Lloyd's Comments on D3 Image Quality

[leping]

Bjørn: Can you offer some comments to what Lloyd Chambers just experienced after few hours with his new D3?

http://www.diglloyd.com/diglloyd/blog.html

Apparently he has been frustrated since the images look “chunked” and there doesn't seem to be fine details at 100%, in comparison to his 1DIII shots.  I guess he would appreciate your direct communication with him so he can better organize a side by side shoot out of the two cameras.

Personally I suspect what he saw somehow related to the D3's new double micro lens, as well as the black point clipping and related tone curve which compress the low-midtone and shadow values.  And a too strong AA filter?  Notice that Rob Galbraith did make the following statement:

"Thanks to all who've written in requesting a comparison of the D3 and Canon EOS-1D Mark III for night sports and other low light photography. We were given a preproduction D3 in late October, but in making the body available, Nikon USA stipulated that we not publish pictures from it. If that weren't the case, we'd show examples of how the D3 stacks up against Canon's EOS-1D Mark III at higher ISOs, and those examples would reveal that, while there are differences in the appearance of image graininess - Canon's grain pattern is tighter - there's no doubt that the D3 produces a less noisy, higher quality file at ISO 3200 and beyond. This is an incredible achievement on Nikon's part.

On the other hand, EOS-1D Mark III files at any ISO - especially CR2s processed through Canon's Digital Photo Professional - are generally slightly crisper and more detailed. This could be because we're looking at photos taken with preproduction Nikon equipment, but experience has taught us that production gear is likely to show the same image quality traits."

http://www.robgalbraith.com/bins/multi_page.asp?cid=7-8745-9153

Thanks,

[nfoto]

Below is the verbatim copy of a an e-mail I sent to my friend Lloyd Chambers, after getting an e-mail from him voicing his concerns about the D3 image quality.


Lloyd wrote amongst other the following "I'm having trouble making acceptable photos with the D3, everything  seems soft and just not "there"."

My answer was as follows,

I wouldn't say "soft and not quite there", but it is not hard to show that D2X at the same detail magnification produces images with better microdetail (however, bigger structures might not be equally well defined). so the overall effect being sometimes the one, sometimes the other, showing the best overall result.

When I shot the first set of comparisons I made a note to remind me how smooth and delicate the D3 image was when the D2X (at 100ISO) looked more grainy. Afterwards I discovered the D3 had been set to 400ISO.

I think the subject will largely dictate the outcome. But we are seeing the same principle that made the Canon 1DS Mk.2 look less sharp than the D2X (my D2x review) if we equalise comparisons on the field of view captured by the respective camera by using different lenses. The same lens on both D2X and D3, and keeping the detail magnification identical, might show almost no difference at all (except for the lower noise of D3).

Pixel density i(photosites per area unit) s lower on the D3 and there are advantages and drawbacks with this compared to D2X.

The first D3 advantage obviously is a vastly improved noise performance, the second that the combination lens + camera can achieve higher image contrast thus imparting a better sharpness for the "big details" that give the overall impression of sharpness of the image, the third (deriving from the second) is that the files stand much better enlargement. The almost total lack of CA and the even illumination of [some, not all] wide-angle lenses might result from the sensor design or the EXPEED processor, or both.

Now, to the drawbacks. It is clear that with high-performance lenses, D3 might have insuffcient resolution due to its lower pixel density. Whether or not this is an issue will depend on the subject and scene contrast, and also the processing of the file done in-camera. Shooting people with D3 should yield remarkable image quality, landscape shouldn't.

Since Nikon positions the D3 as a press, sports, or action photographers camera, and never said it was *the* solution to all needs, I think we. the end user community, have been intoxicated by the remarkable low noise/high-ISO performance and falsely assumed this would be the *ultimate* camera for every conceivable application. It is not. When a high-resolution companion to the D3 arrives, neither will that model be the "best" for sports and action shooting. But it will marvel for studio and landscape work.

My final comment is that the benefits of the DX format still are very much in existence after the arrival of the D3.

