Use of CMOS detector for faint object spectrography

Design, construction, tuning of spectroscopes
Information and discussion about softwares (telescope remote, autoguiding, acquisition, spectral processing ...)
Christian Buil
Posts: 1431
Joined: Mon Sep 26, 2011 6:59 pm
Contact:

Use of CMOS detector for faint object spectrography

Post by Christian Buil »

Un pas significatif... l'utilisation des derniers capteurs CMOS pour l'observation spectrographique faible flux. La cible est le capteur IMX290 qui équipe la caméra ASI290MM (ZWO).

Voici un table qui compare les caractéristiques de caméras populaires. Le gain et le bruit de lecture (RON) ont été mesurés par l'auteur. Noter le bruit de la caméra ASI290, de 1 électron seulement (j'ai mesuré 0,93 électron avec un gain ajusté à 40 dB !) :

A significant step ... the use of the latest CMOS sensors for faint spectrographic observation. The target is the IMX290 sensor that equip the ASI290MM (ZWO) camera.

Here is a table that compares the features of popular cameras. The gain and the readout noise (RON) were measured by the author. Note the noise of the ASI290 camera, only 1 electron (I measured 0.93 electron with an internal camera gain adjusted to 40 dB!).


Image

Une autre table qui résume mes mesures du rendement quantique. Un long et pénible travail, mais cela permet d'y voir plus clair - ce sont des informations rares. Noter le comportement du capteur IMX290, qui est comparable aux meilleurs CCD :

Another table that summarizes my measurements of quantum efficiency. A long and arduous work, but it allows to see more clearly. Note the behavior of the IMX290 sensor, which is comparable to the best CCD:

Image

Le très faible bruit (important en spectrographie), le fort rendement quantitique et la petite taille des pixels changent la perspective...

The very low noise (important in spectrography), the high quantitative efficiency and the small size of the pixels change the perspective...

Le setup, un LISA utilisé avec une fente très étroite (de 15 à 19 microns). Une ASI178 est utiisée pour le guidage, une ASI290 pour l'acquisition principale :

Image

Image

Un spectre de l'étoile Véga (le seul problème est la petite taille du détecteur, mais cela va évoluer). Noter le temps de pose élémentaire : 0,1 seconde (alors que la fente ne fait que 15 microns de large) :

A spectrum of Vega (the only problem is the small size of the detector, but this will evolve). Note the elementary exposure time: 0.1 second (and 15 microns wide slit):

Image

Un spectre de la Nova Sct 2017 (forte turbulence et cirrus malheureusement). Un temps de pose finalement court, et une haute résolution (record) avec le spectrographe LISA. J'ai mesuré R = 2500. Par exemple le doublet du sodium est bien séparé. Le profil P Cygni de nombreuses raies est bien visible et noter les fins détails dans le bleu. Le LISA est poussé loin grace à cette caméra et un peu de méthode (choix délibéré ici d'une fente étroite):

A spectrum of the Nova Sct 2017 (strong turbulence and cirrus unfortunately). Short exposure time, and a high resolution (record) with the LISA spectrograph. I measured R = 2500. For example, the sodium doublet is well separated. The P Cygni profile of many lines is clearly visible and note the fine details in the blue. The LISA is pushed far away - thanks to the camera and some method:

Image

Image

Maintenant le spectre d'une étoile Be de magnitude 7,3. J'ai ici utilisé un gain de 40 dB (multiplié par 10 par rapport à l'observation de la nova Sct 2017). Le détecteur est à -25°C refroidissement très efficace sur ces petites caméra car le détecteur consomme peu de courant). Le temps de pose unitaire est de 6 secondes - je sature si je pose plus longtemps (!), malgré un très mauvais seeing (4 à 5 seconde d'arc) :

Now, the spectrum of a Be star Be at V = 7,3. I used here a gain of 40 dB (multiplied by 10 compared to the observation of the nova Sct 2017). The detector is at -25°C - very efficient cooling on these small camera because the detector work with little current). The unit exposure time is 6 seconds - I saturate for longer exposure (!),despite a very bad seeing (4 to 5 seconds of arc)::

Image

On verra et testera plus avant tout cela au stage spectro OHP de la semaine prochaine...

