A CCD detector for multiple object spectroscopy at the United Kingdom Schmidt Telescope.

A. P. Oates

Department of Physics, Science Laboratories, Durham University, South Road. Durham. DH1. 3LE. U. K.

ABSTRACT

At UKST (United Kingdom Schmidt Telescope, a unit of the Royal Observatory, Edinburgh. U.K. now operated by the Anglo-Australian Observatory, Epping. N.S.W. Australia) a CCD camera has been installed to operate with the FLAIR (Fibre Linked Array Image Re-formatter) fibre system (see Watson). The standard EEV P8603 CCD used in the camera has a poor short wavelength response and so all the CCDs used at UKST are surface coated with a fluorescent dye to partially overcome this problem. In 1988 it was decided that the system's response around could be improved further by replacing the FLAIR fibres with a set of fibres offering superior transmission properties at this wavelength. The introduction of these (larger core diameter) fibres would have meant, however, accepting a reduction in signal-to-noise as the fibres illuminate more pixels on the CCD. The CCD sequencing was therefore modified to enable us to on-chip bin pixels (across the dispersion direction) to alleviate the problem of reduced signal-to-noise which would have otherwise occured. Most recently, modifications have been made to the CCD system to provide detection capability for approximately twice the previous number of objects. This has been achieved by appending a second CCD detector and correlated double sample(CDS) processor to the existing sequencer. Both CCDs are operated via a signal controller which can route clocks and video between the detectors and sequencing electronics. Reduction of galaxy data, obtained during May/June 1988 and August 1989 are to be published in a forthcoming paper. These data have shown that FLAIR, combined with a low noise detector, in both single and dual CCD mode, is easily capable of obtaining cross-correlation red-shifts in the blue with a high success rate.

1. CCD SYSTEM

The CCD camera being used at UKST was developed by Waltham and initially used with FOS 1 (the Faint Object Spectrograph) on the U.K. Isaac Newton telescope. The system is controlled by a DEC PDP11/23 which, at the time the system was built was a fairly standard mini-computer to use for this purpose. The system consists of the components illustrated in figure 1 and described below.

1.1 CCD front-end

The CCD camera utilises an EEV P8603B image sensor housed in a standard liquid Nitrogen filled dewar. The cryogen capacity of the dewar results in a hold-time of hours (this is due to the axial fill tube arrangement and to the fact that we use the dewar horizontally) and maintains the temperature of the CCD at 150K. This temperature was found to be the best compromise in terms of optimising read-out noise, dark current and charge transfer efficiency. A custom PCB at the front of the dewar supports the 30-pin, socketed CCD and the components of a Wheatstone bridge circuit, one arm of which contains a platinum resistance thermometer (T.C. = ; at 273K) which forms part of the temperature servo-circuitry. Bias supplies, (the CCD output) and the horizontal and vertical clocks are all conveyed to the CCD through 9-pin vacuum feed-through connectors; all bias supplies are de-coupled close to the CCD. Along with low read-out noise, blue responsivity was the only other main requirement of the detector (assuming good horizontal and vertical charge transfer, above average numbers of cosmetic defects can be tolerated, though in devices used so far, these have been few in number). The CCD was coated at ESO with an organic polymer to enhance its short-wavelength sensitivity. The coating produces an improvement in quantum efficiency(QE) in the blue of, typically, 15%. Although better detection response has been obtained at shorter wavelengths, the system's response in the wavelength range has been improved still further with the introduction of a new blue-transmitting Polymicro fibre feed; the combination of this with on-chip pixel binning across the dispersion direction has significantly improved the system's signal-to-noise ( S/N) at short wavelengths.

1.2 CDS processor

The CCD output (typically ) is taken, via a constant current source, to the input of a very low-noise operational amplifier within the CDS and provides a first stage signal gain of 37. A constant current load was chosen in preference to a resisitive load as the former resulted in better gain from the on-chip FET amplifier. Analogue switches, driven from the sequencing electronics, enable the following control during pixel read-out:- the CCD output node is RESET, control clock RAMP-UP is enabled for to sample the output node voltage prior to charge transfer, a 3-phase horizontal shift moves charge from 1 pixel in the serial register onto the output node in , control clock RAMP-DOWN is enabled for a further to sample the RESET+VIDEO signal.

