Introduction to Applications of CCDs in Astronomy

COOLED CCD imaging systems are already widely used in astronomy as the scientific detectors of choice. Their role in telescope pointing, control and wavefront correction is much less well known. This information is intended as a general guide to the ways that CCD systems may be used for acquisition, guiding and wavefront sensing applications in astronomy.

Image acquisition and guiding of optical telescopes

WHEN A TELESCOPE is pointed at an object the observer may wish to verify that the field of view is as expected. This is achieved by imaging the field directly for a short time, with an exposure of 0.1 to 10 seconds. When using a spectrograph the astronomer may prefer to view the field reflected by the spectrograph slit rather than to look directly at the field. A relatively rapid read-out and display capability is then needed, and the cycle repeated, until the field is centred accurately or the object of interest is positioned across the slit of the spectrograph. This is the acquisition phase of using a telescope.

The next phase is to keep the field precisely aligned for the duration of the exposure which may be as long as 1 or 2 hours in extreme cases. Here the astronomer will want to select a star in a field offset from the main field of scientific interest. It is then necessary to read out a small area centred on the reference guide star fairly rapidly (0.1 to 10 Hz repetition rate) so that any drift from the starting position of the reference object may be compensated. This compensation may be done manually by having the observer or a night assistant guide the telescope directly, or the image may be processed to compute the change in image position allowing the telescope to be controlled automatically.

The Antares family of CCD controllers allow one (or many) sub-arrays to be defined for fast read-out while discarding other parts of the image rapidly - an ideal arrangement for guiding, and the system can change rapidly between full frame read-out for acquisition and sub-array read-out for guiding. Imaging systems for these applications are normally based on one of our compact, maintenance-free thermoelectrically cooled heads (Click here for details) as they have a dark current that is barely measurable in the short exposure times which are needed.

In selecting a system for acquisition and guiding, the choice is between full-frame CCDs, which require the use of a shutter (normally integrated into our heads) or frame-transfer CCDs which use half the CCD area to detect the image. The image is transferred rapidly at the end of each exposure into a separate storage area covered with an opaque mask to prevent light falling on it as it is read out. The advantages of frame transfer devices are that no shutter is needed (giving better reliability) and essentially all of the time is used to detect light from the sky. The drawback is that the field of view is halved since only half of the CCD pixels are used for imaging.

PixCellent provides a range of software packages that will run the controller in these different modes and display the images on the screen. The system runs either on an IBM/PC compatible under Windows™ 95/98 or on a Sun S-bus workstation under Unix. Should you prefer to integrate our controllers into your own telescope control system then we can provide a comprehensive library of control subroutines. Libraries that run under Linux are also available for the Antares range of controllers.

Fast tilt-tip guiding

A MAJOR COMPONENT of the atmospheric turbulence that limits the detail we can see at a telescope is caused by the rapid random motion of the image on the detector. There has been considerable success in recent years in stabilising this motion with systems that consist of a fast guiding camera plus a small mirror which is tilted rapidly so as to compensate for the image wander. The system requires a reference star close enough to the field centre to act as a guide. Under good conditions using this technique it is often possible to halve the diameter of the seeing disk, allowing fainter limiting magnitudes to be reached. This allows more detail to be seen in direct imaging applications and permits narrower spectrograph slits to be used to improve signal to noise and spectral resolution.

The Antares family of CCD controllers are now used on several telescopes around the world for this purpose. In order to achieve the fast frame rates needed (typically at least 100Hz), it is generally believed that only a small area CCD may be used. In fact we have developed methods of using standard size CCDs, allowing them to be used in an acquisition mode as mentioned previously. Here we position the guide star to be close to the CCD output register and close to the on-chip output amplifier. The CCD is read out over a sub-array centred on the star and binned to give a 2 x 2 pixel quadrant detector. As only four pixels are actually read out per frame a slow pixel rate may be used to minimise read-out noise. In order to avoid flushing the rest of the CCD to remove unwanted charge (which is a slow process for large CCDs which can take tens of microseconds for each parallel transfer) we reverse clock the parallel register to push the charge back up the CCD. As only a few rows are involved and the read-time is short, we do not need to operate the shutter which may be left open for the duration of the tilt-tip run. This way of operating a CCD gives the optimal combination of fast frame rates needed for guiding while allowing long pixel times to minimise the read-out noise of the system. It also permits larger area CCDs to be used, and meets the need for a capable direct imaging acquisition system.

The Antares family of CCD controllers offer the flexibility to allow them to operate in this type of mode. NOTE: only certain three-phase CCDs including some from EEV, SITe and Loral permit reverse clocking - the Kodak two-phase CCDs do not. The read-out rate for this application is sufficiently high enough that dark current will be negligible even with thermoelectrically cooled CCD heads and either air or water cooled heat-exchangers.

PixCellent is able to provide all the necessary software libraries for either an IBM/PC or a Sun S-bus based computer system to control our CCD systems in these modes and enable integration of our cameras into your own tilt-tip system.

Performance of the system

TIP TILT CORRECTION as outlined previously is an effective way of improving the overall resolution of an optical telescope. Higher order correction requires fast measurement of the wavefront phase distribution over the telescope aperture or pupil. The most popular method of doing this is to use an array of small lenses in the pupil plane to give an array of star images, one from each lens or sub-aperture of the incoming wavefront. This is the principle of the Shack-Hartmann sensor. The movements of each image allow the phase gradients across each sub-aperture to be measured. The combined data from the array of star images may be used to derive the phase pattern over the telescope aperture and, with suitable intermediate deformable optics, allow the correction of the errors to give a more compact image.