Do keep in touch.

regards

Bjørn

PS: I'm posting the essence of this e-mail on the NikonGear thread as well. DS

[leping]

I looked at Lloyd's D3 shots (TIFFs from NX) as well as some online sample shots from other photographers (the 14mm wide angle stadium one:

http://www.robgalbraith.com/public_files/Nikon_D3_ISO200_Skydome2_NX_SS.jpg

, for example).  I even asked for some of the sharpest 1DIII files.

My standard is not to just look images at 100% but unsample them to 4800x7200 (16x24 inches at 300dpi), the largest size I would consider printing from the current DSLR files.  Then, since the 100% screen view does not really correspond to what on prints, I look at the unres'ed results (after multiple pass USM with decreasing radius and a final Smart Sharpening 0.2-0.3/150, fade to luminocity after each step) at 50% on screen to give me the idea how the print would look like, which is usually equivalent to around 80% of the original pixel view size.

I found, at least on my dual CRT screens, there are no finer details from the 1DIII files at 100% viewing level on upsampled images in comparison to the D3's.  Actually the D3 images can be little sharper, with the 2 more MP counts, but sometimes also more harsh looking.

The 1DIII's images maybe smooth and at the same time contains a lot of sharp details, right out of the DPP (Canon's RAW processor), and D3's needs a bit more work.  But, as you said, since it came out soft it is easier to enlarge without seeing artifacts since all the sharpenings can be done in Photoshop in a well controled way (masks, etc.)

Also personally I dislike the Canon image's "taste", that renders low contrast relative smooth surface plastic looking, or "Rou", a term they created in China to describe Canon images (especially landscapes and citiscapes through ACR or LightRoom), my country of origin.  The 1DIII's images are less so I have to agree, than the usual Canon's (1DsII especially).  Every time I see images like this I have to pull down the DxO Film Pack filter immediately to add some noise, or "film grain", to make it look better.  Little bit noise dithers out fine details, and fools the eyes and brain to think the images are actually sharper and more three-dimensional.  Not to mention it greatly masks out these ugly digital (mostly demosiacing) artifacts.

I am a landscape photographer mostly still shooting film, so the D3 might not really for me.  But it is really interesting to see side to side comparisons of controlled shots from 1DIII and D3.  The D3's DOF is almost one stop shallower so to be fair it need to stop down more.

[maljo]

This is a very interesting post.
Nothing is as simple as it seems.
All the high hopes and expectations are going to swing
in the other direction, like a sine curve, for a while, as we
discover the limitations of the D3.
I have the luxury of waiting and seeing what others discover
with this camera (because I can't find one).
I wish Nikon had come out with that higher pixel
count landscape camera first.

maljo

[Nikkor]

The news in august were great but now the 12MP High speed demon is showing more clearly its attended audience.

I think an FX cam with around 18-20MP with a CCD sensor and a base ISO of 50 would make a lot of other photogs other than sports shooters and PJs happy. Will the D3x be such a cam? Hope they don't stuff too many pixels in there .

 

[Dallas]

Ya know, pixels are not that important for me personally. I very seldom print my images, preferring instead to present them on the web. Of course if I am working for a client who wants the resolution I have to acquiesce to those demands. However, depending on their needs I seem to be getting by with the 4MP's of the D2H and the 6MP's of the D70.

If I get some kind of windfall I would most probably get the D300, however that D3 is calling my name! I would need to justify the expense at a business level first.

[Nikkor]

Don't get me wrong, I'm not into pixels, I would be happy with 8MB quality pixels in a dx format which i think would have been great. The D3 though is what i consider on the thin side of pixel density and this might be one of the reasons why I'm seeing that plastiky softened look (at least on the images i see on the net). 

[NikosR]

Don't get me wrong, I'm not into pixels, I would be happy with 8MB quality pixels in a dx format which i think would have been great. The D3 though is what i consider on the thin side of pixel density and this might be one of the reasons why I'm seeing that plastiky softened look (at least on the images i see on the net). 

To be honest I don't see any 'plasticky, softened' look. I see files with great potential for interpolation and judicious sharpening and I also see the lack of noise which, to our Nikon trained eyes, might give the impression of plasticky. We are free to add noise to taste also.

[demosaic]

Don't get me wrong, I'm not into pixels, I would be happy with 8MB quality pixels in a dx format which i think would have been great. The D3 though is what i consider on the thin side of pixel density and this might be one of the reasons why I'm seeing that plastiky softened look (at least on the images i see on the net). 

