Call for Action: CMOS Photometry

Hello Ed,

My impression is there is some confusion in this subject. CMOS is just a general micro-electronics technology  used in most of electronics devices at time being. CCD is another much older electronics technology used in the 70's and that has remained only in image sensors years later. The principle of light sensing is the same for both, same physics, same basic principles. CMOS have been adopted in 100% of DSLR, DSLR are also CMOS cameras... And now there are science cameras that are CMOS equipped. The reason was their new functionalities, impossible with CCD, and fast improving performance for usual photography, the barrier being the huge capital investment needed to develop and industrialize such large silicon chips. 

What do you mean by CMOS camera ? being so specific that new practices are needed ? 

The new functions possible in CMOS seems to me not very of use for us: fast reading, selective reading... that's autofocus, face recognition, fast 4K video, any... not very interesting for us, in DSLR photometry we don't use it. 

What is new and interesting in CMOS ? First the massive parallel reading, it permits fast reading with low read noise and (important) without much power dissipation resulting in low temperature increase and low dark currents. Then the reading mode of the photodiode is very linear. The well depth is in fact high if you calculate it per micron-square of sensor: about three times larger than CCD !  But ok this just a difference of some parameters level, high or low, that doesn't change the principles.

What you call "CMOS cameras" are probably a specific class of products often based on ICX or IMX sensors (designed for survey, industrial, security... ) They first existed based on CCD-like sensors (ICX) and were used in astronomy for planetary imaging applying the "lucky imaging" technique. Now they are moving to CMOS, the fast reading of such sensor at low power (low temperature) enabling new lucky imaging performance. This is at the point the technique is now extended to bright deep sky objects.

What to do in photometry with such small cameras ? I don't think the lucky imaging (stack of many very short exposures) is of interest to us, the loss of resolution due to seeing being not critical (we defocus ! ). The sensor is usually very small, FOV not ideal to me. The good points are they are inexpensive, monochrome, have low NJ noise (not so critical to photometry), have fast reading (more time to observe), low power dissipation and in certain cases (not all ! ) very low dark currents. If it's the case the cooling is no more needed and make that solution VERY inexpensive. 

What is unclear with such cameras ? This is the fact they are optimized for lucky imaging: very short exposures with very high gain. The gain setting is the issue. In DSLR  the calibration factor (e-/ADU) is well fixed by the ISO  standard. We know it being about 1 e-/ADU at 200 ISO for an APS-C 14 bits camera. Those cameras have no such calibration factor reference. They have a large range of gain adjustemnt from said 0  to 100, units being sometime dB, sometime not. But the e-/ADU at "0" is unknown and very different between types and manufacturers. That results often in poor photometry settings with regards to SNR and dynamics/saturation. This is even made more critical by the fact the ADC is usually only 12 bits.

Maybe the solution is to make an assessment of those cameras, but I don't remember having seen this for DSLR or classical astro-cam in the AAVSO ? Maybe a problem ?

Clear Skies !

Roger 

I'll be happy to hear more about it. About stacking short exposures - as a person working with this technique, I see more potential than defocusing, but maybe it's just a personal need. Someone on another topic said that single exposures have potential to objects only fainter than 1.0-1.5 magnitude below the saturation limit. As a person looking for NEBs (near eclipsing binaries) for bright transiting candidates, stacked short exposures allow not only to avoid saturation of bright target, but also go deep. Just with 300mm f/2.8 lens, having 1.65"/px scale with IMX178 and R filter, also with expected depth at least the limit (attachment) of ~10.5 mag star, it can easily find a NEB down to ~16.5 mag using 2-minute stacks (12x10s), or 18-19 mag for 15-minute ones, still having main target not desaturated, just in case if that's the variability source. The fainter main target, the less we stack, the smaller delta value will be.

There are no tutorials (at least I haven't seen) how can we filter short exposures from seeing affected frames. The idea I have, is to compare ADU levels of main source to comp stars. If the main target is somewhere between 160000-190000 ADU on 95% of all frames, and the next one shows ~200000, while comparison stars stick to their average values, there's no need to keep that frame. Conditions change, so let's say we compare only 100 previous 3s frames and 100 more forwards. That's probably only for people interested in low budget high precision photometry, as such amount of frames can take A LOT of gigabytes on hard drive. And there's lack of software of that. AIJ fails to work with such large amount of frames and that's the only one program I know, where I can extract those data. Depending what is the resolution of frames, I have to split for every 1000 frames. But if there are 100K of 100ms frames, it can take ages, doing it manually.

