Telescope alignment based on the sharpness function of under-sampled images Download: 879次
The alignment of telescopes is essential for their optimal performance[1]. For those with a small field of view (FOV) and undemanding image quality, one can correct only the on-axis field to achieve the best working state of telescopes, ignoring the effects of aberrations in the off-axis fields. But for telescopes with a wide FOV and a high requirement for image quality, when the on-axis field is aligned, the aberrations in the off-axis fields may be increased, such as coma and astigmatism[2]. As a result, the aberrations in both the on-axis and off-axis fields should be corrected.
There are many approaches for telescope alignment. All these methods can be classified into three categories. One is the direct wavefront sensor method (WFS)[3,4], which is most commonly used in the alignment of telescopes, such as the sensitive table method[5] and Shack–Hartmann wavefront sensor method[6]. This kind of method requires additional optical elements to measure the wavefront errors, increasing the complexity of optical systems. The second method is the image-based wavefront sensor method, for example, the phase retrieval (PR) method[7] and phase diversity (PD) method[8]. Compared with the WFS method, this kind of method only requires one or more detectors, without adding other optical components. However, it does entail more computation to estimate the wavefront errors. The third one is the image-based sharpness function method[9], which maximizes the sharpness function of images using an optimization algorithm to correct the misalignments. Compared with the other two kinds of approaches, this method only requires one detector placed in the focal plane to measure the sharpness of the images in real time, and the misalignments are corrected with the optimization algorithm. For the advantages of concise measurement systems and simple calculations, this method has been successfully used by Yang
However, in imaging systems, especially for telescopes with a wide FOV, under-sampled images are common; nowhere is this problem more acute than on the Hubble Space Telescope (HST)[12]. Imperfect optics, finite detector arrays, and finite individual detector sizes all contribute to the non-ideal sampling of the scene information to meet the Nyquist criterion. The resulting images will be degraded by aliasing effects[13]. The under-sampled images addressed in this Letter are caused by finite individual detector sizes. The information missing from under-sampled images makes it difficult to detect faint sources and accurately represent the aberrations of spots. In the past few years, some research has been proposed[14,15]. All these methods are based on image reconstruction[16] to restore the information lost in under-sampled images. However, few efforts have been made to study the effects of under-sampled images on the accuracy of telescope alignment. This Letter focuses on this issue. The telescope is aligned using a stochastic parallel gradient descent (SPGD)[17] algorithm combined with the sharpness function of different sampling images. According to the Nyquist sampling condition, the largest sampling interval of detectors should be equal to or smaller than the optical cutoff period of the imaging system. Otherwise, the image will be under-sampled and aliasing will appear. Assuming the width of a CCD pixel is
According to the Nyquist sampling condition, the spatial frequency of the telescope imaging system should be equal to or smaller than half the detector spatial frequency:
Equation (
In this Letter, both well-sampled and under-sampled images are simulated to correct the misalignments. The under-sampled images with various under-sampling factors are simulated by the down-sampling method. The larger the under-sampling factors are, the smaller the diameter of the images will be. The diameters of the images are calculated by the full width at half-maximum (FWHM). If the FWHM of the image is smaller than 4 pixels, it is an under-sampled image. The telescope used in this Letter is a Cassegrain system with a full FOV of 0.7° and a central wavelength of 0.65 μm. The other parameters are shown in Table
Table 1. Parameters of the Cassegrain Telescope System
|
Table
For a Cassegrain telescope, the spherical aberration is corrected. However, the coma term grows quickly when the FOV is far away from the optical axis. As a result, aberrations in both the on-axis and off-axis fields should be corrected. In this Letter, the telescope is aligned based on three fields of images, including the on-axis field, which is also called 0field and two off-axis fields. The normalized field point coordinates were 0field (0°, 0°), 0.8field (0°,
For large aperture telescope systems, we set the primary mirror as a reference that is being fixed in place. The misalignments of the secondary mirror related to the primary mirror consist of six degrees of freedom, including decentration along the
The decentrations along the
In this Letter, the telescope system is aligned using the SPGD algorithm, which is a model-free iteration control method. The iterative formula of the SPGD algorithm is[18]
The sharpness function used in this Letter is the RMS of the images, defined as[19]
The alignment of the telescope can be divided into two parts. The first step is the on-axis field correction, in which the RMS of the far-field image is employed as the sharpness function to compensate the spherical aberrations. The second step is the all fields correction, including both the on-axis and off-axis fields to compensate the spherical aberrations, coma, and astigmatism simultaneously. In this part, the RMSs of both the on-axis and off-axis field images are calculated, and the average value of the RMSs is used as the sharpness function. The telescope alignment process based on the sharpness of images with the SPGD method is shown in Fig.
The designed images corresponding to different fields that are obtained when the telescope is in designed working conditions and the misaligned images obtained in misaligned working conditions are shown as Fig.
Fig. 3. Designed and misaligned images with different FWHMs. (a) Designed images. (b) Misaligned images.
Figure
Figure
The telescope is aligned based on both the well-sampled and different under-sampled images. In the alignment processes, the step sizes are same with
Fig. 5. Results of the telescope alignment. (a) Results of the on-axis correction. (b) Results of the all-fields correction.
According to Fig.
The changes of the RMS of the image spots during the correction process are shown in Fig.
Fig. 6. Convergence curves of RMS during the correction. (a) RMS changes during the on-axis field correction. (b) RMS changes during the all-fields correction.
Figures
To verify the accuracy of the alignments based on both the well-sampled and under-sampled images, the residual errors and the wavefront residual errors (WREs) after correction are calculated. The results are shown in Table
Table 3. Residual Errors and Wavefront Residual Errors of Optical System Measured After Correction
|
Table
To verify the accuracy of alignment, WREs in the 0field are calculated using a Zernike polynomial before and after correction. The designed working condition of telescope has few aberrations with WREs of
In conclusion, the misalignments of telescopes can be corrected by maximizing the sharpness function of the far-field images. Whether the images are well-sampled or under-sampled, the correction process can converge stably and rapidly, and the images after correction have similar shapes and sizes even when the FWHMs of the images are different. The accuracy of alignment is analyzed by the RREs and WREs. The RREs show that with the reduction of the FWHM, the accuracy of alignment becomes worse, increasing from 9.5% to 19.5%, and the WREs increase from
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
Min Li, Xin Liu, Ang Zhang, Hao Xian. Telescope alignment based on the sharpness function of under-sampled images[J]. Chinese Optics Letters, 2019, 17(6): 061101.