NRES diagnostic plots

The NRES pipeline produces a set of diagnostic plots for every science (i.e. non-calibration) spectrum that's processed. These plots are bundled into a single pdf file that is included in the tar file of data products. For descriptions of the "BLAZ", "EXTR", and "SPEC" spectra (mentioned below), please read the "Extraction and Wavelength Calibration" section of the NRES pipeline webpage.

The plots on the first three pages of the pdf file are analogues of the "quick" plots produced for Dave Latham's TRES spectrograph. They are intended to help the user evaluate the quality of incoming spectra and judge some astrophysical characteristics of the target object.

  • Page 1, top panel:  This plot shows one order (order index=38, diffraction order = 89, containing the Mg b lines) of the observed spectrum (blue trace), superposed on the same order of the template spectrum (red trace) that is used for the RV analysis.  These spectra are shown in "BLAZ" format, in which a scaled copy of the flat-field (tungsten halogen) spectrum is subtracted from the observed spectrum.  This representation has simple noise properties, and lacks most of the low-spatial-frequency variation that one sees in raw extracted spectra (namely, those in "EXTR" spectra). The BLAZ spectra are the starting point for the radial velocity analysis.  Before they are plotted, both target and template spectra are smoothed with a kernel of about 2/3 the width of the projected fiber diameter, to filter out high-spatial-frequency noise.  The vertical scale is in ADU (= detected photoelectrons in the extracted spectrum).  Note that this representation can appear noisier than one would expect in, e.g. the case of hot stars with low-contrast lines, since almost all of the measured intensity is subtracted out before plotting. The title above the panel contains the target name, the spectrograph site (cut to 2 characters), the UT date and JD of the exposure start, the order index of the plotted order, the target V magnitude, an estimate of the S/N ratio (computed in the continuum near order center), and the pipeline code version number.  The S/N estimate assumes shot noise plus detector read noise appropriate to the extracted order width.  Near the bottom of the plot, one finds various catalog information about the target (RA, Dec, Teff, log(g), [m/H], V_rot (km/s)). 

  • Page 1, bottom left and bottom center panels:  These panels show the cross-correlation between the target and template spectra, plotted against redshift (km/s).  The amplitude is scaled to unity for exact equality between the two spectra.  The center panel shows the raw cross-correlation.  The leftmost panel is intended to show the correlation with a template spectrum that has been smeared by convolution with a stellar rotation profile, but this is not yet done (in V.1.1 of the pipeline). The vertical blue line indicates zero redshift; the vertical black line shows the estimated cross-correlation peak position.  In the upper right area of the panel one finds the correlation peak height, the calculated barycentric redshift (BC) of the target at the flux-weighted mean time of the exposure (km/s), and the estimated redshift (RV) of the correlation peak (ie, the redshift of spectrum features relative to the same features in the template).  The estimated barycentric redshift of the target is therefore RV-BC.  No noise estimates are included in pipeline version 1.1.

  • Page 1, bottom right panel:  This panel shows the raw extracted ("EXTR") spectrum for the same order as the top panel.  The vertical scale is kilo-ADU.

  • Page 2:  These panels show other spectrum orders, containing interesting stellar lines: Ca II H, H-alpha, Li6707, and Na D.  In each case, extracted signal (k ADU) are plotted against wavelength (Å).  Vertical blue lines show the wavelengths of the expected features, assuming zero redshift.

  • Page 3:  This plot shows the Mb b order once again, this time divided by the normalized flat-field spectrum ("SPEC").  Often, this plot is useful for revealing failures in the flat-fielding process.

The plots on the next three pages are intended to help in debugging the processes of spectrum extraction and wavelength calibration. Thus, they are more about hardware and software performance than about astronomy.

  • Page 4, top panel:  This plot shows extracted intensity (k ADU) vs wavelength (nm) for a somewhat redder order than the Mg b order.  For stars with relatively few lines in this order, this spectrum is sensitive to the chirped string of discontinuities that one sees if the extraction box is significantly displaced relative to the actual position of orders on the detector.

  • Page 4, middle panel:  This plot shows the estimated displacement (pixels) between the center of gravity of the light incident on the chip, and the center of the extraction box used to estimate the summed-across-dispersion intensity.  If these displacements show systematic values of more than about 1 pixel in magnitude, then the extraction is probably compromised.  Occasional large values are tolerable.

  • Page 4, bottom panels:  These three plots show (with various normalizations), the 0th, 2nd, and 1st moments of the cross-dispersion light distribution seen in the image, shown as functions of diffraction order index.  All moments are measured in coordinates centered on the extraction box for the given order. Each pixel in the dispersion direction yields an estimate of these moments; plotted are the median values of the individual pixel estimates taken over the central 25 percent of each order.  Thus, from left to right, the plots are intended to show the summed brightness (k ADU), order width (pix) and order displacement (pix) of the measured light distribution, all shown as functions of order index (order number + offset). Order indexes of 0 and 66 correspond to the reddest and bluest orders captured in an NRES spectrum, respectively. In v.1.1. of the pipeline, the width estimate is buggy and does not seem to indicate anything useful.

  • Page 5, top panel:  This plot shows residuals about the fit of a wavelength solution to the positions of ThAr lines, plotted against catalog line wavelengths.  Units in both axes are nm.  Each plotted symbol represents one identified ThAr line.  This plot is useful for identifying failed fits (systematic residuals greater than a few x .001 nm, for visualizing failures of the fitted solution to match the true shape of the variations (particularly visible near the ends of the wavelength range), and possible mis-identified lines. The number of identified lines is printed in the bottom left corner.

  • Page 5, bottom panel:  Besides the cross-correlation of the Mg b order, the pipeline estimates redshifts for each of n_block contiguous segments across each of the orders. (Currently, n_block = 12).  The estimation fits a few-parameter (shift, scale, tilt) transformation of each block in the template spectrum to the corresponding observed block.  This panel shows the transformation redshift parameter (km/s) plotted as a function of block index number.  The block index changes fastest with order number, and slowest with along-dispersion position.  Thus, the plotted symbols span all orders (red to blue) within each of 12 blocks (which are in blue to red order along the dispersion).  Vertical dashed lines indicate block boundaries.  In low-noise cases, the distribution of redshifts within and across blocks is symptomatic of time-dependent changes in the wavelength solution.

  • Page 6, all panels:  These are plots of the observed intensity (k ADU) vs wavelength (nm) for selected regions in the telluric O2 B- and A-bands. Vertical blue lines show the catalog wavelengths for selected molecular transitions.  These plots are intended as sanity checks on the wavelength solution;  at the plotted scale, the telluric lines have essentially constant wavelengths.  Thus, if they are not well centered on the blue lines, then something is amiss with the wavelength solution.