Oh dear....5D3 vs D800

<blockquote>

<p>I think I'll be sticking with the 5D3 for the foreseeable future.</p>

</blockquote>

<p>Dan, I am glad to see that I am not the only one going back and forth on these matters. I would like to give the D800 a sustained try (the kind that is only possible with a purchase and long-term usage), but I am reluctant to knock off a convenience store at this point to satisfy my personal curiosity. Maybe in the future. . . .</p>

<p>[<em>How my camera addiction led to a life of crime</em>, forthcoming]</p>

<p>--Lannie</p>

<blockquote>

<p>Just the other day I tried to do a +2 stop exposure on an accidentally underexposed (heavily backlit, before I dialed in EC) shot from my 5DIII and already saw banding.</p>

</blockquote>

<p>Rishi, that is sobering news. I move exposure compensation all over the place in routine shooting. I wonder if there is a firmware fix for the banding problem. I keep hearing about banding but have not witnessed it in my own Canon shots.</p>

<p>http://www.digitalcamerainfo.com/content/Canon-Acknowledges-Canon-EOS-5D-Mark-II-Banding-Issue-19629.htm</p>

<p>--Lannie</p>

<p>Rishi,<br>

Maybe I'm missing something, but if I look at this :<br>

http://www.dxomark.com/index.php/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/(appareil1)/795|0/(brand)/Canon/(appareil2)/176|0/(brand2)/Canon<br>

then go to the measurement tab and check SNR 18% and Dynamic Range, the curves show a distinct advantage to the 5DIII.<br>

As I said, I have never experienced serious banding / noise problems when lifting shadows to my liking with my 5D. Hearing that in your experience the 5DIII is more vulnerable to that issue is very troublesome. The improved AF and high iso of the 5DIII are compelling reasons for me to upgrade from the 5Dc, but given that a significant amount of what I shoot is in uncontrolled lighting environment (street / documentary), I absolutely need to be able to recover shadows in post. I'm not talking of silly extrem +4ev shadow push, but enough shadow recovery to meet the challenge of average backlit scenes or dark skins shot against a bright light colored background. <br>

Would you care to share some examples ? <br>

Thanks in advance</p>

<blockquote>

<p>Note: I <em>rely</em> on good gear that I use, appropriate to the types of photography I am doing, but I honestly don't think about the gear in terms other than what it does for my photography.</p>

</blockquote>

<p>I sleep with my gear, G. Dan. I know it's sick, but I just can't help myself. I know that some women tend to complain, but let's face it: women are expendable. Glass is forever. Women tend to leave. I have never had a lens or a camera walk away and leave me forlorn.</p>

<p>Let's keep our priorities in order here.</p>

<p>--Lannie</p>

<p>Here's an example of the type of shot where I would have to do a significant amount of shadow recovery. Incidentally, this was shot with film, with significant overexposure and underdevelopment to contain contrast. Still, after a conservative scan, I had to tame the hilights and lift shadows quite a bit in PS.<br>

In the wet darkroom, I had to give the upper right corner 6x the exposure of the faces to get a hilight details in the print. <br>

This is the type of shooting situation I would like to be able to face with my digital system. </p><div></div>

<p>Pierre, both the original capture and the post processing are superb.</p>

<p>--Lannie</p>

<blockquote>

<p>[<em>How my camera addiction led to a life of crime</em>, forthcoming]</p>

</blockquote>

<p>I seriously doubt it. Try audio equipment.<br /> :-)</p>

 

<blockquote>

<p>let's face it: women are expendable. Glass is forever. Women tend to leave. I have never had a lens or a camera walk away and leave me forlorn.</p>

</blockquote>

<p>Guys, stop it. You are killing me.<br>

:-)</p>

<p>Happy shooting,<br /> Yakim.</p>

<p>Lannie,<br>

Thanks for the compliment, it is very much appreciated. <br>

This picture was taken in a remote Vezo village of the southwestern coast of Madagascar. The Vezo, one of the 18 ethnic groups of the island, are essentially semi nomadic fishermen. The southwest of is the only desert region of Madagascar, which is otherwise mostly tropical. The sand there is about as white as snow. In the middle of the day, the temperature shoots well above 40°C. People built canopies out of leaves and gather under them during the hottest hours. Hence the photographic challenge of shooting dark skins in the shade against the background of white sand blindingly lit by the unforgiving sun. <br>

This was not just a photographic challenge. I was fortunate enough to spend two weeks in this village. It was a unique experience.</p>

I haven't seen any banding in the 5D3's output (in several hundred exposures so far). The DR and shadow noise seem

fine so far. Images are sharp, color fidelity is good. It works great with the 580ex. AF is fast and accurate. And now we

learn from Canon that the infamous glowing LCD problem has absolutely no impact on exposure or image quality.

