Speckle Interferometry with CMOS Detector

Abstract

In 2022 we carried out an upgrade of the speckle polarimeter (SPP)—the facility instrument of the 2.5-m telescope of the Caucasian Observatory of the SAI MSU. During the overhaul, CMOS Hamamatsu ORCA-Quest qCMOS C15550-20UP was installed as the main detector, some drawbacks of the previous version of the instrument were eliminated. In this paper, we present a description of the instrument, as well as study some features of the CMOS detector and ways to take them into account in speckle interferometric processing. Quantitative comparison of CMOS and EMCCD in the context of speckle interferometry is performed using numerical simulation of the detection process. Speckle interferometric observations of 25 young variable stars are given as an example of astronomical result. It was found that BM And is a binary system with a separation of 273 mas. The variability of the system is dominated by the brightness variations of the main component. A binary system was also found in NSV 16694 (TYC 120-876-1). The separation of this system is 202 mas.

Appendices

APPENDIX A

THE INFLUENCE OF THE ROLLING SHUTTER EFFECT ON THE CONTRAST ESTIMATION

The rolling shutter effect is a consequence of the non-instantaneous sequential reading of detector lines. Even and odd lines are read at the same time, but each pair of such lines takes some time to read, and each next pair of lines is read with a slight delay compared to the previous pair. In the standard and ultra-quiet reading modes, the time spent on a pair of lines is 7.2 microseconds and 172.8 microseconds, respectively.

For example, taking the angular scale equal to $0\overset{\prime\prime}{.}02/$px and the separation between two stars equal to $1^{\prime\prime}$, we get that the image of one component will lag behind from the image of the second one by 4.3 ms in the ultra-quiet reading mode if the stars are oriented parallel to the reading direction. This value coincides in order of magnitude with the atmospheric coherence time. Therefore, it is necessary to simulate the influence of this effect on the efficiency of speckle interferometric contrast estimation.

To do this, we generated 20 series for three different separations between the components by varying the reading time of a line pair. For each series, standard speckle interferometric processing was applied, from which contrast estimates were obtained. The parameters of the model used to generate the series are as follows:

• $\bullet$ Separations between components are $0\overset{\prime\prime}{.}35$, $0\overset{\prime\prime}{.}7$, $1\overset{\prime\prime}{.}05$, the component flux ratio is 0.05. The stars are oriented in the direction of reading.

• $\bullet$ Two phase screens at distances of 0 m and 10 000 m, moving in perpendicular directions.

• $\bullet$ Wind speed is 10 m s${}^{-1}$.

• $\bullet$ Seeing is $1^{\prime\prime}$.

• $\bullet$ No telescope jitter.

• $\bullet$ Zero defocus.

• $\bullet$ Filter with an infinitely narrow bandwidth.

• $\bullet$ The considered wavelength is 822 nm.

• $\bullet$ Exposure time is 22 ms.

• $\bullet$ The interval between frames increases with the increase in the reading time.

• $\bullet$ Telescope aberrations are taken into account.

• $\bullet$ Photon noise and readout noise are taken into account. RMSD of the readout noise is 0.27$e^{-}$. Conversion factor is $0.11e^{-}/\textrm{ADU}$.

• $\bullet$ Frame size is $256\times 256$. Angular scale is $0\overset{\prime\prime}{.}0205$/px.

• $\bullet$ Sampling in time is equal to the time of reading a pair of lines.

• $\bullet$ 1000 frames were generated for each series.

It can be seen from Fig. 26 that the rolling shutter effect clearly affects the speckle contrast estimation. The longer the readout time and the greater the distance between the components, the more the contrast is underestimated (secondary component appears fainter than it is). However, even if we assume a scenario where the separation between the components is $1^{\prime\prime}$ and the ultra-quiet readout mode is used, then the underestimation of the contrast will be only 10$\%$. For larger separations, the contrast can be obtained without using speckle interferometry by approximating the PSF on the averaged image.

Fig. 26

Estimation of contrast from processing of model series at different reading times of a pair of lines and different separations between components. The vertical dash-dotted line corresponds to the detector’s ultra-quiet reading mode.

APPENDIX B

TELESCOPE JITTER

In the course of comparison (see Fig. 20) of real and model power spectra, it turned out that the high-frequency region of the model spectra, other things being equal, is higher than in real observations. We assumed that this excess is due to the fact that the model does not take into account the vibrations of the telescope. We tested this hypothesis in the following way. In good seeing conditions ($\beta<1^{\prime\prime}$) on different dates at different altitudes and azimuths, bright stars were observed with a low exposure and a high frame rate. The value of the electron multiplication was chosen depending on the brightness of the object. Thus, according to our estimates, the frame rate and exposure required to test the hypothesis should be 500 Hz and 0.002 s, respectively.

The following processing was applied to the obtained series. Cross-correlation of each ($i+1$)-th frame with $i$-th was performed. From here we got the offset of each next frame relative to the previous one, that is, the dependence of the offset on time. Next, filtering was performed by moving average with a window size of 0.1 s, that is, by 50 measurements. Then, the filtered ones were subtracted from the original ones in order to eliminate the offsets associated with atmospheric effects.

Spectral analysis of the data obtained (see Fig. 27) shows that the vibration amplitude in azimuth is indeed greater than in altitude, and reaches $j_{\textrm{tel}}\approx 0\overset{\prime\prime}{.}08$. The graph is dominated by harmonics at frequencies of 20, 40, 60 Hz. No significant correlation of these amplitudes with wind speed was found. Moreover, the observations were made on four different days with different weather conditions.

Fig. 27

Cumulative power spectra (that is, the integral of the power spectrum $|A(f)|^{2}$ over frequency) from zero to the frequency of the telescope jitter in altitude and azimuth (in the plane of the sky), units are the arc seconds squared. 14 graphs are superimposed on each other for different series with different conditions and exposures (from 1 ms to 4 ms).