Christian Buil
Olivier GARDE
Posts: 1243
Joined: Thu Sep 29, 2011 6:35 am
Location: Rhône Alpes FRANCE
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Olivier GARDE »

Christian,

Je pense que ce genre de caméra CMOS peut être utile sur les objets trés faible ayant peut de dynamique, (ou là on aura réellement un gain en terme de temps de pose total sur la cible) mais dans le cas d'objet qui demande pas mal de dynamique comme une étoile Be avec une raie H Alpha en emission intense qui se situe à des niveaux de 10x-20x le continuum, les 12 bits de la caméra risquent de ne pas suivre en terme de dynamique ?

On risque de se retrouver avec des spectres qui saturent vites au détriment d'avoir du signal exploitable dans le bleu ?
LHIRES III #5, LISA, e-Shel, C14, RC400 Astrosib, AP1600
http://o.garde.free.fr/astro/Spectro1/Bienvenue.html
Thierry Lemoult
Posts: 136
Joined: Thu Sep 29, 2011 9:08 am

Re: Use of CMOS detector for faint object spectrography

Post by Thierry Lemoult »

Hello

Whaou, 1.1e- de bruit de lecture ! Bravo Christian pour ces avancées.

Je pense aux spectres de NP super faible... Cela pourrait être bien a condition d'arriver a concentrer le flux sur les petits pixels.
Lors de notre mission a saint veran sur le NP, on utilisait le LISA+ATIK460 en bin 2x2 donc des pixel de 9 microns.
Mais avec le CMOS en "bin 3x3", le bruit sera 3 fois plus grand.. donc 3.3 e-, c'est mieux a que avec l'atik460 à 5.5 e- ?

Quid du non refroidissement ? Tu fait des dark/ pffset dans la meme nuit ?

Donc le spectroscope de mes reves pour les NP: Résolution 1000 avec une fente de 50 micron, F/D=6 en entrée, le tout bien adapté a des pixel de 3 microns.
Ca fait un gros facteur de réduction dans le spectro... Et donc un objectif de devant le détecteur super ouvert..
Christian Buil
Posts: 1431
Joined: Mon Sep 26, 2011 6:59 pm
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Christian Buil »

Olivier, les Be qui ont des raies en emission 100 fois plus forte que le continuum sont plutôt rare. Il est vrai malgré tout que la faible dynamique oblige à jongler sur le gain et à bien faire attention à ne pas saturer. Un usage normal de ce type de composant est de pratiquer des séquences de poses brèves que l'on additionne ensuite - c'est un moyen d'augmenter la dynamique.

La petite taille des pixels est aussi un moyen d'accroitre la dynamique car on sur-échantillonne assez fortement les raies (effet de redondance...)

Thierry, attention en CMOS, on ne réalise pas un binning analogique comme avec un CCD (dans la majorité des situations). Donc faire binning 2x2 ne permet pas de gagner en détectabilité. En CMOS on est toujours en binnng 1x1 (saut au pointage par exemple). Le binning 1x1, 2x2 ... n"apporte pas grand chose.

Christian
Robin Leadbeater
Posts: 1926
Joined: Mon Sep 26, 2011 4:41 pm
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Robin Leadbeater »

I see ATIK have issued the specification for their new Horizon CMOS camera
https://www.atik-cameras.com/news/atik- ... ta-launch/

Robin
LHIRES III #29 ATIK314 ALPY 600/200 ATIK428 Star Analyser 100/200 C11 EQ6
http://www.threehillsobservatory.co.uk
Thilo Bauer
Posts: 37
Joined: Sun May 31, 2015 4:45 pm
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Thilo Bauer »

Christian Buil wrote:...Donc faire binning 2x2 ne permet pas de gagner en détectabilité.
That's right. Reason is given here:
http://www.astroinformatics.de/images/s ... s-2010.pdf

CMOS without doubt is the future of imaging. And it is also Sony, who already announced to no longer develop CCDs. Current DSLR demonstrate what is possible with respect to low read-out noise and large dynamic range.

