DE9304629U1 - Non-contact magnetic torque sensor - Google Patents

Description

&igr; Description

All previously known non-contact torque sensors, which completely or largely enclose the shaft, have the problem that the change in the permeability of the layer on the shaft due to the torque causes a significantly smaller change in the flux density in the air gap or inductance than the fluctuation in the air gap width due to the eccentricity of the shaft with changing torque.
The solution to the problem proposed here and exemplified in the description is particularly suitable for implementation as a semiconductor-based micromechanical component, since the analog and digital sensor electronics, which compensate for the fluctuation of the air gap width and the sensitivity drift of the magnetic field sensors, can be integrated on one chip together with the coils, the yoke and the magnetic field sensors.

The following description is divided into three parts. The first describes the technological steps for implementing the integrated micromechanical torque sensor. The second part deals with the compensation of the fluctuating air gap width. The third part explains the compensation of the sensitivity drift of the magnetic field sensors.

Description, Part 1

The micromechanical torque sensor according to claim 3 can be realized on a double-sided, possibly also single-sided polished silicon wafer.

The wafer must have a crystal orientation in the <100> direction and can be either weakly p- or n-doped (20 &mdash; 30Clcm) .

A cross-section through the sensor is shown in Figure la, and a top view of the top side of the wafer is shown in Figure lb. An epitaxial layer of approximately 6 - 10 μm thick can be deposited on a polished wafer side, the dopant concentration of which is characterized by a specific resistance of 10 - 30 μm . The dopant of the epitaxial layer must be of the opposite conductivity type to that of the wafer.

Sensor elements based on the Hall effect and other electrical components for signal processing can be manufactured in the epitaxial layer using a MOS or CMOS process.

A bipolar or BICMOS process can also be used directly on a wafer for the sensor or circuit elements.

One or more coils are then applied to the back of the wafer, for example by galvanic deposition of gold.
To do this, the back of the wafer must first have a silicon dioxide layer of at least 100nm thick, produced in the previous standard process. A chromium adhesion layer (thickness: 50nm) and a gold starter layer (thickness: 100nm) are then vapor-deposited onto this.

The wafer side with the coils is covered with a KOH (potassium hydroxide) resistant layer. This layer is structured to create pits. The pits are then created using an anisotropic KOH etching process, using the pn junction epitaxial layer bulk as an electrochemical etch stop. A membrane is then obtained in the thickness of the epitaxial layer.

The masking for the KOH etching must now be removed.

The coil and the pit must be electrically passivated, e.g. with silicon dioxide.

A ferromagnetic layer, e.g. NiFe (thickness: 50 - 100 nm) is sputtered onto the silicon dioxide as a starting layer for the electroplating. A photolithography step is then carried out to structure the yoke(s). The yoke material (NiFe) can now be electroplated (thickness: 8 - 100 nm ). The photoresist and the exposed starting layer are removed.

Description, Part 2

A commercially available SD field calculation program is used to dimension the sensor as a whole and to appropriately place the core field and stray field sensors. The program calculates the field curves for common-mode and differential mode excitation.

In the case of common-mode excitation, the flux density lines intersect the symmetry plane YY' perpendicularly, as shown schematically in Figure 2a. The field progression in the case of differential-mode excitation is shown schematically in Figure 2b.
Of particular interest is the component of the flux density which cuts perpendicularly through the interface between the epi layer and the air gap, because it serves to define the area of the so-called homogeneous field distribution and the inhomogeneous field distribution. Below the poles of the yoke, the field is almost homogeneous, ie the flux density lines run parallel, in the edge area next to the poles the field is inhomogeneous. An interconnection consisting of two complementary split-drain MOS field-effect transistors (MAGFET) is used as a flux density sensor. The MAGFET bridges are only sensitive to field components perpendicular to the chip plane.

With the help of the 3D magnetic field simulation, it is possible to position the stray field sensor in such a way that the quotient of the stray field density measured value to the core field density measured value is independent of the applied torque over the entire measuring range of the torque, up to a deviation of 0.05%. If the evaluation of the measurement results is carried out using a digital computer, the placement of the stray field sensors is not critical.
40

The results or the evaluation of the 3D field simulation of the sensor shown here as an example in the description are shown in Figures 3a to 3f.

From Figure 3c it can be seen that the ratio of the stray flux density Bs to the core flux density βχ is almost independent of the permeability _μτ of the layer on the shaft, ie independent of the torque, and depends only on the air gap width d .
For sensor signal processing, the measured voltages Uk and Us must be divided by the sensitivities of the MAGFET bridges Sk (core field) and Ss (leakage field). From Figure 3d, the following equation is derived for the quotient BsJBk in the measuring range of interest:

_U1

B ₃ Ss < , _m

^{= = r+rd (1)}

Sk

with the system constants r\ and r ₂ .