Regards,

Gabriel

Hello Gabriel,

I think there is another confusion here, I am not speaking about single exposure, most of us stack images, either adding images or adding photometry results. You do 10 seconds exposure, this not what is done in lucky imaging where exposures can be as short as 1/100 of a second and the number of images 10000 or more ! 

I personally often use 5 groups of 5 images exposed 30 seconds each, adding the photometry results of each group, the 5 groups making one observing point. This allows to make a deep statistic analysis of various factors. It's also possible to stack the images of each group but with less statistical analysis. In the past I also used exposures of 10 seconds and 10 images groups from a smaller sensor. You are perfectly right, stacking (image or results) increases the dynamics if addition or floatting average is used (not using median). But what is important is the total at end: if you can get the same count of electrons with less images of longer exposures without saturating it's as good or  somewhat better if the NJ noise of the sensor is not very low. With sensors having 10 e- of NJ noise it's even mandatory.

I am not familiar with NEBs processing and I am not sure to understand what you are looking for, I suppose  this is a small drop of luminosity ? But if the objects are not eclipsing (near eclipsing) what makes it ? Ellipsoidal oscillation ?  In any case you would get the same with less images of longer exposure if the end electron count is the same. 

" how can we filter short exposures from seeing affected frames" 

What do you mean here with "seeing" ? is it the blur due to the atmosphere turbulence or the related scintillation ? Photometry is affected by the scintillation, not much the blur. The only way to eliminate the scintillation effect is to integrate the flux long enough. I have never heard about other technique except using a much larger aperture. This is depending the aperture of the scope and also the sky condition. In general the overall added flux (multiple or single exposure) shall result from at least one minute of exposure, two being better for small apertures and poor seeing (there are papers on the subject). 

The scintillation between target and comps is not correlated given the angle separation we usually have. Other problems like faint cloud/haze could be more or less  compensated by the target/comp flux ratio. 

Going to very short exposure (less than 10 s) is not a good idea I think. At a certain point the NJ noise of the sensor will get dominant and the SNR decreases. The only advantage of going short exposure in lucky imaging is to decrease the blur due to seeing and then improving the image sharpness. This has no effect on scintillation. I don't see in what it's useful in photometry, except loosing on SNR. 

Clear Skies !

Roger

Hey Roger:

I freely admit that I am not familiar with CMOS cameras as used for photometry. Obviously, you know much more than I. I am only a simple volunteer trying to teach CCD1 for AAVSO when asked and having done this several times I am now faced with students signing up for the course that have CMOS-chip based cameras. I suspect you know what I mean by CMOS cameras as used in astrophotography; ZWO and QHY cameras come to my mind. Being a good teacher (in my mind at least), I am reaching out to the AAVSO community with the suggestion that those who know something about these cameras relative to photometry might be lead to help develop some "best -practice" protocols that would allow me to understand enough to help these students. 

I am fully aware of my ignorance relative to these cameras and their use in photometry. Please let us focus on the goal of my original post; seeking some informed AAVSO members to help develop a document that outlines best practices for CMOS photometry similar to the existing protocols for DSLR and CCD.

I truth, I do not find any comments that are not directly relevant to this goal productive.

Ed

Hi Roger

There are now available CMOS cameras with quite large chips that are very suitable for long exposures and don't rely on the "lucky imaging" technique that you mentioned.

An example is ASI183MM. I have started using this camera admittedly for spectroscopy but it would also be very suitable for photometry.

There is a very nice review of it here http://www.astrosurf.com/buil/asi183mm/ descibing it's use as a scientific camera. The review is mostly in french but it translates very well with google translate.

Another comparison this time in english is here http://www.astrosurf.com/buil/CMOSvsCCD/index.html

These cameras have very low thermal noise. My 300sec dark frame ar -20degC has just 3 pixels that reaches 1000 . For comparison, my SBIGST10XME has 600 pixels that reach that level.

According to Christian's tests the cameras also have a very linear instrument response.

