I don’t think I ever want to eat pasta again…
I’ve recently come home from the NEON Observing School in Asiago Observatory, Italy. This was a 10 day course designed to give Astro PhD students a chance to experience and learn about observing and data reduction, with a combination of lectures and group work. We also ate more food than I thought was possible; three course meals for lunch and dinner every day is not something I’m used to!
There were 16 PhD students on the course and we all came from different European countries, studying a wide variety of astro topics and at various stages of our PhDs. The thing we all had in common was that none of us had much observing experience, and all wanted to learn.
This particular Observing School was focussed on spectroscopy, which is the process of splitting light from an astronomical target into different wavelengths and comparing how much light we receive at each wavelength. As this is the technique I use for my PhD, although I study X-ray light rather than optical, I was immediately pleased that the focus here would be relevant for me.
We started with lectures explaining the basics (and less basic parts) of telescope optics, CCD technology and spectrometers, and then were put into four groups, with an experienced observer as a tutor (I believe the tutors were all post-docs). Each tutor had designed a project for their group to do over the 10 days, aimed at taking spectral data, reducing that data and classifying their astronomical objects.
I learnt a lot over the 10 days away which can be split into three main sections: instrumentation of spectrometers, observations (prep and at the telescope) and data reduction. My understanding of the latter two sections was greatly improved by the project work we did, and the first of those (observations, both preparation of and at the telescope) is what I will focus on in this blog post.
The project group I was in were set the challenge of observing three objects: an AGN host galaxy, a starburst galaxy and a blazar (a specific type of AGN). We were given three lists of targets (one list for each type) and had to choose one from each. The aim of the project was to determine which object was which type, and then make some scientific measurements if we had time. First we had to determine which instrument set-up to use for the observations, then which targets from each list would be possible to observe.
Preparation of Observations
When choosing the instrument set-up for the telescope we would use, we had to consider:
- Which wavelength range we wanted to observe, and why
- How high a resolution we needed in the spectra
- Which size of the entrance slit to choose
- this is the part that lets light from our target into the spectrometer; the wider the slit the more light we can collect in a given time, but the narrower the slit the higher the resolution we achieve in our collected data
Once our tutor (Jairo, who you can spot grinning at the camera in the photo above) had confirmed that the instrument set-up we decided upon was reasonable (and in fact the same as the one he had already chosen to apply to the telescope earlier that day in preparation for our observations!), we started choosing our targets for the first night of observing.
For each potential target we only had their position information in Right Ascension and Declination, which is a co-ordinate system used for the sky (on the celestial sphere), so we needed to determine which targets would be appropriately visible for this particular night from this particular location on the Earth.
A good way of getting a rough idea whether your object is observable is to first look at the Declination value: this relates to altitude of the target in the sky, so for a northern hemisphere observatory this is better when it’s a positive value, and higher values are better still. This does depend on the latitude of your observatory though, so for locations near the equator it is possible to see objects with negative declination very well.
Right Ascension (RA) relates to the object position on the celestial equator (the Earth’s equator projected into the sky), starting from the Vernal Equinox (spring equinox, in March). RA is measured in hours, minutes and degrees, with 24 hours giving a whole circle, so in March any object with RA of 12 hours is at its highest point, at midnight. There are 12 months in the year and 24 hours of RA so each month is equivalent to 2 hours of RA, therefore in February (as we were) the objects at their highest at midnight are those with a RA of 10 hours (12 – 2 = 10 hours).
Luckily for us, in the age of the internet there are many useful websites that do the hard work of this part for us; we enter the RA and Dec of our potential targets and the position of our observatory and get a visibility chart showing where those objects will be over a particular night. We also had to take into account that the lower any object is in the sky the higher the airmass of Earth’s atmosphere at that level, so we decided an altitude of 30 degrees was the lowest we would find acceptable.
Finally before observing we had to choose a “standard star” to observe so that we could use it for flux calibration in the data reduction process later.
After this prep work we went back to the hotel for dinner, as my group was on the second shift of the night, and then headed back to the telescope to start observing at 10:30pm.
As we were the second group of the night, the first types of calibration images had already been taken at twilight (biases and flat fields), but we had to take some more over the course of our observations (lamps for wavelength calibration and standard stars for flux calibration) as well as our scientific observations.
- Take bias frames, with shutter closed and zero exposure time
- Take an odd number (e.g. 11) so can calculate median later without interpolating
- Take flat field frames
- Start with 1 sec exposure and multiply to get reasonable number of counts to calculate good exposure time
- Number of counts should be high enough to not introduce a large percentage of noise but low enough that the CCD still has a linear response
- Roughly (empirically) about half or a third of the saturation point (in counts)
- Set position angle of the slit
- Important for extended objects: for example for galaxies the slit should be oriented along the semi-major axis of the galaxy if you want to measure the rotation curve of the stars in the galaxy
- This can be done before taking flat fields, but is a second order effect and therefore doesn’t change much
Each CCD has a number of counts artificially added to the detector, so that negative counts (due to noise) are not recorded. This is called the bias and varies slightly over position on the CCD; it can also vary over time, but this is significantly reduced if the CCD is kept at a constant temperature.
