EP1314004A1

EP1314004A1 - Method for correcting angular position measurements, using at least two coded tracks

Info

Publication number: EP1314004A1
Application number: EP01971624A
Authority: EP
Inventors: Gunther Haas; Henrik Siegle; David Heisenberg
Original assignee: Robert Bosch GmbH
Current assignee: Robert Bosch GmbH
Priority date: 2000-08-22
Filing date: 2001-08-21
Publication date: 2003-05-28
Also published as: WO2002016881A1; DE10041096A1

Abstract

The invention relates to a method for the computational correction of angular position measurements, using two coded wheels (1a, 1b). The method allows assembly and production errors, in particular, to be retrospectively corrected. The method is essentially characterised in that during an initial calibration, the errors of the input variables, or signals derived therefrom, are determined and stored by means of a reference angle sensor. The errors (Δα(ζ1), Δβ(ζ1)) are filtered using a Fourier analysis. During operation, the filtered errors are deducted from the measured angle values, depending on the rotational angle. This correction can be iteratively repeated, in such a way that the error angle is reduced to a minimum.

Description

Procedure for correcting angle measurements using at least two code tracks

State of the art

The invention is based on a method for correcting angle measurements by means of at least two code tracks with periodic markings, which are fixedly arranged on a common shaft, according to the preamble of the main claim. Methods and devices for measuring angles with code tracks, for example with a magnetic scale, are widely known. For example, DE 198 18 799 C2 proposes a method and a device for measuring angles, in which two magnetic multipole wheels are arranged on a shaft. A fixed sensor is assigned to each multipole wheel, which decodes the magnetic poles of the two multipole wheels and supplies corresponding signals to an evaluation device. The two multipole wheels differ in the number of their magnetic poles. The number of magnetic poles was chosen so that they are prime to half the circumference of one ring compared to the number of magnetic poles on the half circumference of the other ring. The sensor unit has two magnetoresistive sensors which are operated in the saturated state. In addition, a Hall sensor is provided which is assigned to a coded ring which is an odd one Has number of poles on half the circumference of the ring. This arrangement is relatively complex and does not take into account errors that can occur due to misalignment or an eccentricity of the arrangement of the sensors. Furthermore, no correction of pole pitch errors, amplitude errors and errors of the sensor elements themselves is provided.

Advantages of the invention

The method according to the invention for correcting angular measurements with the characterizing features of the main claim has the advantage over the fact that static tolerance and misalignment errors occurring both in production and in the Motage can be subsequently corrected. This is particularly advantageous since the desired accuracy specifications can be set with certainty by the subsequent angle correction. For example, when determining the torque or pitch angle for a motor vehicle steering system, angle determinations with the greatest precision are required, although the torsion angles that occur are relatively small and difficult to measure.

It is considered to be particularly advantageous that with the aid of the proposed correction methods the angles of rotation can be determined with almost any desired precision, so that the method appears to be advantageously suitable for practically all applications.

The measures listed in the dependent claims allow advantageous developments and improvements of the method specified in the main claim. In this context, it is considered advantageous to first use a preliminary angle determination perform a known vernier method and then correct this angle of rotation according to the invention.

It is considered particularly advantageous that the at least two code tracks are designed as periodically recurring magnetic or optical codes. In particular with a magnetoresistive or light-sensitive sensor, the codings can be easily detected and transmitted to an evaluation unit as input signals for further processing. Such units are simple and inexpensive to manufacture and work particularly reliably.

When the code tracks are attached to opposite end faces of a magnet wheel, lying next to one another or distributed around the circumference of the magnet wheel, a simple arrangement results which is relatively easy to produce. Such an arrangement is also insensitive to contamination and thus achieves a high level of reliability.

For evaluating the input variables measured by the sensors, it is considered to be particularly advantageous that correction values of the input variables are first calculated in a calibration mode with the aid of a precise reference angle encoder, wherein the classic vernier method can be used. These are analyzed and stored accordingly, so that the input variables can be corrected depending on the angle of rotation when the device is used later.

It also appears to be particularly favorable that the actual error of the angle of rotation can be almost arbitrarily reduced by repeatedly checking the input variables with regard to their measurement inaccuracy. In practice, a certain accuracy requirement is specified, which is with With the help of this method, it can advantageously be adhered to or even undercut.

In addition, with the aid of the known modified vernier method, it is possible to further reduce the error of the angle of rotation and thus to determine the actual angle of rotation even more precisely. In this case too, the error can advantageously be reduced further by repeatedly using the modified vernier method.

A particularly favorable solution is also seen in correcting the errors of the input variables dependent on the angle of rotation with the aid of the Fourier transformation. The Fourier components, which were selected according to their amplitude and frequency, are advantageously stored together with the phase in one or two tables. With the help of the tables, the error function that is required for the angle correction can then be calculated.