I also feel that a document explaining how to use these cameras for photometry would be very useful. I'm not up to doing this yet as mine is on my spectrograph but surely there is someone out there that is using one for photometry.

Terry

Terry,

Thanks for your support. The two links were very informative and bring up some of the questions that should be addressed by a committee of the informed. I have saved them to share if needed.

Ed

I would be very interested in this.  I just acquired a color Atik Horizon which is cmos based for an EAA setup, this camera is pretty amazing on my hyperstar rig.  I would be interested in figuring out using it or the monochrome version for photometry.  I'm just learning how to use it for EAA right now but it is so incredibly fast with a large FOV I can't help but think it would be great for variable star observing.  I'm just unsure about filters and such.  Is it more like CCD vs DSLR Filtering like Johnson V Filter vs Trig-G Green and the like.  I am new enough at Variable Star Observing that it gets confusing and the filters are costly to make a mistake on.

Hi Roger,

The sentence about single exposures was adresssed to another user from another topic. I have seen several opinions from astrophotographers about using long exposures with CMOS cameras, about that shorter ones (but more frames, which are stacked to the same integration time) give better results, which is opposite to CCDs. I have no idea why it would be opposite for photometry. Generally I prefer comparison results showing magnitude scatter than mathematic words about noise levels etc. This is what I expect from CHOICE. I do this myself by comparing results on ETD, as I said before.

Finding NEBs, which are near eclipsing binaries, is just detecting false positives of exoplanet candidates, which are a little futher (but still inside the aperture).

I don't have any experience about long exposure photometry due to mount/light pollution limits. Never went above 30s for single ones. But if I keep doing short exposures, for example for this 11.7 mag target with only 4" lens, having satisfying results, something is going on. http://var2.astro.cz/tresca/transit-detail.php?id=1525619181

Also about short exposures (<10s) and small SNR. I guess you meant about targets barely visible, just above noise? No, I meant about bright targets, like this 3s exposure as an example. It is actually just below the saturation limit, for this 10 mag object.

https://exofop.ipac.caltech.edu/k2/files/247098361/Photometry/EPIC_2470…

Regards,

Gabriel

Hello Terry, Ed,

Terry you are right, this camera is very interesting, having good general specifications and a gain setting enabling 0.5 to 2 e-/ADU, the right range for photometry. But not all such new cameras would be a good choice, some have much too high gain, are small, dark current... depending type and brand. 

But the true question is: in what the photometry process with CMOS cameras should be different than CCD ?

Apparently Christian used it the same way he used the CCD from Atik, there are only some difference in RON and other parameters, some better, some worst, but this doesn't change the spectroscopy process he applied. 

In what should the photometry process be different between CCD and such ASI183 MM ? I don't see it. By the way I don't see in what we need a new manual. Just use the CCD one ! 

The issue is the recent excitement about the lucky imaging new fashion, stacking many very short exposure images. Someone think it should offer new possibilities to photometry, that's not obvious to me ! (The post from Gabriel is typical of such thinking, to see my answer). I think we should not encourage this.

To summarize, I think there is no need for a new manual, the CCD one should apply. The need is to advise on camera selection and the way to set the e-/ADU calibration factor. On camera choice it's unclear to me if the AAVSO could  publish something, that has never been the case for CCD or DSLR. For the calibration factor determination there are number of documents on the subject or product datasheet. Christian Buil describes it in several of his posts. Maybe we could make our own, including some more practical details on the way to proceed.  

Clear Skies !

Roger

Hi All:

After reading all of your responses and other forum posts on CMOS cameras, I conclude that CMOS needs its own manual if AAVSO wanted to encourage CMOS photometry. Further, the best education for the beginning CMOS photometrist would be a  course lead by an experienced CMOS photometrist. 

Would I agree to teach CCDI with CMOS beginners included? Sure, I already have tried that. The student was great and I got him some help from an excellent CMOS photometrist. But he had to figure out a lot of things for himself. I would certainly try again, given that I have some information about these cameras (e-/ADU calibration factor, etc.). But would this serve the participant in a manner that is up to AAVSO standards?

