Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: texshooter on June 17, 2012, 03:09:05 am
-
i just saw a noise/iso comparison chart for the 5D MarkIi that shows ISO 640 having less noise than ISO 100. Is this a joke?
-
Yes it was.
-
there is less noise at 640 than 400 though.
-
you can see the chart here. jump to minute 9
http://m.youtube.com/index?desktop_uri=%2F&gl=US#/watch?v=zZRFlKdsHDg (http://m.youtube.com/index?desktop_uri=%2F&gl=US#/watch?v=zZRFlKdsHDg)
someone told me these charts are meaningless because the camera was tested with the lens cap on. beware what you read.
-
someone told me these charts are meaningless because the camera was tested with the lens cap on. beware what you read.
Beware of what someone told you. Dark noise measurement is one proper way to measure sensor response.
I recommend this source:
http://home.comcast.net/~NikonD70/
-
One of the things noticed using the 5D2 during the Great Zacuto Shootouts is that the in between ISO settings render cleaner images. I don't remember the particulars, but and of the multiples on the 5D2, 160, 320, 640, 1280, perhaps higher, gave cleaner vid than the normal ISO progression.
I don't recall if this was for all or just the 5D2. Worth checking their website for the comparison and then testing on your own.
-
I believe the explanation centers around the difference between shot noise and read noise. While one can be better at 640 than 400 (not sure about 100 though), the other type of noise more than compensates, thus the end result has more total noise. If I am not mistaken, the lens-cap-on test measures just one type of noise.
-
Bill Claffs charts explain it well
http://home.comcast.net/~nikond70/Charts/RN_ADU.htm
If you ettr use 100, 200 400 etc for lowest noise if you expose for the scene then the 160 320 etc would be lower noise but not as low as ettr'ing
Marc
-
You never want to PUSH the ISO. And be warned that PULLING the ISO will reduce the dynamic range by 1/3 stop. It is always better to pull than to push. And pulling is better than NATIVE when shooting low dynamic range scenes at high ISOs.
http://shootintheshot.joshsilfen.com/2010/05/13/canon-hd-dslr-native-iso/ (http://shootintheshot.joshsilfen.com/2010/05/13/canon-hd-dslr-native-iso/)
Also, this study shows that ISO 160 is no more noisy than ISO 100 for the 5DMarkII. And it is a myth that ISO 640 is less noisy than ISO 400.
http://www.terragalleria.com/blog/2011/03/22/best-iso-for-low-noise-on-canon-5d-mk2/ (http://www.terragalleria.com/blog/2011/03/22/best-iso-for-low-noise-on-canon-5d-mk2/)
-
Don’t know (or really care) about a chart. But with an image, depending on the camera system and how one exposes (ETTR), a ISO 800 capture shows less noise than an ISO 100 capture:
http://digitaldog.net/files/100vs800iso.jpg
-
Don’t know (or really care) about a chart. But with an image, depending on the camera system and how one exposes (ETTR), a ISO 800 capture shows less noise than an ISO 100 capture:
http://digitaldog.net/files/100vs800iso.jpg
Andrew: Can we assume you used flash as the lighting source in order to keep the exposure settings constant even though there is a 3 stop difference in ISO settings or did you change your raw processing settings?
-
Andrew: Can we assume you used flash as the lighting source in order to keep the exposure settings constant even though there is a 3 stop difference in ISO settings or did you change your raw processing settings?
Nope, no flash. Everything set on manual. Just alter the ISO setting with original aperture and shutter the same.
-
Nope, no flash. Everything set on manual. Just alter the ISO setting with original aperture and shutter the same.
So all you did was change the Exposure setting in Lightroom?
-
So all you did was change the Exposure setting in Lightroom?
Yes (I normalized the exposure as one would do with ETTR).
-
... depending on the camera system and how one exposes (ETTR), a ISO 800 capture shows less noise than an ISO 100 capture:
http://digitaldog.net/files/100vs800iso.jpg
If increasing ISO did not improve noise, in a light-starved situation, it would be a worthless feature. You have successfully demonstrated that the ISO setting does exactly what it is supposed to do. Of course if you had provided 8x as much light to the ISO 100 test, that image would then be the one with lower noise.
-
If increasing ISO did not improve noise, in a light-starved situation, it would be a worthless feature.
True although in this example it wasn’t a light starved situation. The same shutter and aperture was used for both captures.
You have successfully demonstrated that the ISO setting does exactly what it is supposed to do. Of course if you had provided 8x as much light to the ISO 100 test, that image would then be the one with lower noise.
I only adjusted the ISO setting. But yes, had the 100 ISO image had those two stops more photons and had been equality normalized, it would look better than the ISO 800 image.
-
I only adjusted the ISO setting. But yes, had the 100 ISO image had those two stops more photons and had been equality normalized, it would look better than the ISO 800 image.
So let me get this straight. What does the in-camera histogram look like for your two shots? If the ISO 100 shot was exposed to the left and the ISO 800 was ETTR, then I can understand why the ISO 100 is noiser. The arguement I raised and the sources I referenced assume that each compared shot is normalized in camera, not in Photoshop (ie, exposed in the middle of the histogram in camera). If you compare apples to apples, ISO 100 will always have less noise than ISO 800.
