Commissioning report for ESO

Vik Dhillon, version: 19 December 2006

ULTRASPEC was awarded 4 nights of technical time by the Director of La Silla-Paranal Observatory for use on the EFOSC2 spectrograph mounted at the Cassegrain focus of the ESO 3.6-m telescope. The time was scheduled from Dec 1-4, 2006. This document presents a brief report on the commissioning run and describes how we used the telescope time. For an overview of ULTRASPEC and how its hardware and software work, please consult the commissioning plan.

Contents

1. A photo diary
2. Observations
3. Summary

A photo diary

• 2006 November 27: Vik Dhillon (Sheffield), Naidu Bezawada (UKATC, electronics/detectors), Andy Vick (UKATC, software) arrived on La Silla on 2006 November 27. We located our 3 packing crates and moved them up to the detector laboratory on the second floor of the 3.6m. We then cabled up, powered on and verified that the system had survived the shipping to La Silla. The photos below show the lab and its adjoining computer room where we set up our equipment.
  
  
  
  
• 2006 November 28: Our first task was to check the alignment of the CCD, which had to be flat to within approximately 100 microns relative to the front mounting flange of the cryostat, and lie 14.0+-0.5mm from this same flange. The photos below show how Naidu performed the alignment, using a travelling microscope brought out from the UKATC. The one on the right also shows Pascal Robert (ESO), who assisted us during the lab phase of commissioning.
  
  
  
  
  Naidu adjusted the alignment by placing shims on the three posts (shown on the left) on which the chip PCB (shown on the right) is mounted:
  
  
  
  
• 2006 November 29: With the realignment completed satisfactorily, we replaced the cryostat window with the EFOSC2 field lens, as shown below. Note that the cryostat lid had first to be modified by the mechanical workshop at La Silla in order to accommodate the EFOSC2 camera optics.
  
  
  
  The cryostat was then pumped down using an ESO vacuum pump (left-hand photo), and filled with liquid nitrogen using a pressurised dewar (right-hand photo):
  
  
  
  In the meantime, we set up our GPS antenna. The mechanical workshop at La Silla made us a bracket to mount our post to the catwalk outside the dome (left-hand photo). Our GPS cable was routed under this catwalk, through a hole in the door frame (centre photo) and into the room containing electronics racks near the dome floor (right-hand photo):
  
  
  
  Once inside, the GPS signal was converted to run down a fibre, which was routed down to the lab where we were running the instrument. In this way, we were able to verify the proper operation of our GPS timestamping system, which typically picked up 6 GPS satellites from its position on the dome catwalk. The photos below show our setup in the room containing electronics racks near the dome floor.
  
  
  
  
• 2006 November 30: The entire day was spent in the lab characterising the CCD whilst cold. A detailed report on this aspect of the commissioning will be posted on the ULTRASPEC web site in the near future. Note that it was not possible to mount anything on the telescope at this stage, as EFOSC2 was being used in visitor mode for the nights leading up to our run.
  
  The other half of the ULTRASPEC commissioning team - Tom Marsh (Warwick), Chris Copperwheat (Warwick) and Kieran O'Brien (ESO) - arrived on site in the afternoon.
  
  
• 2006 December 1 (setup night): We finally obtained access to EFOSC2 and the telescope. We decided to run the chip warm on the setup night. We did this as it is perfectly possible to take alignment data with the chip warm and it gave us the possibility of rectifying any problems with the chip alignment that might have arisen during the EFOSC2 alignment process.
  
  First, our electronics rack was loaded in the Cassegrain cage (left-hand photo) and mounted on an isolated base made especially for us by the mechanical workshop at La Silla. Then, we mounted the EFOSC2 camera optics on our cryostat (right-hand photo).
  
  
  
  After sunset, we operated ULTRASPEC from a laptop on the dome floor (see photo below), taking images of stars in the sky and adjusting the position of the camera optics until the stars' FWHM were minimised. We also tracked the FWHM of a star across the entire field of view and verified that the value hardly changed, giving a first indication that our CCD had been well adjusted for tilt in the lab. The final position of the camera was 15.5 units on the ruler and 7.75 units on the thumb wheel.
  
