Expected performance and future prospects

The data analysis is not yet finished, with various tests to check the power spectrum estimation still to be performed. Using the complete data processing pipeline and the known noise properties of our data, we have performed simulations for 10 detectors, based on the noise power spectrum (in the time domain) of our best channel. The power spectrum estimated from these simulations is shown in Figure 8. It is the average of power spectra given by each detector, weighted by their correlation matrices. In addition we performed an average over several simulations to check that the method was not biased.

Figure 8: Power spectrum estimated from simulations, using 10 detectors with our best noise level, and averaging their power spectra using their correlation matrices. The estimation is centered on the input model as it is the average of one thousand realizations, showing that the method is unbiased. Error bars are very small at large angular scale before the first acoustic peak at l=200 due to our large sky coverage, but increase quite fast afterwards on small angular scale due to the short duration of our flight.

The error bars are very small (smaller than available measurement) on large angular scale, because of our large sky coverage, and then increase after the first acoustic peak because of the small signal-to-noise ratio due to the small integration time. This is the motivation for a longer duration flight of at least 24 hours, as the signal-to-noise ratio increases nearly quadratically after the first 10 hours (because the sky coverage does not increase further). Two flights are therefore scheduled for next Winter (December 2001 - January 2002), from Esrange, with the same configuration. The expected performance for a 24 hour flight with 10 bolometers is shown in Figure 9.

Figure 9: Power spectrum estimated from simulation : same method as for Figure 8, but for a 24 hours flight. The determination of the first peak is nearly perfect, and the measurement is good until the second peak.

The power spectrum estimated this way (Figure 9) is very accurate up to the end of the first peak and gives good constraints on the second one, allowing precise estimation of cosmological parameters.