Unfortunately current CMOS Imagers for astronomy don't provide 16 Bit converters. Although the more sensitive sensors provide a full-well capacity, that would demand for at least 16-Bit for the A/D converters to obtain best signal in terms of quantum conversion rate (=photon count per digital unit).

Binning on the other side could be thought to be a statistical approach to larger dynamics and overcome stupid limitations of the A/D converter. Like with the myth, that binning could detect fainter signal. Averaging pixels to obtain a better dynamic range based on a statistical consideration of averging by binning is a wrong conclusion, too. But we can expect, to correct for oversampling of our spectrograph while having spread the light over more pixels and add number of photons collected after recording the image. But this requires no arbitrary gain factor of a detector, but a quantum conversion rate of 1 (but still suffers from an increase of noise due to camera read-out and dark current).

BTW: Another good article of how to optimize detector gain: http://www.clarkvision.com/articles/iso/

In the final end, nothing but a high dynamic range of any imager having a quantum conversion rate of 1 (=1 photon count per digital unit) will ever yield an optimal signal.

I really do not understand, why modern cameras like the Atik still come with just 12-bit converters. The only gain one will observe is gain of money for the manufacturer driven by management decisions. :shock:

An interesting approach to count real photons are EM-CCD cameras, like the ones used in fluorescence microscopy (resellers like Zeiss, Germany). Unfortunately, this is still a very expensive choice of a camera. If I had to decide what to buy first, I decided to buy both, a full-featured telescope with a spectrograph plus brand new fluorescence microscope from Zeiss. That was cheaper in the final end, compared to a single EM-CCD imager. Advantage: I can do both, good and bad weather.
Olivier GARDE
Posts: 1243
Joined: Thu Sep 29, 2011 6:35 am
Location: Rhône Alpes FRANCE
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Olivier GARDE »

Thilo Bauer wrote: I really do not understand, why modern cameras like the Atik still come with just 12-bit converters
Hi Thilo,

But all ATIK CCD are all with 16 bits CAN :
https://3ainmfntxe31vi9qd1pxgpd1-wpengi ... Feb-17.pdf

Maybe you mean the ASI/ZWO manufacturer that it only offers CMOS in 12 bits ?
LHIRES III #5, LISA, e-Shel, C14, RC400 Astrosib, AP1600
http://o.garde.free.fr/astro/Spectro1/Bienvenue.html
Robin Leadbeater
Posts: 1926
Joined: Mon Sep 26, 2011 4:41 pm
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Robin Leadbeater »

For CCD, as opposed to CMOS, on chip binning potentially improves SNR in certain circumstances because the read noise is reduced for the same collecting area. There are some caveats to be considered even in CCD binning though.
To be absolutely sure of not saturating any individual pixel the maximum allowable counts per pixel should strictly be reduced by the number of binned pixels. (eg by 4 in the case of 2x2 binning) That is, in the worse case all the light might fall on only one of the 4 pixels. If this restriction is strictly applied, there is then no SNR advantage to binning in the case where there is enough signal to fully expose the unbinned pixels to full well depth, only a reduction in exposure time. This restriction is unnecessarily strict though as for a correctly sampled image the psf of a feature would change only slowly over the binned pixels. Has anyone a recommendation for a reduction factor that should be applied to the maximum counts/pixel when binning?

Robin
LHIRES III #29 ATIK314 ALPY 600/200 ATIK428 Star Analyser 100/200 C11 EQ6
http://www.threehillsobservatory.co.uk
Robin Leadbeater
Posts: 1926
Joined: Mon Sep 26, 2011 4:41 pm
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Robin Leadbeater »

Olivier GARDE wrote:
Thilo Bauer wrote: I really do not understand, why modern cameras like the Atik still come with just 12-bit converters
Hi Thilo,

But all ATIK CCD are all with 16 bits CAN :
https://3ainmfntxe31vi9qd1pxgpd1-wpengi ... Feb-17.pdf

Maybe you mean the ASI/ZWO manufacturer that it only offers CMOS in 12 bits ?
The 12 bits ADC is built into the CMOS sensor. Not chosen by the camera manufacturer.