The bias frames (taken with no illuminating light on the CCD and at zero seconds exposure time) record the value of this bias so that it can be subtracted from the scientific data. Usually more than one bias frame is taken, so that they can be averaged (in our case we took the median) to reduce the contribution of read-out noise from the CCD.
Taking flat field frames of evenly illuminated sources are the next step. These are used in the data reduction to correct for pixel-to-pixel variations in sensitivity of the CCD. There are two types of flat fields that can be taken; dome flats (within the dome) and sky flats (of a reasonably bright twilight sky). The group who observed first on the same telescope as us took dome flats, which are exposures of a lamp illuminating a surface on the dome. Multiple exposures are taken, although in this case it is so that effects of photon noise can be reduced. Photon noise is a poisson process so the signal-to-noise ratio scales with the square root of the number of counts; therefore the higher the counts, the higher the signal-to-noise ratio.
During the night:
- Position the telescope at the scientific target
- Fine tune position of the slit (if needed)
- Choose guide star to make telescope tracking more accurate as the target appears to move across the night sky
- Take scientific exposure
- Take multiple exposures of the same object (if needed/wanted)
- Take exposure of lamp for wavelength calibration, at the same position in the sky as the exposure of the scientific target
- Repeat steps 3 and 4 as desired
- Take exposure of standard stars (this could be done at beginning of the night)
- Take exposure of lamp for wavelength calibration, at the same position in the sky as the exposure of the standard star
When you take scientific spectra the result is a 2D image of the dispersed sky on the CCD, in this case the spatial direction (along the entrance slit) is the y direction and the wavelength (or dispersion) direction is the x direction. Therefore for each position along the slit the constituent wavelengths are dispersed in (almost) a horizontal direction. Initially the image dimensions are in pixels, with the number of counts on each pixel represented by the brightness plotted.
This can then be collapsed into a 1D spectrum, by following the object you want to analyse (almost) horizontally over the CCD and using the integrated flux over the number of lines that object falls on the CCD. The 1D spectrum is then a line plot of pixels vs number of counts, which is not yet helpful for scientific analysis. To convert this 1D spectrum into wavelength vs flux we need to be able to transform pixels into wavelength and counts into flux. For this we have to take extra calibration exposures when we observe.
For the wavelength calibration we need to take exposures of lamps with emission lines (these are different lamps to those used for the dome flats). We used FeAr lamps, which means they give us iron (Fe) and argon (Ar) emission lines. The emission lines have very well known wavelengths, and therefore can be used to transform pixels on the CCD (in the x direction) to wavelength values (in Angstroms) during the data reduction. This transformation can change when the telescope moves, as small ‘flexures’ in the optics and instrument can slightly move the final position on the CCD of each wavelength of light. Therefore we take one exposure of the lamp for each scientific object, at the same position of the telescope.
For the flux calibration we need to be able to transform number of counts measured by our detector into amount of flux, which will not be a linear transformation over all wavelengths, so needs a sensitivity curve showing the conversion over all desired wavelengths. Standard stars are stars where their photometric fluxes (the luminosity that reaches Earth) are known very accurately in many different wavelength bands, so can be used to convert the number of photons our detector counts into accurate flux values. This conversion is different every night (even with the same detector) as it is strongly dependent on the atmosphere, so should always be done on the same night and ideally in the same direction as the scientific observations. The spectrum of the standard star will also have to be wavelength calibrated, so needs its own lamp exposure as well.
After taking all these exposures (and getting some sleep!) it’s finally time to come down to the business of data reduction. There are a number of steps to take to get spectral data ready to scientifically analyse, and I’ve hinted at some of them already.
- bias subtraction
- flat field correction
- cosmic rays removal
- wavelength calibration
- sky subtraction
- flux calibration
Some of these steps are more complicated and time consuming than others, but they are all necessary before you can be confident in your science results.
I have written a lot already on this post, so I will leave the details of the data reduction process to another blog; for now all I will say is that it took a lot of concentration and thought to understand (at least for me!).
In the end we only had one night of clear enough weather to observe, so it was a good decision by the organisers to make sure each of the four groups got some data. This process of planning and taking observations is something I have vaguely known about for years, but never having done it myself means the details were quite fuzzy before this process.
Overall I am very glad I went to this Observing School; the time and effort put in by the organisers, and especially by the tutors into our projects, meant we all left with more knowledge than we arrived with, along with enjoying our time in the snow!