Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: 1DS3 - Questions About Noise  (Read 1982 times)
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6818


WWW
« on: December 06, 2007, 09:31:40 AM »
ReplyReply

I decided to start a new thread taking off from Jack's "mini review" because I'm on a specific tangent about noise performance. In Jack's thread, I raised a question, which I subsequently helped myself to answer about the impact of the HTP option on noise, and I ended-up agreeing with Canon's instructional manual that enabling it causes more noise than not enabling it.

Moving on from there, of course a performance characteristic one tests with a new camera is *noise*. I have read many of the reviews, including Michael's - and in the latter case seen results first hand, so I have every reason to believe that their observations about the camera's noise performance are correct based on their own work. The overwhelming view seems to be that noise performance is excellent in the range of ISO 100 to 800.

I'm only at the early stages of my testing work, but so far I am somewhat less enthusiastic, and this raises questions about (1) me and my expectations, (2) what I'm observing, and (3) whether what I'm observing is normal or abnormal. As this will be a compact evolving process, I thought it useful to start this thread now eventhough my work isn't nearly done, so I could solicit views about how I should consider the issue. I shall make images available one way or another once my work has reached the point where I'm ready to put some rigorous, systematic stuff up on the web for public viewing.

But until then, I need to say a few things in general about the three questions above. Re (1) based on what I've read, at settings in the range of 100~200 I should see no noise. I don't really expect noise to hit me in the face up to ISO 800, but from say about 600 upward I don't expect it to be totally silky-smooth either. In particular, I don't expect to see any noise in skies and skin tones at least to ISO 400. In dark tones upwards of 200 I expect a minor amount of noise no matter how good the camera. These expectations are rough translations of numbers to what I see based on what I've read before I started working with the camera myself. There is no better scientific basis for them, so don't ask. The issue here is whether my expectations are "in the ball park". As for me, I've been told I have a keen eye for small differences in colour and texture, so we'll leave that one at that, but could it mean I see more than I *should* (i.o.w. and unvarnished -  amd I too much of a crank)?

Re (2), in the few tests I've managed over the past few days - haven't had that much time -  I'm noticing slight granularity in blue skies and caucasian skin tones at ISO 400 and above, on the monitor at magnifications of 25%, 50% and 100%, both in the raws and in the rendered images after moderate sharpening in CR 4.3, with color noise reduction at the default 25% and luminance noise 0%.  Noise Ninja is showing noise values in the 14~23% range for these images, much of it Luminosity noise.

Re (3) - that's the question - is what I'm getting "par for the course", or am I doing something wrong in my workflow relative to the needs of this camera, or is my camera somehow not working properly?

I know this is hard to discuss without seeing images, but as I say, I started this before I'm ready to post the evidence because I think there may be some experience out there which can respond to some of this without seeing test images. So it is a tentative discussion and it will evolve. I'd much appreciate any suggestions that will help clarify these issues and provide guidance for my on-going testing work, which I hope to complete by the end of the weekend.

Cheers,

Mark
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Stephen L Starkman
Jr. Member
**
Offline Offline

Posts: 59



WWW
« Reply #1 on: December 06, 2007, 11:23:29 AM »
ReplyReply

Mark,

As you know, with such a new camera there isn't a very good baseline yet of experience out there to determine many of the questions you ask.

I'd suggest one way to approach it now would be to make some of your RAW files available so those interested could open them directly, assess the exposure and apply alternative development and sharpening techniques.

Just a thought.

Stephen
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6818


WWW
« Reply #2 on: December 06, 2007, 11:32:33 AM »
ReplyReply

Stephen - thanks. You're right. At this stage these exchanges can't be determinative, but they can be indicative. Also the case not a great many people have the camera yet! By the way I think 1D-III experience should also be of indicative value.

I shall put up some raw files for downloading - I just need a bit of time to decide which ones and how best to do it given my internet set-up. I'll notify this thread when that's ready.

Cheers,

Mark
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
John_Black
Sr. Member
****
Offline Offline

Posts: 262


WWW
« Reply #3 on: December 07, 2007, 12:39:35 AM »
ReplyReply

Mark - one thing that may be interesting is comparing the 1Ds3 ISO to your Aptus ISO at 50, 100 and 200.  If you assumed the Aptus was best in class for 50 and/or 100, then you could rank your 1Ds3 relative to that scale.  I'm not trying to indirectly request Aptus files   I'm just suggesting that the back may provide a relative benchmark.
Logged

Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6818


WWW
« Reply #4 on: December 07, 2007, 09:09:03 AM »
ReplyReply

Quote
Mark - one thing that may be interesting is comparing the 1Ds3 ISO to your Aptus ISO at 50, 100 and 200.  If you assumed the Aptus was best in class for 50 and/or 100, then you could rank your 1Ds3 relative to that scale.  I'm not trying to indirectly request Aptus files   I'm just suggesting that the back may provide a relative benchmark.
[a href=\"index.php?act=findpost&pid=158858\"][{POST_SNAPBACK}][/a]

I think - unfortunately - you have the wrong Mark - I'm not the lucky Mark with an Aptus   .
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 6818


WWW
« Reply #5 on: December 07, 2007, 09:26:08 AM »
ReplyReply

Further to my initiating post, yesterday I did more shooting and testing, and I think I may have isolated a variable differentiating noise: quite simply put "exposure": where the material sits on the histogram. I shall be putting up comparison images in the near future, but for the time being this is conceptual, based on my observations of display images. I also expect to report on prints after the weekend.

I did a number of sky shots at ISO 200, and used Noise Ninja to measure the resulting noise. The basic exposure ended-up with a histogram quite far to left with the sky below the 50% point - required an exposure increase of +1.9 in CR4.3 to "normalize" the histogram. This image had a noise index of 17 (L=13, C=4), which is display-visible noise. Then I did a variant with plus two stops of exposure compensation. This improved the histogram very cosiderably, requiring a normalization in CR of only +0.5 ( three stops of compensation blew the sky). The Noise Ninja reading on this image was 8 (L=6, C=2). At this level, the noise is not noticeable on display and would not be visible in a print.

The day before yesterday, an image shot at ISO 400 had the sky slightly left of center on the histogram and the noise index was a visible 17 (L=13, C=4). It was the appearance of this sky, along with the previously discussed HTP business, which started to get me interested in a fuller exploration of the noise issue. So what I'm beginning to see here is the notion that exposure plays an important role in noise performance, regardless of the sensor quality and the ISO setting. Perhaps nothing earth-shakingly new in this observation, but it's beginning to tell me that "Expose to the Right" is probably just as important with this camera as it is with any other, when being quite "anal" about noise.  More anon (not more "anal", more results and some images   ).
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad