Observers and support staff are therefore urged to avoid illuminating the Thomson with bright lights either before or during an observing run. Since the dark current also takes several hours to stabilize after the CCD is first powered on, the system should be powered up as early in the day as possible, and observers should be aware that the dark current will increase substantially if for any reason the electronics have to be switched off and powered up again during the night.
As an illustration of the level of illumination which can cause problems, in an early run with the Thomson CCD on the RGO spectrograph several grating changes were made with the cage lights on, so that light entered the spectrograph. Subsequent reduction of the data showed that the dark current increased noticeably after each grating change and remained high for much of the night, seriously degrading the quality of the data. Similar problems may occur with the CCD in direct imaging mode if, for example, the CCD is illuminated with a bright lamp while the shutter is open; or long exposures are taken with the dome lights on so that the chip is grossly saturated.
Since most f/1 observers have a long list of program objects with only a relatively short exposure time on each, some advance planning of the run will help increase efficiency. Most importantly, it is advisable to prepare a co-ordinate list (data catalogue) which can be entered into the telescope computer before the start of the run. This allows the co-ordinates of each object to be recalled automatically when required without the night assistant needing to enter them from the keyboard.
Photometric standard stars must also be chosen carefully, as many are too bright to observe with the f/1 system. The minimum recommended exposure time is 5 seconds, set by the shutter timing because the large shutter of the f/1 system takes a finite time to open and close. Stars brighter than about V=14 may saturate the CCD in less time than this, and are therefore unsuitable.
It is also worth giving some advance thought to how the data are to be flat-fielded. If flat fields are to be taken on the twilight sky at the start of the night, there is only a relatively narrow time window in which this is possible and it is best to forewarn the night assistant that twilight sky flats are needed so that the telescope is ready at the appropriate time (see later section on flat fielding).
If you plan to use filters other than the standard glass V, R and I filters supplied with the system, you should contact your support astronomer well ahead of your run to check that this is feasible.
Details of how to prepare a catalogue are given in section 5 of AAO TM 7 ( Telescope Control System) and in Appendix A. Observers need only create a suitably-formatted file on the VAX - the night assistant can then transfer this to the telescope computer at the start of the night's observing.
Since the autoguider cannot be used with the f/1 system, there is no need to supply guide star positions.
The f/1 system is capable of creating a frightening amount of data on a good night. The telescope staff are aware of this and can cope with it, but again some advance planning is useful. If your program is likely to generate more than about ten Thomson CCD frames an hour (i.e. your average exposure time is five minutes or less), you should make sure that your support astronomer and/or the computer staff at the telescope know this in advance so that they can arrange to keep as much disk space as possible clear for your run.
By default, the OBSERVER CCD system produces Figaro format output files in which the data is held as floating point numbers, a format that requires 4 bytes per pixel, or 4 Mb per full image. The main observing disk, DISK$INST, even when completely cleared, has a capacity of about 800 Mb. While this may sound like a lot of space it corresponds to only 200 full-format Thomson images, a number which can easily be taken over a weekend or even a single winter night. Moreover, large raw data files also mean large processed files, and observers attempting to do on-line processing of their data are likely to find themselves very short of scratch space.
A couple of points are worth keeping in mind:
Copying and archiving the huge amount of data produced by large-format CCDs is a time-consuming process. While the AAT staff do their best to provide tape copies for observers to take with them when they leave the mountain, this may take more than 24 hours if large amounts of data need to be transferred to standard magnetic tapes (a recent single-night run with the f/1 system filled seven 2400-ft tapes at 6250 bpi). Thus, when observers finish at dawn, it may be difficult or impossible to provide them with a `standard' tape copy of their data in time for the afternoon plane.
This problem can be overcome, however, with the use of Exabyte tapes. Both the AAT and Epping VAXes now have high-density Exabyte tape drives which allow a night's observing to be written to a single tape in backup format in 2-3 hours. AAT computer staff can arrange for each night's data to be backed up automatically to Exabyte once the night's observing finishes (with the backup usually starting at 7 a.m.). While Exabyte tapes are not recommended for long-term archiving, observers can transfer their data to `standard' tapes at their leisure, either at the telescope or at Epping.
