulum 14 and incorporating the associated electronics as described below, would be: (1) A measuring range from .+-.1 micro g to .+-.40 g's, with a resolution of 1 micro g; (2) A bias magnitude no more 2 milli g's and long term repeatability of 50 micro g's for 10 years duration; (3) A scale factor magnitude of 0.050 volt per g for an analog scale and 1.6 millimeter per second squared per pulse for a digital scale; (4) A scale factor nonlinearity of 10 parts per million and long term repeatability of 50 parts per million; (5) A input axis misalignment magnitude of 4 milliradians and long term repeatability of 80 microradians; (6) A cross-axis sensitivity of 15 micro g's per g squared; (7) A mechanical shock tolerance in all axes of 250 g's for 11 milliseconds; (8) A size and weight of sensor capsule 10 of 2 cubic centimeters and 4 grams, respectively; and (9) Power dissipation of less than 500 milliwatts. The operating temperature range is from -55 to +96 degrees Centigrade. The estimated mean time between failures is greater than 100,000 hours.

Typical construction characteristics of pendulum 14 would include a length of 0.381 centimeter, a width of 0.254 centimeter, a thickness of 0.015 centimeter, a mass of 0.015 gram, a pendulousity of 3.87.times.10.sup.-3 gram-centimeter, and a polar moment of inertial of 1.3.times.10.sup.-3 gram-centimeter.sup.2. Characteristics of flexures 18 include a length of 0.025 centimeter, a width of 0.025 centimeter, a thickness of 3 micrometers, and an elastic restraint of 76.9 dyne-centimeters per radian. Other specifications of the accelerometer 10 system include a nominal gap width between stops 12 and 16 and pendulum 14 of 2 micrometers, a nominal working gap capacitance of 57 picofarads, a pickoff scale factor of 24,900 volts per radian, a nominal electronics gain of 1000 volts per volt, a damping constant of 2.5.times.10.sup.6 dyne-seconds per centimeter, a torquer scale factor of 15.2 dyne-centimeters per volt, and a digitizer scale factor of 17,500 pulses per volt.

A cross-sectional view of sensor capsule 10 is shown in FIG. 2 having a corresponding numerical designations of FIG. 1. Sensor capsule 10 is a monolithic mechanically integrated silicon structure fabricated using state-of-the-art micro-machining techniques. These techniques combine the precision of planar integrated circuit processing with wet and dry (plasma) etching and multiple wafer-to-wafer bonding to produce a three-dimensional micro-machined device. This approach permits structures to be fabricated with sub-micron tolerances while retaining the potential low cost of batch processing techniques.

Device 10 utilizes a combination of silicon fusion bonding techniques and epitaxial silicon films to form a monlithic single crystal silicon structure, with flexures 18 symmetrically located between two halves of proof mass 14. Flexure dimensions are controlled by high accuracy and patterning techniques. The upper 12 and lower 16 electrodes are formed from standard silicon wafers and bonded to the silicon support 20 using wafer-to-wafer bonding with a dielectric interlayer 22 for electrical isolation.

Cross-sectional views in FIGS. 3a-f illustrate the fabrication sequence of making capsule 10. To begin with FIG. 3a, p+ doped layer 24 of epitaxial silicon is deposited on double-sided polished p- doped silicon wafer 26. Epitaxial silicon 24 is patterned with dry etch 28 to define flexures 18 an pendulum 14. Second silicon wafer 30 is fusion bonded to epitaxial silicon layer 24 in FIG. 3b. The two-wafer stack of wafers 26 and 30 is thermally oxidized to provide dielectric interlayer 22 which sets gap width 34 as noted in FIG. 3c. Pattern dielectric interlayer 22 is a growth of SiO.sub.2 and SiN.sub.3. The two-wafer stack is patterned with an etch resistant material 32 deposited on the stack. Wafer 26 is etched at points 36 and 38 using an anisotropic etch in FIG. 3d. Interlayer 22 at area 44 between points 36 an 38 is removed. P+ doped silicon stop wafer 16 is bonded to the two-wafer stack on the side of wafer 26 as shown in FIG. 3e. Resulting gap 34 exists between the two-wafer stack and wafer 16. Etch resistant material 32 is deposited with the desired pattern on the wafer 30 side of the two-wafer stack. Anisotropic etching is used thereby etching at points 40 and 42 to complete proof mass 14 and flexure 18 definition. Interlayer 22 at area 46 between points 40 and 42 is removed. The flexure material between points 38 and 42 is removed so that the inner portion of two-wafer stack may move as pendulum 14 about flexure 18. Then p+ doped silicon stop wafer 12 is fusion bonded to the two-wafer stack on the side of wafer 30 as illustrated in FIG. 3f. At the stage of construction, sensor capsule 10 is completely sealed from detrimental dust and debris, and is ready for dicing, wire bonding and testing.