The difference between the first and second sampling voltages is a measure of the pixel charge; the dual slope integrator also provides a further signal gain of 5. The video is then digitised by a 16-bit successive approximation analogue-to-digital converter which was selected for its fast conversion time (16-bits in ) and good linearity. Pixel read-out is completed in and this is the rate at which digitised data is available from the A/D converter during a one-line-sequence period (). In 32 pixel binning mode, the whole CCD can be read-out in 17 seconds.

1.3 Driver electronics

A CCD sequencer generates low voltage D.C. which are conveyed to precision voltage references within the CDS processor for the CCD bias electrodes. Table one lists the current settings of the CCD bias voltages and clock read-out parameters. CCD control is provided by the sequencer through use of control EPROMS. These are accessed by the PDP11 which commands a one-line sequence (400 pixels) of the CCD whereby a pixel is read-out, digitised to 16 bits and stored in memory (within CAMAC) for subsequent retrieval by the PDP11. This procedure is repeated upto 578 times to enable a full-frame read-out of the EEV CCD. The observer may select vertical binning sequences of 1, 2, 5 or 10 pixels to ensure that dispersed spectra are correctly sampled when using the large diameter Polymicro fibres (spectral resolution /pixel with the /mm grating). Instructions to achieve the control described above are issued by a FORTRAN program running under RT11. This menu-based program supervises all CCD operations and enables the observer to construct exposure sequences to suit particular requirements. A simple command language is available for this purpose so that custom exposure sequences can be constructed and executed automatically by the sequencer. Provision also exists to store these command sequences on disc for future use. Data read from the CCD is stored on disk and displayed as an 8-bit grey scale on monitors in the dome and at a remote location. Here the observer can more conveniently control the operation of the camera using a vt100 type terminal, away from any extremes in temperature present in the Schmidt dome. Data obtained from earlier runs may be inspected whilst the next exposure is in progress, during which time the dome has to be kept in darkness.

2. PERFORMANCE

The following is a list of observations which have been noted during the general operation of the system over the last 3 years:- Operating a system in a dry enviroment can lead to problems with static charge. Thick earth returns were installed from the dewar, pre-amp electronics and optical table to the star earth within the CCD sequencer. The optical table provides a stable base for the CCD camera and spectrograph and during observations is mechanically and electrically isolated from the dome floor. Only one problem has been encountered with the passive components on the dewar PCB. A decoupling capacitor failed and resulted in spurious operation of the CCD when at operating temperature. The temperature of these components has been measured and found to be 180K when the CCD was at its operating temperature of 150K; it is unknown what long-term effects such temperatures have on these components. No problems have been encountered with the polymer coatings of our CCDs in terms of adhesion to the CCD surface. No adverse effects have been found either with the CCD or with the polymer coating when cooling the CCD from room temperature to its operating temperature in 45minutes (i.e. 3K/minute). Because of the modest aperture of the telescope, the read-out noise of the detector is the limiting factor when attempting to obtain data with good S/N. By exposing the CCD twice to 10 different levels of illumination, it was possible to obtain the mean count by averaging over two regions on each of the two frames and de-biasing from the 11 pixel underscan strip; variance was obtained after subtraction of two identically exposed frames and the result plotted. From this data the system gain (G) was found to be- and the CCD read-out noise ( ) is- In addition to the well known temperature dependence of dark current, there is also a dependence of read-out noise charge transfer efficiency(CTE) and QE with temperature. In the present case it was found that the CCD could be operated anywhere in the temperature range- without any significant change in the read-out noise . The horizontal and vertical CTE of this device appears to be very good. Some tuning of the horizontal clocks and adjustment of the voltage on the CCD output gate ( ) was made in the initial stages and this resulted in, at most, only 1 pixel smearing of cosmic ray events. The 4-pixel overscan was also investigated for residual charge on a low light level flat-field; the mean level of charge was found to have fallen back to the bias level after 1 pixel. The temperature dependence of CTE was not investigated. The 5/10/10BC CCD does not exhibit any measurable dark current in exposures of 1 hour; this being the maximum exposure time used per frame, if the effects of cosmic ray events (CREs) are not to contaminate the data too severely. Our integration times are limited to around 3000s to prevent cosmic rays contaminating our data too severely. This device type usually produces 2 events/cm/minute but this was exceeded in our early data by a factor of 3-4. The source of the excess CREs was investigated and found to be caused by the glass window at the front of the CCD dewar. This window is made of Schott glass some of which, it appears, emit low-level X-rays which can be detected by the CCD. The solution was to replace the original window with one made of fused-silica.