Per degree, a 12MP FX has greater pixel density than an 8MP DX at the same FOV.  Actually, an 12MP anything has more pixel density per degree than an 8MP anything.

Per millimeter, low pixel density is a good thing.  You want the biggest pixels you can get away with.

[cyberean]

... I'm seeing that plastiky softened look (at least on the images i see on the net). 

probably due to the
aggressive use of NR.

[Paul]

Bjørn,

Below is the verbatim copy of a an e-mail I sent to my friend Lloyd Chambers, after getting an e-mail from him voicing his concerns about the D3 image quality.

Lloyd wrote amongst other the following "I'm having trouble making acceptable photos with the D3, everything  seems soft and just not "there"."

My answer was as follows,

...snipped...
I think the subject will largely dictate the outcome. But we are seeing the same principle that made the Canon 1DS Mk.2 look less sharp than the D2X (my D2x review) if we equalise comparisons on the field of view captured by the respective camera by using different lenses.
...snipped...
My final comment is that the benefits of the DX format still are very much in existence after the arrival of the D3.
...

Very many thanks for publishing this. Lloyd speaks so authoritatively that when you have slightly differing thoughts it's hard to be sure. As so often is a case of "horses for courses" which is partly what makes photography so fun Much appreciated.

cheers, Paul.

[diglloyd]

Bjørn,Very many thanks for publishing this. Lloyd speaks so authoritatively that when you have slightly differing thoughts it's hard to be sure. As so often is a case of "horses for courses" which is partly what makes photography so fun Much appreciated.

I'm not so sure Bjorn and I are saying different things.  My blog (http://diglloyd.com/diglloyd/blog.html) will be updated continually, and I share some thoughts there.   I have the utmost respect for Bjorn (nice trip in Norway with him), but he and I have different viewpoints.  We generally agree on most things, and for the rest, it might just be terminology.  For example, a portrait I recently shot is in line with Bjorn's thoughts.  Detailed shots have disappointed so far.

In a nutshell, here is what I'm seeing, keeping in mind that I have owned and used the Nikon D1/D1x/D2x/D200/D70 and Canon 5D/1DsM2 and I've used the Leica M8 and three medium format bodies, and shot a variety of film bodies including 6X17 and 4X5.  So I have a little background.

- the D3 screen and ergonomics are terrific ;
- the D3 can produce quite good detail in technical terms, but it doesn’t “pop” the way my Canon EOS 1D Mark III does.; fine details lack the edge contrast seen with the 1DM3.   Its images require considerably more sharpening the EOS 1DM3 images.  It might (emphasis: might) offer images that can stand more severe post-processing, but "out of the box" I think its psycho-visual impact is "flat" compared to the 1DM3.  And it's more work to make the images look good (the less work the better IMO).
- the D3 is a different animal. It reminds me of the Hasselblad 503CWD in its smoothness, but without the sharpness in fine details.  (see review at diglloyd.com of 503CWD).
- as a whole, images offer high macro contrast and nice dimensionality. Closer in (eg actual pixels on a 30" monitor), D3 images just fall a little flat.  They don't make me suck in my breath and go 'wow!".  I've shared some images with some expert friends, and they have all noticed the same effect.   By comparison, the Canon 1D Mark III continually impresses me.  The 21 megapixel Canon 1Ds Mark III has another story to tell, as yet untold.

The foregoing is highly subjective. Ignore it if you like, I'm working on objective comparisons.  But feel does matter, or Leica and Zeiss would not exist.

In short, don't dismiss the D3 as a great camera, but expect that it might take some learning, especially in post-processing, to produce the results you want from it.  I have learned in a few days that it will require weeks, and perhaps months to really understand the D3 files.  It is the most challenging DSLR to evaluate that I've yet seen.

[cyberean]

D3 images just fall a little flat. They don't make me suck in my breath and go 'wow!". 

don't know about you (... and don't mean to start a pissing
match here, as frankly all of this is quite subjective anyway)
but the images depicted in the following link take my breath
away. 
(of course our opinions probably differ on what "wow" is.)

http://www.russellrutherford.com/nikon_d3/

(... granted there's generous amount post work going on here.
though nothing unusual when it comes to a demanding client,
regardless brand of equipment used.)


still though, thanks for taking the time to share your perspective
and observations on the topic of the d3 image quality.

