S/N : Actual vs. the Hiedelberg ETC

There is some concern over the accuracy of the predictions of the Hiedelberg ETC. Here we try to address some of the issues.

First we begin by examining a specific test case. The spectrophotometric standard star HR718 was observed during the FEROS-II commissioning during November 2003. Observations of this star demonstrated that a system efficiency of ~19.5 had been achieved (very close to the theoretical maximum of ~21) (see figure 1)

Efficiency HR718
Figure 1: Measured efficiency. Spectrum is a 120sec exposure of HR718. (Image frame is FEROS.2003-11-15T03:46:08.507.fits).

On the basis of this observation and a second (FEROS.2003-11-15T03:49:49.307.fits) in which an equally good efficiency was achieved, we assume that FEROS is performing very close to its theoretically expected best.

Hiedelberg ETC calculations for HR718

The Hidelberg ETC give the results tabulated in table 1 for HR718, both for the default parameters of the ETC and the actual parameters measured during the FEROS.2003-11-15T03:46:08.507 exposure.


 U
 B
 V
 R
 I
Default Parameters
 93.5
 703.7
 658.9
 597.3
 438.5
Actual Paramters
 113.4
 835.1
 774.5
 699.1
 510.4
Table 1: Expected S/N calculated with the Hiedelberg ETC. For the Default Parameters only the V magnitude (4.28mag), the spectra-type (B since HR918 is a B9III) and the exposure time (120sec) are adjusted from the default values set by the ETC. For the Actual Parameters in addition to the above the following settings are made: the moon brightness to dark, the seeing to 0.95 arcsec, the airmass to 1.27, the temperature to 12.5°C, the atmospheric pressure to 772 hPa, the humidity to 23%, the CCD ReadOut Noise to 5.0 electrons

Measured S/N

The S/N measured in the spectrum.


 U
 B
 V
 R
 I
Standard reduced spectrum using snrUBVRI.prg 5
 283
 288
 217
 174
Standard reduced spectrum using snrscanUBVRI.prg 5.6
 857.7
 731.3
 661.1
 452.7
Non Flatfielded spectrum using snrscanUBVRI.prg 36.6
 729.6
 742.1
 612.1
 157.6
Table 2: Measured S/N calculated.  The snrscanUBVRI.prg script calls the snrscan.prg MIDAS script of the FEROS-DRS. The snrUBVRI.prg and sncscan.prg scripts just mentioned call the snr.prg MIDAS script of the FEROS-DRS for the actual S/N calculation.

The snrUBVRI.prg script calculates the S/N for each reference filter (UBVRI) using intervals of 7, 10, 10, 18 and 16 nm respectively centered on the central wavelengths of the filters (361, 441, 551, 658 and 806 nm).

The snrscanUBVRI.prg script scans the S/N over a range of wavelengths using a interval of 4nm (which includes approximately 135 data points) to calculate the S/N to find the maximum S/N in this region. The intervals are in general offset from the central wavelengths of the UBVRI filters in order to sample the spectrum in the middle of the blase function of the relevant order, since of course the S/N is a function of where on the blaze function of each order the measurement is made.

The snrscanUBVRI.prg script is thus MUCH less sensitive to absorption/emission lines which strongly bias the S/N calculation made by snr.prg.

Discussion

The S/N for BVRI computed using the snrscanUBVRI.prg script on the FlatFielded spectrum compare quite favourably with the predictions from the Heidelberg ETC.  The S/N for U does not, but this is to be expected because the current Calibration Unit delivers almost no light at this wavelength (i.e. the bluest order), hence the S/N in the Flatfielded spectrum for U is dominated by the devision by the Flatfield with zero signal.

We see from the S/N calculations on the Non Flatfielded spectrum using the snrscanUBVRI.prg script that the S/N in U is considerably better in fact, though still somewhat less than predicted.