ATIK have an interesting comparison discussion here.

https://www.atik-cameras.com/news/diffe ... s-sensors/

I found this statement particularly interesting

"It is important to note, though, that read noise on a CMOS sensor is linked to full well depth, and using the sensor at its lowest read noise settings is usually at the expense of well depth. At the full well depths we tend to use for deep sky imaging, there’s actually little difference between CMOS and Sony CCD sensors"
LHIRES III #29 ATIK314 ALPY 600/200 ATIK428 Star Analyser 100/200 C11 EQ6
http://www.threehillsobservatory.co.uk
Thilo Bauer
Posts: 37
Joined: Sun May 31, 2015 4:45 pm
Contact:

Re: Use of CMOS detector for faint object spectrography

Post by Thilo Bauer »

Hi Olivier,
Olivier GARDE wrote:But all ATIK CCD are all with 16 bits CAN :
https://3ainmfntxe31vi9qd1pxgpd1-wpengi ... Feb-17.pdf
This is true for the CCDs, except the Atik CMOS camera linked by Robin. This is announced to have just 12 Bit, like the ASI cameras. That is what I'm referring to.

I think the statement of Atik might be true for this unique sensor introduced by Atik and probably others. Designs vary, however, and certainly special designs for scientific applications shall be expected.

The CMOS sensors found in latest DSLRs (EOS 100D, 40D, 60D, 77D), which I have personally in use for astronomy and fluorescence microscopy, continue to constantly improve. The EOS 77D is a very quick one and I found problems to reproduce exposure times of the 60D, because the 77D again provides less noise and is also more sensitive compared to the 60D. So I have to reduce exposure time and ISO number of 100 it surprisingly too sensitive for certain illumination in microscopy. Less intensive LED illumination means protists are better surviving microscopic illumination stress. The story continues to improve for the CMOS imagers.

CCD is dead end. So I currently ask myself if it is wise to buy a CCD today and even tomorrow or if it is better to continue to go with CMOS sensors to replace the famous EOS 60D / 77D for applications in spectroscopy.

I can't follow the reasons given by Atik. And I personally feel they hurt themselves writing such statements.

There is no amplifier glow found with latest CMOS imagers in DSLRs. Full-well capacity is much higher and ADCs provide 14 bit. Fixed pattern noise is caused by the special designs of addressing pixel individually, but can be perfectly corrected by dark frames. So the ASI and Atik ones may have the advantage of better gain of signal with the monochrome sensors. Unfortunately, there is no direct possibility to compare my 60D with these cameras (not yet...). The most significant disadvantage of Canon + Alpy is probably related to the Barlow lens required for the backfocus, not the sensor itself (if binning is a requirement, then adaptation is wrong, but cannot be avoided in this case).

I still hesitate to buy any of the new CMOS imagers for astronomy, because they are limited to 12 bit. My last Canon with 12 bit ended up as a present for a nice lady doing brilliant nature photography. She is happy since that time, and so am I. ;)

As CMOS imagers continue to improve, I bet there will be any next generation DSLR sensors coming with 16 bits. Of course cooled sensors work better even in case of a CMOS, which is typical for silicon. However the low dark current of a modern CMOS is fascinating. It was Kodak which first provided low dark current CCD imagers. Not quite sure, if this was a first approach of a hybrid mix of CCD and CMOS technology (I remember to have heard such rumors at the time when these Kodak chips were launched, but found no written confirmation in application notes).

In astronomical literature CMOS imagers are worth just a footnote (McLean, 2008). But I think, this book is almost ten years old and history. Improvement in design and electronic details lead to almost perfect noise reduction. Success of development over the past two decades is overwhelming. There has been changed a lot in the meanwhile. Such properties of the CMOS sensors weren't expected from the beginning and they still suffer from a myth that CMOS will introduce lots of noise.

I currently think about which of the current astro-CMOS cameras might perfectly fit to Alpy. So I appreciate any feedback from you, which will drive my decision which one to buy. :D I certainly won't go for CCD any longer, maybe wait a few more months. The many disadvantages of CCDs are perfectly summarized in McLean (2008). Nothing to add here.

Christian, again thanks for the nice article!

Best regards,

Thilo

Literature:
Ian S. McLean: Electronic Imaging in Astronomy: Detectors and Instrumentation, Springer 2008.
Post Reply