ISIS 5.9.6 + limit magnitude eval of a spectrograph

[Christian Buil]

A new release of ISIS (5.9.6) :

http://www.astrosurf.com/buil/isis-software.html

Many internal improvements and also, addition of a tools for an easy name conversion for FITS files coming from disparate sources (MaximDL, ASIair, ...), a simple drag and drop operation (see Miscs tab - > Rename) :



Also addition of in-line commands for help to compute the limit magnitude of a telescope+spectrograph system. Look a complete tutorial here (sorry in French) :

http://www.astrosurf.com/buil/detectivity/

Christian Buil

[Olivier GARDE]

Génial la fonction pour le calcul de la magnitude limite !!!!

[Robin Leadbeater]

Hello Christian,

In your limiting magnitude calculation you say:

"Note : cette relation fait l’hypothèse que le bruit est proportionnel à la racine carré du temps d’intégration, ce qui n’est pas strictement juste (le bruit de lecture est en effet constant avec le temps), cependant l’approximation est correcte quant le temps de pose est long et l’objet faible, le bruit thermique et le fond de ciel finissant par dépasser le bruit de lecture dans les situations limites."

"Note: this relationship makes the assumption that the noise is proportional to the square root of the integration time, which is not strictly accurate (the read noise is indeed constant over time), however the approximation is correct when the exposure time is long and the weak object, the thermal noise and the sky background eventually exceed the noise of reading in the limit situations."

Is this approximation true for low thermal noise CCD cameras ? It is if course true for an infinite exposure time but for example with my ATIK314 the read noise is larger than the thermal noise in a 1 hour exposure. (I have not checked the sky noise level at high resolution though)

EDIT: I have just checked the background noise in some of my LHIRES high resolution spectra and the read noise is much larger than the thermal and sky background noise in 20min exposures

Cheers
Robin

[Robin Leadbeater]

Also I do not see the signal photon noise contribution anywhere. You have measured the noise in the star spectrum by measuring the background.
http://www.astrosurf.com/buil/detectivi ... ee-855.png
Should this not be measured in the same region as the star to include the star photon noise ?

Cheers
Robin

[Christian Buil]

Robin, if the readout noise dominate thermal noise and background noise for a one-hour exposure, for a precise estimation of the detectivity I recommend to take a pair of 3600 seconds exposure (in real conditions, with the background illumination) and compute the difference. In the différence, check if the readout is really dominate.

By definition, the signal noise of the star is very near zero in limit magnitude conditions observation (choose a faint star of course).

Christian

[Robin Leadbeater]

Hello Christian

Robin, if the readout noise dominate thermal noise and background noise for a one-hour exposure, for a precise estimation of the detectivity I recommend to take a pair of 3600 seconds exposure (in real conditions, with the background illumination) and compute the difference. In the différence, check if the readout is really dominate.

I am not sure I will get 2 hours of darkness here until autumn now but I do have a comparison at 1200s. Attached are typical bias, 1200s dark and 1200s star exposures. (I have not measured the differences yet but I think the effect is already clear) The measurement is in the blue using the LHIRES with 1200l/mm grating and ATIK 314 camera 2x binned. (The sky background would be even lower with a 2400l/mm grating unbinned)

Bias sigma = 14.5 ADU (which at 0.28e-/ADU = 4.1e- ie the read noise of the ATK314)
1200s dark sigma = 14.9 ADU (The thermal signal is very low with the ATK314 and makes almost no contribution to the total noise in 1200s)
1200s sky sigma = 15.9ADU. (Here we see a small contribution to the noise from the sky in 1200s but the read noise still be significant)

Even at 3600s I think it is clear that the read noise will still be significant

By definition, the signal noise of the star is very near zero in limit magnitude conditions observation (choose a faint star of course).

Perhaps I do not understand properly but your example shows ~3000 ADU counts/channel from the star and ~50 ADU/channel sky noise. I do not know the gain of your CMOS camera but if it was 1 e-/ADU for example the photon noise would be sqrt(3000) = 55 ADU which is as big as the sky noise.

Cheers
Robin

bias_noise.png (239.82 KiB) Viewed 5794 times

1200s_dark_noise.png (227.58 KiB) Viewed 5794 times

1200s_sky_noise.png (230.21 KiB) Viewed 5794 times

[Christian Buil]

Robin,

The calculation consists of an extrapolation between the target star used for the calculation and a hypothetical star, with a much lower brightness associated with an SNR of 1.

It is important not to consider the noise for the target star. This star only serves to measure the signal (the "gain" of the instrumental chain), not the noise. The photon noise in the hypothetical star (very low SNR) is almost zero. This is the basic principle of the algorithm.

Christian