The iterative repetition can also minimize the error in this case.

A particularly favorable application can be provided for determining the steering angle of a motor vehicle. Here the handlebar can be equipped with the multipole wheels and the sensors. In particular when additionally using a torsion element which is arranged between the two multipole wheels, an angle of rotation of the two multipole wheels can additionally be determined, so that a torque acting on the handlebar can also be calculated precisely.

drawing Two embodiments of the invention are shown in the drawing and explained in more detail in the following description.

Figure 1 shows a first embodiment of the invention with two multipole wheels, Figure 2 shows a second embodiment of the invention with a multipole wheel and two code tracks, Figure 3 shows a first diagram with two

Error angle curves, FIG. 4 shows a first flow diagram, FIG. 5 shows a second flow diagram, FIG. 6 shows a further error diagram and FIG. 7 shows two diagrams for Fourier analysis.

Description of the embodiments

Figure 1 shows a first embodiment with two code wheels la, lb, which are fixedly arranged on a common shaft 3. In a vernier evaluation, each code wheel carries at least two code tracks 6a, 6b with uniformly distributed codes 2. The code tracks 6a, 6b are preferably arranged on an end face of the code wheel 1a, 1b. For example, 48, 50 or other numbers of code pairs 2 can be selected. In an alternative embodiment, the code track la, lb can also be arranged on the circumference with a corresponding design of the code wheel la, lb. A sensor 5 is assigned to each code wheel 1 a, 1 b in such a way that it detects the alternating codes 2 of the rotating shaft 3 and supplies a corresponding input signal for a schematically illustrated control 10. In a preferred embodiment, the codes 2 are magnetized alternately with north and south poles and the sensors 5 are magnetoresistive, so that when rotating the shaft 3 each sensor 5 emits a sine and cosine function as input signals for the evaluation unit at its two outputs. The sensors 5 are preferably arranged on a printed circuit board 4, which also receives the evaluation unit 10. In order to achieve the smallest possible design, recesses 8 are provided, into which the code wheels la, lb can plunge more or less deeply. To protect against contamination and damage, the entire unit is surrounded by an appropriately stable and tight casing. A preferred application is seen for steering angle measurement in a motor vehicle in which this arrangement is attached to the steering column.

If, in an alternative embodiment, a torsion element with known stiffness is arranged between the two pole wheels la, lb, in addition to measuring the angle of rotation φ, a torque acting on the shaft 3, for example a steering shaft of a vehicle, can also be generated due to the angular displacement between the two multipole wheels la, lb be determined.

Instead of the magnetic embodiment, it is alternatively also possible to use optical codings 2 in the known embodiments, for example bar codes, color markings, impressions, which are scanned by corresponding optical sensors 5. The optical code tracks 6a, 6b can, for example, also be applied directly to the shaft 3. They scan the optical codes 2 and deliver corresponding input signals to the evaluation unit 10, which carries out the correction of the angle of rotation analogously according to the inventive method.

FIG. 2 shows a second exemplary embodiment of a multipole wheel 1c with a simplified embodiment, in which two code tracks with the on the two end faces Pole pairs βa, 6b are arranged. Accordingly, the two sensors 5 are arranged on the top and bottom of the multipole wheel 1c and fixed by the circuit board 4. With this device, only an angle of rotation, but not a torque, can be measured.

The further description and evaluation of the signals corresponds to the exemplary embodiment according to FIG. 1.

The evaluation method according to the invention for the angle signals supplied by the sensors 5 is explained with reference to FIGS. 3 to 7. When the shaft 3 rotates, the angle-dependent sine or cosine voltages arise at the output of the two sensors 5 and are supplied as input variables to the evaluation unit 10. By forming the arctangent function from the sine and cosine values, the evaluation unit 10 forms the periodic angles, the so-called saw teeth α (φ), β (φ), which are error-prone for a variety of reasons and are corrected using the method according to the invention. For example, pole pitch errors, amplitude errors, errors in the sensor elements themselves, or errors which arise from eccentricities of the pole wheels are corrected. The actual angle of rotation φ is then determined.

According to the invention, it is therefore proposed that an error .DELTA..alpha. Or .DELTA..beta. Of each pole wheel is recorded and stored as a function of the rotation angle .phi. For this purpose, a calibration device with a high-precision angle encoder is required, with which the shaft 3 is rotated up to 360 ° in a sufficient number of steps and the input data (actual angle, measured angle of rotation φ) are recorded step by step and the error angle Δα or Δß of the step two code wheels la, lb is determined. Relating to one Rotation with an angle of up to 360 ° for the shaft 3 results in a sinusoidal error curve (error function) for the angle measurement, as shown in FIG. 3. The error curves for an eccentricity of the magnetic tracks and the code wheel itself were shown here as examples. The upper error curve shows the error distribution on a code wheel with 48 pole pairs and the lower curve on a pole wheel with 50 pole pairs. The dashed curve is obtained by arithmetic adjustment with a sine function. As shown by way of example in FIG. 3, the maximum error is approximately + 0.25 °. With the help of these curves, the input variables α, ß can now be corrected for every angle of rotation φ. This makes it easy to correct eccentricity errors. As can be seen in FIG. 3, the period corresponds precisely to the circumference of the code wheels la, lb.

