Dispersion Question

[Keith Graham]

Hi

I am having an issue understanding a dispersion discrepancy I am having with my LHIRESIII/ST8/f/10 SCT setup, and I am hoping someone can help me solve it.

The dispersion as shown in Vspec is .057A/px. When I divide a wavelength range by the number of pixels in that range, the dispersion is verified at .057 A/px. However, the dispersion value given in SimSpec 4 is .114 A/px, exactly twice the .057 value.

Now, the fwhm of my emission line profiles is 4 pixels. I am using a 35u slit which allows a throughput of four 9u pixels. So if my dispersion is .057 A/px, this would result in a .228 A throughput. Using the 6503A neon line, 6503/.057 results in a 28784 resolving power (R). However, if I use the .114 A/px dispersion, the 4 pixel throughput would be .456 A resulting in a much more reasonable R of 14260.

So, can someone please help me understand this discrepancy? The .114A/px dispersion gives an R value much closer to what I get when processing my spectra in ISIS, but the .057 value appears to be the proper dispersion value.

Thanks,

Keith

[Ken Harrison]

Keith,
It comes down to dispersion v's resolution. They are not the same thing.
The resolution (and FWHM) will depend on the optics and pixel size (Try changing these two variables in SimSpecV4 to see the impact).
The Nyquist sampling calls for at least two pixel to be used for resolution. If the dispersion is 0.57 A/pixel then the "best" resolution would be around 0.114A.
Measure your actual FWHM of the neon lines after calibration. If it is 4 pixel then the result should be around 0.45A.
R then equals the wavelength/ FWHM ie 6503/0.45 = 14260.
Does this help?

[Robin Leadbeater]

Hi Keith,

~0.114A/pixel dispersion at H alpha with 9A pixels and LHIRES 2400 is correct (for comparison I get 0.107A/pixel with 8.3 pixels with the ATK 16 IC-S) How was the spectrum produced? If it was generated originally using ISIS then the final bin size is not neccesarily the original pixel size as ISIS up samples internally to aid precision. (check then number of bins. I suspect you will find it is 2x the actual number of pixels of your camera)

Dispersion in VSpec should really be labelled A/bin rather than A/pixel, though if you use Vspec to reduce the spectrum, this ~equates to A/pixel (the actual value will vary across the spectrum if equal bin size and a non linear calibration is used)


Cheers
Robin

[Keith Graham]

Hi Ken and Robin,

Thanks for your replies. I have been going through both of your explanations and both helped me to resolve my issue.

Ken, I do understand that dispersion is measured as A/px and resolution is measured in A. I also understand that dispersion is the number of A in one pixel whereas resolution is expressed as the number of A in the FWHM of the neon line. But I could not understand why Vspec was giving me a dispersion of .057A/pixel when .114 A/pixel was given in Simspec and it was also the value that gave the proper R value of 14260.


Robin, I do use ISIS to process my spectra. But I also use Vspec for some post processing checking. At your suggestion, I checked the number of pixels in my profile, and you were right - there are 2830 pixels along the X axis whereas the camera has 1530. I do not know why the spectra come up 230 pixels short of 3060, but when 2830 is divided by the wavelength range, the dispersion comes up as.057 A/px.

Anyhow, problem solved. I just knew I was missing something .Thanks a bunch to both of you for putting me on the right track.

Keith

[Christian Buil]

I confirm, I have programmed ISIS for oversample (0.1 -> 0.05 A) the original dispersion after processing.
It is an abritrary choice (for prevent future alteration if the spectrum is reprocessed).

For a correct info about original dispersion look the disersion equation (coefficient #1) - Profile tab.

Also a tip. From the command line enter the command : dispersion (no parameter).

A file "dispersion.lst" is produced in your current working directory. Edit the file : the true actual dispersion
is given in function of the wavelength.

Christian

[Keith Graham]

Hi Christian,

Thanks for the explanation. Previously I could never find where in ISIS the dispersion was located. Admittedly I have always used Vspec for this because it shows the dispersion at the top of the screen where it is easily seen. As I said earlier, i had always wondered why Vspec showed the dispersion to be 1/2 of the actual, but thanks to Robin's explanation I now know why.

I now see the dispersion in the A1 box in the ISIS Profile tab as you pointed out, but I do not see a command line. Could you please direct me to where that is?

Thanks,

Keith

[Robin Leadbeater]

Hi Keith,

The ISIS comand line function is found under "tools" "comand line"

The "dispersion tab" when you display a profile is to calculate the dispersion and shows the last used dispersion equation parameters, not the native dispersion for the displayed spectrum. This is not recorded in the fits header in ISIS either so there is no way to know the original dispersion with spectra generated by ISIS unless you have the accompanying txt file produced during the reduction.

Note also that in any case the dispersion in A/pixel is not constant along the spectrum as most spectrographs are non linear. The first (ie slope)term in the dispersion fit is only an approximation as is the "A/pixel" given by VSpec when it is used to produce a non linear fit.

Professional software like IRAF have modes where the the original pixel values can be stored in the fits file along with the calibration equation and this is used to calculate the spectrum only at the time you display it. This has the advantage that you do not need to upsample as ISIS does in an attempt to avoid resampling artifacts when you convert to a constant bin size. Unfortunately this makes them impossible to read by any amateur software or even most professional fits file readers unless they are specifically designed for spectroscopy

This is a problem that is beginning to worry me as I get more involved with professionals. I often cannot read the files they produce but I do not want to have to learn IRAF It would be nice if ISIS included features to read the various IRAF/MIDAS fits formats and convert them to constant bin size files

Cheers
Robin

[Robin Leadbeater]

Professional software like IRAF have modes where the the original pixel values can be stored in the fits file along with the calibration equation and this is used to calculate the spectrum only at the time you display it. This has the advantage that you do not need to upsample as ISIS does in an attempt to avoid resampling artifacts when you convert to a constant bin size. Unfortunately this makes them impossible to read by any amateur software or even most professional fits file readers unless they are specifically designed for spectroscopy

This is a problem that is beginning to worry me as I get more involved with professionals. I often cannot read the files they produce but I do not want to have to learn IRAF It would be nice if ISIS included features to read the various IRAF/MIDAS fits formats and convert them to constant bin size files

Attached is an example of a fits header produced by IRAF in "multispec" format used to store echelle spectra. The data is based on the original binned pixel values (ie no resampling) and you can see all the wavelength calibration fit terms for each order. Ideally I think we should be generating something similar.

Cheers
Robin

[Keith Graham]

Thanks for this additional information, Robin.

I am glad you pointed out the inconsistency of dispersion along the spectrum due to its non-linearity. This does make perfect sense to me.

I agree with you about learning another software program. So far, ISIS and Vspec do everything I need at this point in my spectroscopic career. In fact, these programs have features I have yet to explore, so I still have my work cut out for me. Actually, as I looked through the IRAF header you sent, I I have to admit it did not make a whole lot of sense to me. But that is for another day.

Again, many thanks for your input.

Cheers,

Keith