 

It's a great camera, folks. If you liked the 5D2 you'll like this one even more.

 

Lannie, good call on lenses versus women. Glad you have your priorities straight. :-)

<p>First off: I need to stop and compliment Landrum Kelly as his posts literally have me cracking up to no end here :) Also, Landrum, that link you provided was for a different problem (with sRAW). We're talking about banding in RAW in general, which I doubt Canon has officially acknowledged. But most of us who own Canon bodies implicitly acknowledge its existence. Funny, my Nikon friends don't even know what the term 'FPN (fixed pattern noise)' is... :)</p>

<p>Second: Pierre, I don't think you'll have any trouble getting more DR than film with Canon or Nikon... the SNR of negative film is so low on the low end that it eats up overexposure & almost requires it to get good shadow detail. Digital SLRs just require a different philosophy where you expose more for the highlights than for the shadows. You then recover shadow detail. With Canon, you'll just have to compromise more, maybe blow more highlights than you would with Nikon to preserve shadow detail. Or use noise removal software... someone on dpreview successfully removed a lot of the 5D3 banding... and even though you may cringe at the possibility of removing detail by using noise removal, your SNR in the shadows will start off so much higher than it would be with negative film that it will really be moot. For more on this, read Roger Clark's treatment of it <a href="http://www.clarkvision.com/articles/dynamicrange2/">here</a>.</p>

<p>As for the 5D vs 5D3: I'm sorry, I was comparing the 5D <em>Mark II</em> to the 5D Mark III; <strong>my bad</strong>. The 5D Mark III has slightly lower DR than the 5D Mark II, but both the Mark II & Mark III appear to have slightly higher DR than the 5D classic. At the pixel level, the 5D III only has 0.2EV more DR than the 5D classic, so I don't know if that translates to any real world advantage. But you do have more pixels with the 5DII & 5DIII so there is likely some DR advantage (indicated by DXO's normalized 'print' scores).</p>

<p>Either way, though, I wouldn't call your 5D classic 'lowly' at all compared to the Mark II & Mark III. Now, the entire 5D line compared to the D800? Maybe :)</p>

<p>Also, I do believe there's some degree of inter-unit variability. So that's why some may see more banding than others.</p>

<p>Here's an example of where I did a +2 stop exposure & +100 Blacks in LR 4:<br /> <img src="http://rishisanyalphotography.com/ForumPostFiles/photo.net/5D3-Noise&Banding.jpg" alt="" width="800" /><br /> <a href="http://rishisanyalphotography.com/ForumPostFiles/photo.net/5D3-Noise&Banding.jpg">Link to full-size image</a></p>

<p>View full image at 100%. Original exposure on left, adjusted on right. You can see a thumbnail of the full original shot at the upper left. After a few shots I dialed in EC b/c I didn't care too much about the sky & wanted a more blown-out overexposed look to the shot. Of course, to do this properly, I should've brought my off-camera flash & umbrella, but this was supposed to be a landscape shoot & what happened was the sunrise was pathetic b/c of too many clouds at the tulip fields so I just ended up taking some shots of my friend in the fields. Given how many clouds there were in the sky, this scene doesn't have an unreasonable amount of DR (and the skies were still blown & unrecoverable in my underexposure). But try to lift those shadows, & it gets pretty noisy, as you can see.</p>

<p>I'm not saying the 5D Mark III is too noisy to work with, I'm just saying that if this were the D800 or D7000, this much shadow lifting would not be a problem. In general, this<strong> isn't</strong> a big problem for me when it comes to <strong>people</strong> shots, where I generally don't care about <strong>a ton of DR</strong>... & when I do, I use <strong>off-camera flash</strong> to get this sort of look:</p>

