Article about 5Dmark2

[Scott Martin]

Since the 5Dmark2 has a few exceptional features that set it apart from the rest for night photography, I thought it would be appropriate to write a short article on it. If you are interested, it can be read at:

http://www.on-sight.com/2009/05/29/5dmark2...-photography-2/

[digitaldog]

Since the 5Dmark2 has a few exceptional features that set it apart from the rest for night photography, I thought it would be appropriate to write a short article on it. If you are interested, it can be read at:

http://www.on-sight.com/2009/05/29/5dmark2...-photography-2/

My initial comment, based on something from Eric Chan of Adobe is that unless you're shooting JPEG, you'd be better off setting the camera to 3200 ISO and upping exposure in ACR/Lightroom resulting in less noise. In a nutshell, settings above ISO 3200 on the 5D MII are only useful for shooting JPEGs. Might want to test that.

[Scott Martin]

Above 3200 ISO the noise does get nasty so I can see Eric's point. But in this article no one is suggesting that anyone shoot final images at anything above native ISO. The whole point about shooting throw away test frames at 6400 is to determine the optimal exposure before committing to a long final exposure. It's convenient that the exposure time in seconds at ISO 6400 is the same as the exposure time in minutes at 100 ISO. Shooting at native ISO produces far cleaner images than anything higher.

[jeffball]

Hello Scott,
Thanks so much for the article.  I really enjoyed your website and will frequent it again.  I just recently acquired the 5DMKII as a move from the original 5D.  I utilize a modified 450D for astrophotography and then use the 5D for night landscape exposures.  Your article has inspired me to explore the 5dMKII possibilities next time I am at a dark sky site.  I am experimenting with no dark frame subtraction on my astophotos with the 450D and sticking with LR2 and PS3/4 for processing my astrophotos.  We stack multiple exposures that help to improve the signal to noise ratio.  I have found that my processing through LR2 produces better results than calibrating the image through dark frame and flat field subtraction that specific astronomical imaging software provides.  I am just not confident in the RAW conversion qualities inherent in these astronomical software packages.  I have posted some of the astronomical work at my blog earthandskyphoto.blogspot.com.  Thanks again for the article and I look forward to more information from you on this and other subjects.  

Jeff Ball

Since the 5Dmark2 has a few exceptional features that set it apart from the rest for night photography, I thought it would be appropriate to write a short article on it. If you are interested, it can be read at:

http://www.on-sight.com/2009/05/29/5dmark2...-photography-2/

[Scott Martin]

I am experimenting with no dark frame subtraction on my astophotos with the 450D and sticking with LR2 and PS3/4 for processing my astrophotos.

That's the way to go for long legnth image stacking. FWIW, Shawn Peterson captured one of my favorite image-stacked images last year at a workshop in Bodie: You can see it at http://www.on-sight.com/newsletter/29.html and http://www.shotnthedark.blogspot.com/

Great looking work on your site! Thanks for your post!

[ejmartin]

Above 3200 ISO the noise does get nasty so I can see Eric's point. But in this article no one is suggesting that anyone shoot final images at anything above native ISO. The whole point about shooting throw away test frames at 6400 is to determine the optimal exposure before committing to a long final exposure. It's convenient that the exposure time in seconds at ISO 6400 is the same as the exposure time in minutes at 100 ISO. Shooting at native ISO produces far cleaner images than anything higher.

Actually, if you are doing long exposures you are much better off using a higher ISO gain, and stacking multiple shorter exposures.   Read noise in photoelectrons drops nearly a stop for each doubling of ISO from ISO 100-400 on Canons, and substantially from 400 to 800.  Stacking N images results in read noise that is sqrt[N] times the read noise of an individual image (and substantially improves nonfixed pattern noise).  So for instance four one minute shots at ISO 400 stacked will have about half the read noise of one shot of four minutes at ISO 100.  The shadows will be much cleaner in the stacked image.  Photon capture will be the same, but camera electronic noises will be lessened.

On the other hand, at ISO 1600 and beyond read noise on Canons levels off, and so the sqrt[N] penalty for stacking N images is not compensated by any gain in reduced read noise per image at higher ISO.  There is a sweet spot, and I believe most good astrophotographers are shooting at around ISO 400 or 800.  Intermediate ISO's are a waste of time on Canons (zero benefit in terms of read noise).

