PDA

View Full Version : MkIII V D200 ISO640 crops


Steve Saunders
July 7th, 2007, 11:08 AM
Quick test this. Similar shooting conditions and at similar distances from the subjects.
Canon 300 f2.8 IS USM L on the 1DMkIII and Nikon 300 f2.8 VR AFS on the D200. Jpegs as they came from the cameras, no NR or sharpening was applied either in-camera or afterwards and I just took a 500x500 pixel from each image and saved at quality setting 11 in Photoshop.
The D200 is generally accepted as being slightly less noisy than the D2X, ie it's currently Nikon's best performer at higher ISO. The Nikon sample holds up well, but clearly not as well as the Canon;

http://www.goldwingfacts.com//images/nonWingimages/Mk3samples/croptest.jpg

Jack_Flesher
July 7th, 2007, 04:06 PM
Hi Steve:

Nice comparison post! FWIW, it is a pretty well accepted reality that Canon produces its cleanest files at whole ISO's -- IOW, 200, 400, 800, 1600. I've not personally tested it on the 1D3, but for some unknown reason, and testing confirmed this on my 1D2, 1Ds2 and 5D, using the intermediate ISO's INCREASES noise beyond the next highest full integer value ISO. You would need to confirm it for the 1D3, but I suspect it will hold and in your above example, ISO 800 should be cleaner than the 640 shots(!) I now only use my Canon's at the full ISO settings.

Cheers,

Steve Saunders
July 7th, 2007, 04:28 PM
I didn't know that Jack, thanks for the tip.

Asher Kelman
July 7th, 2007, 05:45 PM
Great point Jack! This was discussed here extensively in relation I believe to the Rebel XTi and the 5D, I believe, but I didn't realize it applied all the way through the flagship 1D series.

Asher

Paul Bestwick
July 7th, 2007, 05:50 PM
Good information Jack, I was not aware of that. Is iso 200 better than100 ?

Thanks,

PB

Asher Kelman
July 7th, 2007, 06:34 PM
Great point Jack! This was discussed here extensively in relation I believe to the Rebel XTi and the 5D, I believe, but I didn't realize it applied all the way through the flagship 1D series.

Asher

Actually it was the Canon 30D and the 5D that we discussed at length with explanations by John Sheehy based on the combing of histograms at intermediate ISOs well worth reading, here (http://www.openphotographyforums.com/forums/showthread.php?t=546&highlight=intermediate).

Maybe we should have a new look at this with the 1DIII? I'll start a new thread, LOL! That will be fun!

Asher

John Sheehy
July 7th, 2007, 09:13 PM
Hi Steve:

Nice comparison post! FWIW, it is a pretty well accepted reality that Canon produces its cleanest files at whole ISO's -- IOW, 200, 400, 800, 1600. I've not personally tested it on the 1D3, but for some unknown reason, and testing confirmed this on my 1D2, 1Ds2 and 5D, using the intermediate ISO's INCREASES noise beyond the next highest full integer value ISO. You would need to confirm it for the 1D3, but I suspect it will hold and in your above example, ISO 800 should be cleaner than the 640 shots(!) I now only use my Canon's at the full ISO settings.


That's because the read noise in Canon DSLRs does not scale with ISO, for its real gain-based ISOs. It is only 2.27x as high (20D/30D) to 4.4x times as high (400D) at ISO 1600 than it is at ISO 100, and the smallest differences are in the 100 - 400 range. That means that at the low and mid end of the ISO scale, the 1/3-stop ISOs have 1.25x and 1.6x the read noise as the base 1-stop ISO from which they are pushed. A 5D, for example, might have a read noise of about 2.0 ADU at ISO 100, and 2.1 ADU at ISO 200. ISO 160 is created by under-exposing 100 by 2/3 stop, and amplifying the original signal after sensor readout by 1.6x, for a read noise of 3.2, higher than the 2.1 for ISO 200. Most of the Nikons have read noise that pretty much scales with ISO, so it doesn't matter how they accomplish the ISOs; they are nearly linear with respect to read noise. ISO 1600 generally has about 15x the read noise of ISO 100 on most of the older Nikons. The D40, however is more "Canon-like", so Nikon may have figured something out, which you might see in future cameras.

The 160/320/640/1250 group are not "bad" in all Canons, though; in the 30D, they are pulled from the next higher ISO, and have lower read noises. On my 30D, 160 has the most DR of any ISO, and is better than either 100 or 200 on my 20D, even though they use the same sensor and readout circuitry.

John Sheehy
July 7th, 2007, 09:19 PM
That's because the read noise in Canon DSLRs does not scale with ISO, for its real gain-based ISOs.

The other main noise is shot noise, which is a property of light, and not a creation of the camera. Shot noise on any given camera depends only on the light intensity of the sensor (as determined by Tv, Av, and the subject's luminance) , and has nothing to do with the camera's electronics, including how it achieves ISOs.

Jack_Flesher
July 8th, 2007, 07:04 AM
Good information Jack, I was not aware of that. Is iso 200 better than100 ?

