JPH10276065A - Digital filter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To calculate low frequency component signals for input signals in a short time without lowering accuracy by detecting a weighted average by using a multi-level multi-integration(MLMI) method. SOLUTION: In the digital filter processing of a processor 3, to the discretization detection signals (discretization signals) of an inputted cross section curve, the MLMI method is applied, weighting is performed by using a Gaussian function and a waviness curve is obtained. The obtained waviness curve is turned to an average line, a roughness curve is detected from the difference of the discretization detection signals of the cross section curve of a roughness meter 1 inputted through an A/D converter 2 and the average line and the parameter of desired surface roughness is calculated from the roughness curve. Thus, since processing time required for waviness curve calculation is greatly reduced while maintaining the accuracy of the filter waviness curve as the low frequency component signals, the processing time is greatly shortened without lowering the accuracy of this digital filter.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an improvement of a digital filter such as a phase compensation type digital filter for detecting a undulation curve based on a detection signal of an roughness meter when measuring surface roughness.

[0002]

2. Description of the Related Art Conventionally, when measuring the surface roughness of an industrial product, as is standardized in "JIS B0601 Surface Roughness-Definition and Indication (Revised 1994)", for example, by using a roughness meter or the like. From the cross-sectional curve of the measured object, a component of surface roughness shorter than a predetermined wavelength is removed by a phase-compensated low-pass filter to obtain a filtered undulation curve, and a deviation between the filtered undulation curve and the cross-sectional curve is calculated. A roughness curve is obtained (FIGS. 3A and 3B).

When this processing is performed by digital signal processing, the detection signal of the cross-sectional curve of the object to be measured is converted into a digital value, and the discrete value of the cross-sectional curve at each sampling point within the measurement range of the reference length L is obtained. For the detection signal
For example, weighting is performed using a Gaussian function as shown in FIG. 4 within the range of Gaussian half width λ _C around each sampling point, and a weighted average is calculated to obtain a swell value at each sampling point. The filtering undulation curve is obtained by performing the processing for each sampling point.

This operation is equivalent to obtaining a filtered undulation curve without waveform deformation by a low-pass filter having a cutoff value λ _C indicated by a wavelength at which the gain of the amplitude transmission ratio becomes 50% as shown in FIG. doing.

[0005] The filtered undulation curve thus obtained is used as an average line of the roughness curve as shown in FIG. 3 (b), and deviations from the average line at each sampling point to the cross-sectional curve are successively obtained. As a result, a roughness curve without waveform deformation is obtained. Then, based on the roughness curve, the average roughness Ra, or, and calculates each parameter representing the surface roughness of such a maximum value R _MAX of average roughness.

[0006]

However, in the conventional digital filter for measuring roughness, the swell value at a certain sampling point, for example, a sampling point n ₀ = n shown by a point A in FIG. The weighted average is obtained by using the data of the discretized detection signals at all sampling points n ₀ = _{0 to} N within the range of the Gaussian half width λ _C on both sides from the target sampling point n ₀ = n. Seeking. The number of data used for the weighted averaging varies depending on the cutoff value λ _C , sampling interval, and the like, but generally ranges from several hundreds to several thousand points. Therefore, an extremely large amount of calculation is required to determine the swell value at the entire measurement area, that is, at each sampling point within the range of the reference length L, and to form a swell curve from these swell values.

[0007] A filtering program for performing a calculation for obtaining the undulation curve is created, and for example,
Normal commercial 32 bit, clock 16 MHz
Personal computer and MS-DOS (registered trademark) version N88 Japanese BASIC as a programming language
(Registered trademark), the calculation time of 30 minutes to 2 hours or more is required, and it takes more time to complete the surface roughness measurement. There is no problem.

In order to avoid this problem, the sampling number of the discretized detection signal of the sectional curve is thinned out, or
Although a method such as reducing the Gaussian half width λ _C is considered,
When this method is applied, it is possible to shorten the calculation time, but at the same time, there is a problem that accuracy is lowered, and it is not possible to effectively reduce the time.

In particular, when measuring the roughness of a plane, that is, two-dimensional roughness, the number of calculations further increases, the calculation time increases drastically, and the surface roughness is determined. There is an unsolved problem that it takes an enormous amount of time to obtain each desired parameter.

Therefore, the present invention has been made in view of the above-mentioned conventional unsolved problem, and is a digital filter capable of calculating a low-frequency component signal with respect to an input signal in a short time without reducing accuracy. It is intended to provide.

[0011]

In order to achieve the above object, a digital filter according to the present invention detects a weighted average of a plurality of input discrete signals over a predetermined range before and after each discrete signal. In a digital filter adapted to detect a filtered discrete signal, the weighted average is detected using a multi-level multi-integration method.

[0012]

Embodiments of the present invention will be described below. FIG. 1 is a schematic configuration diagram showing an example of an embodiment of a roughness measuring device to which a digital filter according to the present invention is applied.

