Flux standard stars (STD) are observed on a regular basis with UVES on Paranal. Their purpose is to monitor the overall instrument efficiency (DQE; telescope+instrument+detectors) and to provide a measurement of the spectral response. They are taken with the slit width set to 10 arcseconds.

The flux standard table is delivered as part of the Service Mode package. It can be downloaded as flxstd.tfits.Z (576 kB). It is based on the ESO library of standard stars. The extinction table (also delivered) can be downloaded here.

Strategy for acquisition

Two kinds of standard star observations exist: daily measurements close to morning or evening twilight and long-term measurements during night-time.

Daily measurements: The acquisition strategy for daily STD measurements has been changed in November 2002. Now, standard stars are observed in three dichroic settings (346+580, 390+564, and 437+860, 1x1 binning) only, independently of the settings used for science observations during the night. These three dichroic settings are measured every Service Mode night close to twilight with standards taken from a small set of stars with few spectral features. They are used for daily performance checks and as input to create master response curves for flux calibration of science spectra.

Long-term measurements: Since April 2002, a dedicated set of standard stars is oberserved about twice a month during night-time under photometric and dark conditions and at low airmass. The reason for this strategy is to avoid the large intrinsic scatter (see below) of the daily measurements taken close to twilight. These measurements are used for long-term trending of the overall instrument efficiency.

QC1 parameters: chromatic efficiency

The pipeline processes the raw STD data (de-bias, flatten, average extraction), divides the flux table into the result and finally transforms into a fractional efficiency (where 0.1 means 10% of all incoming photons are registered on the CCD). The pipeline delivers an efficiency table in tfits format (called UV_PEFC). Find more about the reduction of STD files here.

The efficiency table can be plotted, per order, over wavelength. The blaze wavelength of each order and the corresponding efficiency value there is written as QC1 parameter into the FITS header. The corresponding numbers for the central order are stored in a database as BL_WLEN and BL_EFFIC. The BL_EFFIC parameter is available for trending, assuming that all efficiency curves per setting are similar apart from a scaling factor. Their trending with time primarily reflects efficiency variations of instrument components.

 

Trending: Chromatic efficiency DQE

The monitoring of the BL_WLEN and BL_EFFIC values for all standard settings is used to

• monitor the instrument performance in all standard settings over time
• construct the overall UVES chromatic efficiency function by combining the available values BL_WLEN and BL_EFFIC per setting, for the whole quarter investigated.

It has turned out that the proper selection of data is crucial. Accepting all BL_EFFIC values as produced by the pipeline produces a large scatter easily in excess of 10%. Reasons for this are:

• there are non-photometric nights
• there are standard stars visible from Paranal at relatively high airmass
• often twilight shots have a high SKY contribution
• sometimes data are taken under bad seeing
• there are sometimes STD frames taken with less than 10" slit width
• some STD stars have a flat continuum, others have a rich absorption line spectrum

High-airmass data are observed to have the spectrum in the blue exposures shifted off-center which, especially in combination with a high sky level, brings the pipeline algorithm to assume an unrealistic background, with odd results for the efficiency.

The same effect is observed under bad seeing since then the SKY level is unreliable as well.

Stars with a rich absorption line spectrum tend to produce a larger scatter since BL_EFFIC is monochromatic.

The trending procedure therefore rejects all values taken at AIRM > 1.4 for all 346 and 390 nm settings. Also, all entries with slit width < 10 arcs are discarded. For all other effects, a statistical approach to reject outliers has been adopted. The distribution histogram of all points per setting is created. The maximum of this histogram is chosen to give the nominal average BL_EFFIC value. This approach takes into account that several selection effects work in an asymmetrical way to distort the true distribution. E.g., bad nights always produce lower numbers. Bad SKY subtraction also tends to reduce efficiency numbers.

The trending plots (see figure) show the individually measured efficiency values as small circles: the results of the daily twilight measurements are indicated as red open circles (for 1x1 binning) or as filled blue circles (2x2 binning), respectively. Filled magenta circles denote measurements during night-time for long-term trending. The latter show much smaller scatter then the twilight observations. The maximum of the histogram, the rms error of the average, and the number of points used are also indicated on the trending plots.

There exists also a separate plot for long-term trending of the night-time STD observations (see current plot as example).

All average efficiency values per central blaze wavelength are plotted in diagram 2. They construct the chromatic UVES efficiency curve. Dots are all entries, full blue circles are their averages. A more detailed version of this curve is available further below.

Here are some typical values observed for the UVES DQE in the period JAN-MAR 2002. These numbers are taken from the trending plots.

Note: Routine recording of these numbers started with Oct 2000. The 2000 values, however, show a large scatter since at that time no SKY subtraction was performed by the pipeline for STD frames. Hence these numbers are not available here.

The above trending tracks the evolution of the UVES efficiency at certain isolated wavelength points, namely at the blaze wavelength of the central order of the selected settings. From the data points for the blaze wavelengths of all orders, it is straightforward to construct a more complete UVES efficiency curve.

For this process, it is even more important to avoid systematic errors, the most important of which would be bad nights. Hence the following selection criteria are applied:

- use the histogram maxima (see above) for the selected period as starting value;
- select only those efficiency curves which have a BL_EFFIC value which is within 5% of the histogram maximum;
- for these measurements, average for all orders the efficiency value at blaze maximum.