To backup files on the Exabyte yourself, mount a tape in the drive, log in
on the VAX 3800 and type the following command sequence:
$BACKUP/LOG/REWIND/VERIFY/NOASSIST *.DST MUB0:save.BCK
where `save' is the name of the saveset.
Since the last edition of the CCD manual was published, AAO has begun using a new system of instrument control software, the OBSERVER system. The system is described in detail in an AAO software document called The OBSERVER software system, a draft copy of which should be available in the AAT control room.
The following command sequence should be all that is needed to start up the CCD system:
If you want to use a smaller window, WINDOW THOMSON_CENTRE will give just the central quadrant of the chip.
You are now ready to start taking data. The GLANCE command is a useful (and quicker) alternative to RUN while setting up and testing the system when you are not sure whether you will want to keep the data. GLANCE 5 will give a 5 second exposure, and after the data is displayed you can type KEEP GLANCE if you decide you want to save the data. Note that if you do a large number of GLANCES, the memory may fill up and give an error message (``no image memory'') when you try to do your next exposure. DIR will list the GLANCE frames kept in memory, while DEL *;* will remove them all and allow you to continue.
Online help is also available for the observer system.
Each new image is displayed on the XMEM immediately after readout. Occasionally the XMEM may hang up and need to be reset. This can be done by pressing the black reset button on the panel behind the observing terminals. more detail needed here!
After the f/1 system is on the telescope it is useful to check the bias level and dark current. No other setting up is necessary during the afternoon, though the cautious may wish to take a few dome flats to be sure that everything is working.
Two steps must be done during twilight - focussing the CCD and checking that objects are well centred in the CCD field. It is also advisable to take several flat field exposures on the twilight sky 40 minutes or so after sunset or before sunrise (see section 3.11).
The Thomson CCD dewar has only a short hold time, and will need refilling every six hours. Normally this will be done by the night assistant or daytime technical staff, but support astronomers may be asked to lend a hand at weekends.
To fill the dewar, move the telescope to prime focus access and connect the liquid line of the grey liquid nitrogen dewar to the thick black hose which emerges through the floor of the prime focus cage. The clear plastic nitrogen line should also be connected to the dewar (via a snap-on connector). Once the liquid nitrogen dewar is pressurized to about 10psi, open both valves to allow nitrogen to flow to the CCD cryostat. This will take about 10 minutes to fill - when liquid nitrogen starts to spill noisily from the CCD end it is time to stop. Close both valves and disconnect the grey dewar (using the hairdryer to unfreeze the frosted-up connector if necessary).
The f/1 system has a filter holder which accepts up to three 50.6mm diameter circular filters, and V, R and I filters are currently provided. Filter specifications, along with transmission curves, are given in Appendix B. There is also provision for a single filter to be placed below the shutter.
Filter selection is via a small metal box located in the AAT control room. The box has a switch with four positions (generally V, R, I and CLEAR). There are also three meters which provide a check on the position of the three filter arms, and two sets of shutter lights which indicate whether the shutter is open or closed (one set for each half of the two-part shutter).
At present, the filters are not encoded, so you should be very careful that you have selected the filter you want, and that it is recorded correctly in the data log. It is a good idea to incorporate the filter name into the header of each of your observations (by typing e.g. OBJ NGC 123 R as the object name in OBSERVER).
Because the two-part shutter of the f/1 system is large and takes a finite time to open and close, very short exposures are not recommended. Shutter timing tests using dome flat fields show the timing over the whole frame is accurate to better than 0.5% over the whole frame for a 5s exposure. For exposures of 3 seconds or less, flat fields show a noticeable brightening in the centre relative to the outer part of the frame, due to the finite opening/closing time of the shutter.
The minimum recommended exposure time is 5 seconds, but observers who plan to do accurate photometry of extended objects at low light levels might prefer to use minimum exposures of at least 10 seconds and should make sure that their flat field exposures (if done on twilight or dark sky) are also at least this long.