With the simulation results from Figure 3a and Figure 3b, an equation for Bk with the system constants 7 ^- 3, r ₄ , r ₅ and r ₆ can be set up:

^ = r3 + ^ + r ₆ .d+2 (2)

Sk μ΄α α

In order to determine the relative permeability _μτ and thus indirectly the torque, one must measure Bs and Bk , divide the measured values Us and Uk by the sensitivities of the MAGFET bridges, insert these into equation (1), solve equation (1) for d , insert the air gap width d into equation (2) and solve equation (2) for _μτ .

The procedure for calculating µ _Γ is shown in the block diagram in Figure 4. It can be seen that the algorithm is suitable for implementation in an analog computing circuit.

The accuracy of the calculation of µ, _τ and d depends on the offset and the sensitivity drift of the MAGFET bridges. The zero point offset and the low frequency noise (eg 1// noise) of the MAGFET sensor bridges can be eliminated by periodically switching the magnetic field on and off and using the well-known autozero technique. The sensitivity drift can be continuously compensated by applying a so-called self-calibration during operation.

Description, Part 3

Magnetic field sensors whose operating principle is based on the Hall effect are characterized by a particularly good linearity. One disadvantage, however, is the sensitivity drift, which is caused by temperature fluctuations, among other things. Therefore, a method must be found that enables the sensitivity drift of the sensor elements to be automatically eliminated. First, the basic considerations that led to the invention according to claim 2 are explained.

Equations (1) and (2) represent a system of equations with the two unknowns d and _μτ , which enables the unambiguous determination of d and _μτ if Sk and Ss are known. If the sensitivity Ss and Sk are also assumed to be unknown, at least two further equations with the unknowns Ss, Sk, d and _μτ are required. This requirement is met by using one and the same core field or stray field sensor in different sensor geometries (field profiles), but with the same _μτ and the same d.

In the invention, this requirement for different sensor geometries (field profiles) is met by changing the qualitative and quantitative magnetic field profile by operating the two coils or the series-connected partial coils of the coil with center tap alternately in common-mode and differential-mode. The direction of current flow, the basic arrangement of the core and stray field sensors and the roughly schematically indicated course of the flux lines are shown in Figure 2a for common-mode operation and in Figure 2b for differential-mode operation.

The two different courses of the magnetic flux density lines resulting from the two operating modes imply two different sensor geometries with the same µ, _τ and d.

We start with a total of 4 equations with the uniquely determinable unknowns /i _r , d, Ss and Sk :

&rgr; _ Uk ₃ I _ &ggr; { &igr;&igr;&igr; \ /&ogr;&khgr;

°Kgl &mdash; ~~~7, — /l [µg] _&tgr; , U, Kl...K _n ) [&oacgr;

JK

Us ₃ I

Bk ₃₃ = -7;&mdash; = j3{Mr,d,ki...k _n ) (5)

JK

£>S ₃₃ &mdash; ~5 = JA {μ&tgr;, «j Ki-..K _n ) (&thetas;]

JS

The indices gl and gg stand for common mode and differential mode operation respectively. k\ to k _n are system constants that are constant during operation of the torque sensor.

Equations (3) and (4) by themselves allow the unambiguous determination of μ _τ and d if Sk and Ss are known, as was already shown in part 2 of the description. The same applies to equations (5) and (6). Equations (3) and (5) as well as (4) and (6) are pairwise independent of each other so that if μ _τ and d were known, Sk and Ss could be determined.
Overall, a system of equations is available that not only allows the calculation of the torque by means of the change in the relative permeability μ _τ , but also enables the determination of the air gap width d and the sensitivities Sk and Ss of the flux density sensors.

The micromechanical sensor presented in this description serves as an example for the calculation of the four unknowns.

The starting point is equations (1) and (2), which are set up both for common-mode operation with the system constants r _n to r^e and for differential-mode operation with the system constants r ₂₁ to r ₂ ß.

Us ₃ ,

S _K

Uk ₉ I _ , !il , . , ^r i6 S _K μ _&tgr; d

Usnn

(8th)

^Ss "2i + T ₂₂ -d (9)

aa

Sk
Us ₉₃ r ₂₄ r ₂₆

-77&mdash; = »"31 + +T ₂₅ -C?+-- (10)

Ss μ _δ α

These equations can be solved for the four unknowns _{μ τ} , d, Ss and Sk . For example, if equations (7) and (9) are solved for the quotient Ss/Sk and eliminated by equating (7) and (9), an equation for the distance d is obtained:

j_

^U Sgg

U _Kg i Uk ₉₃

&ggr; — r ₁₂ - &mdash; _r22

USgl Us ₉₉

From equations (8) and (10) the unknown Sk can be eliminated and an equation for the permeability μ _τ is obtained, with the distance d as a variable: 35

&ldquor; Uk ₉ IT ₂ A Uk ₃₃ Vu , *

/X ₇ - &mdash; j \ ^&iacgr;&Dgr; )

Uxggr-13 &mdash; Uk ₉ IT23 + d [Uk ₃₃ Ti ₅ - Uk ₉ I^s] + -, [Uk ₃₃ Tw &mdash;

If one inserts the equation for the distance (11) into equation (12), one obtains a unique relationship for the permeability μ _&tgr; .

