I want to share with you the scheme for observations we have used lately.
Of course, you can follow your own procedures. The only critical point we
should agree is the flattening of the continuum. Any fitting of the
continuum of eta Car is harmful. This must be done always for the comparison
star (A-Type).
regards
Augusto
------------------------------------
1-Daylight calibrations:
We take 100 bias and immediately after 20 dome flats. We calculate the
median combination of each set, subtract the median bias from the median
flat and divide the result for the average (or mode) of the image. We use
that normalised flat to correct for the pixel to pixel response and also for
the illumination pattern which arrives to the CCD. We have seen some low
frequency wiggles in the flatfield image.
We have seen that with this CCD, a smaller number of bias images introduces
noise, limiting the final S/N of the spectra.
When we divide the A-type star image spectrum by the normalised flatfield,
the continuum is also a straight line and can be fitted by a low order
polynomial: Legendre 3-5. Of course, we skip the stellar lines to fit the
continuum, just stating the number of sigmas to reject (normally sigma=3).
2- Scheme for observations:
-eta Carinae
3 short exposures - 90 seconds- to avoid saturation of the H-beta line plus
3 deep exposures - 600 seconds (H-beta saturated) to get high
signal-to-noise at the stellar continuum. With our observational setup, in
a clear sky we get S/N~400. One HeAr or NeAr spectrum before and one after
the set of exposures.
Standard star
- 3 exposures of the A-type star. The S/N doesn't need to be high, since the
continuum will be fitted. The most relevant thing is to put the comparison
star exactly on the same pixel as eta Car, to be affected in the same way by
the illumination distortions. we take one HeAr spectrum after the standard
star images.
3- Data reduction
We use IRAF, but it any other software using the same mathematical approach
would work well.
- We trace the light dispersed in the wavelength direction and co-add the
the columns to extract the spectral vector. Tracing the aperture is relevant
when the spectrum is not exactly parallel to the CCD rows. We subtract the
sky from both sides of the dispersed light, by fitting a 1rst order
polynomial, interpolated to the position of each CCD row, along the
dispersion direction. Only then we add up the rows and extract the spectral
vector (intensity X pixel).
We extract the spectrum of the HeAr lamp following exactly the same aperture
used to sum the stellar spectrum.
Then, we calibrate in Angstroms with a polynomial (Legendre) degree 2-3
(equivalent to Chebyshev 3-4th order).
Since the spectrograph works in a inclined position, there internal, flexure
impact the precision of the wavelength calibration. However, for eta
Carinae, this is not relevant, since we know the wavelength of some FeII
lines, like the one at 4727A and shift the zero point of the spectrum to
that of that line (measured with a Coude spectrograph, which is much more
precise). For the moment, I use the peak of that line as 4727.5 - but I'll
do a better job later...
After dividing the eta Car spectrum by the polynomial fit of the A-type
star, we get the spectrum normalised in a very precise way. The continuum
does not follow a straight line, but we leave it in that form. In this way,
all other spectra, taken by different observers, can be plotted one on top
of the other just by dividing each spectrum by a constant.
In the past, I felt on temptation to fit a polynomial through eta Car's
spectrum, to get it straightâ€¦ however, I see that this is not a good
procedure, since it depends on where you believe the stellar continuum is.
For eta Car, we never know where is the continuumâ€¦. and specially close to
HeII4686 there are places which looks like a continuum, but whose level
changes along the periastron, indicating that they are affected by broad
line wings. Far from the periastron, the line wings are stable.