Gamma Velorum Drizzle vs Normal Combining

[sjastro]

In the 3C-273 thread there was no difference in the quality of the spectra for drizzle and normally combined sub exposures.
This was due to the telescope/camera set up not resulting in undersampled images and a limited degree of dithering between exposures.

Instead of using my BRC-250 scope, I used a 300mm lens which gave a resolution of 4.7 arcseconds/pixel resulting in undersampled images.
Each exposure of Gamma Velorum (10 X 20 seconds) was dithered using a spiral routine and displacement of 4 pixels.
These exposures were drizzled combined.

As a comparison another set of 10X20 second exposures without dithering were normally combined.

Since the drizzling process resamples the combined image by 2X, the normally combined image was also resampled 2X.

The results speak for themselves.

Steven

[Atmos]

That's quite interesting Steven. It is in agreement with what I see when dealing with non-spectra images as well.

[Merlin66]

Hmmm
One other option to consider is the dithered stack without the drizzle....

I'd also check the alignment and stacking technique....

This will allow a better analysis.

[sjastro]

For some reason the normally combined image didn't resample.
This has been fixed and there is now a more realistic comparison.

Steven

[Merlin66]

;)

[sjastro]

Yes it seemed too good to be true, the 2:1 ratio in the peak heights in the original attachment was a dead give away that something went wrong with resampling.:)

The other issue is that the Gamma Velorum spectrum sticks out like a sore thumb, in fact the major emission line reached the saturation limit of the CCD.
I might try this again at much shorter exposures where a lower SNR might lead to a greater differentiation between the combining techniques.

Steven

[robin_astro]

Hi Steven,

The apparent difference in the continuum between the two techniques is also of concern. This suggests either a non linearity in the process or alternatively a difference in the binning region or background subtraction. (The binning region should include the full width of the spectrum, including all the rows with spectrum data, to avoid distortion of the continuum shape due to changing focus along the spectrum and the background region should completely exclude the spectrum to avoid cross contamination)

Cheers
Robin

[sjastro]

Robin,

I think the main culprit is the overexposure.
My CCD does not use anti blooming gates.
The spectrum was not only saturated indicating the CCD response was no longer linear but the main C emission line was also exhibiting blooming.

The trouble is the CCD capture software automatically stretches the image and undesirable features such a blooming spikes are lost in the image.
What I need to do in the future is to take test images of spectra and measure the pixel count in the emission lines to ensure I am in the 30-50% operating range for the CCD.

It's all part of the learning experience for me.

Steven

[Merlin66]

Steven,
It should be possible to find a compromise - between sub exposures and full well depth.
I typically use 5 to 10 min subs with a C11 using an ATiK314L and AstroArt.
An average combined stack of 20 subs usually gives the SNR I need.

[Merlin66]

I was re-reading the thread....
Why are you using Registax??

[sjastro]

That's an impressive spectrum Ken.
It will be my next step to make the transition to higher resolution spectroscopy.

Registax was only used to combine two separately PixInsight stacked images of 3C-273 spectra collected over different nights.
For what ever reason, there is not much latitude for alignment errors when aligning and stacking spectra when compared to deep sky images.

For Gamma Velorum keeping the alignment error to a minimum and tweaking the settings in PixInsight there was no need for Registax for stacking spectra images.
Since dithering was used, autoguiding was necessary to control the dithering.
The maximum alignment error in this case was 4 pixels which PixInsight was able to handle.

Steven

[Merlin66]

Steven,
That's where the custom designed spectroscopy software comes in....
Programs like ISIS or BASS Project will do all the stacking and alignment for you and provide a solid means of selecting and binning the target Spectral image as well as the necessary background sky removal.
I'd suggest running your data through one of these programs.
I use BASS Project for the processing of all my Spectral data.
Ken
https://uk.groups.yahoo.com/neo/groups/astrobodger/info
http://www.astrosurf.com/buil/isis/isis_en.htm

[sjastro]

Ken,

I downloaded BASS and experienced the same registration problems on the 3C-273 spectra.

[Merlin66]

Steven,
Hmmmm
Which alignment method did you use for the frames?

[sjastro]

Ken,

I used the align images option.
Like the other software packages I have tried for stacking and aligning spectra, if the imaging session is interrupted one needs to be careful in relocating the object in the same position on the chip, otherwise aligning images is not accurate.
The margin of error seems to be less for spectra than for deep sky objects.

Steven

[Merlin66]

Steven,
Ok.
If you select sufficient alignment points BASS usually is smart enough to do a fair job with the stacking.....

[sjastro]

Given the blooming issues the test was repeated using 30 X 2 sec sub-exposures.
This gave a pixel count of around 37,000 on the C emission line, well within the saturation limit.

The drizzle combined spectra being closer to the X-axis indicates dithering has reduced the background noise as expected.
On the negative side drizzling does introduce a small amount of noise which might explain the smaller peak heights in the drizzle combined image.

This is supported by SNR calculations.
The main C emission peak in the drizzle combined spectra has a maximum SNR of 195, in the normally combined spectra the maximum SNR is 228.

The conclusion is that there are no earth shattering improvements in using drizzle combined undersampled images.

Steven