iraf.net

I am using sbands for the determination of the equivalent width of an absorption line in a stellar spectrum.Using the mean location and width of the line and setting the pseudo continuum parameters on the left and right of the line,
gives me in 1 case (out of 112) a negative e.width.
How can this happen?In the out file, it is visible, that the main problem is the determiantion of the continuum - it is too low and wrong.Please help me ASAP
Thank you very much.

Hey - does no one have any idea how it is possible to explain these negative equivalent widths in the SBANDS output file?Any help is appreciated, it is kind of urgent - thanks.

Frank would be the best person to reply but is away currently. The equation for the eqwidth is given in the help page as[code:1:45542f1917] eq. width = (1 - flux1/flux2) * width[/code:1:45542f1917]so obviously you get a negative if the ratio is greater than one or you've specified a negative width.Having a listing of parameters would be helpful, especially to see if you've turned on the 'normalize' flag as recommended, as would some idea of how the spectrum looks or how you've specified the bandpass (as one, two or three bandpasses?). The 'flux2' in the equation above is actually a continuum value computed differently depending on the number of bandpasses and the choice of normalization. The code is available in noao$onedspec/t_sbands.x if you'd like to trace it yourself, otherwise please post more information.-Mike

Thanks for the data, but I'm not sure there really is a problem in the task. Your data all have values of order 1e12 and I'm guessing you're trying to measure that dinky little Li line at 6708 in the midst of some really noisy points? Note that a negative eq. width isn't always an error, by convention it means that an emission feature was measured rather than an absorption feature. Putting the cursor on the line I think you're after shows it at 6709.6 and a width of 16A seems pretty huge to this data. Have you checked your dispersion and/or played with the bandpass width at all?I've forwarded your data and files on to Frank in case he has more enlightened comments.-Mike

Hi,I have looked at your data and what Mike said is correct. A negative equivalent width means the flux in the line band is greater than the estimated continuum based on the continuum bands.I think this is happening because you have a very wide line band wdith such that most of the band is in continuum. It looks to me like the line band should be something like 8 Angstroms in width and not 16. This would then be 6705 to 6713. Even so the region around 6720 seem lower than what appears to be the edges of the line wings around 6705 and 6713. I don't know if there is something funny about the way the data was generated. The fact that the minima in the noise see to be fairly constant while the maximum vary a lot more looks like some kind of clipping was done and what appears to be noise is not noise, or at least, is not Gaussian noise.So I think all we can tell you is the algorithm is doing the right thing with your data as described by the mathematically definitions of the measured quantities as documented in the help page.Yours,
Frank Valdes

Thank you for your help.The width of the line bandpass (16) is chosen because of theoretical reasons.
However, the main question is how SBANDS determines the continuum flux. Since we use three bandpasses there are two continuum bands. Does sbands simply add up the counts of both bands or are the two indiviudal continuum fluxes averaged?What does the "normalize" flag do since it is not normalizing the continuum in the general sense (does it only normalize for a filter?).So, could you explain the following two output files? The second case has smaller continuum bands but a larger absolut continuum flux - how is that possible?# bands = test.ew, norm = no, mag = no
# band filter wavelength width
# line none 6708. 8.
# contblue none 6692. 8.
# contblue none 6724. 8.
#
# spectrum band flux band flux index eqwidth
test.fits(1) line 1.23200E14 cont 1.22369E14 1.00679 -0.054309# bands = test.ew, norm = no, mag = no
# band filter wavelength width
# line none 6708. 8.
# contblue none 6692. 4.
# contblue none 6724. 4.
#
# spectrum band flux band flux index eqwidth
test.fits(1) line 1.23200E14 cont 1.29974E14 0.947882 0.416943
Thanks again,
Chris & Boris

Hi Chris and Boris,First I have to suggest you read the help page carefully [code:1:205a1645eb]cl> phelp sbands[/code:1:205a1645eb]. From the help page:[quote:205a1645eb]
If there are three bandpasses then a continuum bandpass flux is computed as the interpolation between the bandpass centers to the center of the first bandpass. The special bandpass identification "cont" will be reported.
[/quote:205a1645eb]What this means is the flux in the two continuum bandpasses are assigned to the center wavelength and then the two (wavelength,flux) values are linearly interpolated to the line bandpass center. If the positions of the continuum bandpasses are equidistant on either side of the line center (as in your case) then it is the same as the average.Because you have such noisy data it is possible that making the bandpass smaller makes the flux higher. This would happen if the flux in the part excluded when you make the bandpass smaller is, on average, lower.Concerning the normalization. This only applies to the bandpass function. Without a shape it is simply a box function. Without normalization each pixel in the bandpass is added with a weight of 1, that is a simple sum. With normalization the weights are 1/width, that is the sum is divided by the width to give an average flux per unit wavelength.Yours,
Frank