In FIG. 1, reference numeral 1 denotes a reflection type, contact type, etc. roughness meter for measuring the unevenness of the surface of an object X to be measured. A measurement value of the roughness meter 1, that is, a cross-section curve detection signal corresponding to the shape of the surface of the measurement target is converted into a digital signal by the A / D converter 2 and input to the processing device 3. The processing device 3 performs digital filter processing based on the input discretized detection signal of the cross-sectional curve.

In the digital filter processing of the processing device 3,
First, an MLMI method (Multi-Level M) is applied to the input discretized detection signal (discrete signal) of the cross-sectional curve.
By applying a multi-integration method, weighting is performed using a Gaussian function as shown in FIG. This undulation curve corresponds to the output signal of the digital filter as a low frequency component signal.

The obtained undulation curve is defined as an average line,
A roughness curve is detected from a difference between the discretized detection signal of the cross-section curve of the roughness meter 1 input via the A / D converter 2 and the average line, and a desired surface roughness parameter is determined from the roughness curve. Is calculated.

Here, the MLMI method is an algorithm for shortening a calculation time without reducing accuracy.
As an algorithm for speeding up the calculation of the amount of elastic deformation, “Ph.D. Thesis, Univ.
e (1991); Venner, C.E. H. "It is described in.

Therefore, according to the above embodiment, the processing time required for calculating the undulation curve can be greatly reduced while maintaining the accuracy of the undulation curve as a low-frequency component signal. The processing time can be significantly reduced without lowering the accuracy of the processing.

Further, when the digital filter according to the present invention is used, the time required to calculate the filtered waviness curve of the cross-sectional curve is greatly reduced, so that it is extremely effective when analyzing two-dimensional roughness. is there.

In the above embodiment, a case has been described in which the digital filter of the present invention is applied to a phase compensation digital filter for measuring surface roughness.
The present invention is not limited to this. For example, if a low-pass filter process is performed on a vibration waveform as shown in FIG.
It can be applied arbitrarily.

Further, in the above-described embodiment, the case where the present invention is applied to a digital filter for extracting a low-frequency component signal has been described. However, the present invention is not limited to this. The present invention can also be applied to a high-pass digital filter that extracts a component signal.

[0021]

Embodiments of the present invention will be described below. Table 1
Is the cut-off value λ _C = 0.08, 0.25, 0.8
Calculation time when filtering is performed for the three types of [mm] using the conventional undulation curve calculation method using the phase-compensated low-pass digital filter and the undulation curve calculation method using the digital filter to which the MLMI method according to the present invention is applied. And the calculation accuracy.

At this time, the Gaussian half width is 1.0 × λ _C , the sampling interval is 0.5 μm, and the cross-sectional shape measurement length is 4
[Mm], which is a commercially available personal computer of about 32 bits and a clock of about 16 MHz, which is processed by a compiled BASIC.

The calculation accuracy is based on the undulation curve obtained by the conventional undulation curve calculation method.
Deviation of the undulation curve by the LMI method application method, that is,
And the range of error is the square root R _MS mean square value.

[0024]

[Table 1]

As shown in Table 1, the calculation time for calculating the undulation curve is reduced to 1/40 to 1/160, which is within a practically required time range. The calculation accuracy at this time is as follows: when the cutoff values λ _C = 0.8 and 0.25, a roughness meter having a resolution of 0.01 [μm] (such as Rank Taylor Hobson Surtronic 3+) is used. Is sufficient, and even when the cutoff value is λ _C = 0.08, the error is about 1/10 of the resolution, so that it can be used sufficiently. When the cut-off value λ _{C is} 0.08, the calculation speed is reduced by 1/40. At this time, even in the conventional swell value calculation method, the calculation time is shorter than other cut-off values. Since it is originally short, a sufficient effect can be obtained even with such a shortening rate.

Therefore, as shown in Table 1, both the calculation accuracy and the calculation time can be satisfied. Therefore,
The time required to calculate the waviness curve can be reduced without lowering the calculation accuracy, so that the time required to calculate the roughness curve and analyze the surface roughness can be significantly reduced. In addition, even when analyzing two-dimensional roughness, that is, when obtaining planar roughness, the processing time can be significantly reduced as compared with the related art.

[0027]

As described above, according to the digital filter of the present invention, when obtaining the weighted average, M
By applying the LMI method, the processing time can be significantly reduced without lowering the accuracy of the digital filter.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram showing an example of an embodiment of a roughness measuring device to which a digital filter according to the present invention is applied.

FIG. 2 is an explanatory diagram illustrating a method of detecting a swell value.

FIG. 3 is an explanatory diagram illustrating a method for detecting a roughness curve.

FIG. 4 shows a weight function applied to the filter processing.

FIG. 5 is an example of a transfer characteristic of a digital low-pass filter.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Roughness meter 2 A / D converter 3 Processing device

Claims (1)

[Claims] 1. A digital filter for detecting a filtered discrete signal by detecting a weighted average of a plurality of input discrete signals over a predetermined range before and after each discrete signal, wherein the weighted average is To multi-level
A digital filter characterized in that detection is performed using a multi-integration method.