CMOS vs CCD for spectroscopy applications

Design, construction, tuning of spectroscopes
Information and discussion about softwares (telescope remote, autoguiding, acquisition, spectral processing ...)

Re: CMOS vs CCD for spectroscopy applications

Postby Andrew Smith » Fri Sep 08, 2017 8:38 am

Interesting work as always Christian.

I had not come across the term Telegraph noise before but it seems to be the "small scale" cause of 1/f noise - http://www.nii.ac.jp/qis/first-quantum/ ... apter9.pdf.

I am currently making a scaled down version of your VHIRES with a target R ~ 20,000 using a Borg 55FL f/4.5 telescope (250mm focal length). If you were purchasing a camera for it to day what would you recommend?

Regards Andrew
Andrew Smith
 
Posts: 237
Joined: Fri Sep 30, 2011 12:23 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Christian Buil » Fri Sep 08, 2017 5:21 pm

Hi Andrew,

Make a VHIRES spectrograph type on the base of a 55mm Borg is really an excellent initiative. It is a financially reasonable project.

The first key is a good refractor. The Borg optics with its high f/d is a good choice for me (the relative aperture is an important point in addition to the optical quality - apochromatism or semi-apochromatism).

The second key concern the echelle grating of course. Have you made your choice ? Are you going to use a grating that covers the 55 mm pupil?

You select a VHIRES version (an order) or a VHIRES-MO version (multi-order).

(note, in a few days, some interesting results about fibers will be presented ... !)

Difficult to conclude on the CCD. The ATIK460EX camera remains a good choice, but I must admit that the increase in the spectral range (in VHIRES-MO version) by using the ASI1600MM is attractive fo me, and some new processing are applicables. Of course a large format CCD with a noise less than 5 electrons (important), would be good solution.

Christian
Christian Buil
 
Posts: 1022
Joined: Mon Sep 26, 2011 6:59 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Andrew Smith » Sun Sep 10, 2017 11:58 am

Hi Christian, I do have an echelle grating. I went, copying you, for the same 110 l/mm grating with a ruled area of 56 x 142mm.

My thoughts was to focus just on H alpha using a filter (VHIRES) at least to start with. I have my temperature controlled MRes echelle with R ~ 8000 and was looking for a simple fiber fed spectrograph that could contribute at the resolution of a LHires III at least in H alpha where I can use the telluric lines for accurate calibration without the need for temperature control. In this case the CCD may be best.

My aim is to be able to switch between spectorgraphs but just switching the fiber over at the quide head so as to leave the telescope in a fixed state so that the automatic operation did not need recalibrating. (I have too few clear nights to waste time on calibrations.)

I look forward to your work on fibers but hope it will not lead to to much extra expense on my part to continue to follow the master!

Regards Andrew
Andrew Smith
 
Posts: 237
Joined: Fri Sep 30, 2011 12:23 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Christian Buil » Sun Sep 10, 2017 5:54 pm

(Hi Andrew, the infos about fibers concern evaluation of optical transmission near UV spectral domain vs type of fiber, the cost is not really impacted)

A recent spectrum of Be star QR Vul taken with VHIRES-MO + ASI1600MM:

Image

The increase of the noise on the right part of each order is caused by the small 90-deg. folding mirror placed just after the optical fiber. It causes a beam obstruction on one edge of the ASI1600MM camera detector, wider than the ATIK460EX camera detector.

After a new adjustment of VHIRES-MO obstruction disappeared. I can also capture in the same image frame the halpha line
and Na sodium doublet. Thanks to the large surface of the ASI1600MM Panasonic detector! Here, a daylight spectrum:

Image

Because VHIRES-MO capture only a partial part of the spectrum at the same time, the Hbeta line (for example) is just outside the sensor (:-() - the
overside of high spectral resolution. But now, the presence of Na lines in the captured domain is good news for astrophysical research.

Christian B
Christian Buil
 
Posts: 1022
Joined: Mon Sep 26, 2011 6:59 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Thilo Bauer » Sun Sep 17, 2017 7:29 pm

Christian,

I strongly discourage the use of a median filter for pixel binning. Reason is, that the median is a non-linear filter. Pixel intensities are sorted ignoring information about their position. Hence, the "middle" intensity picked up by the median filter is not the same as the expectation (mean) value at the observed pixel position. This will become even worse using a larger median kernel. This arbitrary scatter of pixel positions caused by the median filter is visible as image artifacts and thus will impact further image analysis and position accuracy like radial velocity measurements. For pixel binning a simple average of 2x2, 3x3 or 4x4 pixels is best to preserve the signal.

