spectro-aras.com

Hi Christian (and others) I have some questions on the use of ISIS to help ensure the most accurate processing I can manage. If documentation already exists please just point me to it.

Firstly, I was making a master dark and noticed that after processing them using under Masters Tab the master dark had a number of zero level pixels. So it seems ISIS does not use a "pedestal" value to avoid negative numbers e.g. when the offset is subtracted from the darks. So, if I am correct, does this not lead to some error in the processing?

Secondly, I have a gradient on my LISA flat from the calibration unit so I think the recommended solution is to use the Vertical gain Correction setting with the Ymin & Ymax set just inside the extremes for the illuminated flat field. As this removes the PRNU from the calibration flat should I use a separate PRNU flat generated from a more uniform source directly illuminating the CCD?

So thirdly, if this is so, when do you apply the PRNU flat as I can't seem to locate it in the automatic processing?

Thanks for any advice you can give.

Regards Andrew.

Dear Andrew,

It is not normal to have negative pixels in the dark master. It has not physical meaning, really. Can you indicate the camera type used.
You do make offset the images in a real darkness?

It is possible to use a flat-field at the CCD level. This is the procedure used for the eShel spectra processing (see eShel documentation). But the conditions
are special: only the PRNU is adressed here. Always a second flat-field for the most important low-frequency variations is necessary. These are complex operations.
I recommend using the standard procedure.

Make a flat-field by illuminating the entrance of the telescope for a very "flat" illulmination difficult situations.

Christian

Hi Christian, I think the issue with the negative numbers is due to the very low dark current on my SXVR H694. I checked the offset and repeated the measurement in the dark and got the same values. For an offset master using 16 images I get and average of 2080 ADU with and SD of 6 ADU. A single dark exposed for 900s gave an average of 2092ADU with an SD of 25 ADU. ( I used an area free of hot, cold or warm pixels to make the measurements.)

I am not sure if ISIS averages the Darks before subtracting the Offset or subtracts the offset then averages the darks. However, either way the very low dark current plus the noise could cause negative numbers when the offset is subtracted.

I will look at substituting the offset master with a constant master (2080 ADU) as the stats from the whole frame and randomly selected areas differ by less than 1 ADU

Thanks Andrew

I intended to add for others with this camera that with my example if I take a series of bias frames one after another the temperature rises and I get a higher offset than if I let the temperature recover before taking the next frame.

On readout the temperature rose from -10 to -7 but recovered in a few seconds. It is not an issue in normal use.

Regards Andrew

Hi Andrew, Christian,

Negative numbers can appear because of the noise in the darks and offsets and for best accuracy these should be handled correctly, otherwise you risk getting a false zero. The error is quite small however provided enough averaging of darks is done. The problem is more severe however when dark correcting a spectrum image with a low background. The noise in the image background will be much higher in a single exposure than in the averaged darks and significant negative values will be generated. If these are not treated correctly, for example if they are clamped to zero, the result is an offset in the background level and an error in the final spectrum. How does ISIS treat these negative values?

Robin

Bonjour à tous

Je rencontre un probleme similaire de pixels négatifs vers le bleu sur l'image traitée. En regardant de plus près, les FLAT semblent être en cause, je n'ai plus de problème avec un FLAT synthétique.
Avec l'Alpy çà remonte tragiquement le spectre et donc inutilisable sous les 4000A
J'ai trouvé une "bidouille" qui ne semble pas fausser les résultats en ajoutant une constante au FLAT. Est-ce correct ?
J'ai aussi remarqué que le dark de flat qui dure 2s est primordial, je crée un dark avec coef pour le ramener à dark2s.

Qu'en pensez vous ?

Google trad:
Hello to all

I encounter a similar problem of negative pixels to blue on the processed image. Looking more closely, FLAT seem to be involved, I have no problem with a synthetic FLAT.
With Alpy ca goes tragically spectrum and therefore unusable in the 4000A
I found a "hack" does not seem to affect the results by adding a constant to FLAT. Is this correct?
I also noticed that the quality of flat dark 2s duration is important, I create a dark with coef to bring dark2s.

What do you think?

Pierre

Pierre, Adding a constant to the flat will cause errors in the final result as you are no longer dividing by the correct number. How big an error will depend on how big the pixel to pixel variations are and possibly on any vignetting. This is why packages like Maxim DL do the calculation using real arithmetic and store the files with a "pedestal" value added to the true count to avoid the negative numbers being truncated to zero. The value of the pedestal is recorded in the FITS header and removed before any processing is done.

However, I would have thought that your master flat should have a high ADU value and low noise so as not to cause negative values when the offset is subtracted.

Regards Andrew

Andrew Smith wrote:I intended to add for others with this camera that with my example if I take a series of bias frames one after another the temperature rises and I get a higher offset than if I let the temperature recover before taking the next frame.

On readout the temperature rose from -10 to -7 but recovered in a few seconds. It is not an issue in normal use.

Regards Andrew

Hi Andrew,

it is quite normal (with any camera as far as I know) that temperature rises during readout. In actively cooled cameras this heat is normally quickly transported away (it may be a problem with uncooled DSLR), but heat may build up if a large number of biases are taken after each other. I personally always use 30 sec. in-between each bias exposure, although that is probably way overdone.
But I have noticed with my camera that bias may change during the night if ambient temperature changes too much, so a problem with negative values might arise if the biases to correct the master dark are too far away in time (and if the dark current itself is small).

Apart from that it would be better if ISIS didn't trunkate at zero, especially for faint (noisy) spectra. I guess in this case one should first add spectra and then subtract background (is this what ISIS already does with measurement series? Or doesn't it trunkate to zero internally?)

Cheers,
Sander

Hi all,

please correct me if I am wrong but I think when the master dark is calculated the number format is converted to floating point (this is why the processed images have double memory size), which can also handle negative values correctly. Negative values in the flat however should be avoided. As far as I know negative values are not truncated. This was an issues in Audela eshel back around 2010 when I encountered this problem (wrong subtraction of background), but this has been corrected since. In normal usage if the background is subtracted from two parallel strips on the side of the spectrum, a small residual offset from a drifting bias should pose no problem, but for extended sources the situation may be different.

Regards, Martin

Hi all - I attach a screen capture that show that ISIS seems to be correctly using -ve numbers with my darks (note the negative minimum value). However, in the FITS file the BZERO scale is set to zero and no Pedestal values specified so other packages, Maxim in this case, sets the lowest value is zero.

Christian, would you be kind enough to comment on this? My assumption would be that ISIS is internally using a "Pedestal" value and hence getting the computations correct.

Regards Andrew