Thanks,

PB

You actually need to test this on your individual camera. It isn't all a noise difference either, but one of finding best total DR with lowest noise -- finding your camera's true base ISO. I found ISO 160 superior to ISO 100 on my 1D2 for DR with no noise, but my 1Ds2 liked 125. My 1's also had about the same noise doubling their base ISO they did at 400 with no difference in DR... So I used only the base ISO and 200 or 400 on the 1Ds2 and added 800 for the 1D2 when it was needed. My 5D defintely has more DR at 200 than 100, yet the noise is indistinguishable up to ISO 400 -- and 200 seems to be the true base ISO. So with it, I'll use 200 or 400 a lot, and 800 or 1600 when needed. (And I carry a 6-stop ND filter for when slower shutter speeds are needed, and still use ISO 200 or 400 with it.) As a general rule, pushing up from base ISO adds noise while pulling down reduces DR.

Cheers,

Jack Joseph Jr
July 8th, 2007, 07:13 AM
Gentlemen. Reality check here. Let me politely play the skeptic here. I'm not so sure that all of this mega-pixel peeping with charts and graphs and scientific measurement applications really relates to photography in any real sense. It strikes me a pseudo science when non-engineers (I'm not aware of formal training here) start throwing around numbers, charts and graphs obtained outside of the lab.

It seems to me that if a trained photographer cannot see the difference between the IQ of two cameras then any measurable difference that is there just does not matter. Sometimes these discussions here and at dprevew go on for days. Even though I shoot at a lot of low light venues would my photography and income be significantly improved by replacing my 1D2N with a Mark III any more than it was by replacing my 1D2 with the N?

Are the pictures in Sports Illustrated going to get "better" than when they were shooting the old-fashioned 1D2Ns and the 1D2s before that? For IQ purposes does it really matter if an example downsized and posted to the Web was shot with a Mark III or a 10D?

I just don't see the point of all of this. Most likely the actual engineers at Canon and Nikon have already done all of this stuff. Other than something to sit at a bar and BS over I'm having trouble with appreciated the obvious hours of work that some of the posters have spent on this subject.

Steve Saunders
July 8th, 2007, 07:55 AM
It seems to me that if a trained photographer cannot see the difference between the IQ of two cameras then any measurable difference that is there just does not matter.


That's why some of us post sample images instead Jack. Some of us prefer to see "real world" images instead of just (or in addition to) graphs and charts. I know many people like to study graphs and numbers, but they mean nothing to me personally without some images from the camera. That way we have the best of both worlds and both the data and images help us more.

Ray West
July 8th, 2007, 07:55 AM
Hi Jack,

You are right in what you say, but I remember when magazines were not printing colour photos, until engineers, formally trained or otherwise, pushed the boundaries of development, made things, got them broken, and progress? was made.

In most things, it usually helps if you know a bit more than you need to know, it sort of gives a factor of safety. If you push your equipment to its limits, then it helps to know where it is likely to fail. In the analogue world, things tend to fail gracefully. Not so in digital.

Personally, I don't see the point in football, real football, or whatever it is that is played in the USA, but if folk want to be involved in either or some other version, then good for them. I'm not sure if the world is a better place because of pixel peeping or football.

Best wishes,

Ray

Ben Rubinstein
July 9th, 2007, 01:34 AM
That is weird, my 5D's are considerably less noisy at iso 1000 than 1250 or 1600 and iso 500 is just a bit worse than 400, certainly noticeably better than iso 800. This what I'm seeing with day to day use of these iso's under work conditions, am I missing something here? Could it be something to do with the jpg's? I only shoot RAW and have found ACR excellent at totally killing all colour noise (colour noise reduction=10) while leaving the more natural luminance noise intact.

Steve Saunders
July 9th, 2007, 01:51 AM
I think this goes to show how different cameras even from the same manufacturer address noise differently. There seems to be no magic cure and the engineers tweak and refine as each new model is being worked on. Yesterday evening I used ISO800 on the MkIII in the rain and I was delighted with the results. I always dread using ISO800 with D2X or D200, but it is much more workable with the MkIII. I've put a sample jpeg on that link in my other post, link here;

http://www.goldwingfacts.com/images/nonWingimages/Mk3samples/CanonMk3samples.htm

I've now put a brief caption under each thumbnail with ISO, shutter speed, aperture and lens used.

Whoops, I just noticed the updated page isn't showing yet. I'll fix that when I get home from work later on.

John Sheehy
July 9th, 2007, 05:37 AM
That is weird, my 5D's are considerably less noisy at iso 1000 than 1250 or 1600 and iso 500 is just a bit worse than 400, certainly noticeably better than iso 800. This what I'm seeing with day to day use of these iso's under work conditions, am I missing something here? Could it be something to do with the jpg's? I only shoot RAW and have found ACR excellent at totally killing all colour noise (colour noise reduction=10) while leaving the more natural luminance noise intact.

If you are only seeing a simple relationship between noise and ISO, then you are shooting in such a way that you don't have much in the way of shadow areas. The anomalies of in-between ISOs are only in the read noise, which is prevalent only in the deep shadows.

Jack_Flesher
July 9th, 2007, 07:06 AM
The anomalies of in-between ISOs are only in the read noise, which is prevalent only in the deep shadows.

Indeed, you need to compare noise in the shadows, that's where the 'tweener' ISO's show their weakness...

Steve Saunders
July 9th, 2007, 08:01 AM
Okay I just updated that page;

http://www.goldwingfacts.com/images/nonWingimages/Mk3samples/CanonMk3samples.htm