falconeyes: The real message here is that a crash of the Photos app can corrupt the library.
That should never happen with a well designed app (hint: transactional safety).
I think that's a real advantage with LR: not only did I never see a corrupted library after a LR crash (I've seen quite a few after fresh major releases). But because it is an SQLite database file, it could be repaired using SQLite tools. And then, LR really cares to keep backups of that file (of course, Time Machine does the same).
To understand how serious this is: Just imagine an application crash would leave your file system emptied ...
It is hard to accept Apple released such a poorly designed piece of software. Apple seems to have changed a lot recently.
Iliah, I suppose we just have different inherent sensitivities to such issues. As a firmware engineer there have been countless hardware bugs whose manifestation was attributed to my logic's handling of the issue rather than the underlying hardware bug itself. It works both ways, but a measured and objective evaluation of who is truly at fault for such issues is rather nuanced. If a pilot crashes a plane due to his handling of a malfunctioning instrument for a scenario where training should have allowed him to safely fly the plane, is the cause of the crash a malfunctioning instrument or pilot error?
@Iliah, Leica may not yet have root cause the issue down to the detail of malformed EXIF data, assuming that is indeed the cause. I don't read Leica's statement as attributing fault to Apple - it's worded rather diplomatically, saying the problem is the result of an "incompatibility". If I were a Leica+Photos user and were given a choice of which issue I would prioritize for resolution, I would choose Photos' systemic issue instead of one of likely many input triggers that causes the systemic issue to manifest.
Even with malformed EXIF (or image) data, proper code design should limit the extent of aberrant application behavior to the specific dataset rather than result in systemic corruption.
"The key thing to remember though is that, although shot noise and electronic noise stem from very different sources, with the exception of a few special cases (such as 'banding' pattern noise), they are visually indistinguishable."
This is correct above a certain signal level because the Poisson distribution of shot noise is close to Gaussian. However at low signal levels (deep shadows) the Poisson distribution of shot noise is distinctly different (visually and statistically) than Gaussian read noise.
ThrillaMozilla: The Web site blows the picture up to about 60" wide on my monitor. Why is this a surprise that you can see noise? If you enlarge it instead to 16"x24" size, you have to look fairly carefully to see any noise at all.
Now, I suggest that you all get ahold of that Nikon image and blow it up to 60" wide, and see if you can find any image defects.
Here you go ThrillaMozilla, full-sized images from a D7000 (even less DR than D810):
The Tulip photo sums up the Canon system in a single image; class-leading lenses with mediocre sensors. Fortunately bracketing can be utilized to overcome sensor limitations in most scenarios, which I use anyway even with better-performing sensors.
mpgxsvcd: I literally cannot find anything not to like about this announcement. This is Canon simply providing a good product at an extremely competitive price.
Kudos to Canon on this one.
I figuratively disagree with that literal statement.
ttran88: The Canon makes the D800E look like a killer deal for half the price. But if you have Canon lenses than you will have a nice body.
Used D800E's in mint condition are going for $1500 on Fred Miranda vs a (new) 5DSr for $3900.
photogeek: Things are not as bad as I thought they would be. Color me impressed. OTOH, I do not believe that many lenses in Canon lineup can actually get anywhere close to the resolution of this sensor. They will have to overhaul their entire lens lineup.
Canon has replaced nearly its entire modern lens lineup over the past past 5 years. 11-24mm f/4, 16-35 f/4 IS, 24mm f/2.8 IS, 28mm f/2.8 IS, 35mm f/2 IS, 24-70 f/2.8 II, 70-200 f/2.8 II, 300mm f/2.8 II, 400mm f/2.8 II, 500mm f/4 II, 600mm f/4 II
Digimat: hmmm sample the iso12800 raw files down to A7s resolution, put them in photoshop and switch between the layers...the sony(and nikon) cameras show a very ugly yellow color cast and less contrast than the canon. the downsampled canon files show a bit more sharpness than the A7s and in terms of noise i cant find a difference.
The 5DSr does very well but most of the tones in this studio scene are midtones so it's expected to perform on par with other modern FF sensors - the A7s doesn't show much advantage vs the D8xx/A7r either at these levels. If you look at the shadows around the Jackson-Triggs bottle you'll see the shadows a bit cleaner on the A7s. As you scale up from ISO 12,800 those shadows start to become midtones and that's where the A7s will start pulling ahead of other FF sensors.
