Spring 1999
Nearby Supernova Campaign
Spectroscopy Guide





For the nearby campiagn, spectroscopic observations have two different components: The observing requirements are different for each of these. During the screening process determination of the redshift is the principle goal, and can be derived from either the host or the SN itself. Determination of SN type is necessary if the redshift is less than the canonical cut-off, but not absolutely neccessary for SNe beyond the canonical cut-off. Since redshift is the primary goal, the slit should be rotated to cover both the SN and the host. The weekly follow-up of confirmed supernovae below the canonical redshift cut-off is much more demanding; the required signal-to-noise is much higher and systematic errors must be minimized. Minimizing systematic errors is more difficult for spectroscopy than for photometry due to the broader wavelength coverage and the hard limits imposed by the spectrograph slit. The keys to obtaining data such that systematic errors are minimized are:


Usually spectrograph set-up is very specific to a given instrument. Local documentation should be consulted, but the basics are to tilt the grating(s), or in the case of grims translate the CCD, to obtain the desired wavelength range, and then obtain the best overall focus for the slit over this wavelength range. Do remember that for follow-up observations blue coverage is of paramount importance, so set the spectrograph to work as far into the blue as is practical. A good range for most spectrographs is 3500 to 7000 Angstrom. When setting up for red spectra on single-channel spectrographs, use an order blocking filter to make sure that 2nd order light does not contaminate the red portion of the spectrum.

Note that since the spectrograph focus is set for the slit, a star will be in focus if it is in focus on the slit jaws. To ensure good telescope focus, make sure the slit is in sharp focus in the guider/acquisition camera. Then, the telescope focus that looks the best on the guider will be the best telescope focus for acheiving good focus on the CCD. It is usually possible to view the slit in the guider during the day by illuinating the slit with internal or dome lights. Intensified guider cameras can not tolerate bright light, so be careful when using them! Note that the quality of the guider focus will also affect the faintness to which objects can be seen on the guider camera.



During the afternoon, and again in the morning, obtain about 11 bias frames. Make sure the dome is dark or that the instrument is absolutely light-tight; for typical CCD readout times light leaks can be a problem even though the shutter is never opened. If the bias has little or no 2-D structure this calibration will be sufficient. Most CCD's have an overscan region that will track the DC component of the bias. In the worst cases the bias will have significant 2-D structure, this structure can vary with time, and there might not be an overscan region. If you suspect this is the case, take bias frames periodically throughout the night for safety. It is not uncommon to see a little periodic time-variable 2-D structure in bias images cased by electronic pickup. Assess whether the level of this noise is significant enough to affect the spectra, and if so, try to get a technical assistant to check it.

Internal Flats:

During the afternoon and throughout the night, obtain internal or dome flats. Exposures should be set so the brightest parts of the image (not necessarily individual pixels) are about 2/3 of the way to saturation. These flats not only help correct for pixel-to-pixel sensitivity variations, they correct for non-uniformities along the slit (nicks, dust, etc.). Spectrograph flexure can cause movement in both the spectral and spatial directions, therefore it is wise to obtain such internal calibration for each supernova set-up, unless only screening observations are being obtained. Note that dome flats are most useful if they can be obtained without re-pointing the telescope since otherwise flexure will change the result somewhat.

Pinhole spectrum:

The spatial scale of any spectrograph with refractive optics will vary as a function of wavelength. Moveover, the CCD may be rotated with respect to the dispersion axis. For both of these reasons it is necessary to obtain spectra of point sources along the slit to determine the rotation and relative scale. This can be accomplished using a slit with a column of pinholes, available at better observatories. Ask about availability, since sometimes these are considered engineering slits and are not advertised to guest observers. With the pinhole slit in place, take an internal flat with exposure that does not saturate the pinhole spectra. A less desirable, but sometimes necessary, alternative is to take spectra of a spectrophotometric standard star at regular intervals along the slit. The spectrograph should be rotated to the parallactic angle during such observations. In the absence of this type of calibration, it is common to "trace" the supernova spectrum during extraction. However, this can introduce systematic errors in the presence of background galaxy light. For all follow-up except for screening please do not skimp - obtain this calibration!

