Spectral response and global efficiency
The total throughput of ESPaDOnS as estimated from the measurements of the individual
optical components (full line on graph) should peak at about 19%
around 500nm (telescope and detector included), dropping down to about 2% at 370 and 1000nm. The combined efficiency of the
telescope (at Cassegrain focus), polarimeter, fiber link and slicer (dashed curve
on graph) is roughly flat down to 400nm and equal to about 40% on average, while that
of the spectrograph and ccd detector (dotted curve on graph) peaks at about the same
value but strongly drops towards both ends of the spectral domain (the red drop
reflecting mainly the decrease in ccd efficiency).
In addition to this, one must take into account the light losses at instrument aperture
(about 10% in median seeing conditions) and through the atmosphere (about 10% for an
average airmass of 1.5), bringing the peak total efficiency in average observing conditions
at a level of about 15%.
To compare the above estimate with actual measurements, the best way is to proceed in 2 steps.
The first step consists in measuring the normalised spectral response of the instrument through
observations of stars with known surface temperature. In this purpose, we used observations
of solar-like stars, with temperatures ranging from 4,500 to 6,000 K, and derived a model
spectral response by matching the variations with wavelength of the signal to noise ratio
(S/N, reflecting directly of the amount of detected photons) within the collected spectra.
Our result is shown on the right, the full line depicting the measured
spectral response of ESPaDOnS,
compared to the spectral response predicted from individual components (dashed line).
Both curves are scaled to the same throughput in the red part of the domain. It shows that
the instrument is less efficient in the visible than predicted. One possible origin for this
loss is that the EEV CCD has a lower efficiency than expected in the blue.
Once the relative spectral response is modelled, the second step is to evaluate the absolute
efficiency of ESPaDOnS by observing a few selected stars (with known temperature and brightness)
at different epochs. For instance, the figure on the right side presents an observation of the
cool star HR 1099 (temperature 4,800 K, magnitude 6.0) obtained on 2004 Sep 03 (the clearest
enginneering night with ESPaDOnS). By fitting the modelled instrument response
(full line) to the measured signal to noise ratio (per 2.6 km/s bin) as a function of wavelength
(open dots), we obtain that the instrument is about 20% less efficient than the model response
(10% smaller signal to noise ratios). It implies that the instrument peak
efficiency is of order of 12%. The origin of this additional loss is not completely
clear yet, but may be due to a construction problem in the optical fibre bundle.
© Jean-François Donati, last update Nov 16 2005