<p><img src="http://rishisanyalphotography.com/Singles/Andrew-DiscoveryParkSunset.jpg" alt="" width="800" /><br /> Where I <strong>do</strong> care, however, is when it comes to <strong>landscapes</strong>. Take the following photo, e.g.: <br /> <img src="http://rishisanyalphotography.com/ForumPostFiles/photo.net/Seattle-RizalBridgeSunset.jpg" alt="" width="800" /><br /> That's already 3 exposures merged by hand using layers in PS, & I used a Daryl Benson 3-stop Reverse Graduated ND filter. The most overexposed file was used for the patch of trees at the lower right & still, if I brighten them any more than what you see here, I see vertical banding in a large print. Granted, this was shot with my 5D classic, but a little more DR would really have been welcome. Yes, you might say that I should've taken another +1 or +2 shot in the field... but to that I say: hindsight is 20/20 :) Though I must say, the 5D certainly made this scene easier to work with than good ol' Velvia 50... here's the same scene a couple years ago shot on my EOS-3:<br /> <img src="http://rishisanyalphotography.com/ForumPostFiles/photo.net/Seattle_RizalBridgeSunset_Velvia50.jpg" alt="" width="800" /></p>

<p>Back then I had no choice but to wait till 45 minutes after sunset to even attempt this shot, even though I was also using the Daryl Benson 3-stop Reverse GND back then (else the band of sky above the mountains would be blown-- actually, in the original slide you can already see it getting close to blown or losing color... some highlight recovery on the scan was used to get color back in the sky above the mountains).</p>

<p>Just to state my position: I'm torn, because so far I love the 5D III AF, & the 600EX-RT flashes... the system 'just works' now, my major gripes with the 5D Mark II/580EX II being solved (poor AF, flashes frying & poor reliability when used w/ Pocket Wizard). But the sensor on the D800 is a game changer, as is the 14-24 f/2.8 lens (I still can't find a copy of the 16-35 or 17-40 Canon wide zooms that doesn't require me to go to f/16 to get edge-to-edge sharpness... e.g. that Rizal Bridge shot shows softness on the left side compared to the right when printed at 20x30 even though it was shot at f/18). And b/c I love 18-point sunstars & circular OOF highlights down to f/2.8 on primes, I'm tempted to invest in Nikon for their consistent use of 9-blade apertures in their prime/pro lenses.</p>

<p>So it's not *just* disappointment in the 5D3 sensor that has be on the cusp of switching. It's just the last straw... though a rather heavy one, I think.</p>

<p>-Rishi</p>

Maybe the banding is caused by the merge process?

<blockquote>

<p>Maybe the banding is caused by the merge process?</p>

</blockquote>

<p>No, it's in the original +2 exposure, & I used luminosity masks for the merge, & made sure the area with the trees got 100% of its pixels from the highest exposure file (so I did some hand painting of the mask). So I can't fathom any mechanism that would make the banding result from the merging process.</p>

<p>Actually, now that I go back & look at the image, it's the general noise in the shadows that bothers me even more than the slight bit of vertical banding. But the original point remains: this type of shot would benefit from a significantly cleaner sensor.</p>

<p>Some people won't care. Some people don't even care in the 24x36 print of that shot. The point is: I care.</p>

<p>Just haven't decided yet if I care enough to switch. I'm going to have to go on a number of fast-paced people/event shoots with my friend's D800 + fast primes to see if it can keep up with the 5D Mark III AF, which has been pretty awesome so far (using the side cross-type points; i.e., not having to focus & recompose).</p>

<p>-Rishi</p>

Why was +2 selected if that exposure was noise prone? Perhaps you could have overexposed the shadows at +4 or so and then

adjusted exposure before combining files.

<p>I meant +2 from the previous exposure, which was already +X from what the meter thought. Obviously I was shooting in Manual mode. Here is the unprocessed overexposure:<br>

<img src="http://rishisanyalphotography.com/ForumPostFiles/photo.net/Seattle-RizalBridge_OverExposure.jpg" alt="" /></p>

<p>I didn't take an even higher exposure because I wanted to use this overexposure for everything other than the mountains & the sky (i.e. the buildings, highway, trees, etc.). If portions (e.g. the border of the buildings against the water & sky) become significantly overexposed & bleed into other areas, merging with finesse becomes very difficult (& I dislike all automated HDR software... their output is usually too 'baked' IMHO).</p>

<p>You could argue, though, that I should have taken an even higher exposure just for the trees. To which I say:</p>

