Previous: Summary of Requirements.
Up: Requirements for new CCDs on ING Telescopes.
Previous Page: Summary of Requirements.
Next Page: Existing ING CCDs.

Some Comments on Performance Specifications.

Current CCDs have a wide range of performance characteristics, and somehow we must judge what may be realistic to achieve in any proposed new devices. As an aid to this process we list below the most important factors that determine the scientific usefulness of a CCD.

• Format (i.e. number of pixels, pixel size). This is adequately discussed elsewhere in this report.

• Readout noise. We assume that a low readout noise is one of the most important parameters, especially for spectroscopic sensors. Currently our best CCDs have noise figures in the range ; we should aim for figures in the range in future ( has at least been demonstrated by EEV).

• Quantum efficiency. (Spectral response as a function of wavelength). This topic is discussed in more detail in other sections. In summary, a minimum peak response of 70% is expected. Response at other wavelengths is partly a function of technique, and AR coating. A higher peak may be possible.

• Dark current. The dark current rate must be low enough so that it's contribution to total read-noise is not significant in the longest anticipated exposure (). With cryogenic cooling we expect to achieve , so that this should not be a real problem.

• Cosmic ray sensitivity. The incidence of cosmic rays can be the main factor that limits the duration of an exposure to 1 hour. For low-background spectroscopy it is particularly important that the CCD package does not generate additional events, so that the limiting factor is the canonical rate of true cosmic ray events ()

• Dynamic range and linearity. We should expect that new devices perform well over a wide dynamic range. The full-well signal is mainly a function of pixel-size (although an MPP design can reduce it further). We expect a full-well of at least for pixels, although our CCD electronic systems with 16-bit ADCs may not always cope with this range. We expect that the CCDs will be linear over a wide signal range, with no threshold of sensitivity.

• Charge transfer efficiency (CTE). For CCDs that are intended for operation at low signal levels, it will be important that they exhibit excellent CTE. Adequate performance levels have been demonstrated in several current devices made by EEV, Loral and Tektronix, where CTE's of have been achieved.

• Uniformity of response. It is important that the uniformity of response is quite good, and this performance is maintained over the whole imaging area; this should ensure that subsequent data calibration can be made with confidence. We should therefore look for a global non-uniformity of better than 10%, with a small-scale non-uniformity better than 2%; this may be a function of wavelength.

• Density of defective columns and pixels. This cosmetic classification is hard to quantify. Current top-grade devices have no defective columns and pixel defects (traps). For proposed larger devices we should expect a small number of columns () and a small number of traps (). For some (spectroscopic) applications it may be more important that the central zone exceeds the above specification.

• Temporal stability of response. We need to ensure that the spectral response (and electronic gain) are stable with time. Changes in QE, over a period of weeks, has been a concern for some previous flash-gate thinned devices.

• Undesirable effects. (Eg. Image retention, slow clocking speed, non-flat surfaces). Some devices (notably those produced by Thomson) have exhibited some image retention even after clearing the CCD of charge. This is an un-wanted feature and should not be present in new devices. Tektronix thin CCDs require slow parallel clocking, which would only be a limitation in high-speed read-out applications; these devices also suffer from non-flat image planes. We would expect to acquire devices with a flatness of .

• Long-wave interference fringeing. It may be that all thinned devices exhibit this effect; we will aim to have this minimised if possible.

• Readout architecture. The most useful read-out architecture is that provided by a 4-corner design. We aim to get devices with several outputs to provide flexibility, redundancy, and the highest read-out speed possible consistent with other performance characteristics.