Costos directos

The SuperWASP instrument is currently achieving levels of precision very close to it’s optimal performance over the entire magnitude range of sensitivity. The hardware upgrades and implementation of detrending algorithms has reduced the e↵ects of systematic errors on single exposures to a negligible level. This is also partly true on time scales comparable to the duration of planetary transits where for faint and moderately bright stars the sky background and photon noise from the targets dominates. However, the stars of interest for this survey are the brightest, where follow up studies are more likely to be successful. The bulk of these targets are reaching precisions of around 4 mmag on hourly time scales and at this level of precision there is evidence of a small fraction of residual systematic noise.

The current flat fielding strategy employed in the SuperWASP instruments is introducing unwanted systematic e↵ects into the light curves due to the wavelength dependence of the CCD chips. The broadband filter used in the telescopes makes them susceptible to e↵ects of this kind and the wavelength di↵erence between the twilight light and the stellar light is an example of a possible source of noise. An analysis of the SuperWASP detector maps has not only revealed this problem but also served as a diagnostic study for the performance of the CCD chips in terms of non-linear pixels and issues with the pipeline software. It can be used as a basis for selection of further pixels to reject from the software reduction and has revealed the presence of features of unknown origin that a↵ect large portions of the chip and that warrant further study.

The amplitude of the systematic e↵ects these features could introduce in the light curves has been found to be negligible under the typical observing strategy of the instrument, where on average 16 exposures are taken every hour for a given field. However, a recent staring observing strategy employed in the northern facility has increased this number to over 90 exposures per hour in order to increase the chances of finding longer period planets, as well as improving the photometric precision on time scales comparable to the planetary transit duration. This has generated a data set with the potential to reach photometric precision similar to the typical amplitude of the noise generated by the detector map features on time scales of one hour. However, the use of a fractional RMS plot as a function of magnitude shows a clear photometric limit inconsistent with the expected modelled noise, suggesting the existence of systematic noise sources at the 1mmag level. Using the detector maps as a decorrelating step has revealed no significant improvement on the brightest targets in the field and was therefore deemed unsuitable at this time. The results show a

distribution that is consistent with the implementation of a random multiplicative function of negligible amplitude compared with the existing noise levels. This can be explained by a number of scenarios:

• It is possible that an unknown source of systematic error is limiting the preci- sion at the 1mmag level, thereby causing the detector map to be an ine↵ective means of compensation.

• The typical fractional RMS of the detector map series shown in Figure 2.22 for a staring field appears to be larger than that of Figure 2.20. This is likely to be related to the fact that the non-staring data was taken in 2009 whilst the staring fields are all observed in 2011, where only half a year of data were used for the creation of the detector map. The detector maps require a very large number of measurements to achieve the S/N necessary to visualise the features seen, typically entire seasons. The noise present in the 2011 detector maps is likely larger that any other previous year, which leads to a larger fractional RMS of the detector map series generated from it. It is possible that a map created with the full data set for 2011 (not yet processed completely) would provide a more accurate measurement of these features and would potentially serve as the basis for an improvement in the photometry of bright stars.

• Using the detector map as a basis for decorrelating every star may be unsuit- able in principle, since this would be compensating for a feature resulting from a wavelength dependence, whose amplitude is likely to be di↵erent for every star. A more profound understanding may be required in order to scale the detector map series to the color of every star and make this decorrelation a viable step.

Despite the fact that at this time systematic e↵ects visible on the detector maps appear to be negligible in the bulk of SuperWASP light curves, this analysis is relevant to any survey using CCDs for optical astronomy. Knowledge of such noise sources has had an impact in the design and operation strategy of the Next Generation Transit Survey (see 1.4.3). They have motivated the need for precise guiding to take place and the choice of 600nm as the low wavelength cut-o↵for the instrument filter in order to reduce the e↵ects of wavelength dependent flat-field noise. These results are also applicable to any project aiming to reach sub mmag photometric precision.