To measure the fastest phase variations a frame rate of several hundred Hertz is required, and the number of lenslets needed depends on the telescope size and the site conditions. A 4-metre class telescope will need at least eight across a diameter and therefore at least a 64-element lenslet array. This format will give an array of 8 x 8 star images which can be imaged conveniently onto a 64 x 64 pixel detector area, implying a pixel read-rate of several MHz.

Careful design of the CCD controller can greatly reduce the pixel rate needed to allow the positions of the star images to be read at high frame rates. Controllers and systems that do this are described in the next section. However, it is important to record wavefront images rapidly so the site conditions may be monitored routinely and that telescope movements and dome effects may be investigated. This 'open-loop' operation is a vital aspect of achieving a good understanding of the operation of the telescope, the dome seeing and the characteristics of the seeing on the site.

The Capella family of controllers may be operated at pixel rates from 0.5 to 5.5 MHz, with 8 MHz capability for set-up purposes with 2-phase CCDs only such as the Kodak devices. Site monitoring generally allows relatively bright stars to be used so the ultimate noise performance is not required and the CCDs can be read out extremely rapidly. Liquid crystal solid-state shutters may be used to give very short exposure times (as low as 100 microseconds) although they do suffer from a peak transmission of only 30 percent. Alternatively frame transfer CCDs may be used. Capella systems are normally used with CCDs that have two-stage output amplifiers such as those from Kodak which give read noise levels of 7 to 10 electrons at 1 MHz and 25 to 50 electrons at 5.5 MHz. CCDs with single-stage amplifiers may be used provided their output gain is high enough. These include the EEV CCD39-02 with 80 x 80 pixels and the CCD37 and CCD47 families of frame transfer devices.

PixCellent is able to supply, in addition to the Capella controllers, complete software and hardware to give sustained data capture at the full speed of the Capella systems. The read-out rate for this application is sufficiently high that dark current will be negligible even with a thermoelectrically cooled CCD heads and either an air or water cooled heat-exchanger. This may be done with either IBM/PC compatible system or with Sun S-bus computer systems.

Wavefront error measurement systems

THE Capella has the somewhat higher read-out noise that is an inevitable consequence of the high pixel rates. However the lates generation of fast, low-noise CCDs that are now being made (such as the EEV CCD-39 with 80 x 80 pixels) allows many wavefront error experiments to be carried out with the Capella controller.

The Capella is an excellent system for fast recording of wavefront data in open loop mode for subsequent analysis. The latest generation of fast ultra-low noise CCDs also allows it to be used as the main wavefront sensor in closed loop systems.

For the very lowest noise read-out, the methods developed for tilt-tip operation of the Antares family of CCD controllers may be extended to two-dimensional arrays of star images such as those produced by a lenslet array for a Shack-Hartmann sensor system. The star array image has to be manipulated so that stars nominally lie on CCD pixel boundaries. The controller is operated in binned mode to synthesise four super-pixels around each star image so as to act as a quadrant detector. The system can compute the star image offset error correction interleaved with the read-out either in the transputer microprocessor contained in the camera controller itself or in the host computer, reducing the time between gathering the light from the star and generatingthe error correction signals for passing to the phase compensation optics. Minimal time delay between gathering the light and producing the error corrections is vital in allowing the slowest possible operation to optimise system sensitivity without compromising correction bandwidth.

With the Antares controller, frame rates of up to 180 Hz may be achieved for a 64 x 64 pixel CCD and an array of 8 x 8 star images. Proportionally faster frame rates are achieved with smaller arrays. Even higher frame rates may be obtained with the Antares Quad controller which has the capabilities of the Antares Duo controller and has the benefit of four independent signal channels so that CCDs constructed with four separate outputs may be driven at four times the pixel rate. In fact it can operate even faster since each channel has the choice of a 12-bit high speed digitiser or the 16-bit converter used in the Antares family of controllers. Image/store section drivers are also faster, so faster parallel transfers are possible. This allows frame rates for a quad output 64 x 64 frame transfer CCD with an array of 8 x 8 star images well in excess of 1 kHz. With a Capella controller and the CCD39-02 from EEV it is possible to operate at 500 frames/sec at full resolution.

PixCellent imaging components for your system

WE REALISE THAT an initial investment may already have been made, and welcome working with you to develop and supply components to assist with your own research programme. One example is in the CCD detector heads used for these applications - with fast frame rates there is no need for the low temperatures achieved with the liquid nitrogen cooled head commonly used by astronomers.

Few observatories have experience of thermoelectric cooler design. PixCellent offers a range of thermoelectric heads that can be customised for the CCD of your choice. They are dry air filled, maintenance free units that are rugged and compact. The convenience of thermoelectric cooling is considerable and well matched to the applications mentioned in this document.


The pages on this Web Site are Copyright of PixCellent Imaging Ltd., Cambridge, England. Reproduction in part is permitted with acknowledgement to PixCellent Imaging Ltd., and of this Web Site Address (
Comments, please, to
Site last updated: 23 June, 1999.