CCD#40 on EFOSC2

Table of Content

Summary of Parameters

Type Loral/Lesser, Thinned, AR coated, UV flooded, MPP chip
Controller ESO-FIERA
CCD Size 2048 x 2048
Image Size 2060 x 2060 (overscan inadequate for bias subtraction)
Pixel Size 15 microns  x 15 microns ;  0.12 arcsec x 0.12 arcsec
Field Size 4.1 arcmin x 4.1 arcmin  (vignetting in corners, ~8% of detector area)
Full well capacity 104,000 electrons/pixel
Dark Current 7 electrons/pixel/hour
Saturation 65535 ADU
Linearity regime 0.25% (maximum deviation)


Readout Mode
Bias  (ADU)
Readout Noise (electrons)
Gain (electron/ADU)
CCD readout time (1x1 binning)
CCD readout time (2x2 binning)

Note : The CCD is monitored regularly and the latest results, as well as the evolution of its properties (bias, gain, linearity, noise, etc.) may be seen on the CCD tests pages below.


CCD Tests

Following the links below will lead to results from the latest tests conducted on the CCD using a Beta light (radio active source). The values monitored include bias, gain, readout noise, linearity, shutter error, etc. The pages also display evolution of the values over the last couple of years. The first set of links will take you to the new dynamical pages, with the most recent results, while the other links are for historical information only.

Current CCD#40 Left Amplifier - Dynamic Monitoring Pages on wiki internal page
Current CCD#40 Left Amplifier -
CCD#40 Right Amplifier. Not in use since 21 March 2003


CCD#40 has (but is not just) a pretty face! Apart from a sprinkling of bad pixel clusters, most of which are just 1-5 pixels in size, all the others are concentrated either on 3 (and a quarter) columns or are on the margins. The Bad pixel masks page has a more detailed description of the CCD cosmetics and downloadable masks.

Cosmic Ray Events

The measured cosmic ray impact is about 900 events per hour over the full CCD. A hit typically covers several pixels (radius of 2-3 unbinned pixels). Exposures in excess of 20-30 minutes are not recommended for imaging programmes and 45-60 minutes is the upper limit for spectroscopy. Observers should plan on taking at least 2 and preferably 3 exposures per field to detect and eliminate cosmic ray hits.

Dynamic Range and Linearity

The CCD has a saturation limit of 65535 ADU (dynamic range of 16 bits ADC) which is considerably lower than the actual well saturation corresponding to 80000 ADU. The CCD is known to be linear to better than 0.2% over the entire range up to the software saturation. Thus the entire available range up to 65535 ADU can be used for scientific purposes.

Even in case of very strong "super-saturation" during a particular exposure the CCD recovers quite well with little remanence. In the worst case, in our experience, taking 3-5 biases will completely remove any lingering residuals.

Field & Orientation

Each observation produces a 2060x2060 (unbinned) image but about 25-30 columns/rows on the margins cannot be used. The scale on the CCD is 0.12 arcsec/pixel and so the useful extent of the CCD field is 4.1 arcminute on each side.

The default orientation produces an image with north towards the top and east towards the right (i.e. flipped in the x direction compared with the sky). See the Adaptor page for more details and instructions on how to change this orientation. Note that the real-time display can flip the image back to North up and East left for comparison with finding charts, although the interaction with MIDAS (e.g. for centering targets on slits) must be done in frames displayed as they appear on the CCD.

A spectrum is always produced with the slit (or slitlets) aligned parallel to the horizontal axis (x-axis) and the spectral dispersion along the vertical axis (y-axis). The wavelength increases as one goes from bottom to top.


The thinned CCD chip shows variations in its thickness and this is manifested in a fringe pattern when the CCD is illuminated by monochromatic light. Basically multiple internal reflections below the chip surface lead to interference patterns when the thickness is of the order of the coherence scale of the incident light. Thus red grisms lead to a considerable amount of fringing beyond 8000 A. The presence of night sky lines in the red also results in fringes seen across the CCD in the red filters (R-band, i-band and z-band).

In spectroscopy, regular internal flat fields are sufficient to take out the fringing from the science spectra. In imaging, one needs sky/dome flats as well as a fringe flat made from combining the science frames (the super flat) to eliminate the fringes.

Quantum Efficiency

The CCD quantum efficiency as a function of wavelength is shown below. The back surface of the CCD is periodically flooded by UV light to improve its blue efficiency.

Click on the image for a larger version


The shutter is not strictly a part of the CCD but since their roles are so intimately connected we will discuss the shutter here. The shutter used on EFOSC2 is of the iris type and so the illumination is not uniform across the CCD. CCD tests show that the shutter error is 20-30 milliseconds including a 25 millisecond shutter delay and a +/-10 millisecond exposure variation across the CCD.

So a 1 second exposure will reduce the position dependent error to less than 1% while a 3 second exposure will reduce the shutter delay error to the same level. At this stage the final photometric errors (typically 5-10% for EFOSC2) will be dominated by contributions from other factors.

Thus flat fields should not be taken with an exposure of less than a second - in fact, the observing procedure does not allow it. However, we recommend a minimum exposure of 5 seconds, and for dome flats generally use exposures times > 10 secconds.

Operating Modes


CCD #40 has a scale of 0.12 arcsec/pixel. The best seeing at the NTT is  ~0.4 arcsec while the typical value is in the range 0.8 - 1.2 arcsec. Consequently, the image is adequately sampled in almost all cases even with the CCD binned 2x2. The only case where binning 1x1 is possibly required is for spectroscopy with a 0.5 arcsec slit. Soon EFOSC2 will also have new 3x3 and 4x4 binning modes for increased S/N in poorer seeing conditions.

Unless the scientific goal specifically requires otherwise we strongly recommend use of 2x2 binned mode - the main advantage is the higher read-out speed. The smaller image sizes is also an advantage though not a major issue in these times of disk space profligacy!  It should also be noted that one can execute a sequence of 5 twilight imaging flats in only 1 band if 1x1 binning is used.

Unequal binning along the X and Y directions is possible. However use it at your own risk as routine CCD tests are not yet done in this mode. Make sure that enough calibration data is taken. It is planned to offer unequal binning with various combinations up to 4x for spectroscopy with EFOSC at the NTT, but this has not yet been commissioned.

Read-out Speed

The read-out noise during the Slow mode is only 1 ADU lower than for Normal mode (6.4 and 7.4 ADU, respectively), and this difference is insignificant when, as is usually the case, the the image is sky noise limited. Since there is a considerable difference between the time taken to read the CCD in the 2 modes there is almost never a need to use the Slow read-out mode.

The Fast mode is about a factor of 2 faster than the Normal mode and the read-out noise is only a couple of ADU higher. However, two different amplifiers are used during the read-out and the 2 halves in effect behave like different chips with different calibrations (bias etc). It may be noted that the time difference between the Normal and Fast read-out modes (~15 seconds) is only a very small fraction of the total time overheads and so it usually doesn't save much time in practical terms.

However the fast mode with CCD windowing is an option for fast photometry of variable stars.


Windowing the CCD is an option but unless the window is defined close to the upper edge  - the CCD is read upwards - there is no significant saving in read-out time, especially in the normal readout mode.  Further, EFOSC2 is a focal reducing instrument and the image quality is considerably worse at the edges and so it is better to use the central region for observations without windowing.