Temps de poses VS Nombres de poses

[etienne bertrand]

Bonjour,

Difficiles questions pour moi, pour faire des objets faibles que préférer ?
Avec Alpy600 la fente faisant 23µm il faut utiliser le bin1x1 pour bien échantillonner (actuellement j'ai monté la fente 18µm), mais dans ce cas pour avoir assez de signal avec mon c8 qu'est ce qui donnera le meilleur résultat ? Je penses à des objets de magnitudes 15/16 :
- faire des poses vraiment longues comme 1800s ou 2400s unitaires ; quand c'est possible je fais 4 ou 5 x 2400s ?
- faire des poses plus courtes mais plus nombreuses comme 10 x 1200s ? (mais avec 1200s, pas sûr que j'ai grand chose comme signal...).

Deuxième question, sur l'Alpy600, j'utilise une Atik414EX, mais quand je fais des poses longues comme 2400s j'ai l'image remplie de petits pixels chauds, avec l'Atik314L+ sur des poses de 2000s je n'ai quasiment pas de pixels chauds. Mon Atik414EX a-t-elle un problème ? Est ce que cela me dégrade les spectres ?

[Robin Leadbeater]

Hi Etienne,

I have been thinking about this too. For measuring the faintest objects at low SNR from a dark site, read noise is the biggest source of noise after the photon noise so anything to reduce this helps. This means concentrating the spectrum on the minimum number of pixels (binning if necessary) and taking the minimum number of exposures. The problem is that this increases the risk of damage in the spectrum from hot pixels, cosmic ray hits. Also if your spectrum is not well sampled, it can be difficult to remove sky lines. ( I found this with the ALPY 200 and a 2x2 binned ATIK 314. (12.9um) This gives <2pixels/slit width. Removing the sky lines is much easier now I have changed to the ATIK 428 2x2 binned (9.1um) at 2.5pixels/slit width.) I have also thought about binning 4x2 to reduce the read noise further.

I have also noticed the increase in warm pixels with each generation of Sony CCD. Fortunately these are only warm, not hot and the dark correction seems to correct them well. There is a post discussing them here for the ATK460
http://www.spectro-aras.com/forum/viewt ... 919&p=9871
and also on the BAA forum here
https://britastro.org/node/15502
Is this what you are seeing with your ATK414 ?

Cheers
Robin

[Peter Somogyi]

Hello Etienne and Robin,

I have both the ATIK 414 (topic opener) and the 428 (characteristics very similar to the 460), so have some practical experience.
The 414 does have certain dark signal, whilst the 428 almost nonexistent.
However, for the 414 I was measuring 4 electron readout (bin 1x1) vs 6 electron for the 428 (bin2x2).

If you stay below 20 minute, my vote for low res would be definitely the 414.
However, the 428 can go up to 40 minuter (a few times i did), then my insight it might jump over the 414 (at around -15C - however, my the 414 chip does appreciate more cooling than my typical -15C).
By practice, the weather/wind/star loss conditions, start/stop limits, and star loss possibility renders me to stick with 20 min - rarely, going up to 30 min, but above a factor 2x exp time, the darks ought to be redone for the 414.
For creating dark lib, don't stop at 10 (it's very few). Just put it out on a rainy night, and gather at least 30 or so (the higher is better). The 414 is more sensitive to this.

Fortunately, in case you setup the Alpy's core lens firmly, then the exposure time won't limit you.

Peter

[etienne bertrand]

Thanks Robin and Peter for yours responses.
I make reflexion since 3 days for this question...
I have a number of darks 7 or 8 numbers by time , like you say Peter your Conseil is have ~20 darks by time exposition ?
The 414EX works good so with my little diameter of telescope (8'') I think I continue with long exposure. Normally the signal is coerent, on QSO I have good continuum and redshift. I think too with the 414EX is better function with captor at -20 I think, -10 it's a minimum.

Robin you use 428EX CCD now ? No the 414EX ?
When you take spectra of faint object you always use bin2x2 ?

[Olivier GARDE]

Etienne,

Ce qui est important c'est que sur une pose unitaire tu arrives à capter l'info nécessaire, si ce n'est pas le cas alors tu n'obtiendras guère d'infos supplémentaires sur N poses. C'est la technique que nous avons adopté pour la confirmation de NP ou l'on commence à faire une pose unitaire de 1200s, puis l'on regarde à l'écran ce que l'on obtient en poussant les seuils à fond, si l'on a du signal alors on continu à réaliser des poses pour améliorer le rapport signal bruit, par contre si 'lin obtient rien sur une pose unitaire alors on passe à une autre cible.

Après au niveau traitement il faut prendre également en compte :
- Le nombre de Dark et d'Offset : plus il y en a mieux c'est. Perso sur mon spectro eShel, je fait une trentaine de Dark et une cinquantaine d'offset
- Le nombre de flat : également moi j'en fait plus de 40, cela permet d'améliorer le flux dans le bleu du spectre ou une lampe tungstène et la réponse du CCD n'est pas des meilleurs.
- LA température du CCD

[etienne bertrand]

Bonjour Olivier,

Merci pour ton message.
Je fais pareil pour le signal, si je n'ai pas de continuum ou des raies d'émission visible je passes à d'autres cibles.
Pour le nombre de darks, je vais revoir ma bibliothèque. Dès que le temps de poses est long, je me suis arrêté à 7 ou 8 darks, si je comprends il en faudrait plutôt une quinzaine ou vingtaine.