Thank you all for your responses,

Ed

I'm glad that Ed started this thread!  I've done some preliminary tests with Gary Walker's new FLI Kepler sCMOS camera, along with the ZWO ASI178mm-cool and ASI183mm-cool CMOS cameras.  The former is a true science-grade camera with some flaws; the two latter cameras are consumer-grade items (the price can't be beat!).  I have a lot more work to do on the ZWO cameras, but they look very promising.

The current set of commercial-grade CMOS cameras do have several defects that make processing their images, and observing, different from CCDs, and so I second the motion to consider a separate CHOICE course for their use.  Christian Buil's ASI183 article highlights some of these differences:  amplifier glow, 12- or 14-bit ADC, funny dark current images, tiny pixels, etc.  At the same time, these CMOS cameras have multiple advantages:  low cost, nearly perfect cosmetics over many MPix, high QE, fast readout.  The fast readout part is what interests me right now, as it means you can do either high cadence work, or take sub-second images to work brighter without saturation.  Understanding their limitations, and finding/writing software to take advantage of their unique capabilities, are my short-term goals.

Toward those ends, does anyone know of public-domain software that will do real-time stacking (not later off-line stacking) of ZWO camera images as they are taken?  I hate to save thousands of subframes if it can be avoided.  A plugin for MaximDL would be ideal, but I'm willing to try other approaches.

Arne

We could suggest a feature for SharpCap about recording multiple FOV areas at once. The camera shows the whole frame, we only click on stars of interest (100x100px is enough) and the program would save them all in sequence, in one file. This would reduce the amount data a lot. For example, if we have only three reference stars in 3096x2080 file (IMX178 example) and we need only 100x100 areas around them, we capture only those four objects (including variable). Total area of interest is equal to 0.6%, making the file size of ~70 kb instead of 11 megabytes only (the final frame is 400x100 pixels). The rest 99.4% area/data size, which is background, is useless. The program shouldn't guide the star and all FOV areas must be static - that would make calibration frames usable.

Then I'm pretty sure all photometric programs would work well with such large amount of frames, even if there would be ten of thousands. Would be very useful for bright targets if comparison stars are really far.

Regards,

Gabriel

EDIT: Attaching a concept idea from a few months ago.

I think what Ed was looking for when he started this thread was some basic recipies for doing photometry with the CMOS cameras. The new CMOS cameras should be useable for basic photometry which is the goal of the CCD CHOICE course.

I am working to develop instructions for this basic use case as I have an ASI178 which will soon be mounted on my scope.

The other comments in this thread focused on some of the extraordinary capabilities of these cameras. High cadence, near video frame rates and stacking strategies to match these capabilities; these use cases should be best addressed in another forum thread. Exciting stuff but a distraction to the goal of supporting the basic CHOICE course material.

If anyone has experience with calibrating and collecting filter photometry with CMOS cameras, please contribute here!

George

Dear Ed, dear Fellows,

    I can list three papers dealing with observations of Delta Scorpii using CMOS camera. It is SANYO CG9 commercial camera, and the second magnitude was well reachable. The problem of stability of magnitude to do reliable photometry is heavy. The camera has a fixed 0.5 s of integration, it should overcome most of scintillation, but it does not. I used a similar camera SANYO HD1010 to follow N Scorpii eclipse last June: same problems even if the aperture of the camera is larger and the limiting magnitude around 5-6. The corresponding data are on AAVSO database at SGQ code. Here the 3 papers on Del Sco.arXiv:1107.1107   arXiv:1109.5865 arXiv:1112.2356

In summary: naked eye was always more precise.. and I could not suggest using photo/videocameras for making photometry at didactic level... nor smartphone.. what do you think?

Greetings

Costantino

Tried to talk a seller into letting me test a CMOS for this purpose to no avail! I suspect that these will work fine once the differences and techniques are well studied. I would think the technique will be very similar especially if the numbers match up!

George, thanks for reminding all of the basic intent of the thread. You have defined the issue with clarity.  I look forward to seeing the results of the work you, Gary, Arne and others are doing to  bring CMOS cameras into our programs.

Ed

Hello All

I finally got a chance to reduce the first 13 nights of data on my new setup.  It uses the excellent sCMOS camera from FLI, called the Kepler K400.  These are all time series results from the excellent SRO site, but  I do have a collimation issue with the scope, as the fwhm are 8 arc seconds.  Attached is a screenshot from the LCG1,  and hi-lighted in blue are my observations.  The object is the symbiotic campaign object J1444107-074451, and the exposures are all 180 seconds.  Needless to say, I am very pleased with the results.

