This invention relates to servo control mechanisms and methods of operating servo controlled mechanisms.
The present invention relates to the control of servo loops and, in the preferred embodiment, the simultaneous rapid open and closed loop control measurements obtained using a binary pseudo-random sequence.
The use of a binary pseudo-random sequence to test an electrical or mechanical system for the purpose of computing its power spectrum by discrete fast Fourier transform (DFT) techniques is well known, for example, as described in U.S. Pat. No. 4,093,988. In accordance with U.S. Pat. No. 4,093,988, the properties of binary pseudo-random noise and the discrete fast Fourier transforms are utilized to, in effect, simultaneously inject a closely spaced set of equal amplitude sign waves over a prescribed frequency range into the system under test and to separate the response of the output. For its discussion of binary pseudo-random sequences and filters therefore, and discrete fast Fourier transform techniques employed therein, the U.S. Pat. No. 4,093,988 patent is incorporated herein by reference. Still, the '988 patent does not describe the use of binary pseudo-random sequences within feedback devices, for simultaneous rapid open and closed loop bode plot measurement in servo loops, or in general, any loop calibration application.
Within the context of servo loop control, existing techniques to generate a bode plot measurement inject a swept sinusoid, random white noise, or a chirp into the system under test in order to obtain the bode plot measurements. A method and apparatus for tuning motion control system parameters using random noise excitation, for example, is described in Yutkowitz, U.S. Pat. No. 6,259,221. In the '221 patent, the system comprises a random noise signal generator in communication with the servo controller, and a frequency response analyzer in communication with the servo controller. The random noise signal generator is configured to provide a random noise signal to the servo controller during controlled motion of the actuator. The frequency response analyzer is configured to generate frequency responses for the random noise signal. The '221 patent does not specify the type of random noise used beyond the generalities, “white noise,” “pink noise,” or “band limited white noise signals.”
Thus, the previous binary pseudo-random sequence test apparatus, such as U.S. Pat. No. 4,093,988, uses only the output measurement of a system under test and cannot generate an open loop or closed loop bode plot, which would require both input and output of the system under test. Other devices, such as U.S. Pat. No. 6,259,221, describe the calculation of open loop and closed loop responses, but fail to recognize the value of binary pseudo-random sequences used to generate a bode plot.
The present invention injects a binary pseudo-random sequence, using both the input and the output of the system under test to generate bode plots. The invention has the advantage of generating the open and closed loop bode plots while the system under test is operating closed loop. This allows real-time bode plots to be computed for in-situ applications, such as on-orbit applications, while simultaneously maintaining control of the system under test. The advantages of the present invention are significant. The invention produces measurement results approximately 1,000 times faster than a swept sinusoid approach and has superior dynamic range as compared to using random white noise as a test input source.
The discrete fast Fourier transform of each 1,023 point block is then computed in the data analysis section 18. As noted in
In the closed loop condition, G(s) is
Assuming that X(t), Y(t) and Z(t) denote the discrete fast Fourier transforms of a 1,023 point block of the signals test, test1, and test2, respectively, then the open loop bode is computed from the gain and phase of Z/Y by averaging over the ten blocks. The closed loop bode is similarly computed from Z/X or can be computed directly from the quantity Z/Y as G/(1+G) where G equals Z/Y. Of course, numbers greater or less than the ten blocks discussed can be employed in alternative embodiments.
Thus, the test2, test1, and test signals are input into, respectively, A to D converters 15, 16, and 17, the output of which are input to the data analysis block 18 via 16-bit data lines. In the data analysis block 18, digital fast Fourier transforms operations are performed on the test2, test1, and test signals at blocks 19, 20, and 21. The output of the DFT blocks 19-21 are input, as appropriate, to gain and phase blocks 22 and 23 such that the open loop bode result 26 from the computer gain and phase block 22 and average computation block 24 is derived from the signals Z(f) and Y(f), respectively referring to the test1 and test2 transforms. Similarly, the gain and phase bloc 23 and average block 25 yield the closed loop bode result 27 using the Z(f) and X(f) transforms of, respectively, test1 and test.
The programmable band pass filter 12 allows the bode plots to be generated in distinct sub-octave frequency bands to eliminate aliased signal errors in the subsequent DFT processing in blocks 19-21. The distinct sub-octave frequency bands can later be concatenated to produce a continuous bode plot over the entire frequency range of interest. Segmenting into bands also allows different test signal amplitudes to be used for each band, optimizes the signal-to-noise ratio of the measurement, and prevents circuit saturation.
Because each spectral line generated from the 1,023 bit pseudo-random noise sequence is at the exact center of the 1,023 point discrete fast Fourier transform channels, all 1,023 spectral lines arc processed with total orthogonality relative to each other. The DFT guarantees this because all other spectral lines other than the channel center line will coincide with a “zero” of the sinX/X shaped DFT response. This produces a considerably superior measurement over techniques that use random noise generators other than pseudo-random noise sequence synced to a DFT.
The present invention has application in any embedded system bode plot generation system, for example, on in-situ self-testing of any servo control mechanisms. One example in-situ system may be an on-orbit servo control mechanism where control of the servo may be maintained while the servo is under test.
Other applications that may employ the present invention to advantage are test equipment of any sort such as dynamic signal and FFT analyzers. Thus, fast transfer function measurements performed by network analyzers requiring simultaneous measurement of gain and phase will be advantaged by application of the present invention therein.
While the invention has been described in connection with what is presently considered to be the most practical and preferred embodiment, it is to be understood that the invention is not to be limited to the disclosed embodiment, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
The inventions claimed herein were made with support from the United States government, which has certain rights in the inventions claimed.
Number | Name | Date | Kind |
---|---|---|---|
4067060 | Poussart et al. | Jan 1978 | A |
4093988 | Scott | Jun 1978 | A |
4306113 | Morton | Dec 1981 | A |
4527116 | Sorba et al. | Jul 1985 | A |
5406496 | Quinn | Apr 1995 | A |
5446648 | Abramovitch et al. | Aug 1995 | A |
5517594 | Shah et al. | May 1996 | A |
5687733 | McKown | Nov 1997 | A |
5880415 | Colby et al. | Mar 1999 | A |
5909661 | Abramovitch et al. | Jun 1999 | A |
5929400 | Colby et al. | Jul 1999 | A |
6204988 | Codilian et al. | Mar 2001 | B1 |
6259221 | Yutkowitz | Jul 2001 | B1 |
6278898 | Shah | Aug 2001 | B1 |
6281650 | Yutkowitz | Aug 2001 | B1 |
20020019715 | Cohen et al. | Feb 2002 | A1 |
20030176983 | Josselson et al. | Sep 2003 | A1 |
Number | Date | Country |
---|---|---|
1345102 | Sep 2003 | EP |
Number | Date | Country | |
---|---|---|---|
20030176983 A1 | Sep 2003 | US |