FIG. 4 first shows a flow chart for the calibration of the sensors 5 before the start-up. The sensor 5 is calibrated in position 20. The reference angle transmitter 19, which works very precisely, is used for this purpose. At each step, the reference angle α _re f (φ) is subtracted from the measured angle α (φ), ß (φ) of the two tracks of a code wheel in order to obtain the error angle Δα (φ) and Δß (φ), α and ß are the angles of the respective code track. The error curves Δα (φ), Δß (φ) (position 21) are analyzed with a Fourier transformation (item 18) and the relevant components (amplitude, frequency and phase) are stored in one or two tables (item 17) so that they are available for later operation.

In the right part of the flowchart in FIG. 4, it is shown how, after the calibration of the sensors 5, the evaluation of the input variables or the determination of the angle of rotation φ he follows. In position 22 the sensor elements 5 are activated and in position 23 measure the input variables α1, βl in a first step. These input variables α1, β1 are then corrected for their respective errors Δα (φ) and Δß (φ). For this purpose, however, preliminary knowledge of the angle of rotation φ is necessary in every situation, which is initially determined as a preliminary angle φl. Since the provisional rotation angle φl is still relatively inaccurate, the classic vernier method is used to determine it due to the greater fault tolerance according to item 24. This provides the provisional rotation angle φl according to position 25. The modified vernier method is not able to do this if the errors of the input variables αl, ßl exceed a certain error value. With this angle φl, the errors Δα (φl) and Δß (φl) are calculated according to the table (item 179) and subtracted from the input variables αl, ßl (position 27) The corrected angles are then called α2 or β2 This step is repeated in accordance with the jump back to position 24 until the input variables have the accuracy required for the modified vernier method, so a certain number of repetitions is specified in position 31. In many In some cases, a single calculation of φ with the classic vernier method is sufficient: After reaching the error limit, the modified vernier method is used in position 28, which in turn has a strong error-reducing effect and outputs the angle of rotation φ2 according to positions 29, 30.

If a further improvement in the measurement accuracy is desired, the error correction according to position 26 is repeated in accordance with the specification in position 33, so that the errors of the input variables α1, βl are minimized almost arbitrarily can be. This eliminates systematic errors.

However, if the accuracy requirements are so low that the modified vernier principle can be dispensed with, alternatively iterative execution of the classic vernier method is also sufficient to achieve an error reduction. In addition, it should be pointed out that the classic vernier method or modified vernier method is known, for example, from DE 195 06 938 AI.

FIGS. 5 to 7 propose an alternative embodiment of the invention for optimal correction of the errors Δα (φ), Δß (φ) for the input variables α (φ) and ß (φ). This correction method described below is particularly suitable for correcting long-wave disturbances due to the eccentricity of the sensors 5, the harmonics in the sensor signal and pole pitch errors of the multipole wheels la, lb. This correction method essentially exploits the properties of a Fourier transformation.

As already explained, in connection with the calibration of the sensors 5 the error signal Δα (φ) or Δß (φ) depending on the angle of rotation is determined. For example, the multipole wheels la, lb are rotated with the shaft 3 in steps of 0.1 ° -2 ° over an entire rotation angle of 360 ° and the signals α (φ) and β (φ) from the sensors 5 (magnetic field sensors) are recorded. An absolute angle of rotation φl is then determined using the classic vernier method. The Fourier spectrum of the error is then calculated for the quantities α (φ) and ß (φ). These periodic quantities α (φ) and ß (φ) are also called saw teeth. The correction method of the angular error with the iterative Fourier filtering is now explained on the basis of a second flow diagram according to the figure. According to position 41, the input signals α (φ) and β (φ) are calculated by the evaluation unit 10. With the two saw teeth α (φ) and ß (φ) (position 42) the classic vernier method is used in position 43 and an absolute angle of rotation φl is calculated from this (position 44). It was found during the calibration that the step size to be selected for determining the angle of rotation depends on the required accuracy of the

Absolute angle φ is dependent. The higher the required accuracy, the higher the frequency in 1 / ° of those Fourier components that have to be subtracted from the sawtooth. The maximum frequency of the spectrum is the inverse step size. Experience has shown that the last band in the Fourier spectrum with a relevant amplitude was around 0.6 / °, which would correspond to a step size of 1.66 ° (cf. FIG. 7). The subtraction of the long-wave Fourier components from the value of the sawtooth is carried out in position 45 and then the modified vernier method is then used in position 46, so that the absolute angle of rotation φ2 is obtained in position 47.