This example shows that the calculation of permeability from the system of four equations is analytically possible and that no numerical determination of zeros is required. Therefore, the algorithm can be implemented as an analog calculation circuit.

The description shows how an integrated micromechanical torque sensor can be implemented:

&bull; By using stray field sensors, the problem of fluctuations in the gap width between pole and shaft is solved.

&bull; The zero point offset and the low frequency noise of the magnetic field sensors is eliminated by periodically switching the magnetic field on/off and using the well-known autozero technology.

&bull; The operation of two field coils in common and differential mode causes a qualitative change in the magnetic field, whereby the sensitivity of the magnetic field sensors can be calibrated.

· The stray field measurement and the calibration of the sensitivity of the magnetic field sensors enable an accurate determination of the torque.

There follow five pages of drawings:

Figure la: Cross section through the micromechanical torque sensor.

Figure Ib: Top view of the top with pits, yoke and coils underneath.

Figure 2a: Principle of operation in common mode

Figure 2b: Operating principle in push-pull operation

Figure 3a: Characteristic map of the core flux density with common mode excitation (β&khgr; ₃ &igr; = /(μgr; _&tgr; ,&aacgr;)) Figure 3b: Characteristic map of the stray flux density with common mode excitation (Bs _a i = /(μgr; _&tgr; ,&aacgr;)) Figure 3c: Change in the relative stray flux density (Bs _g i/ β&khgr; ₉ &igr; = /(μgr; _&tgr; , d)) Figure 3d: Relative flux density over the distance (Bs ₉ i/ Bk ₉ I = f(d)) Figure 3e: Characteristic map of the core flux density with differential mode excitation [Bk ₉₉ = f(ß _r ,d)) Figure 3f: Characteristic map of the stray flux density with differential mode excitation (Bs ₉₉ = /(μgr; _&tgr; ,&aacgr;)) Figure 4: Block diagram for the elimination of the distance from the equation for µ _&tgr;

Claims (1)

&igr; Protection claims
&igr;. Contactless magnetic torque sensor, consisting of at least one magnetic circuit, which is represented by at least one coil, at least one ferromagnetic yoke, as well as a ferromagnetic layer on the shaft or a ferromagnetic shaft with a relative permeability dependent on the applied torque,
characterized in that the torque sensor, consisting of at least one coil, at least one soft magnetic yoke and at least one magnetic field sensor, is manufactured using the methods known in micromechanics, semiconductor technology and thin-film technology as a so-called integrated component on a so-called chip made of semiconductor material, characterized in that the torque sensor is manufactured on a semiconductor chip, with the coil(s) and the soft magnetic yoke being arranged on the back of the chip and the magnetic field sensor(s) being arranged on the front of the chip,
characterized in that the torque sensor is manufactured on a semiconductor chip, whereby the functional principle of the magnetic field sensors is based on the Hall effect and the magnetic field sensors are arranged on one side of the chip using a MOS process known in semiconductor technology together with other electronic components that can be used to operate the torque sensor, characterized in that the shape of the yoke and the position of the poles of the yoke are determined by etching pits into the semiconductor material, characterized in that the field coils below the NiFe yoke are made of gold and are buried in the semiconductor material. Torque sensor according to claim 1 with at least two magnetic field sensors for measuring the magnetic field in the gap between the sensor head and the shaft characterized in that the magnetic field sensors are arranged both in the area of the almost homogeneous field profile and in the area of the inhomogeneous field profile, characterized in that the magnetic field sensors are arranged at at least two locations in the area between the ferromagnetic yoke and the shaft, whereby there is an arbitrary but specific functional relationship between the magnetic field sizes of both locations, which may depend on the torque and the gap width between the poles of the yoke and the shaft, characterized in that at least two coils or one coil with a center tap are used to generate the magnetic field, whereby the qualitative course of the magnetic field can be changed. Torque sensor according to claim 1 and 2 or 1 or 2 , characterized in that additional electronic circuit components for sensor signal processing and for the operation of the sensor are arranged on the semiconductor chip, characterized in that the torque sensor is manufactured together with electronic circuit components for sensor signal processing and for the operation of the sensor as a hybrid module, characterized in that the torque sensor is operated together with an electronic circuit that determines the unknown torque, the unknown distance between the sensor head and the shaft and the unknown sensitivity of the magnetic field sensors during operation of the sensor.