If the article is correct, it seems that the buffering for LENR is rather idiotic.  The whole point is to take a dark frame image to subtract thermal noise.  If you've got four images in the buffer, why take a separate dark frame for each rather than a single dark frame and subtract it from all of them?  It would drop the downtime substantially!

[Scott Martin]

Actually, if you are doing long exposures you are much better off using a higher ISO gain, and stacking multiple shorter exposures.

For astrophotography, yes. For exposures (let's say up to ~30 minutes) where light painting is involved it's out of the question. Some people will always prefer the simplicity of having the shot in a single RAW frame/file. Different techniques for different types of photography.

If the article is correct, it seems that the buffering for LENR is rather idiotic.  The whole point is to take a dark frame image to subtract thermal noise.  If you've got four images in the buffer, why take a separate dark frame for each rather than a single dark frame and subtract it from all of them?  It would drop the downtime substantially!

A good suggestion for the camera manufacturers! This long exposure stuff is pretty far off their radar but if they hear from more of us perhaps it will get more sophisticated....

[Panopeeper]

The whole point about shooting throw away test frames at 6400 is to determine the optimal exposure before committing to a long final exposure

This is really a smart way of "metering".

Shooting at native ISO produces far cleaner images than anything higher

This is not so. Shooting with the 5D2 up to ISO 1600 makes the image cleaner when using the same exposure as with a lower ISO.

In other words: if the shot does not come close to clipping, it pays to up the ISO. As the long exposure itself tends to cause noise, it might be interesting to evaluate, from which shutter time it pays to up the ISO instead of making the exposure last longer. I am pretty sure that for example using ISO 100 is inferior to ISO 200 with somewhat shorter shutter (not necessarily half the shutter time, up to clipping).

[Scott Martin]

This is really a smart way of "metering".

There is more to it than that of course. As with daylight photography shooting for a right biased histogram is also optimal for night photography.

In other words: if the shot does not come close to clipping, it pays to up the ISO. As the long exposure itself tends to cause noise, it might be interesting to evaluate, from which shutter time it pays to up the ISO instead of making the exposure last longer. I am pretty sure that for example using ISO 100 is inferior to ISO 200 with somewhat shorter shutter (not necessarily half the shutter time, up to clipping).

I've done a lot of testing. Keeping the ISO at native and increasing the exposure time is definately the way to go. The noise levels between 1 and 30 minute exposures are indistinguishable on the 5Dmark2. Longer exposures are often more desirable for night photography anyway because they allow time for light painting and for stars to trail.

[Panopeeper]

The noise levels between 1 and 30 minute exposures are indistinguishable on the 5Dmark2

Do you mind sharing some of such raw files? Some with and some w/o long exposure NR are particularly interesting.

[Guillermo Luijk]

I never tried long exposure shots, but when dark frame substraction is used to eliminate/minimise hot/dead/pattern noisy pixels, I wonder if stacking programs really take advantage of those cameras with a shifted black point, i.e. a non-zero RAW level corresponding to the expected absence of light (I have seen that on Canon and Olympus RAW files so far). A dark frame built from averaging several dark frame shots should statistically be more robust and precise than performing the same procedure on a camera with 0 clipped deep shadows (like a Nikon).

Does anyone know if these programs work at a RAW level to enjoy this advantage?

BR

[feppe]

I never tried long exposure shots, but when dark frame substraction is used to eliminate/minimise hot/dead/pattern noisy pixels, I wonder if stacking programs really take advantage of those cameras with a shifted black point, i.e. a non-zero RAW level corresponding to the expected absence of light (I have seen that on Canon and Olympus RAW files so far). A dark frame built from averaging several dark frame shots should statistically be more robust and precise than performing the same procedure on a camera with 0 clipped deep shadows (like a Nikon).

Does anyone know if these programs work at a RAW level to enjoy this advantage?

BR

I use TufusePro for stacking, which reads .cr2 files (and possibly others). The limited documentation doesn't specify whether it uses the data as you described.

[Scott Martin]

Due to limited buffering and a desire to have subsequent shots taken with minimal delay, those doing image stacking leave long exposure noise reduction disabled and deal with the noise later. Its a lot of work but worth the wait!