Pendulum 14 may be substituted with diaphragm 15 as illustrated in FIG. 4. The fabrication process is the same except flexures 19 are provided on four sides of square diaphragm 15. Solid state accelerometer 10 is functionally the same as accelerometer with pendulum 14 except that the diaphragm accelerometer is more rugged though less sensitive. Typical specifications of solid state accelerometer 10, utilizing diaphragm 15, having flexures 19, and incorporating the associated electronics as described below, would be: (1) A measuring range from .+-.20 micro g's to .+-.100 g's, with a resolution of 5 micro g's; (2) A bias magnitude no more 10 milli g's and long term repeatability of 1 milli g for 15 years duration; (3) A scale factor magnitude of 0.20 volt per g for an analog scale and 0.64 millimeter per second squared per pulse for a digital scale; (4) A scale factor nonlinearity of 100 parts per million and long term repeatability of 500 parts per million; (5) A input axis misalignment magnitude of 10 milliradians and long term repeatability of 1 milliradian; (6) A cross-axis sensitivity of 100 micro g's per g squared; (7) A mechanical shock tolerance in all axes of 20,000 g's for 7 milliseconds; (8) A size and weight of sensor capsule 10 of 2 cubic centimeters and 4 grams, respectively; and (9) Power dissipation of about 500 milliwatts. The operating temperature range is from -55 to +96 degrees Centigrade. The estimated mean time between failures is greater than 100,000 hours.

Pickoff of accelerometer is achieved by measuring the capacitance change as pendulum 14 or diaphragm 15 moves towards and away from silicon stops 12 and 16 under external acceleration. Rebalance is achieved by applying electrostatic forces to force pendulum 14 or diaphragm 15 to a null position. Silicon stops 12 and 16 are used, not only for mechanical supports and signal pickoff, but also for rebalance by separating the pickoff and rebalance frequencies. The electronics for accelerometer 10 include analog rebalance loop 50 of FIG. 5 and digitizer loop 60 of FIG. 6. Analog rebalance loop 50 includes sin wave generator 54 connected to differential amplifier 56 and switches 58. The output of amplifier 56 is connected to a capacitance bridge consisting of ultra stable metal-oxide-semiconductor (MOS) capacitors and the variable capacitances of the gaps between pendulum 14 or diaphragm 15 and stops 12 and 16 of accelerator 10. The common connection of the accelerometer capacitances, which is pendulum 14 or diaphragm 15, is connected to ultra-stable voltage reference 52. The output of amplifier 56 also goes to peak detector 66. Differential amplifier 68 has its inputs connected to the output of detector 66 and to voltage reference 52. The output of amplifier 68 is fed back to generator 54. The outputs of the capacitance bridge involve connections to stops 12 and 16 which are capacitively coupled to the inputs of differential amplifier 70. The differential outputs of amplifier 70 go the switches 58, which in turn have outputs connected to the inverting input of differential amplifier 72. Differential amplifier 74 has a bias adjust input and an output connected to the non-inverting input of amplifier 72. The output of amplifier 72 goes to the non-inverting inputs of differential amplifiers 76 and 78. The outputs of amplifiers 76 and 78 are fed back to the inverting and non-inverting inputs, respectively, of amplifier 70. The outputs of amplifiers 76 and 78 are the positive analog scale factor and the negative scale factor voltages, respectively. The analog scale factor (SF) is governed essentially by equation SF=Kd.sup.2 /e.sub.r V.sub.b where d is gap width 34, e.sub.r is the dielectric constant of the gap, V.sub.b is the bias voltage used to linearize the scale factor and K is a constant of the geometry of the accelerometer 10 pendulum 14 or diaphragm 15 in combination with stops 12 and 16.

The output of circuit 50 of FIG. 5 is connected to circuit 60 of FIG. 6. Circuit 60 is a digitizer loop that takes the analog output from amplifier 76 of circuit 50 and converts it to a digital output. The analog signal of circuit 50 goes to summation device 80 and to buffer amplifier 82. The output of summer 80 goes to integrator 84 which in turn provides an output to comparators 86 and 88 of circuit 90. Outputs of comparators 86 and 88 go on to logic circuit 92 which outputs delta voltage signals in a digital compatible format indicating acceleration measurements of capsule 10. A feedback signal from circuit 92 is sent to analog switches and drivers circuit 94. Circuit 94 has a positive precision pulse (PPP) and negative precision pulse (NPP) inputs. The output of circuit 94 goes to summer 80. Various controlling and adjusting inputs include the sample and clamp signals to logic circuit 92, the analog gain adjust to amplifier 82 and the threshold adjust to comparator circuit 90.

FIG. 7 shows the total package of the solid state accelerometer system as fabricated. The support electronics 102 which include analog rebalance loop 50 and digitizer loop 60 may be a nonolithic or hybrid integrated circuit situated in package 100. Sensor capsule 10 and electronics 102 may be fabricated on a monolithic chip if desired. However, the present embodiment having capsule 10 and electronics 102 not on the same monolithic chip is presently less expensive than a monolithic chip for the complete accelerator system including support electronics 102.