[rotarycooling]

Rutherford's photos take my breath away, as well.  I rarely look twice at sports photography, preferring wildlife; but these are extraordinary photographs of a species, homo sapiens.  Rutherford is a genius, and if the D3 is good enough for him...   Still waiting for mine (with which I hope to make passable photographs of ducks, otters, and my granddaughter)...

[leping]

Rutherford's shots are heavily processed with LucisArt.  Actually there are quite some artifacts (or artistic preference) in these images like the horizontal light strips.  You can download a trial version of LucisArt and play with your images to see the effects and the artifacts.  I played it a little but found it is not for landscapes.

As a faithful Nikon shooter and tireless promoter, I have to say the newly published 1DsIII images take my breaths away.  I don't care about noise (and as mentioned often add noise to my printing files) but really into details.  The 1DsIII's 21MP are obviously putting into very good use.  Just check out the test images on Imaging-Resource.com.  Actually the D3 test image there (the ISO100 one) is quite good in details in comparison to 1DIII and 5D -- it just needs more sharpening at my standard 4800x7200 size (16x24 at 300dpi), something like USM 2/50/0 + USM 0.3/100/0 + Smart Sharpening Gaussian Blur 0.5/500/0 + USM 0.2/200/0 to see that the level of really fine details are better than the other two.  Of course they are from different lenses, with different focusing errors, etc.  But the 1DsIII images are really something to see for which I can go to 24x36 with very little problems.

[macusque]

As much as I like Rutherford's images, we can't judge actual details at those sizes.

I have high respect for Lloyd Chambers (I read most of his reviews when I switched to Nikon) and I believe him, even if I would have much preferred to hear from him a "wow!" feeling about rendering of fine details from the D3...

But maybe it's only a sharpening issue, as also per Leping's findings.

[NikosR]

I would not at this stage take what diggilloyd writes in his blog at face value. He himself seems to suggest we shouldn't. To give two examples:

1. He seems to be criticising some D3 files for softness and then he finds out about what he considers an NX sharpening bug (it might well be). Now does this make all his previous blog comments irrelevant? (Sidecomment: he seems to be testing his D3's raw output using a raw converter he does not seem at all familiar with. Perilous path IMHO).

2. He talks about suspicions of aliasing effects although at the same time he talks about a too-strong AA filter while he displays an actual pixels example of what he considers aliasing which, to my eyes, looks suspiciously like raw conversion artifacts. (Of course the 100% picture size is too small to judge accurately but this is my strong impression judging from this single image he displays as aliasing evidence). If this is actually artifacts from the bayer interpolation and not aliasing I would have expected a person who talks with such authority to be able to recognise the difference

I will wait  for his final review and in the meantime I'm not going to bother with his blog.

(PS. I have the feeling that Bjorn is approaching matters in a much more scientific way than Lloyd does, although he can often sound cryptic and pythean)

[leping]

Just realize among the test images posted today at Imaging-Resource there is one converted with ACR from RAW.  You can put it side by side with the JPEG version to see how much sharper the RAW converted can be, after more sharpening of course.

Lloyd, one of my highly respected mentors, always tries to be objective and unbiased.  But if is difficult to really rate image qualities since different companies may emphasis different kind of "looking" and when artistic style comes into the equation.

[macusque]

I tried to convert some D3 NEFs (from a polish site) with RawDeveloper 1.7.1 and it does an outstanding job IMO.

Lloyd, maybe you could have a look at it (it also runs perfectly fine under Leopard):

http://www.iridientdigital.com/products/rawdeveloper.html

[macusque]

Just realize among the test images posted today at Imaging-Resource there is one converted with ACR from RAW.  You can put it side by side with the JPEG version to see how much sharper the RAW converted can be, after more sharpening of course.

Indeed, quite astonishing difference...

For a joke I upsampled that file to 1DsIII size and compared both. While the Canon shows its superiority as far as absolute resolution, the D3 IMO holds its own. 

http://www.pbase.com/marco/image/89922822/original