Very accurate Photometry with a DSLR

[Terry B]

An interesting article

http://astrobites.org/2015/08/05/pro...nsumer-camera/

or the original article
http://arxiv.org/abs/1506.03097

[rustigsmed (Russell)]

thanks Terry - weirdly today I found myself re-reading some chapters of Bruce Gary's exoplanet observing for amateurs before seeing this thread.

[Atmos (Colin)]

I have done some astrometry studies with my DSLR over the years looking at variable stars within globular clusters. Found that to work quite well to within 0.1 mag accuracy. More recently got a CCD to which I want to see whether it can out perform the DSLR

[rustigsmed (Russell)]

Originally Posted by Atmos

I have done some astrometry studies with my DSLR over the years looking at variable stars within globular clusters. Found that to work quite well to within 0.1 mag accuracy. More recently got a CCD to which I want to see whether it can out perform the DSLR

i'm sure it probably will.
the interesting thing about this experiment was that it was for the purpose of doing widefield sky surveys with just a camera lens and it was able to detect hot Jupiter sized transits. the paper mentions that dslr users like to use a defocus method but this isn't that appropriate for survey work.

there is a whole bunch of interesting points on DSLR's themselves eg part of the sensor should be cropped out, their linear nature. also interesting they used a low ISO.

[rally]

Interesting find

This technique caught my eye.
Looks like the mount has a "drizzle" ability - they call it "PSF broadening" which occurs during a single sub exposure. Enlarges PSF by a factor of 1.5

They say that it ". . . improves photometric precision substantially"

I had trouble reconciling that the CCD had an almost perfect linear response in all 4 channels RGBG across the full dynamic range to saturation, but then they say that the antiblooming is built into the chip. So maybe the antiblooming is really antisaturation rather than a bleed off as saturation is approached ?

Does the Canon include CCD temperature into the metadata of each sub ?
Thats a great option for having darks in a DSLR
H - do you know ?

[rustigsmed (Russell)]

Originally Posted by rally

Interesting find

This technique caught my eye.
Looks like the mount has a "drizzle" ability - they call it "PSF broadening" which occurs during a single sub exposure. Enlarges PSF by a factor of 1.5

They say that it ". . . improves photometric precision substantially"

I had trouble reconciling that the CCD had an almost perfect linear response in all 4 channels RGBG across the full dynamic range to saturation, but then they say that the antiblooming is built into the chip. So maybe the antiblooming is really antisaturation rather than a bleed off as saturation is approached ?

Does the Canon include CCD temperature into the metadata of each sub ?
Thats a great option for having darks in a DSLR
H - do you know ?

did you mean dslr or ccd for perfect linear response? I thought I read the ccd was good for a range then wasn't so linear.

the temp is recorded -Backyard EOS records the temp in the file name
eg here is a sub of mine
https://www.flickr.com/photos/803366...posted-public/

btw I have no photometry experience but eventually want to get into exoplanet detection.


also I would like to know what they mean by adding the channels together. I assume they stack each channel separate perhaps after using 'superpixel' type setting (as in DSS) for each channel.

[rally]

Russell,

Have a look at the paper
They were referring to the Canon DSLR response

They plotted all 4 G,R,G,B channels. (but only three lines because both greens were identical.
They used dcraw to extract the 4 channels from the raw file (ie the bayer matrix channels) where each green is treated separately.

CCDs that have antiblloming gates are usually non linear once the well's electron charge starts to reach a level that is near full and gets bled off.
So yes - they start off being linear then at a certain point then become non linear.

Sorry couldnt get your link to work of your file - is that CCD temperature or is that internal camera temperature ?

My understanding is :
Usually the A/D converters or whatever other circuitry that is built into the CCD chip - usually on one side or corner tend to start heating up and so the chip ends up with a side or a corner that gets hot and the excess heat results in stray electrons that add noise to the signal.
So my uncooled KAF8300 OSC chip suffered from purple "amp glow" that grew out form one corner and progressed further down one side and into the CCD the longer I used it on long exposures - sufficient to make my McNaught comet photos barely useable without some fancy footwork.

The problem is since the chip is not being actively cooled - the way in which this affects the image is very unpredictable at a pixel level - ie how fast is the heat being generated, what is the ambient temperature, how much residual heat was generated by the last image and any images before it etc - which can make getting accurate and reliable Darks hard to achieve and therefore darkframe subtraction is not always useful as it is with a chilled CCD where almost all the pixels are actively being cooled faster than they can heat up and so are able to be held at a constant temperature across all pixels, or alternatively any gradient is constant.