Twilight Flats:

Just after sunset obtain about 3 dithered twilight spectra. Make sure the telescope is close to the correct focus since this will affect how well dust and nicks along the slit can be flattened. These spectra are not used to perform pixel-to-pixel corrections, but rather to determine the illumination correction that must be applied to internal or dome flats. The illumination correction is usually a very low-order function of slit position, and can have some wavelength dependence. However, sometimes illumination correction is needed on the scale of slit non-uniformities. So, try to obtain good signal-to-noise in the twilight flats. The twilight sky is not intrinsically "flat"; in general the flattest part of the twilight sky is about 15 degrees from the zenith in the direction opposite the sun. Adjust the telescope pointing around this canonical location to avoid the Galactic plane and bright stars (consult Norton's Star Atlas). The brightness of the twilight sky changes by a factor of 2 roughly every three minutes; this can be used to estimate the appropriate exposure time based on previous exposure times. Note that spectral twilight flats must be started much sooner after sunset, and can be taken much closer to sunrise, than imaging twilight flats.

Arc spectra:

Arc spectra are used to obtain wavelength calibration and to linearize the projection of the slit on the CCD. For our purposes only modest wavelength calibration accuracy, good to ~ 1 Angstrom, is needed. The requirements for linearization of slit are more stringent, since this partly affects how well strong sky lines can be subtracted. Therefore, even if a seemingly adequate wavelength solution can be obtained from a handful of arc lines, try to get many lines with good signal-to-noise. This may come at the expense of saturating a few stronger lines in the arc spectrum; these are often blends so saturating such lines is usualy not much of a sacrafice.

Often more than one arc lamp must be used. Sometimes an exposure time can be found that works well with all lamps considered. More often it is best to obtain separate arc exposures for each lamp both to obtain good signal-to-noise and to avoid too many line superpositions. If such full-arc calibration is too time consuming, and there is not too much spectrograph flexure or the rotation isn't changed much, it may be possible to obtain more limited calibration during the night. For screening observations, the arc calibration requirements are not as stringent.


Each CCD will have a dark current. For spectroscopy the flux per pixel can be fairly low, therefore the dark current can be a non-negligable source of spatial noise if not corrected. In any event there can be hot pixels which must be corrected or masked Most of our runs are short (1 night) so it is difficult to obtain good quality dark frames, since a good meaurement requires several (for cosmic ray rejection) long (to build up enough signal that isn't lost due to charge-transfer inefficiency) exposures in a completely dark dome. Certainly if the night is cloudy and a given CCD has been/will be used more than once, take several 3600 second dark frames just so they are available.


External calibration is obtained by observing stars with known spectra. These observations provide the flux zeropoint as a function of wavelength. Unlike the case of photometry, we are much more concerned about getting the correct relative calibration rather than the absolute zeropoint. The absolute calibration changes as the seeing changes since the slit width is fixed, so the absolute zeropoint is not particularly stable. To do this, observe spectrophotometric standards with the slit rotated to the parallactic angle and the star well centered in the slit. Obtain a signal-to-noise of at least 25 per pixel.

The system response as a function of wavelength depends on the optics, including the telescope mirrors, spectrograph collimator and camera, and grism or grating, the dectector response, and the absorption and scattering properties of the atmosphere. The response variations due to the optics generally vary slowly with wavelength, although gratings can exhibit shape features known as Woods anomolies. The dectector response can vary quite strongly with wavelength due to fringing. Atmospheric scattering varies smoothly with wavelength, but atmospheric absorption includes sharp molecular absorption lines longward of 6800 Angstroms. It is the strongly varying response features which can cause the most noticable calibration difficulties, but smoothly varying response components are just as important for our needs. If any of these system or atmospheric properties is changed, additional calibration data should be taken.

Since standard stars are fairly bright, they can be observed during twilight. It is important to plan standard star observations in order to obtain a good range of airmasses and stellar spectral types since the network of good spectrophotometric standards is relatively sparse on the sky . A range of stellar spectral types is useful to check for 2nd order light and to discriminate between atmospheric absorption features and stellar absorption features. In order to be able to determine cross-telescope calibration very precisely, every telescope must at some point obtain observations of one or more of the HST white-dwarf spectrophotometric standard stars, G191B2B, GD71, GD153, and HZ43, under photometric conditions.

Our preferred list of standards, obtained with modern linear detectors at good resolution, are listed below. Finder charts for these standards are available from ESO or from APO . The standard star names link to IRAF compatible calibration files prepared by Greg Aldering. Old spectrophotometric standards, having only very low resolution calibration, are still commonly listed at many observatories but should be avoided.


