Natural guide star adaptive optics systems at LBT: FLAO commissioning and science operations status S. Esposito*a, A. Riccardia, E. Pinnaa, A. T. Puglisia, F. Quirós-Pachecoa, C. Arcidiaconob, M. Xomperoa, R. Briguglioa, L. Busonia, L. Finia, J. Argomedoa, A. Gherardia, G. Agapitoa, G. Brusac, D. L. Millerc, J. C. Guerrac, K. Boutsiac, P. Stefaninia a INAF – Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50125 Firenze, Italy; b INAF – Osservatorio Astronomico di Bologna, Via Ranzani 1, 40127 Bologna, Italy; c LBT Observatory, University of Arizona, 933 North Cherry Avenue, Tucson, AZ 85721, USA. ABSTRACT This paper summarizes the activities and the principal results achieved during the commissioning of the two Natural Guide Star (NGS) AO systems called FLAO#1 & 2 installed at the bent Gregorian focal stations of the 2x8.4m Large Binocular Telescope (LBT). The commissioning activities of FLAO#1 took place in the period February 2010 – October 2011, while FLAO#2 commissioning started in December 2011 and should be completed by November 2012. The main results of the commissioning campaign are presented in terms of the H-band Strehl Ratio values achieved under different observing conditions. We will also describe the automatic procedures to configure and set-up the FLAO systems, and in particular the modal gain optimization procedure, which has been proven to be a very important one in achieving the nominal performance. Finally, some of the results achieved in two science runs using the near infra-red camera PISCES are briefly highlighted. Keywords: Large Binocular Telescope, high-order adaptive optics, pyramid sensor, adaptive secondary mirror.

1. INTRODUCTION The Large Binocular Telescope (LBT) [1] features two co-mounted optical trains with 8.4m primary mirrors. Each optical train has three bent-Gregorian focal stations, one of these stations on each optical train being equipped with a First Light Adaptive Optics (FLAO) system [2], dubbed FLAO#1 & FLAO#2. The FLAO systems rely on Natural Guide Stars (NGS) to probe the atmospheric turbulence, and feature two key components: an adaptive secondary mirror and a modulated pyramid sensor (Figure 1). The rationale behind the selection of those components was motivated by several reasons. On one hand, an adaptive secondary mirror (a) allowed us to have a reduced number of reflections; (b) features a large number of actuators; (c) attains a performance that has a low sensitivity to a small number of damaged actuators; and (d) provides AO correction to all LBT focal stations. On the other hand, a modulated pyramid sensor allowed us to: (a) have a larger sensitivity than a Shack-Hartmann (SH) sensor in the faint end; (b) have a reduced aliasing effect in the bright end with respect to the SH; (c) have an easily adjustable subaperture size via on-chip binning; and (d) use a small (80×80 pixel) CCD for 30×30 subapertures. This paper first reviews the main results of the on-sky commissioning of FLAO#1 that took place in the period of February 2010 – October 2011, together with the results achieved up to now in the ongoing commissioning of FLAO#2 which started in December 2011 and that should be completed by the end of 2012. An important part of the commissioning was dedicated to setting up the automatic configuration of the FLAO systems. In this paper we will in particular describe an important aspect of the automatic configuration procedure, namely the modal gain optimization procedure, which allows us to automatically tune the FLAO system to obtain the nominal performance under different observing conditions. An assessment of the reliability of the overall FLAO automatic initialization will be also presented. Finally, we will present some closed-loop results using extended objects as reference sources, and we will review some of the first scientific observations performed with the FLAO#1 system using the near infrared (NIR) camera PISCES [3].

*[email protected]; phone +39 055 2752 309; fax +39 055 2752 292

Adaptive Optics Systems III, edited by Brent L. Ellerbroek, Enrico Marchetti, Jean-Pierre Véran, Proc. of SPIE Vol. 8447, 84470U · © 2012 SPIE · CCC code: 0277-786/12/$18 · doi: 10.1117/12.927109

Proc. of SPIE Vol. 8447 84470U-1

Figure 1. Key components of the FLAO system of the LBT; (Left) pyramid WFS board. (Right) Adaptive Secondary Mirror.