<ol>

<li>If you saw that overexposure I posted above on the back of your LCD, with no indication that blacks are clipped in the histogram, wouldn't you have thought you'd be OK?</li>

<li>You are correct :) But hindsight is 20/20</li>

</ol>

<p>Like I said, in retrospect, it probably would've been a good idea to take yet another +1 or +2 exposure, then dial in -1 or -2 & then merge it... that way at least the shadows in those trees would've been <em>clean</em>.</p>

<p>But the point I was trying to make is that with the >2 stops better DR of the D800, it would've been clean to begin with.</p>

<p>Just like the noise in the hair/neck area of my friend in that +2 shot at the tulip fields also would've been clean.</p>

<p>Yes, I'm cherry-picking bad examples, which account for a very small percentage of my photography. That's why I said this DR thing is more of a 'last straw'. But I don't think anyone <em>wouldn't</em> appreciate even those few % of photos being recoverable, because sometimes there's a gem in that few %. </p>

<p>-Rishi</p>

<p>Here's another example from my shoot the other day with the 5D Mark III:<br>

<img src="http://rishisanyalphotography.com/ForumPostFiles/photo.net/5DIIIBandingUponVignettingCorrection.jpg" alt="" width="800" /><br>

<a href="http://rishisanyalphotography.com/ForumPostFiles/photo.net/5DIIIBandingUponVignettingCorrection.jpg">Link to full-size image</a></p>

<p>Please view it at 100% by following the link.</p>

<p>I did a partial vignetting correction (this was shot w/ the 35/1.4L at f/1.4, ISO 100), and a modest shadow lift, then some contrast adjustments to get the image on the left. Original on right. <strong>I don't think that's an unreasonable shadow lift</strong>. Yet it's already noisy, with pattern noise evident.</p>

<p>Yes, you'll probably only see it on a large print or at higher magnifications on screen. But you won't see it at all on a D800. With Canon, you pretty much have to accept that your files will often be riddled with this. Especially given Canon primes' propensity to vignette when wide open (the 24/1.4 vignettes more than 3 stops!)...</p>

<p>Whether or not you care is the crux of the matter. I personally feel that Canon should've done better. <br>

-Rishi</p>

<p>Rishi,<br>

Thanks for you long and informative answer. In the examples you posted, the first one, the portrait is really where I see something problematic taking place. Noise in the tulips is probably noticeable only on very large prints. I generally print no larger than 30x40cm, to match the maximum I can deal with in my wet lab.<br>

I browsed through the link you posted and will have to read it over more thoroughly as it deals with technical concepts I'm unfamiliar with (transfer function, among others). The contention that digital cameras have much better DR than film is something I've first read not long ago and which surprised me, as I was under the opposite impression. But some reasons why I may "feel" that film has better DR are : 1. in much of my shooting, highlights contain more crucial information than shadows; 2. out of many years of experience I have become more skilled at getting the most out of film, hence compensating for the limitations of the medium; 3. the blooming that happens in digital overblown highlights is something I particularly dislike, as opposed to how film reacts to drastic overexposure. <br>

On the other hand, the concept that digital cameras have better DR, based on better SNR makes sense when you consider that digital cameras fare so, so much better in low light situations than film.<br>

The bottom line is that I probably need to improve my skills in digital photography. I'm still doing much of my shooting with film, so that my experience with the digital format very much lags behind.</p>

 

<blockquote>

<p>But some reasons why I may "feel" that film has better DR are : 1. in much of my shooting, highlights contain more crucial information than shadows; </p>

</blockquote>

<p>You raise a good point: that in digital, the 'rolloff' into highlights is poor when compared to film. Blowing out the sun looks so much better on even my Velvia shots... I believe this has to do with a number things that goes on in film -- with negative film, for example, the more a portion of film is exposed, the less likely future photon encounters will be productive (due to negative charge build-up at 'sensitivity specks' & other factors). Furthermore, there is potential for diffusion of photons within the emulsion-- the absolute amount of which increases with increased exposure. This essentially 'blurs' harsh highlights. </p>

<p>I also don't like the harsh transition to highlights in digital; neither do cinematographers in Hollywood (in the Zacuto 'Great Camera Shootout', filmmakers pay special attention to digital cameras that have a smoother rolloff in the highlights). However, I have found that certain RAW converters have evolved to handle these highlights better... I bet with some more intelligent software processing, it could be made even better (think: function that introduces some blur as a function of pixel intensity above some critical threshold).</p>