  
  
  With the camera optics aligned, we then installed the cryostat on EFOSC2. As shown in the photos below, we also installed the SDSU CCD controller, mounting it on a frame using insulated pads modified for us by the mechanical workshop at La Silla. The cables connecting the cryostat/controller with the electronics rack were fed through a hole in the bottom of the Cassegrain cage (and protected with foam) to form a simple cable twister.
  
  
  
  With ULTRASPEC now properly mounted on EFOSC2, the next step was to align the collimator. We did this using a Hartmann test, illuminating a pinhole mask in the slit wheel with an internal lamp and adjusting the collimator position until there was no more shift in the spot positions when switching from one Hartmann mask to the other. (Note that we could not use a slit, arc lamp and grism to perform the Hartmann test, as the Hartmann shutters are located in the same wheel as the grisms). The final position of the collimator was 1.0 mm, which gave tilts of 0.6 pixels from left-to-right across the chip, and a similar amount from top-to-bottom. These values are far too small to adjust by shimming the CCD, and would not be noticeable in our data.
  
  The collimator alignment was completed just as morning astronomical twilight began. With no apparent problems with our chip alignment (rotation and tilt), we then felt confident enough to start cooling the cryostat with liquid nitrogen (preceded by an hour of vacuum pumping) in preparation for the first night of observation.
  
  
  
  
• 2006 December 2 (first night): During the day, we set up the ULTRASPEC data reduction PC in the 3.6m control room, from where we could then control the instrument via its GUI and run the ULTRASPEC data reduction pipeline software. We used a pre-existing fibre routed from the Cassegrain cage to the control room to run our own internal network.
  
  
  
  We then checked the readout noise in the normal output channel. In the lab, we had achieved a readout noise of 4.2 electrons, with no discernible pickup noise patterns. After much experimentation with cable positions and earthing arrangements in the Cassegrain cage, we ended up with the same value on the telescope, although we did notice that this was dependent on the Cassegrain rotator angle. The reason for this is probably because the SDSU controller and its cables then rotate relative to the various electronics racks in the cage; some of these racks must be interfering with ULTRASPEC, causing the pickup noise we observe.
  
  We spent the rest of the day aligning the spectrograph slits and grisms, which were aligned to the rows and columns of the CCD to better than 0.1 pixels across the chip. We also checked the dispersion and wavelength range of our various slit/grism combinations. For example, the wavelength range, central wavelength and dispersion of grism 8 is 2040 A, 5340 A and 0.99 A/pixel with the standard EFOSC2 CCD (2048 pixels on a side, each of 15 microns), but with the ULTRASPEC CCD (1024 pixels on a side, each of 13 microns) the corresponding figures are 909 A, 5260 A, and 0.89 A/pixel. With a 1 arcsecond slit, we achieved an arc-line FWHM with grism 8 of 5.7 A, slightly better than the figure of 6.2 A when using the standard EFOSC2 CCD, verifying our good spectrograph focus.
  
  By sunset of the first night, we were completely ready to go. The photo below shows the commissioning team with ULTRASPEC on EFOSC2. From left to right: Emilio Barrios (ESO), Naidu Bezawada (UKATC), Kieran O'Brien (ESO, standing), Vik Dhillon (Sheffield), Chris Copperwheat (Warwick), Tom Marsh (Warwick), Andy Vick (UKATC, standing)
  
  
  
  

Observations

• 2006 December 2 (first night): A log of the observations taken on this night can be found here. The aim for this night was to fully commission ULTRASPEC on the sky, so the following tasks were performed:
  