3. RECENT DEVELOPMENTS

During 1988 two proposals were made in an attempt to up-grade the system. The first related to the problem of the system's lack of sensitivity at the blue end of the spectrum (around ) and the second related to how we could modify the front-end so that we could increase the number of objects observed during a given exposure.

3.1 Improving sensitivity

To try and overcome the problem of the system's poor response at short wavelengths, a new fibre feed was installed whose transmission properties in the blue were superior to the existing FLAIR m fibres. To ensure that a comparable S/N could be maintained with the new Polymicro fibres, pixel binning across the dispersion direction was introduced. The new fibres have a core diameter of 100m, the optics introduce a magnification of 1.7 hence it was thought at first that a 7+3 (170m/22m & 66m/22m; the EEV pixel size is 22) pixel binning sequence should be introduced; the inter-fibre gap on the CCD is 3 pixels. It was, however, impossible to ensure that the output of the fibres would keep exactly in step with pixels binned in this fashion along the entire 400 pixel row (it was felt that the fibre output would get out-of-step with the pixel pitch). A less critical 3-pixel binning scheme was therefore introduced which resulted in a fibre coverage of 2 binned pixels (6 CCD pixels) and an inter-fibre gap of 1 binned pixel (3 CCD pixels). Initially, single pixel read-out was accomplished by having charge on the output node sampled as and of the serial clocks were changing state. The transition times for the clocks, and hence the pixel integration time was therefore relatively short, , see table one. As any 1/f component of read-out noise () is dependent on this time, it is kept as short as practicable. However, to compare in un-binned and binned CCD frames, the sequencing was changed so that a complete shift of the serial clocks occured when reading out a single pixel; the pixel time () now being . The read-out noise was measured again and found to be comparable with that obtained using the old pixel read-out sequence. From here it was a (relatively) simple matter to increase the number of complete serial shifts during a pixel read-out, to enable any number of CCD pixels to be binned. Although some clocking flexibility is available from the sequencer, a downloadable sequencer would have enabled modifications to be made to the clocking patterns without re-course to blowing sets of EPROMS each time a new sequence pattern was required. Fortunately this operation was only carried out 2 or 3 times before we were satisfied that the sequence pattern was correct. Again the read-out noise of the CCD was measured for the 2 schemes which were tried, 3 and 5 pixel binning, there being no change in read-out noise compared with that from single pixel read-out.

3.2 Improving detection capability

The second proposal was that we could, at reasonable cost, append a second detector head to the existing sequencer, permitting us to increase (from 35 to 71) the number of objects observed during a given exposure. The primary motivation for this was that it was the most expedient way we could realistically commence work on a large survey of galaxy redshifts.

3.2.1 PC CCD system

To achieve the above we required a second detector, this time to be dye coated by John Barton at the AAO, a new cryostat, pre-amp and electronics designed to perform the switching operation between the 2 CCDs. The system would be tested in Durham using a laboratory CCD camera which had been built during 1988/89 using front-end CCD hardware similar to that in Australia. It was decided to develop the new system interfaced to a PC running GEM Windows 3.01. The 20Mhz 80386 PC, supplied by the U.K. firm Elonex, also included 1Mbyte of RAM, an 80387, 20MHz co-processor, a 44Mbyte hard disk, DOS 3.3, VGA and serial mouse. To manage the operation of the camera, a new application was written in 'C' and 80x86 assembler, the opportunity being taken to add new features to the system software. Extra hardware (figure 2) was also required in the form of- a 48-bit I/O card to communicate with the CCD sequencer, permitting 16-bit control, 3-line interrupt and 16-bit data capability through a new CCD interface. The I/O card was supplied by Amplicon in the U.K. an opto-isolated CCD interface, approximating the function of the CAMAC crate in the system at UKST. Designed and made in Durham, and new image display hardware providing pixel resolution at 8-bits per pixel grey scale or false colour, hardware pan and scroll, image processing, zooming etc., significantly better than the image display at UKST. This was supplied by the Canadian based firm Matrox.