Hier j'ai fait la comète 64P et pour avoir un meilleur signal sur le flat dans le bleu j'ai posé sur les flats de manière à arriver à 57000 ADU, pas plus. Tu valides ce nombre d'ADU ou ça fait trop pour un flat ? (j'en fait toujours au moins 41).
Même si le flat commence à être bruité un peu en dessous de H&K CaII le signal final est quand même bon après traitement.

[Peter Somogyi]

Etienne,

"Hier j'ai fait la comète 64P et pour avoir un meilleur signal sur le flat dans le bleu j'ai posé sur les flats de manière à arriver à 57000 ADU, pas plus. Tu valides ce nombre d'ADU ou ça fait trop pour un flat ? (j'en fait toujours au moins 41)."

My measurement for the ATIK 414, that it's linear up to - very roughly - 48000 ADU.
(As for my earlier camera Orion G3, this was 32000 ADU !!! although its max is also 64K).
I'm a bit unsure about ATIK 428, haven't done it precisely. But I see severe linearity problems at 60K ADU, hence doubt that this was fully linear. Therefore setting my flat limits to around a safe 50K ADU.

I agree with Olivier, if blue/UV has low SNR, that can only be treated by numerous flat repeats.

When I read Demetra's handbook, filtration was mentioned. That might be an alternative when you're low on flats/darks/biases.
Of course, it all matters more and more when high SNR is targeted (e.g. a bright star with tiny features to measure).
When the target is faint, gathering tons of calibration frames won't help much. Only thing when doing jumbo exposure (> 20 min), to do 2-3 frames of dark, but just only for cosmetics processing (not absolutely important, but it helps identifying + removing defects more accurately). Doing more just only to elaborate cosmic rays. Otherwise, filtration techqniques on darks/flats/biases might help (never tried).

- Peter

[Robin Leadbeater]

Robin you use 428EX CCD now ? No the 414EX ?
When you take spectra of faint object you always use bin2x2 ?

Hello Etienne

I do not have a 414EX. just a 314EX which I use with my LHIRES and a 428EX for the ALPY. Yes I always 2x bin with faint objects at low SNR. I take at least 20 darks and flats but I take them on cloudy nights and reuse them until I change spectrographs.

What is important for keeping the noise low is quite different for bright objects and high SNR compared with faint objects and low SNR.

You could use Christian Buil's calculator to explore the effects
http://www.astrosurf.com/buil/etcl2/etcl_en.htm
(it is for the LHIRES but the effects of changing the camera parameters and exposures are the same for all spectrographs)

At high SNR, a good well exposed flat is important, particularly at the blue end, but is not so important if the SNR is already low.

For faint objects at low SNR the read noise becomes very significant so it is important to keep it low by using a low number of long exposures and not over sampling, binning if necessary (note this is for CCD, not CMOS)

Thermal noise for faint objects at low SNR can be significant but it depends on the camera. It is not significant for my 314 or 428 but it might be for your 414

Sky background can also be important with faint objects so it is important to keep the sky background included with the signal to a minimum (The slit width matched to the star image size, good seeing, perfect focus and minimum guiding errors in both directions) and to take a wide sky zone so the sky subtraction does not add much noise.

Cheers
Robin

[etienne bertrand]

Hello Peter & Robin,

My flat of yesterday at 5700ADU it's more hard marybe, normally i take flat with 44000ADU and 41 numbers.
How many times during yours banks of darks ? 1 year ? 2 years ? And offset ?

Now i use the 18µm slit on Alpy, it's better for cometes the Résoltuion is ~610. For Nébulae i can separate [NII] and Ha...

For the desicant : you had change it in yours caméra ?

[Peter Somogyi]

Dear Etienne,

I typically do 11-15 flats for the Alpy + ATIK 414 (depends on how much signal expected in blue).
There's an ambient part of the light in UV/blue (cannot substract a const setting background to 0 due to other division problems close to 0).
Note that the ATIK 428/460 having its usual ripples at low resolution (R~50A ?) detected with LHires + 600/mm (R~1200-1800 average) being present even between the 3550-4000A region. This means: it's a less severe problem for the 414 to correct the blue end appropriately, even when the background light is coming up.

I haven't redone darks yet for the 414 since 2015 (just adding like +10-20 every year), but maybe it's just my neglection. For 428 (active project with LHires III for me) I do +20 every year for -6, -9, -12 and -15C occasionally and weighted-averaging with the old darks ("running average"). But, also having a smaller -20C library, needed because sometimes encounter -18C or less on winter... However, I stop activity below -20C ambient temperature
As for biases, I'm redoing (33 for 414, 44 for the 428) per every observational night (+/- few days e.g. at weekend only once) for every temperature. The ATIK 428(/460) is the most sensitive on readout noise, so why not recreate every time (takes only a few minutes). Deleting from disk as soon as processed in ISIS.

An alternative could be filtration of new darks/biases, but never tested.
The highest priority is getiing high enough SNR (as Robin mentioned above) + interesting enough observations/project followups, as an amatur interest these are what I'm focusing on...

- Peter