Another factor to consider is that tripod-mounted shots provide an unfair advantage for higher-density sensors for High ISO comparisons, since in the real-world High ISO is utilized in shutter-speed limitation scenarios and those require a higher shutter speed on higher density sensors to achieve the relative resolution advantage vs lower density sensors.
PhotoKhan: If I click all over the scene, with RAW selected and different sample ISOs, what I see is that a $4,000 camera delivers the same overall highres quality of a camera that costs more than double (as a system, it even exceeds it, if we look at corners).
It doesn't matter that I am not the target for this camera. I can recognize success in a project when I see it.
No downsampling or upsampling - native comparison.
If you look even closer you'll see a $1,400 camera (used) that produces the same overall quality as a $4,000 (new) camera.
Greg VdB: Recommendation for pixel peepers: put it on "print size" and investigate how much difference you really see with your preferred camera... Personally, I'm quite comfortable in the comparison to my feeble Eos 70D, especially since I realize they are entirely different tools for different needs.
That said... Let the pixel peeping war commence!
12MP vs 36MP:http://www.dpreview.com/forums/post/55750050
The WB on all the 5DSr images is too cool. That may be why you think the A7s image has a yellow color cast. Here are the ISO 12,800 raws processed in PS CC, neutral profile, WB matched, downsampled to 8MP:
At first glance the 5DSr ISO 100 image looks to have a brighter ambient exposure even though the camera exposures are the same. Upon closer examination it appears ACR is applying a much steeper default tone curve to the 5DSr, with much darker shadows, slightly darker midtones, and highlights about the same. Here are sampled luminosity values in PS CC using the KODAK gray scale card in the scene. All raws processed in PS CC using Camera Neutral profiles and matched WB:
Gray Wedge (A):5DSr: 212.185DM2: 212.51D810: 211.63
Gray Wedge 3:5DSr: 168.795DM2: 170.69D810: 168.48
Gray Wedge 7 (M):5DSr: 105.885DM2: 109.22D810: 111.76
Gray Wedge 11:5DSr: 54.365DM2: 61.94D810: 65.50
Gray Wedge 16 (B):5DSr: 22.965DM2: 35.75D810: 32.82
Gray Wedge 19:5DSr: 16.985DM2: 30.68D810: 25.47
I think "own" has the wrong connotation. Unlike competitors which do own their markets by locking in customers to proprietary, closed mounts, Sony has to earn their semiconductor customers anew for each and every generation of their sensors. That's the benefit of open competition, where business is earned through innovation rather than strangulation.
falconeyes: Interesting and important article.
However, it should have used fewer words. The article makes a simple matter look more complicated than it really is. And may discourage some to read it.
Everybody thinking that noise is (mostly) a camera artefact should read the article tough.
Theory aside, I think the Coolpix images above support the notion that 14 bits is enough for a fully ISOless implementation through -7EV, which is about the limit of useable images for shot noise at the equivalent nominal ISO 12,800.
Not sure it's the ADC bit depth Rishi. Might be more a matter of getting the precise black levels right in the camera and/or raw processor. For example the Sony 16MP APS-C Exmor is the best ISOless sensor I've tested so far, at least for >= APS-C size - there might be some smaller sensors that do even better. Here's that sensor in the Cookpix A - look how close ISO 100 -7EV is to the equivalent nominal ISO 12,800. The latter still looks a little better but it's very close.
Source of above image, including instructions on fixing the A6000's ISOless tint:
Horshack: I use ETTR extensively on my original 5D bodies. Due to JPEG histogram issues mentioned in the article I recommend that everyone calibrate their camera's raw highlight clipping by shooting typical scenes with varying overexposure and then processing the images in their preferred app to see correlate actual clipping to what they see on the histogram. Another method is with Iliah Borg and Co. RawDigger app; there's a great 3-part article on this here: http://www.rawdigger.com/howtouse/rawdigger-histograms-what-is-the-raw-histogram.
My field ETTR metering technique is to leave the camera on spot metering. I meter the brightest highlight I want retained, which puts that highlight at middle grey, then dial in an exposure +2EV to +3EV above that depend on the reflectivity of scene and the camera I'm using.
Here are some 5D ETTR examples (before+after):
Thanks zsedcft. I used a combination of LR digital GND and local adjustments. For me the biggest benefit of ETTR is not lower noise per se but greater flexibility in applying contrast in post. Extreme contrast adjustments can sometimes yield more visible noise than directly raising shadows.
semorg: WOW...Dpreview finally discovered ETTR?
Looks like Richard forgot to write about an additional source of noise - the comment section of dpreview articles :)