-
So let me get this straight. What does the in-camera histogram look like for your two shots?
I have no idea. Didn’t look, don’t care. I’m capturing raw data, the in-camera histogram is immaterial in this case.
If the ISO 100 shot was exposed to the left and the ISO 800 was ETTR, then I can understand why the ISO 100 is noiser.
Yes, exactly! You got it. The ISO 100 was exposed to the left but based on what the incident meter (which hasn’t a clue about raw or ETTR) recommended.
But the important point here is the concept which some believe, that in all cases, an image captured at ISO 100 will always have a lower level of noise than one shot at ISO 800. That isn’t necessarily the case.
If you compare apples to apples, ISO 100 will always have less noise than ISO 800.
Agreed.
-
But the important point here is the concept which some believe, that in all cases, an image captured at ISO 100 will always have a lower level of noise than one shot at ISO 800. That isn’t necessarily the case.
This must be a misunderstanding. Anyone who believes that ISO 100 is always less noisy than ISO 800, would have to believe that camera makers are maliciously supplying an adjustment knob which you can use to make your pictures worse. The real situation is quite simple: when you have enough light to expose properly, ISO 100 will require 8x (three stops) more light than ISO 800 and will be less noisy. Conversely, when you do not have enough light, setting ISO 800 will reduce noise, compared to simply underexposing ISO 100. This is precisely the case in which you would use ISO 800, as a tool to reduce noise.
-
Anyone who believes that ISO 100 is always less noisy than ISO 800, would have to believe that camera makers are maliciously supplying an adjustment knob which you can use to make your pictures worse.
I don’t know what they believe but I hear this all the time.
-
True although in this example it wasn’t a light starved situation. The same shutter and aperture was used for both captures.
I only adjusted the ISO setting. But yes, had the 100 ISO image had those two stops more photons and had been equality normalized, it would look better than the ISO 800 image.
And this is the point. ISO 800 has less read noise in photon equivalents than ISO 100 on many cameras. The only disadvantage of ISO 800, then, is that it clips 3 stops earlier than ISO 100. In your example, you had the headroom for the particular exposure that the image didn't clip at the higher ISO.
-
i just saw a noise/iso comparison chart for the 5D MarkIi that shows ISO 640 having less noise than ISO 100. Is this a joke?
Be careful of what units are being used in the measurement. The relevant thing is how much noise there is relative to a given light signal being captured (ie the amount of photometric exposure). An image exposed for middle gray at ISO 640 captures 6.4 times less light than one exposed for middle gray at ISO 100. So one has to do a careful calculation of the noise (both the photon noise and the electronic read noise, and I suspect your comparison chart included only the latter) to determine which ISO has less noise. As a rule of thumb, when exposing according to the meter (which takes into account the ISO setting) generally the lower ISO wins on noise, because the factor of overriding importance is the amount of light captured, and the lower ISO setting allows more light to be captured without clipping in the raw data (assuming that opening up the exposure doesn't cause motion blur or too shallow DOF).
-
What I hear all the time is that ISO 160, 320, 640 is better than ISO 200, 400, 800. It turns out that is only sometimes true. Pulling the ISO from 800 to 640 will reduce noise, but could clip your highlights due to loss of dynamic range. It depends on what's more important to you: noise or clipping.
-
The example of Andrews isn't surprising nor should it be. For the ISO 100 image to improve noise, exposure needed to increase so the histograms/data collected would have been equivalent. But his test does show that on a dSLR, increasing ISO to get more information/expose further to the right can be a benefit if you can't increase exposure with aperture/shutter speed, as long as the reduced dynamic range for the higher ISO isn't a problem.
-
This seems to be one those things one can test to one's own satisfaction to see the capabilities of the specific camera model you use, any in camera settings you prefer (like ADL for Nikons) that have an effect on a raw file, and your choice of raw processing program and settings.
-
This seems to be one those things one can test to one's own satisfaction to see the capabilities of the specific camera model you use, any in camera settings you prefer (like ADL for Nikons) that have an effect on a raw file, and your choice of raw processing program and settings.
Exactly so! I remember testing this issue years ago with my Canon 20D. A full exposure at ISO 800 produces noticeably less noise in the shadows, lower midtones and upper midtones than the same exposure at ISO 100 (which is 3 stops underexposed). This is old hat.
However, this situation does not apply to all sensors. Recent Nikon sensors have such superb dynamic range at base ISO, the noise in absolute terms is about the same at all ISOs. Not exactly the same. There may be some marginal benefit in using a high ISO in preference to underexposing at a lower ISO, but the advantages of avoiding any risk of overexposure, even of specral highlights, can be interesting. Sometimes one can even see the filament in a switched-on light bulb.
As Emil Martinec has pointed out, we should distinguish between absolute noise and relative noise in relation to the signal srength. With Canon DSLRs the absolute noise is less at higher ISOs, at a given exposure.