ISO/exp time verses noise DSLRs any info.

[Garyh]

Hi everyone...
Just thought I would start a thread about noise in DSLRs as regards to random noise in long exposures and a way to calculate which iso setting to length of exposure would be most suitiable and at what temperature?

Example: say its 20 deg C , would it be better to have a 4 min exposure at 1600 iso or a 8 min at 800 iso. Which would show less random noise/temp considering all the other variables are the same..(same model camera..etc)
compared to same settings at 30 C, would shorter exposures on higher iso make any difference?
I have always wondered exspecially in the warmer weather which would be the better way to go? Any one done tests or have seen a web site with any info regarding this?
Would be interesting to see any differences between different model/brand cameras as well in regards to noise.
Cheers Gary...:thumbsup:

[TidaLpHasE]

:)Hi Gary, i am still learning myself, so i can only direct a couple of links that might help, plenty of info in this onehttp://www.imatest.com/docs/noise.html

[jase]

There are many variables to consider not just the ISO settings, exposure length and ambient temperature. Something as simple as the brightness of the object you are shoot can also play a integral part in the quantity of noise or the lack off. I do like the mathematical formulas provided in the link. Though I'm still crunching the numbers.

Slightly off topic, attached is the US S&T article from the November 2006 issue (which subsequently I received yesterday). The article talks about going deep with DSLRs with renowned astrophotographer Chuck Vaughn. http://astrophotography.aa6g.org/

As quoted from the article:
"I consider 3 hours to be my minimum duration per subject when imaging under dark skies. As with cooled CCD cameras, creating long cumulative exposures is best done with many shorter ones combined; due to the
greater sensitivity of digital sensors, they reach the skyfog limit far quicker than film. From my location, I’ve standardized my exposure length to 10 minutes per frame at ISO 800. Lower ISO values cause undesirable
posterization (visible steps between brightness levels) in the faintest areas of my images, and I don’t see much advantage to ISO 1600, with its decreased dynamic range. I save all my images in RAW format, to preserve the entire 12 bits of data produced by the camera. I also turn off the automatic noise reduction and instead record dark exposures (images of the same duration as the light frames, recorded with the camera’s lens
covered) at roughly the same ambient temperature so I can calibrate my images later. Of course, accurate guiding of the telescope is a must for any image longer than a few seconds, so I use an old SBIG ST-4 autoguider to ensure round stars every time."


Enjoy:thumbsup:

[jase]

Two items to pick up on in the article is the use of image processing to further reduce noise i.e splitting the R G B channels and using other progams such as noise ninja - http://www.picturecode.com/

Also a comment around the EOS 20Da no longer being manufactured - interesting news, so buy up now or get a hutech modified rig.

[Merlin66]

I'd be interested in comments re max exposure for each frame. In the ol' days I was limited to about 20-25min with Hyper film and that was about that.
With the digital addition of shorter exposures the inference is that a short exposure still has enough info to be added together to give the "long exposure outcome" ie if the absolute faintest star image recorded in say a 1 min exposure was 16 mag and you add ten images... where does the 17 mag star image come from??? I'd think there must be a minimum "short" exposure to get some digital data to be added together???????

[Garyh]

Thanks for the link Trev, rather technical but interesting...:thumbsup:.
Yeah read that article Jase in the Aust S&T , I usually do a hour of data if I can, I should do more... 3 hours is quiet some data set and I hear that the 20da has much less noise than my 300d. maybe I need 4 hours!!
But would be interesting to see if 40 frames at iso 1600 or 20 frames at 800 iso would give in ways of noise comparision vers ambient temp. I know that the colder weather makes a real difference but more interested for the warmer summer season? I usually image the Ds stuff so noise in the image is always a problem when you push the processing..I use neatimage but I rather not use it if possible..
Cheers Gary

[Garyh]

Another interesting question Merlin66...I suppose that is also noise dependant (as well as other things like QE..etc..etc) on what is the faintest recordable star? The more you stack the less noise so the recorded photons can be more easily identified..? How many to stack before the benefits become to small to see...?
I have no idea on the faintest Mag my gear would pick up but my newt 6" f/5.6 I can go 10 min iso 1600 before skyfog hampers the image...
Be interesting what others have to say about your question merlin..
Cheers

[rogerg]

In your example you won't get the mag 17 star if you're adding the two 1 minute exposures together using software (stack using Add algorithm). That's where you need to do a single longer exposure. Of course the objects that are visible in the 1 minute exposure will be doubled in brightness when you add 2 images of the same. You're effectively shifting the exposure brightness up but you're not filling the gap left at the faint end with anything.

For my Hubble Deep Field South image I ended up doing 40 minute exposures. Adding 4 10 minute exposures would not produce the same result because some of the faint detail simply wasn't in the data of the 10 minute exposures. 20 mins vs 40 mins was showed very slight differences but still some.

Roger.

[Merlin66]

Think you right. So it really means that if the "detail" isn't in the image on the short exposure; it won't be there after added half a dozen others!!
Need to think in terms of what is the longest exposure you CAN get without excessive noise, sky glow etc then add them. ie 4 at 10 mins MUST be better than 10 at 4mins.

[Lee]

Intuitively I would suspect 4 @ 10min would be better for depth and 10 @ 4 min would be better for noise..... I guess you need some sort of compromise between the two......

[jase]

After performing some research on this topic over night, I have come to the conclusion that the duration of an exposure is not the primary concern when it comes to noise. There appears to be a direct correlation between ISO speed and noise. The higher the ISO, the greater the noise. While a higher ISO provides greater sensitivity in low light conditions, thus shortening the exposure duration, it is not generally better.
The resolution and noise reduction obtained at ISO's 800 and lower seem to be the best option for DSOs. Many CCD and CMOS image sensors are usually calibrated so that they give the best image quality (greatest S/N ratio) at its lowest possible ISO speed. The trade off to this is sensitivity so the exposure duration would need to be increased to compensate.

For dim DSOs, a longer exposure will still provide a greater signal to noise ratio, than stacking smaller exposures. (The opposite is true for bright objects such as planets). However, the exposure duration must be determined by finding your locations sky fog limit. If you consistently come close to reaching this limit (Say for example 8 minutes for urban locations, 10-15 minutes for rural locations) then you would be maximising the signal over noise. Over the limit, regardless of how many frames you stacked/combine there would continue to be too much noise with little improvement. There are no hard and fast rules for the exposure duration, one must determine these. Obviously the optical system focal ratio can play a significant role.

Jerry Lodriguess has a great article on determining your sky fog limit. The article is film centric, but can easily be adopted to DSLRs.
http://www.astropix.com/HTML/I_ASTROP/I05/I05.HTM