Spectral shift during sequence acquisition

[Olivier GARDE]

With my LISA and the ST8 camera I can slew all over the sky and the neon lines don't move. I have tried taking many sets of exposures throughout the night and they are all identical. This makes it much easier.
Cheers
Terry

I don't think that it's possible to compare LISA with LHIRES and the shift problem with Neon lines :
- LISA have a resolution about 600 to 1000
- LHIRES have a resolution about 18000

So in fact LHIRES is 18 times more sensitive than the LISA

[Robin Leadbeater]

Hi Olivier,

As far as I can tell. the shift (in pixels) for the LHIRES is independent of resolution (This is expected I think as it is caused by a certain amount of deflection in the optics which moves the image on the CCD and is not related to wavelength.) If this is so then the comparison with the LISA should still be valid. There are many differences in the construction of the two instruments of course but perhaps the LISA having a fixed grating is significant.

Robin

[Keith Graham]

OK, it is now clear to me how to use the new Wavelength Registration feature. But I am still unclear regarding the resolution issue on shorter sequence runs when the Wavelength Registration feature is not used. If I am correct, ISIS uses the neon spectrum to determine resolution, not the target star spectrum. Is this correct Christian? If so, then it seems to me that if there is a shift between the first and last neon image, the resolution of a combined neon spectrum image would be lower than the resolution of just one image due to the “bloated” line in the combined image. I have tried running sequences using just one neon image and the combined neon image, and there is a difference in the resolution.

If the above is correct, then I have thought about another approach. Could I find the center of each of the neon images (first and last), average those values, and then enter that averaged value into ISIS for calibration. I would then use only one of the neon images in the processing to get a truer resolution. I would then tweak the final calibration result using telluric lines. It seems to me that the only issue might be a greater shift correction using only one neon image compared to two images. And the resolution would not be based on a “bloated” neon line.

Would this work, or am I missing something that would prevent this from working properly?

Cheers,

Keith

[Robin Leadbeater]

Hi Keith,

If you use a common calibration for all spectra (based on either a single or combined neon) and do not re-register the individual spectra then the (bloated) combined start and end neons best represents the resolution in resulting the star spectrum. Note that using the neon to estimate the resolution is only approximate in practise though in any case as the FR of the light from the neon is different from the scope FR so the effect of aberrations will be different. This does not really matter though as knowing the resolution with a high degree of precision is not generally critical. (If you want the most accurate figure then the FWHM of the tellurics in the actual star spectrum will give the best result.)

Cheers
Robin

[Keith Graham]

Hi Robin,

Perhaps I am being overly concerned about resolution. My setup theoretically should be close to 17000, but I have been getting more typically resolutions of 14000 to 16000 depending on varying conditions, The resolution is based on the final outcomes using ISIS and the bloated neon lines, and this is what are reported to BeSS. Your comment about re-registering each individual spectra makes perfect sense. So now I am wondering if ISIS performs the calibration in this way. I know if the Wavelength Registration option is checked then it does (hence the proxxx files). But Christian recommends this process only for long sequences. I am still not clear as to why the Wavelength Registration option would not work for all other sequences that can be checked and tweaked with telluric calibration. To me that would then solve both the calibration and resolution issues. Perhaps Christian can clarify this and whether ISIS does register each image individually to the neon spectrum.

Robin, I am not familiar with the “cross correlation” algorithm. Would you mind explaining what that is and how it is implemented? It soulds like something I would like to try.

Cheers,

Keith

[Robin Leadbeater]

Robin, I am not familiar with the “cross correlation” algorithm. Would you mind explaining what that is and how it is implemented? It soulds like something I would like to try.

Hi Keith,

Essentially, you calculate the correlation coefficient between the two spectra, then incrementally shift them relative to each other and repeatedly recalculate the correlation. The shift where the correlation reaches a maximum is the offset required to bring them into line. For best results you would need to resample the spectra to sub pixel level. Christian may have a more efficient algorithm to do all this eg by using Fourier transforms.

Cheers
Robin

[Paolo Berardi]

If I correctly remember, in one of the latest ISIS versions was adopted a new algorithm for sequence spectra wavelength registration. Christian told us that accuracy has been improved (on condition that the single spectra show numerous and contrasted lines).

I have previously used quite regularly the ISIS WR option on Be star spectra anyway with good results. I understand that the cross-correlation method requires several contrasted lines but it seems to me that old algorithm could usefully be used with hot stars h-alpha region (hi-res). With WR I obtained more sharp lines and a better wavelenght calibration before H2O check and shift correction.

With the new algorithm seems that often enough is introduced a moderate wavelength calibration error on Be star spectra (but is perfect with cooler stars).

Here is an example with the same 5 Cnc data just after process end (note H2O reference). Release 5.0.2 vs 5.2.1:





With 5.2.1 I very often see a high value of first Delta X (strange?), followed by more "regular" intervals. Note this behavior repeats similar for other hot stars.

I wonder if it would be useful to have the old algorithm option or, generally speaking, to implement an algorithm (available as an option) that works with profile shape matching (low frequency intensity variations, not sharp lines, as is often observed in Be stars Ha region). I know we have the "manual option", that is, re-sync by eye the individual sequence spectra but it seems probably less accurate than a mathematical fit.

Mechanical flexure and thermal effect on Lhires III are important problems. It's a pity not be able to use the ISIS WR powerful function to solve. Wavelength calibration is correctable thanks to telluric lines (if visible) but spectral resolution remains compromised.

Ciao
Paolo