When calculating sensitivity functions from observations of spectrophtometric standards, continuum locations affected by atmospheric absorption features should be avoided. Subsequent to flux calibration, the spectra should be corrected for atmospheric absorption. Here are links to a atmospheric transmission plot and a atmospheric transmission textfile


Spectroscopic observations of our supernovae will be challenging at most telescopes on the schedule, mainly because their faintness will make them hard to see on acquisition cameras requiring accurate offsets to place the SN down the slit. We expect to have finder charts, with offsets, and suggested exposure times available on the web, and we hope to be able to provide suggested nightly schedules customized for each telescope. If a nightly schedule is not available then you will need to plan the night's observations during the afternoon. During the afternoon, print-out the finder charts (or go prepared with them) and the latest version of the follow-up status since you never know when the Internet will bulk.

Start by calculating the parallactic angle for the position and observing time; some telescopes must be near zenith to rotate the spectrograph while other telescopes allow the spectrograph to be rotated while moving to the SN coordinates. Move to the nominal coordinates of the offset star, adjust the focus if necessary, then center the offset star in the slit; try to use more than one position along the slit over the course of a night to help reduce systematics. At telescopes with closed-loop offseting, a guide star should now be acquired and the offset performed. For open-loop offsets, perform the offset and immediately acquire a guide star and begin guiding; try to minimize the time for this precedure to prevent the SN from drifting out of the slit.

Note that many guiders are normally run without filters, so the guider may be guiding on red light while your are trying to take a spectrum in blue light. The difference between blue and red locations due to atmospheric dispersion will be in the direction of the parallactic angle. If you are not observing at the parallactic angle, it may be prudent to use a filter with the guider that covers the wavelength range for which a spectrum is being taken.

At most observatories, important observational parameters are stored in the image headers. For quick reference it can also be useful to record some of these - especially those that are changed often or which must be closely monitored - in a log. Make sure the object name, right ascension, declination, exposure time, universal time, grating, filters, and some running number are recorded. It is usually useful to also record the hour angle, airmass, telescope focus, telescope (truss) temperature, CCD temperature, relative humidity, and atmospheric pressure. For telescopes used often, it can also be useful to record the guider coordinates of a good guide star so it can be reacquired on subsequent runs.

If you are able to use WhatsUp, use the Candidates tool to check-out observations you plan to make and to check-in whether the observations were completed successfully. This is vital to avoid redundant observations. WhatsUp also provides an exposure time calculator and other tools to help plan and monitor observations.


Since many of our runs are only a single night, plan carefully how to archive the data so that you don't have to stay up all the next morning writing tapes or FTP'ing data. At some observatories the data can be automatically written to tape when the detector is read-out. In this case beware of power glitches or other errors that could cause the tape to be rewound, and subsequently overwritten, by accident. Always make two copies of the data, preferable on tape, and store them separately.

It is wise to check that the data you think is on tape is really there. A good approach is to use the log of tape files to construct a list of images to delete and then delete only those images; if there are science images left over that means they didn't make it on the tape! Label all tapes with the date, telescope, detector, observer name, file format (e.g. FITS), archive format (e.g. tar), and number of images. Especially note if the data for a night/run spans more than one tape, or if there are multiple tars on one tape.

For instructions on how to FTP the data to LBL, click on FTP Instructions.



As the number electrons per pixel approaches the CCD full-well depth, the CCD response will become non-linear. Do not assume that the A/D convertor saturation and CCD full-well staturation are the same. For some CCD's the A/D convertor saturates before the CCD full-well saturates, well sometimes the opposite is true. If the CCD full-well saturates before the A/D convertor saturates, it is important to know where non-linearity begins to occur. At some observatories this information has been determined and recorded in the instrument documentation. For spectroscopy it is generally easy to keep spectrophotometric standard and flat spectra well below the non-linearity range. See the analogous section of the photometry guide if you wish to determine the CCD linearity curve.

Shutter Function:

Due to mechanical latency, true exposure times and requested exposure times can differ by a constant offset or even a gradient. At better observatories the importance of such effects will be known. Try to choose exposure times (i.e., for spectrophotometric standard stars) at least 100x longer than this latency time. Since even standard star exposures are at least 30 seconds, this issue are generally unimportant for spectroscopy.

Amplifier Cross-talk:

It is becoming more common with large-format CCD's to use more than one amplifier to speed up the readout time. It is not unusual for cross-talk to occur between amplifiers. One problem that can occur in this case is that when a bright feature is clocked through one amplifier it perturbs the signal reported by the other amplifiers. Since single-object spectroscopy usually uses only part of the CCD in the spatial direction, use of multiple amplifiers is not as common. If electronic "ghost images" are severe it might be nessary to use only one amplifier. This should be the exception rather than the rule. Again, ask the local experts or consult local documentation if you suspect cross-talk problems.