3.2.2 Implementation

It would obviously have been extremely useful to have the dual mode of operation managed from the CCD program, however there was both insufficient time to modify the FORTRAN code on the PDP11 to cope with the dual operation of the system and there was a problem in that the system up to this modification was operating at its limit due to the (now) somewhat out-dated computer hardware. Switching between the two CCDs would therefore be accomplished manually, either from the dome or from the remote location. The additional electronics providing control of the two CCD camera heads being housed in the box containing the new CDS processor and CCD bias supplies. Connections to both the old and new electronics boxes were changed to military specification devices and each CDS box made identical so that it would be a simple matter to revert to the old system configuration should this be required.

Two problems were identified,

(1) how charge on one CCD could be maintained whilst the other was being read-out after charge integration and (2) as horizontal and vertical clocks from the sequencer are supplied to both CCDs, but can only be optimised for one device, it was unclear whether this would compromise the performance of one CCD or the other.

3.2.3 Installation

A simple solution to the first problem was found during discussions with John Barton at the AAO. Charge could be maintained on the CCD not being read-out by supplying both devices at all times with high impedance quiescent voltages (of -9 and -7 volts, similar to those available from the vertical clock driver card during charge read out). The low impedance vertical clocks being easily capable of supplying enough drive to overcome the quiescent state of the CCD electrodes during charge read out. So in addition to the switching function, the new electronics also provides high quality buffered supplies to both CCDs vertical transfer electrodes. During an exposure each CCD simultaneously receives dispersed light from upto 35(CCD I) & 36(CCD II) objects, whilst maintaining the quiescent voltages on the vertical electrodes of both CCDs, sequencing waveforms are applied first to one CCD and then switched to the second CCD and charge read-out from this device.

The system was put together and tested (as far as possible) here in Durham during the Summer of 1989, with the laboratory PC CCD camera acting as the first detector (see figure 3). Due to shortage of time (we were due to observe with the completed system in August/September 1989) we had to wait until the camera was installed at UKST in August, before the second of the problems, mentioned above, could be resolved. In the event it turned out that only slight adjustment to the horizontal transfer clocks was needed to get the new CCD working straight away. There did not appear to be any detrimental change in the performance of the old CCD when running with the new device attached to the sequencer clock drives.

Both CCDs were tested for changes in read-out noise () and CTE when being driven separately or together from the sequencer. The original device showed no adverse changes in either of these parameters when the new CCD was running from the sequencer at the same time. The new device, appeared to be somewhat sensitive to the upper level of in terms of ; there being a factor of 2 improvement in (from to ) for a 0.1 volt downward change in . Small changes in did not appear to have any significant effect on . The new CCD also exhibited good CTE. There did not appear to be any change in these parameters when both CCDs were operating from the sequencer concurrently.

Presently, investigations are in hand to determine whether we could duplicate the PC based CCD system at UKST. The plan has much to commend itself in terms of:-

the PC system would completely replace the ageing PDP11, RL02 disk drive, CAMAC, and tape drive. It is faster, cheaper, more reliable and much easier to obtain spares if these should be required,

better image display hardware, pixels by 8 bits/pixel, full frame buffering, image filtering, panning, zooming etc.,

more flexible system control, the second CCD camera can be controlled from the CCD program. Other features, such as spectrum extraction and manipulation which already exist on the system in Durham, could be added as required,

interaction with the program is via a front-end graphical user interface; the system is therefore very easy to control. There is also a command processor and compiler (for the die-hards) to enable custom exposure sequences to be constructed.