Results for different periods are available through the query interface. Four examples are shown here:

The following table shows average efficiency values for all settings during different periods:

MJD-51000 = 1300: February 2002, after recoating
MJD-51000 = 1791: June 2003, before cleaning and recoating
MJD-51000 = 1830: July/August 2003, after cleaning of M1 and M3 but before recoating
MJD-51000 = 1901: October 2003, after re-coating of M1-M3
MJD-51000 = 2187: 01JUL-10OCT 2004, before exchange of blue CCD
MJD-51000 = 2292: 14Oct-30Nov 2004, after exchange of blue CCD

From a comparison of the June 2003 with the February 2002 values, an overall efficiency loss of about 9% per year can be determined. The table shows also that the mirror cleaning accounts for 80% of the gain in efficiency and the recoating for 20%.

Response curves

The UVES pipeline produces response curves from STD calibration frames. The results from pipeline versions earlier than 1.6 were, however, of limited value for direct use since they were not binned to the resolution of the reference flux table (typically 50 or 16 A), were not normalised for exposure time, gain, and binning, and did not contain an extinction correction. These shortcomings have been removed with version 1.6.0 of the pipeline. However, all response curves suffer frequently from high SKY background (since they are taken during twilight) and from high exposure level so that the extraction quality is low. There is the same scatter of average level as observed for the efficiency data.

Therefore, a set of master response curves has been created. This set can be used for flux calibration of the reduced science data. Presently these curves exist for the period after 2001-10-01. Earlier data taken with the same filter and grating have the same shape of response curve, but scaled down.

Major events that affected the shape and the level of the response curves:

• The blue filter CUSO4 was replaced at the end of November 2001 (see above).
• At the end of January 2002, UT2 mirrors M1-M3 have been recoated.
• Gratings #1B and #4B have replaced gratings #1 and #4 in 2001 and 2000, respectively.
• Since pipeline version 1.3.3, the optimum extraction algorithm has been changed so that the flux of the extracted spectra is a factor of 1/sqrt(2) lower compared to previous versions. This affects all delivered extracted spectra from Service Mode programmes observed after 2002-07-10. For these spectra, a new set of response curves has been created from standard stars observed after 2002-08-01.
• The mirrors M1-M3 have been cleaned on 2003-07-06 and recoated early September 2003.
• The blue CCD has been exchanged on 2004-10-13.

Note: Some response curves have been edited to replace artificial features where the standard stars have strong absorption lines. In the 346 and the 860REDU curves, some points have also been extrapolated in the UV and IR part, respectively, since the flux tables do not extend into these regions.

Note: Standard stars are reduced using optimal extraction (see science recipe for details). The resulting master response curves can also be applied to average extraction (including slicer observations).

Note: Accurate absolute flux calibration cannot be achieved but relative variations of the response curve can be effectively corrected. For science data reduced in the same way as the standard stars the estimated absolute flux calibration accuracy is perhaps 10% but almost certainly NOT better than this and quite possibly worse due to changes in the optical system (e.g. dust on the mirrors) with time and variations from night to night of the atmosphere itself.
In order to make an accurate flux calibration to higher precision than is possible with the master response curves, you need to consider standard stars observed as close as possible in airmass and time (certainly within the same night) as the science and with the same instrument setup as the science, in particular the slit must be the same and normally a 5-10" slit should be used to minimise slit-losses for both science AND standard star.

download fits files view plots (listed is name of order separation filter and grating ID which were in place) applicable range in time: blue arm 1) red arm 2) BLUE 346 BLUE 390 BLUE 437 REDL 564 REDL 580 REDL 860 REDU 564 REDU 580 REDU 860 effect of recoating     example example example example example example example example example 2001-06-01... 2001-09-30 UV_MRSP_010601_BLU.tar n/a CUSO4/CD#1 n/a CUSO4/CD#2             2001-10-01...2001-11-28 UV_MRSP_011001_BLU.tar UV_MRSP_011001_RED.tar CUSO4/CD#1B n/a CUSO4/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B 2001-11-29...2002-01-31 (blue filter replaced) UV_MRSP_011129_BLU.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2         after 2002-02-02 (mirror recoating) UV_MRSP_020202_BLU.tar UV_MRSP_020202_RED.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B after 2002-08-01 (pipeline V1.3.3) UV_MRSP_020801_BLU.tar UV_MRSP_020801_RED.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B before 2003-07-05 UV_MRSP_030501_BLU.tar UV_MRSP_030501_RED.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B 2003-07-06 ... 2003-09-06 (mirror cleaning) UV_MRSP_030706_BLU.tar UV_MRSP_030706_RED.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B after 2003-09-19 (mirror recoating) UV_MRSP_030919_BLU.tar UV_MRSP_030919_RED.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B before 2004-10-12 (blue CCD exchange) UV_MRSP_040928_BLU.tar UV_MRSP_040928_RED.tar HER_5/CD#1B HER_5/CD#2 HER_5/CD#2 SHP700/CD#3 SHP700/CD#3 OG590/CD#4B SHP700/CD#3 SHP700/CD#3 OG590/CD#4B after 2004-10-14 (blue CCD exchange) UV_MRSP_041130_BLU.tar   HER_5/CD#1B HER_5/CD#2 HER_5/CD#2

Use of solutions for earlier periods in time:
1) blue solutions usable for periods earlier than 2001-06-01, except for some scaling up or down
2) red solutions generally usable for periods earlier than 2001-10-01, except for some scaling up or down

Flux calibration

To give an idea about the impact of the flux correction on the large-scale spectral slope, and about the precision that can be achieved, we have prepared a tutorial about flux calibration. Here you can also find a description of how to flux-calibrate your UVES data.

tutorial on flux calibration