Rally

Originally Posted by rustigsmed

did you mean dslr or ccd for perfect linear response? I thought I read the ccd was good for a range then wasn't so linear.

the temp is recorded -Backyard EOS records the temp in the file name
eg here is a sub of mine
https://www.flickr.com/photos/803366...posted-public/

btw I have no photometry experience but eventually want to get into exoplanet detection.


also I would like to know what they mean by adding the channels together. I assume they stack each channel separate perhaps after using 'superpixel' type setting (as in DSS) for each channel.

[rustigsmed (Russell)]

Originally Posted by rally

Russell,

Have a look at the paper
They were referring to the Canon DSLR response

They plotted all 4 G,R,G,B channels. (but only three lines because both greens were identical.
They used dcraw to extract the 4 channels from the raw file (ie the bayer matrix channels) where each green is treated separately.

CCDs that have antiblloming gates are usually non linear once the well's electron charge starts to reach a level that is near full and gets bled off.
So yes - they start off being linear then at a certain point then become non linear.

Sorry couldnt get your link to work of your file - is that CCD temperature or is that internal camera temperature ?

My understanding is :
Usually the A/D converters or whatever other circuitry that is built into the CCD chip - usually on one side or corner tend to start heating up and so the chip ends up with a side or a corner that gets hot and the excess heat results in stray electrons that add noise to the signal.
So my uncooled KAF8300 OSC chip suffered from purple "amp glow" that grew out form one corner and progressed further down one side and into the CCD the longer I used it on long exposures - sufficient to make my McNaught comet photos barely useable without some fancy footwork.

The problem is since the chip is not being actively cooled - the way in which this affects the image is very unpredictable at a pixel level - ie how fast is the heat being generated, what is the ambient temperature, how much residual heat was generated by the last image and any images before it etc - which can make getting accurate and reliable Darks hard to achieve and therefore darkframe subtraction is not always useful as it is with a chilled CCD where almost all the pixels are actively being cooled faster than they can heat up and so are able to be held at a constant temperature across all pixels, or alternatively any gradient is constant.

Rally

Hi Rally,

Thanks for the post, I could see they are using dcraw to extract but i'm not sure that really explains what they are doing with the file and how the channels are reconstructed eg DSS allows the following: Bilinear, Adaptive Homogeneity-Directed Interpolation (AHD), Bayer Drizzle and 'Super Pixels'. Perhaps one of those methods is akin to their "PSF broadening".

I'm not too familiar with chip intricacies they do say on page 5

Due to anti-blooming features implemented on
the chip, charges are simply lost after saturation; they
do not overow to neighboring pixels, and thus can not

be measured to provide precise photometry.) But then go on to say avoid saturation in capture.

Could it be a software issue or a function of it being a CMOS chip rather than CCD?

For canon's the temperature reading is from the PCB board 2mm from the sensor recorded in the EXIF data. so it is possible to get a reasonable indication of what the temps are for darks although you would need quite a library but doable I guess it would just be a bit more time consuming. Something like this would be better http://www.centralds.net/cam/?p=7372

Cheers

Russ

[rally]

Russ,

I got the impression that they are drizzling by microshifting the mount whilst in the middle of subexposure in a circular fashion. But that is only reading between the lines, its not a fact !

My query with regard to the antoblooming and saturation commenst, is they put up a graph showing almost perfect linearity across almost the full dynamic of the CCD - just seemed like a contradictory set of information.
Unlike Kodak etc, Canon and Nikon etc dont really publish a lot of information about their own proprietory chips - so nobody knows absolutely what is going on at a chip level.
We, they included can only guess and experiment to see what happens under different circumstances.
I am sure Clarke Vision knows !

[rustigsmed (Russell)]

Originally Posted by rally

Russ,

I got the impression that they are drizzling by microshifting the mount whilst in the middle of subexposure in a circular fashion. But that is only reading between the lines, its not a fact !

My query with regard to the antoblooming and saturation commenst, is they put up a graph showing almost perfect linearity across almost the full dynamic of the CCD - just seemed like a contradictory set of information.
Unlike Kodak etc, Canon and Nikon etc dont really publish a lot of information about their own proprietory chips - so nobody knows absolutely what is going on at a chip level.
We, they included can only guess and experiment to see what happens under different circumstances.
I am sure Clarke Vision knows !

Ahhh yes, i think you are right actually on the psf broadening (real world drizzle!). I'm confusing myself, so i guess we really don't know how they reconstructed the channels.
amazing that these chips can still keep their secrets when being tested scientifically.