  
  1. Telescope focus: This was achieved by simply minimising the FWHM of a standard star on ULTRASPEC, with the CCD running in a fast TV/acquisition mode and the slit and grism out of the beam. The optimum position of the secondary mirror (14.05mm) was very close to that used for standard EFOSC2 runs, further verifying the good alignment of our chip.
  2. Rotator centre: This was measured by rotating the Cassegrain rotator through 360 degrees - we verified that the rotator centre (and hence the telescope pointing) fell very close to the chip centre (pixel 570, 457).
  3. Orientation and plate scale: These parameters were measured by stepping the standard star by known amounts in RA and Dec. The resulting plate scale (0.138 arcsecs/pixel) and orientation (East up, North left, as seen on the chip at a slit position angle of 270 degrees) were required for object acquisition.
  4. Throughput test: We observed the ESO spectrophotometric standard G158-100 using a blue and a red grism and compared the resulting counts (in normal, not avalanche, mode) with those predicted for this object by the EFOSC2 exposure time calculator. We found that the calculator predicted the ULTRASPEC counts almost perfectly (in fact, the ULTRASPEC count values slightly exceeded the estimates), implying no problems with throughput.
  5. Acquisition test: We experimented with different ways of acquiring objects onto the slit, including the use of a script (as we do with ULTRACAM on the VLT). In the end, it became apparent that the simplest and quickest method was to acquire using a "trial and error" approach: the chip was set up in a fast framing mode, the initial position of the star was determined, and then the telescope stepped by approximately the correct amount (using our knowledge of the plate scale) to place it close to the known position of the slit on the CCD. The slit was then placed in the beam, and the position tweaked to centre the star in the slit. Two-star acquisition was also tested, and again it was found that it was simplest to use a rough first guess at the angle/position, followed by manual tweaking of the angle/position by the telescope operator, checking the results in real-time using ULTRASPEC.
  
  The second half of the night was spent observing standard stars in normal and avalanche mode in order to characterise the performance of the chip. Unfortunately, it soon became apparent that there was a problem with the positions of the slits and grisms that we had so carefully aligned that afternoon. This was later traced to a problem with the EFOSC2 LCU and the way it initialised the slit and grism wheels. We continued to take (poor) standard star data in various modes until an hour before morning twilight, when the telescope operator reported that the dome was stuck whilst slewing to a cataclysmic variable star. We spent the remaining time observing this star (as the dome had slewed to close to its target position before sticking), but the data were of poor quality due to the dome vignetting, which increased as the star set.
  
  
• 2006 December 3 (second night): A log of the observations taken on this night can be found here. The day was spent sorting out the slit and grism alignment problem, which was solved just before sunset. With ULTRASPEC now well aligned, we spent the night repeating the standard star tests we had attempted on the previous night. We observed stars of magnitude 15, 17 and 19 in order to probe the different regimes of the EMCCD in ULTRASPEC, in particular the normal output and avalanche output with different gains (with and without photon counting). As expected, the improvement in signal-to-noise when using the avalanche gain was tremendous. We intend to submit a short article to the ESO Messenger in January with further details on our initial results, as well as preparing a paper for a refereed journal once we have fully reduced and analysed all the data.
  
  The last three hours of the night were spent observing an example science target. We chose the well studied cataclysmic variable OY Car, which is faint enough (magnitude ~16) for photon counting and shows bright emission lines with rapid shape and flux changes due to the periodic (~1.5 hour) primary eclipses. Once again, the data quality far surpassed what would have been possible using a conventional CCD on EFOSC2.
  
  
• 2006 December 4 (third night): A log of the observations taken on this night can be found here. The day was spent taking calibration frames, performining noise tests and ironing out bugs in our chip formatting. Since we had fully characterised the chip when observing standard stars on the previous night, we decided to spend the final night focussing on example science targets. Our aim here was to demonstrate the improvements possible in terms of signal-to-noise and time resolution when using EMCCD's. We spent the first half of the night observing the AM CVn (cataclysmic variable) star ES Cet, which shows strong Helium emission lines and rapid variability due to its short (~10 minute) orbital period. The second half of the night was spent observing more primary eclipses of the cataclysmic variable OY Car.

Summary