The existing UKST sequencer would plug directly into an interface card which has already been built and tested, hence there would be no changes required to the front-end CCD hardware. The only problem which remains to be solved is that of data archival. Two solutions to this would be:-

(1) to Ethernet data directly to the UKST micro-Vax and, after observing, TRANSFER all CCD data frames to the AAT VAX cluster for data reduction (there is currently no Ethernet link from the UKST micro-Vax to the AAT VAX). Plans are already in hand in Durham to run thin Ethernet around the Physics Dept. with the purpose of enabling user PCs to be interfaced to our VAX cluster. This facility is eagerly awaited as, at the moment, the only way to transfer data to our VAX is via Kermit. This takes minutes (at 4800 baud) per full CCD frame and was one of the main reasons for installing some spectrum extraction routines into the CCD program. This enabled us to select the parts of the data frame which were to be transferred to the VAX, speeding the transfer time by a factor of for the work currently in progress.

(2) install a PC tape drive which are widely available, can hold large quantities of data and would be easily transportable, though it would require the observer's home institution to support such hardware on their main-frame computer.

4. CONCLUSIONS

Reduction of galaxy data (to be published in a forthcoming paper, see also Watson), obtained using the system in both single and dual CCD mode has shown that we can now obtain cross-correlation red-shifts with a high success rate. It was clear from this data that CaII H & K absorption lines, absent from previous un-binned CCD exposures using the original FLAIR fibres, were being detected in single exposures of 3000 seconds (to obtain the desired S/N (), 3 or more exposures are made on the same field which are subsequently added together during the data reduction process). Figure 4 is a representitive spectrum taken from the June 1988 run using FLAIR with the new Polymicro fibres and on-chip pixel binning. This spectrum illustrates the data quality obtained for an E/S0 spiral galaxy after adding 5, 3000 second exposures; the detection of the Ca H & K lines in the blue can be clearly seen. The data were inspected after each 3000s exposure using some rudimentary facilities available during execution of the CCD program. This gives the observer some indication of how the S/N of individual galaxy spectra is progressing as frames are added together. There are no facilities on the PDP11 to remove CREs which become progressively more of a problem as frames are added. Data are archived to a FITS format tape when the RL02 hard disc becomes full, or at the end of the night's observing, enabling data reduction to be subsequently undertaken on one of the U.K. S.E.R.C. Starlink VAX nodes.

5. ACKNOWLEDGEMENTS

The author would like to thank Tom Shanks (Durham) for providing the initial impetus for the project and Fred Watson (R.O.E., Edinburgh) for the work he has done on the UKST FLAIR system. Dick Fong (Durham) provided financial support during the latter stage of the project, thanks go to him and to Mike Breare and John Webster (both at Durham) for their continued support (technical & otherwise) during the project. I would also like to express my gratitude to John Barton at the AAO, Epping for many useful discussions relating to the operational performance of CCDs in general and to the particular problems encountered with our own devices. This work was supported by the UK SERC to which due thanks are also extended.

6 REFERENCES

1 Watson F.G., Proc. S.P.I.E., 627, 787. 1986.

2 Cullum M., Deiries S., D'Odorico S. & Rei R., Astron. & Astrophys., 153, L1. 1985.

3 Deiries S., The optimisation of the use of CCD detectors in astronomy. p. 73, June 17-19. ESO. Ed. Baluteau J.-P & D'Odorico S. Pub. E.S.O. 1986.

4 Waltham N.R., The development of a detector system for faint object spectroscopy on the Isaac Newton telescope, Ph.D. Thesis. University of Durham. 1987.

5 Breare J.M., Cox G.C., Ellis R.S., Martin G.P., Parry I.R., Purvis A., Waltham N.R., Webster J., Fosbury R.A.E., Gellatly D.W., Jorden P.R., Lowne C.M., Lupton W.F., Powell J.R., Thorne D.J., van Breda I.G., Worswick S.P. & Wynne C.G., Mon. Not. R. astr. Soc., 227, 909. 1987.

6 Thorne D.J., Jorden P., Waltham N.R. & van Breda I.G., Proc. S.P.I.E., 627, 530. 1986.

7 Mackay C.D., Charge-coupled devices in astronomy. Ann. Rev. Astron. Astrophys., 24, 255. 1986.

8 Watson F.G., The FLAIR wide-field multi-object spectroscopy system , these proceedings.

Table one - Operating parameters for the 5/10/10BC CCD V refers to a bias electrode voltage; refers to one of the horizontal, reset or vertical CCD clocks; refers to the amount of clock overlap within a 3-phase group; u,l refer to upper and lower clock levels & r,f refer to clock rise and fall times.