Andrew,

I found many of the papers describing the effects of 1/f (telegraph) noise very confusing. It is getting even more contradictory, if one tries to compare CCD and CMOS. Certain authors describe the effect typical for MOSFET, others describe the effect typical for CMOS sensors. In the final end, the 1/f noise energy also depends on the chip designs of the CCD and CMOS devices. There is also correlated double sampling used to suppress certain noise of the imagers. Certain CMOS cameras (also DSLR) show temporal line intensity variations in the image. If the effect is similar with the ASI 1600, the root cause is the same. I suggest this is caused by internal measure of an arbitrary line bias before further processing of the sensor signal. The effect is also seen as a double peak dark current histogram (noise by line bias variations). There is also fixed pattern noise found in dark signals when averaging many (>10) dark frames.

The following paper shows a broad pixel histogram of the noise (this is not found with modern sensors, however).
http://harvestimaging.com/pubdocs/103_2006_dec_IEDM_random_telegraph_noise.pdf

Keep in mind: The paper was state of the art in the year 2006. Compare this with image histograms of modern CMOS or CCD imagers.

There is also an interesting note found in this paper: "Illuminating the sample with infrared photons has the same effect as increasing the temperature for the thermal activation mechanism..."

It seems even more complicate: Image noise and detector (1/f) noise in spectroscopy is expected to also depend on the observed wavelength.

To overcome 12 bit limitation: Use an average of 10 or more dark and flat frames.

Best regards,

Thilo
Thilo Bauer
 
Posts: 36
Joined: Sun May 31, 2015 4:45 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Robin Leadbeater » Sun Sep 17, 2017 9:20 pm

Thilo Bauer wrote:To overcome 12 bit limitation: Use an average of 10 or more dark and flat frames.


Hi Thilo,

Generating spectroscopic flats for wide wavelength band spectrographs using Halogen flats, (like the ALPY, LISA for example) needs many more averaged frames than this because of the very high dynamic range required (from the low output of the lamp and low sensitivity of the camera in the blue compared with in the yellow.) At least 20x16 bit frames are needed so 12 bit would need at least 320 frames. In fact depending on read noise it might not even be possible to generate a satisfactory flat using CMOS in these circumstances (Christian Buil's results with CMOS to date are on narrow wavelength regions where this flat problem does not arise.)

Cheers
Robin
LHIRES III #29 ATIK314 ALPY 600/200 ATIK428 Star Analyser 100/200 C11 EQ6
http://www.threehillsobservatory.co.uk
Robin Leadbeater
 
Posts: 905
Joined: Mon Sep 26, 2011 4:41 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Christian Buil » Mon Sep 18, 2017 12:28 am

Robin, my actual solution for cover the dynamic problem of CMOS cameras during flat-field acquisition consist to capture two flat-field sequences: one for the red/green spectral domain (exposure time = t), and one for the blue region (exposure time = 10 x t). The spectra are processed by using the two flat-field separately (with echelle spectrograph, I separate red/green orders, and blue orders). See here:

http://www.spectro-aras.com/forum/viewtopic.php?f=8&t=1850

Christian Buil
Christian Buil
 
Posts: 1022
Joined: Mon Sep 26, 2011 6:59 pm

Re: CMOS vs CCD for spectroscopy applications

Postby Robin Leadbeater » Tue Sep 19, 2017 10:40 am

Thanks Christian,

I missed that. I suppose a similar technique could be used to make a flat for the ALPY etc, cutting and pasting the high and low regions together with a suitable scaling factor. It reminds me of a technique I experimented with years ago to increase dynamic range in 8 bit webcam images which I had completely forgotten about !
http://www.threehillsobservatory.co.uk/ ... m_8bit.htm

Cheers
Robin
LHIRES III #29 ATIK314 ALPY 600/200 ATIK428 Star Analyser 100/200 C11 EQ6
http://www.threehillsobservatory.co.uk
Robin Leadbeater
 
Posts: 905
Joined: Mon Sep 26, 2011 4:41 pm

Previous

Return to Spectroscopes and Softwares

Who is online

Users browsing this forum: No registered users and 1 guest

x