2. THE FLAO SYSTEMS ON-SKY PERFORMANCE The main purpose of the on-sky commissioning of the FLAO systems is to verify that the closed-loop performance under different observing conditions, namely different illumination levels and seeing values, is in agreement with the performance estimates found with numerical simulations [4]. Figure 2 shows the summary plot of the commissioning results attained by the FLAO systems. A total of 597 and 696 Strehl Ratio (SR) estimates in H band (1.60µm) measured with FLAO#1 and FLAO#2, respectively, are included in this plot. The SR was measured from long-exposure PSFs acquired with the LBT’s InfraRed Test Camera (IRTC) [5] having a pixel scale of 10 mas/pixel. AO telemetry data was used to estimate the equivalent star R-magnitude and the seeing value [6]. By comparing the on-sky performance estimates with the simulated ones shown in this plot, we can conclude that both FLAO systems attain the performance requirements. It is important to mention that the system configuration during most of these observations was completely automatic. Indeed, a good fraction of the commissioning time was dedicated to setting up the system so that it could be started, configured, and operated with the minimum intervention of the telescope operator. We will discuss in more detail the automatic initialization of the FLAO system in Sections 2.1 and 2.2. An evaluation of the reliability of the automatic configuration will be presented in Section 2.3. 2.1 Automatic configuration of the FLAO system In order to start an observation with the FLAO system, the user needs to specify only two parameters: the position and the expected R-magnitude of the AO guide star. From that point, all subsequent procedures are executed automatically within three sequential phases: 1.

The Preset AO phase. The system is pre-configured based on the AO guide star magnitude specified by the user. All AO system parameters (sampling frequency, binning mode, number of controlled modes, pyramid modulation, etc.) are selected using a configuration table built with the aid of numerical simulations (Table 1).

2.

The Acquire Reference phase. The system acquires the AO guide star based on the position specified by the user. Once the NGS is acquired the real flux is measured and the AO parameters are updated if needed according to configuration table (Table 1). The position of the four pupil images on the WFS camera is verified and adjusted if required. At the end of this phase the AO system is ready for loop closure.

3.

The Start AO phase. The AO loop is closed and a modal gain optimization is performed. We will discuss in more detail this optimization in Section 2.2.

The division in three phases has been designed for both efficiency and flexibility: the Preset AO phase does not need any light from the AO source and can be executed in parallel with other telescope setups and slew. The Acquire Reference phase needs the telescope tracking (but not guiding) on the target, and leaves the system ready for AO loop closure. The three phases require from 2 to 3 minutes to complete.

Proc. of SPIE Vol. 8447 84470U-2

Figure 2. (Color on-line) Summary plot of the commissioning results obtained with the FLAO#1 and FLAO#2 systems. The plot shows the measured Strehl Ratio in H band versus the guide star R-magnitude. The color of the different points gives an estimate of the seeing value (points for which the seeing could not be estimated are shown in black). The symbols indicate the binning mode used. The dashed lines refer to the expected performance values found with numerical simulations for a seeing of (top to bottom) 0.6, 0.8, 1.0, 1.2, and 1.5 arcsec.

Table 1. System configuration table as a function of the equivalent star magnitude in R band.

Equivalent star R-mag MR 7.4 8.4 9.4 10.0 10.0 10.9 11.4 11.9 12.4 13.4 13.4 14.4 14.4 15.4 16.4 17.5

WFS camera (CCD39) Binning fs mode (Hz) 1 990 1 990 1 990 1 990 2 990 2 990 2 990 2 990 2 625 2 400 3 500 3 200 4 300 4 200 4 105 4 105

Expected photons / subap. / frame 500 213 85 48 192 82 54 34 34 21 37 38 45 27 21 8

Num. of controlled modes (nmod) 400 400 400 400 153 153 153 153 153 153 66 66 36 36 36 10

Pyramid modulation (±λwfs/D)

Proc. of SPIE Vol. 8447 84470U-3

3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 6.0 6.0 6.0 6.0 6.0 6.0

Acquisition camera (CCD47) Binning fs mode (Hz) 4 4.3 4 4.3 16 4.3 16 4.3 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6 16 12.6