<p>Anyway, yes, different philosophies of shooting are required. It's nice to hear that people are still shooting film though... every now & then I feel the urge to break out a roll of 120 Velvia just for the gorgeous look of those colors on a lightbox!</p>

<p>-Rishi</p>

<p>Rishi,<br>

What you say about film behavior in the hilights is consistent with the curve of film which show a smooth "shoulder" when you reach higher illumination values. There is also a subjective aspect: as light diffuses within the emulsion in the adjacent area of the burned out spot, it reveals the grain structure of the film, hence adding "matter" to a part of the picture that has little tonal attractivness. That makes burned hilights on film even more acceptable than in digital images, which is why I find it much more difficult to do night photography with a digital camera. The burned out spots left by electrical lighting are much less attractive than with film.<br>

As for my using film: part of it has to do with many years of experience. My eye is simply better educated in this medium. Often I find it harder to anticipate the final print result when in the shooting stage with digital. Also, even though I have seen very beautiful digital prints, especially with cotton-like papers, I still think there is something very special about carefully crafted BW fiber base traditional prints. Finally, about ten years ago, I built my dream lab at home and I enjoy tremendously working in it !</p>

Rishi, have you added any noise reduction. I typically add a bit to 5D2 files when shadow detail is important. LR4 lets you

do this locally.

 

I noticed that the D700 had less shadow noise than the 5D2, but also less detail. This is a guess, but that leads me to

believe that Nikon opted to add default NR to their processing algorithm, whereas Canon provides a true raw file with less

processing. Maybe the same is true with the next generation as well.

<p>D700 image softness is due to the sensor optics, not "noise reduction" applied to the data. But because of the low noise the images sharpen ok. Noise reduction activates only above ISO 1600 if you've turned it on and doesn't affect raw data apart from tags. In any case it's pretty difficult to apply any sensible NR to bayer sensor raw data since it would first have to be converted to RGB then noise reduction applied, then back to the original raw format - doesn't make much sense to do that.</p>

 

Illka,

 

I would think that 5d2 has more detail, simple because of higher resolution. Can you explain what "sensor optics" is ?

Thanks in advance.

<p>My NR-in-RAW theory is just a theory, but I think it's quite reasonable.</p>

<p>I don't buy the argument that raw files are completely unprocessed. Those EXPEED, BIONZ, and DIGIC chips are doing SOMETHING to the images, even raw files. For instance, I've read that the D800E applies some moiré reduction in firmware. </p>

<p>Another example. The Nikon D3X was based on the same sensor as the Sony A90 (I think that was the model). The Sony had a reputation for noisy output; the D3X did not. How can that be? It's the same sensor, right? Yes, but not the same in-camera processing and firmware.</p>

<p>Canon's strategy seems to be to let the noise pass through and let the photographer clean it up in post production. Canon's JPEG images are far, far cleaner when NR is applied, cleaner even than what I can accomplish in LR3 or LR4. And they remain quite detailed, not blurred.</p>

<p>Just a theory. I accept the fact that I may be completely off base with this, but it still seems feasible.</p>

<p>If someone do need to print A1 or A0 sizes then the D800 is a must-have. For speedy FPS and if one is not in need of A1 size - Mark 3 is a perfect camera. I have shot JPEGs in Canon, Nikon and Olympus and after Olympus skin tones and JPEG quality I prefer Canon. <br />D800 is a strong Nikon's desire to find its niche in Med. Format group of potential users. </p>

<blockquote>

<p>"If someone do need to print A1 or A0 sizes then the D800 is a must-have.<em></em>"</p>

</blockquote>

<p>Uh, no it isn't. </p>

<p>Dan</p>

<blockquote>

<p>Uh, no it isn't.<br>

Dan</p>

</blockquote>

<p>Dan,<br>

Technical control of your customer might not accept your 22 Mp shot printed on A1 or A0 because of poor details and you won't be paid. Period. Take 270-300 dpi and re-calculate them to the lenght of A1 print - go maths. <br>

With Nikon you get 50% more resolution across the frame and about 20% more linear resolution. There is a makret and tasks for digital Med. Format, still. <br>

But I like Canon better for <em>my</em> use. </p>