The simplest way to focus the telescope and CCD at the start of each night is to point the telescope at a star and take a series of exposures in multiple mode.
A suggested recipe is as follows:
The f/1 optics cause some vignetting at the edges of the field, so flat fielding is necessary not only to remove pixel-to-pixel sensitivity variations in the CCD but also to correct for this vignetting. Appendix C shows typical flat-field contours for each of the currently-used filters.
Experience shows that flat-fielding of images taken with the focal reducer can be done to better than 0.5%, but this requires some careful work to created a master flat-field frame from the median of twenty or so sky-limited frames. Observers whose fields contain few bright stars or extended objects can achieve this by suitably scaling a number of data frames and taking the median.
If most of the data frames contain large extended objects such as bright galaxies, this approach is not usually feasible and there are two possible alternatives. One is to take twilight sky flat fields just after sunset or before sunrise, the other is to take a number of `flat' frames on the dark sky, offsetting the telescope by several arcseconds between each. Which of these is chosen will probably depend on the requirements of the observing program.
There is no significant night-sky fringing with the current (thick, uncoated) Thomson CCD.
As a guide, exposure times can be estimated from the typical count rates given below.
Total counts for a 20th magnitude star (e-/s) with f/1 system (summed over the entire image of a 20th magnitude star near the zenith. The area over which the sum is taken depends on the seeing.)
Typical sky brightness (new moon) at Siding Spring
Typical sky levels in e-/second/pixel
The signal to noise ratio is then given by given by
is the total number of counts in the object,
is the number of pixels covered by the object (i.e. the size of the seeing disk if the object
is a star),
is the number of counts per pixel from the sky background, and
is the CCD readout noise in electrons.
For the f/1 system, where even the shortest exposures through a broad band filter are sky limited, this simplifies to:
For example, a star with R=23 has an expected total count rate of about 8.8 photons/second, ie. Nobj = 8.8 and from the table above, for a dark sky, Nsky = 72 photons/second/pixel in R. Assuming the seeing is 1.5 arcsec FWHM and the star image is spread over 9 pixels, we have S/N = 6.0 after 300 seconds. From this, we can estimate that a five minute exposure with the Thomson CCD in 1.5 arcsec seeing will reach a limiting magnitude (5 detection) slightly fainter than 23 in R.
Observations of several standard fields were carried out in Director's time on 21 July 1990 in order to determine the colour and extinction terms for the f/1 system. While the data have not yet been fully analysed, it should soon be possible to provide this information to observers. As with the RCA CCD, it should then be possible to obtain reasonably accurate photometry by observing just a few standard stars each night to check the zero point.
As mentioned earlier, standard stars must be fainter than about V=14 to avoid saturating the CCD in less than the 5s minimum recommended exposure time. Most of the Graham E-region fields (reference??) are suitable and can be observed in a single f/1 frame. The Landolt (1992) CCD fields also now contain stars sufficiently faint for use with the f/1 system.
Two standard HST fields were also observed in July, and should yield a number of secondary standards once analysis of the data is complete.
As a guide to repeatibility, Brian Boyle provides the following RMS errors (frame-to-frame) for stellar images in a 300s exposure with a broad (V+R) passband filter, based on a comparison of 5 pairs of V+R CCD frames taken one month apart (detection limit 4 of sky):
Experience shows that accurate astrometry is possible with the f/1 system. The residuals in the astrometric solution for a standard open cluster observed in May 1990 were of order 0.1 arcsec: in practice the errors in individual program fields are likely to be dominated by uncertainty in the positions of secondary reference stars such as those in the HST Guide Star catalogue.
The Starlink program ASTROM can be used to measure astrometric positions - more details will be given in the next draft of this document.
The orientation of the Thomson CCD at f/1 is such that the columns run east-west, and (row1, column1) is northwest. Note, however, that the XMEM will display the data the opposite way up from FIGARO, i.e. data will apear on the XMEM screen with North to the left and East to the top.