2.2 Modal gain optimization It is well known that the gain of an integrator controller can be optimized mode-by-mode since each mode may have a different Signal-to-Noise Ratio (SNR) [7]. In brief, the turbulence variance and the propagated measurement noise variance decrease with a different rate as the radial order increases; hence, the gain minimizing the residual modal variance (i.e. the sum of the variances of the un-rejected turbulence and the propagated measurement noise) may differ for different radial orders. At LBT, we have decided to optimize the gains of three groups of modes: tip-tilt (TT), mode #3 to #120 (aka midorders), and modes higher than #120 (aka high-orders). The modal gain optimization procedure (also refer to as autogain) is an iterative procedure launched in the Start AO phase that looks for the gain minimizing the sum of the modal variances in each group. AO telemetry data (mainly WFS signals and the reconstruction matrix) are used to estimate the modal time series, from which the variances can be computed. The autogain procedure sweeps the group gains in pre-defined intervals, acquires AO telemetry data, and searches for the optimal gains. The pre-defined intervals can be fine-tuned automatically if no minimum group variance is found. Without a-priori knowledge on the turbulence parameters, modal gain optimization is crucial to get the expected performances at the telescope before starting a scientific observation. Furthermore, since TT modes are usually prone to telescope vibrations, TT gain optimization can help reduce the effects of vibrations as long as the vibration’s frequency lies within the control bandwidth of the AO system. This is actually the case at LBT for which the dominant vibration occurs at ~13Hz, and so it can be partially attenuated by increasing the TT gain when the FLAO system is operated at fast framerates (fs > 800Hz) [8]. In this paper we will focus on the gain optimization analysis for mid-order and high-order modes. Figure 3 shows the group gains versus the seeing value (estimated from AO telemetry data) for a subset of acquisitions obtained under highflux conditions (7.0 < MR < 9.7) with the FLAO#2 system. First of all, note that, as expected, the SR achieved (coded in color in the plot) decreases as the seeing worsens. Also, note that there is a trend to increase the gains for poorer seeing values, improving in this way the turbulence rejection capabilities of the AO system. However, there is a large spread of gain values found by the autogain procedure for the same seeing value which, as we will show below, is related to the different temporal characteristics of the turbulence (i.e. different median wind speeds across the particular line of sight) encountered in different observations. As an example, when the seeing is 0.8-0.9” the range of mid-order gains spans from 0.7 to 1.8, and from ~0.5 to 1.1 for high-order gains (Figure 3). The parameters of two particular acquisitions within this subset are reported in Table 2. The performance measured in these two examples is equivalent (65-70% of SR in H band) for a seeing of ~0.85 arcsec.

Figure 3. (Color on-line) Modal gain analysis for a subset of acquisitions obtained in high-flux conditions (binning mode #1, 1kHz, ±3λ/D). The plots show the group gains found by the autogain procedure versus the seeing value. (Left) midorder modes; (Right) high-order modes. For each acquisition, the symbol indicates the number of controlled modes used, and the color indicates the range of SR in H band achieved.

Proc. of SPIE Vol. 8447 84470U-4

Table 2. Comparison of modal gains found by the autogain procedure for two different acquisitions obtained under high-flux conditions and with a seeing of ~0.85 arcsec. All other system parameters are the same (binning mode #1, 400 modes, 1kHz, ±3λ/D). An estimate of the cut-off temporal frequency of the 13th and 23rd radial-order modes is also given. Acq. ID 1 2

Acq. Date (Date-UT) 20120505 115822 20120503 063006

MR

Seeing (arcsec)

Elevation (degrees)

%SR (H)

TT Gain

Midorder gain

Highorder gain

fc [Hz] n =13

fc [Hz] n =23

9.0

0.85

72

70.6

1.3

0.7

0.6

9.0

15.0

7.6

0.84

61

64.8

1.8

1.2

1.1

14.0

23.0

Figure 4. Examples of temporal power spectra of the input disturbance (estimated from AO telemetry data) for the two particular acquisitions reported in Table 2. (a) Mid-order group example: averaged temporal spectra of 13th radial-order modes. (b) High-order group example: averaged temporal spectra of 23rd radial-order modes. The vertical dashed lines in both plots indicate the cut-off temporal frequency of each spectrum (vales reported in Table 2).