Gary

Gary's keynote address will be on CMOS, and I hope Arne and George (and perhaps others) will also present what they have learned.  I think there will be a lot interest and a lot of questions, but the time for questions after a presentation is often very limited.

Would it be possible to set aside a block of time at the meeting for questions and open discussion of CMOS?  This could be during a lunch break, in the evening after formal presentations, or even Sunday morning. 

I'm sure that this would happen informally at the meeting in any event, but I think it it would be good to set it up in advance so people can make their travel plans accordingly.

Phil

So now CMOS camera (12-bit ZWO ASI-183mm) is used on one telescope of BSM network.

Can anybody describe how it performs compared to classic CCD? Does it give good photometry?

Regards,

Max

I have monitored this forum for a few months, ever since I decided to buy a ZWO ASI1600MM-COOL CMOS astronomical camera, and filter wheel with V and B filters. Attached is a document with the results of testing done so far, limited to linearity and precision. Although I have acquired transformation coefficients, I have not yet tested the accuracy of transformed magnitude measurements. That's a project for the next clear nights.

To summarise (see the attached for detail):

1. The camera shows good linearity to near full well capacity (checked with regression analysis and residuals plots).

3. The camera allows high precision photometry of bright stars, to 10th magnitude, as shown by (a) time series photometry of the 8th magnitude delta Scuti star RS Gru and its check star, and (b) measurement of the non-transformed V magnitudes of 4 pairs of 10th magnitude Tycho-2 stars.

4. Precision drops off at 11th magnitude, based on testing of pairs of Landolt standard stars.

5. All testing was done on slightly defocussed images, at unity gain. I could not achieve good precision on sharply focussed images.

Note that, although this camera has 12 bit ADC, full well capacity is found to be 65,504 (see Figure 1 in the attached). That is because the camera software multiplies the native ADUs by 16 and outputs those values as the results.

Roy

Looks like the attachment did not load.

I'll try again.

Roy

Hi Roy,

Your data look good!  What was the seeing for your RS Gru nights (arcsec and in pixels), and what was the size of your measuring aperture?

Arne

Arne,

Re the measuring aperture - something wrong with the scale. The scale bar at the top of the image shows 96 pixels. If that's correct, the aperture would be way bigger than 32 pixels.

Need to sort this out, but won't be tonight (9:27pm here in Brisbane).

Roy

Arne,

The aperture for the RS Gru photometry was 64 pixels (diameter). As noted, the images were significantly defocussed. Even so, the highest pixel counts were just above the range of linearity, and just below saturation for RS Gru at the peak of the light curve. Counts for the comp and check were a little lower.

Roy

I've recently bought a simple ZWO ASI120MM planetary camera and tested it as a photometric device.

The camera has 12bit ADC. I used 2*2 binning (software binning) to get virtual 14bit precision (Arne Henden wrote in one of his posts that BSM_NH2, equipped by ZWO ASI-183mm (12-bit) uses 2*2 binning to achieve 14-bit precision).

Here are the results of a test run against XX Cyg. This is unfiltered CV photometry. Exposure=10s.

The first plot is for 10s exposures, the second -- binned (by groups of 4 points) measures with estimated uncertainty (standard deviation by 4 points). Looks not worse that DSLR photometry which I practice now (of course, with a filter applied the noise would be bigger). However, now I'm seriously thinking to buy a somewhat better camera (say, ASI183 cooled or ASI1600 cooled) for photometry.

I have not made linearity testing yet, I will try.

*** UPDATED *** Slightly better results with lower noise can be achieved using a smaller aperture (r=5px instead of 6px, at least Muniwin shows the noise minimum here. Measurements were done with AstroImageJ yet Muniwin was used to check the result). The third figure shows the light curve measured with the aperture radius 5px, the 4th one -- binned results (averaged by 4 points)

Chart    X24817BFV
Comp Name    106 [000-BJV-171]
Comp V    10.606
Check Name    118 [000-BJV-173]
Check V    11.757
Air Mass    1.1 - 1.3

Best regards,

Max