To determine the absolute angle of rotation φ, the table (item 17) is divided into two tables, one for the low-frequency Fourier components up to approx. 0.06 1 / ° and one for the higher-frequency Fourier components 0.06 1 / ° to 0.6 1 / °. The tables show the amplitudes, frequencies and phases of the Fourier components, the amplitude of which exceeds a certain limit. Basically, the limit depends on the

Precision requirements for the angle sensor specified. This step takes place in position 48. In order to further improve the accuracy of the angular error, iteratively according to Position 49 reapplied the modified vernier method and in position 50 a new absolute angle of rotation φ was calculated. This iterative correction procedure for angle measurements can be used until a predetermined limit value for accuracy is reached.

The mode of operation of the correction method is illustrated on the basis of a third error diagram according to the figure. As already explained above, the angle error Δα (φ) or Δß (φ) is plotted over the angle of rotation φ. Curve 1 shows the angular error of a single sawtooth α (φ) or ß (φ). Curve 2 shows the error angle after correction using the classic vernier method in accordance with position 43 (FIG. 6). Curve 3 shows the correction function with the modified vernier method and curve 4 shows the angular error for the saw teeth after deduction of the correction function. A further improvement in the angular error is achieved with curve 5, in which the angular error for the sawtooth was calculated iteratively after two correction functions.

FIG. 7 shows two diagrams, the phase of the Fourier transformation of the error signal being plotted in the upper diagram and the corresponding amplitudes being plotted in the lower diagram.

The lower diagram in particular shows which errors can be identified at which frequency. For example, the amplitude is particularly high in the frequency spectrum at a low frequency in the range of 0 1 / °. This is an indication of an eccentricity error. At around 0.25 1 / °, the high amplitude indicates a pole length error, etc. With the Fourier spectrum method, the type of error can also be identified in a simple manner and targeted remedial measures can be initiated.

Claims

Expectations

1. Method for correcting angle measurements by means of at least two code tracks (6a, 6b) with periodic markings (2) which are fixedly arranged on a common shaft (3), the two code tracks (6a, 6b) differing by at least one pair of markings , and when the shaft (3) is assigned, sensors (5) recognize the markings (2) and send corresponding analog periodic electrical input signals to an evaluation unit (10), characterized in that each code track (la, lb) has only one sensor ( 5) that the error angles (Δα (φ), Δß (φ)) of each code track (6a, 6b) are assigned for the initial calibration with the aid of a reference sensor when the shaft (3) is rotated step by step up to 360 ° ) are determined and stored as a function of the angle of rotation (φ) and that during later operation the input signals measured by the sensors (5) or quantities derived therefrom for determining an actual rotation angle (φ _n) can be corrected using the stored errors (ΔαC, Δß (φ)).

2. The method according to claim 1, characterized in that the measured angle of rotation (φ) is determined using a vernier method.

3. The method according to claim 1 or 2, characterized in that the at least two code tracks (6a, 6b) are formed as periodically recurring magnetic or optical codes (2) and applied to the shaft (3) or at least one pole wheel (la, lb) are.

4. The method according to any one of the preceding claims, characterized in that the at least two code tracks (6a, 6b) on opposite end faces of the magnet wheel (lc) are arranged side by side or on the circumference of the magnet wheel (lc).

5. The method according to any one of the preceding claims, characterized in that the classic vernier method is used to determine the preliminary angle of rotation (φl).

6. The method according to claim 5, characterized in that the associated stored error angles (Δα (φl), Δß (φl)) are added to the provisional angle of rotation (φl).

7. The method according to any one of claims 5 or 6, characterized in that a repeated error determination for the angle of rotation (φl) is carried out until a predetermined accuracy limit is reached.

8. The method according to claim 7, characterized in that in a subsequent step the modified vernier method is used to calculate the actual angle of rotation (φ).

9. The method according to claim 7 or 8, characterized in that the modified vernier method is used repeatedly to reduce the errors of the input variables.

10.The method according to any one of the preceding claims, characterized in that the errors of the angle-dependent input variables ((Δα (φl), Δß (φl)) are corrected with the aid of Fourier filtering.

11. The method according to claim 10, characterized in that a table is created for the low-frequency and higher-frequency Fourier components, in which the amplitudes, frequencies and phases of the Fourier components are stored, the amplitudes of which exceed a limit value specified by the accuracy requirement.

12.The method according to claim 10 or 11, characterized in that the error reduction can be carried out repeatedly.

13.The method according to any one of the preceding claims for use in determining the steering angle in a motor vehicle.

14. The method according to claim 13, characterized in that a torsion element with known torsional rigidity is inserted between the multipole wheels (la, lb) and that the angle of rotation between the two multipole wheels (la, lb) is determined for torque determination.