Occasionally, there may be reason to suspect that the performance of UCLES is abnormal, but there may be no clear way to prove this or to diagnose the problem. This Web page provides a set of "benchmarks", against which the current performance of the system can be compared. If you suspect a problem, please alert your support astronomer, the night assistant and/or the afternoon technician, and the instrument scientist Stuart Ryder.
Count rate is suspiciously low
- If you are not getting the count rate that you expected on a star (and it really is photometric out there), then first of all check with the UCLES Signal-to-Noise Calculator. The number Nobj should agree, to within 20% or so, with the total counts in a single pixel (as a first approximation, take the peak height above the background, and multiply by 1.07 x the FWHM of the spatial profile). Also, be sure to input the magnitude of the star nearest the wavelength you're observing at (e.g., if observing near 6500 Å, use the R magnitude). If you only know the V magnitude (say), then consult a table of colour as a function of spectral type and luminosity class (e.g. Table 15.7 of Astrophysical Quantities, 4th edtn.)
- Don't forget that due to camera vignetting, throughput estimates must be based on wavelengths near the centre of the detector. This is especially true with the MITLL3 and EEV2 2K x 4K devices, which suffer as much as 50% vignetting at the edge of the full readout window, and at least 10% in orders more than halfway from the centre to the edge of the array.
- Since the slit losses due to seeing, etc. are based on Gaussian models, the only definitive way to measure the UCLES throughput is to observe one of the Hamuy et al. spectrophotometric standards. Make sure you observe with a wide (>5") slit, and with the beam rotator removed. Measure the counts in a pixel near the array centre at a known wavelength, and compare with the tabulated fluxes to derive the equivalent AB magnitude of the star which would result in one photon/sec/Å at zero airmass (see the UCLES throughput history for a description of the procedure).
- If it is cloudy, or not yet dark outside, you will have to rely
on internal lamps to test the throughput (this can also be used
to try and isolate a problem to either the spectrograph, or the
telescope). Unfortunately, this is not as straightforward as it
may seem, as all lamps change their output over time. For this
reason, only representative values can be given. The table below
lists the observed count rates (ADU above bias, at the peak of the
order nearest the quoted wavelength, per unbinned pixel,
per second) from the quartz lamp, with a 1" slit and NORMAL readout
Date Detector Grating Wavelength Count rate 991022 MITLL2 31 6182 586 991022 MITLL2 31 3946 9 000122 MITLL3 31 6690 1440 000122 MITLL3 31 4550 64 000814 MITLL2A 31 4307 12 000815 MITLL2A 79 5126 127 000905 TEK 31 4897 61 000905 TEK 31 4211 15 000905 TEK 31 4180 14 000906 TEK 31 4180 14 000907 TEK 31 4180 14 000908 TEK 31 4180 14 000909 TEK 31 4211 15 000909 TEK 31 4211 15 000910 TEK 31 4211 15 000911 TEK 31 4211 15 000913 TEK 31 4211 15 000913 TEK 31 4180 14
The lines don't seem to be in focus
- Since the slit width can be set to a specific number of pixels on the detector, the FWHM should be comparable to that specified. However, if the wrong pixel scale was set on startup, or on-chip binning was not allowed for, the conversion will not be done correctly; check the Detector and UCLES Wavelength Setup tabs for these.
- The large depth of field of UCLES (and consequent small range of travel of the collimator) means the instrument is unlikely to get subtantially out of focus. If the focus is suspect, then carry out the proper focussing procedure using the Hartmann mask.
I don't recognise any of these lines...
- One of the problems with UCLES is that it produces a bewildering number of lines, especially with the MITLL or EEV CCDs. First of all, make sure that the grating really is the one you asked for - the echellograms for the 31.6 and the 79 l/mm gratings are quite different. Setting up is easier to do with the 79 l/mm grating (since this gives the greatest inter-order spacing, and potentially the least confusion), even if the 31.6 l/mm grating is to be used. Once the appropriate offsets have been defined for one grating, they should serve equally well for the other. But it pays to check that the grating in use matches that from which the arc atlas was produced.
- Apparently, if the polarity of the ThAr lamp is reversed, a quite different spectrum will result, so check this too.
- Given the size of the current CCDs, it is unlikely that the desired central wavelength is not somewhere on the detector. Even so, it may be best to reset all the grating angle offsets to 0.0, as the previous user may have used an unconventional setup.
- There aren't that many obvious patterns of lines in the blue
(there is an image of the ThAr lines near 5009 Å with the Tek
in Appendix C3 of the
UCLES manual), but
out beyond 7000 Å, there are a number of lines which are bright
enough to saturate in just 1 second. Set the central wavelength to 8115
Å with the 31.6 l/mm grating, and see if you can't recognise
this line (circled below) and surrounding lines.
- A set of reference images captured from Ximtool is also given in the Table
Date Detector Grating Wavelength Window Image 991022 MITLL2 31 6182 MITLL_PLANET_BX2 991022_19 991022 MITLL2 31 3946 MITLL_PLANET_BX2 991022_55 000122 MITLL3 31 6690 MITLL_PLANET 000122_10 000122 MITLL3 31 4550 MITLL_PLANET 000122_16 000814 MITLL2A 31 4307 MITLL_PLANET_BX2 000814_15 000815 MITLL2A 79 5126 MITLL_PLANET_BX2 000815_01 (a)
Stuart Ryder, sdr -@- aao.gov.au