Figure 4 compares the open-loop (turbulence) temporal power spectra (estimated from AO telemetry data) averaged over the 13th and 23rd radial-order modes, chosen as examples of mid-order and high-order modes, respectively. An estimate of the temporal cut-off frequency fc (i.e. the “knee” of the spectrum from which it starts to roll off rapidly at a rate proportional to f-11/3) is given in Table 2. Recalling that fc is proportional to the median wind speed (e.g. for Zernike polynomials fc =0.3(n+1)V/D, where n represents the radial order, V the wind speed, and D the telescope diameter [9]), we can affirm that the median wind speed during acquisition #2 was ~1.5 times faster than during acquisition #1. This fact is taken into account by the autogain procedure while searching the gain that minimizes the residual modal variances, finding indeed higher gains (a factor ~1.7-1.8 higher) for acquisition #2. Note that these higher gains also produce an overshoot at ~100Hz. However, at least in the presented high-flux case, the turbulence + noise power amplified in this frequency range does not dominate in the residual modal variances. Finally, note that the spectra of acquisition #1 flatten up to higher values due to the slightly higher propagated photon noise (star mag 9.0 versus 7.6). We should make a comment on the different number of controlled modes used in these acquisitions (Figure 3). As specified in Table 1, the default number of controlled modes for the considered range of star magnitudes is 400. Hence, those acquisitions using a different number of controlled modes required the loading of a different look-up table. When the seeing was very good, we tested the performance of the 500-modes reconstructor, whereas when the seeing was bad, we had to use the 250-modes reconstructor to avoid Force saturation on the ASM. In fact, as we will discuss in the following section, closing the loop with 400 modes under bad seeing conditions may cause the autogain process to fail. From these considerations, it is clear that the automatic configuration procedure of the FLAO system could be improved by having as input, in addition to the expected star magnitude and star position, the expected seeing value delivered by the LBT’s Differential Image Motion Monitor (DIMM). The software interface to implement this is already in place, the new look-up tables are being updated, and they will be tested during the last part of the FLAO#2 commissioning. The autogain procedure remains crucial as no a-priori knowledge on the median wind speed can be at present retrieved.

Proc. of SPIE Vol. 8447 84470U-5

2.3 Analysis of reliability of the FLAO system automatic initialization We have quantified the percentage of successful AO system initializations (i.e. the successful automatic completion of the three phases) during two scientific observation periods: the Science Demonstration Time (SDT) held in November 89th 2011, and the AO science run held in June 17th-22nd 2012, both using the FLAO#1 system. A total of 56 and 103 Preset requests were executed on each run, respectively. The results of this analysis are presented in Figure 5, showing the percentage of successful/failed presets, specifying the problems encountered. About 60% of all requested AO initializations were successfully completed in both runs. Regarding the failed initializations, the most frequent problems encountered were: •

Sequence/script errors. The observing script either issued a command in the wrong sequence (for example, a Start AO command before acquiring the AO reference star), or with incorrect parameters (for example, an AO reference star magnitude outside the limits defined in the AO configuration table). This was the most common source of error during the first scientific run of SDT.

•

Autogain/bad seeing problems. The modal gain optimization procedure (also refer to as autogain) failed to converge. This problem often occurs in bad seeing conditions, for which high modal gains are usually required. The application of high gains (in particular for high-order modes) may lead to forces exceeding the adaptive secondary safety thresholds, causing the AO system to open the loop.

•

No star on camera/cannot center. The AO guide star is not visible on the AO acquisition camera. This may be caused either by telescope pointing problems, input of wrong AO guide star coordinates, or convergence problems of the star centering algorithm (particularly in crowded fields).

Figure 5. Analysis of successful/failed AO initializations during two particular scientific runs.

2.4 Loop closure on extended objects Pyramid wavefront sensor gains in sensitivity with respect to other focal plane wavefront sensors such as the ShackHartmann (SH) thanks to the shrinking in closed loop of the guide star PSF at the vertex of the pyramid. While the dimension of the spots on the SH is defined by the lenses size in the array, in the pyramid case the signal benefits from the AO partial correction obtaining in this way an improved sensitivity. Because of this reason the ability of the pyramid sensor to close the AO loop efficiently on extended sources was often discussed, especially related to laser guide stars considerations. During commissioning and science observations time we had the opportunity to test the performance of the pyramid sensor using extended sources of different sizes (from diffraction limited to sizes of up to 2.5 arcsec, the pyramid diaphragm limit size) and shapes (full disks such as solar system objects or extended sources with halo such as galactic nuclei).

Proc. of SPIE Vol. 8447 84470U-6

Regarding solar system observations, we have obtained diffraction-limit resolution images of the asteroid 45 Eugenia (resolved on an elliptical PSF ~0.21×0.12 arcsec) and its moon Petit Prince at