Nikon D3 high ISO commentsPERMALINK

I’m steadily accumulating experience and images for an in-depth look at the Nikon D3 (links) and Canon EOS 1Ds Mark III (links), but I’m so excited about the Nikon D3 high ISO performance that I feel compelled to share some thoughts.

The image below is a good example of what the Nikon D3 is capable of (keep in mind that my comments apply to the original 16-bit TIF; the medium-quality web JPEGs seen here lose something in the dumbing-down). I’ve seen various high-ISO noise comparisons on the web and well...just let me say this: nothing matters like shooting a variety of real images, and the Nikon D3 stands head and shoulders above anything I’ve ever used (including the EOS 1Ds Mark III and EOS 1D Mark III).

It’s not just the amount of noise, it’s that it’s subjectively pleasing noise, something one can’t evaluate with graphs and charts. If you regularly shoot in low light, the D3 is the only camera to consider—end of story. It is the camera I pick up now in the winter whenever daylight is limited and/or it’s walk-around photography—it’s a serious breakthough for available-light work. The noise in the image below is so attractive (as noise goes), that I’m not even sure I’d have wanted to make the image at a lower ISO—I like how the noise compliments the image.

Nikon D3 + Zeiss ZF 28/2 Distagon, handheld, 1/30 @ f/4, ISO 3200
(click image to see larger and crops)

The machine seen above is used to pound rice into elastic material for making Japanese “mochi”. I can testify that it does an impressive job, so long as the operator is skilled enough to avoid crushing his hand in the process. There is a lot to be appreciated about such elegant machinery, and the skilled person who can operate it.

The image was made using the new Zeiss ZF 28/2 Distagon. My Zeiss ZF Lenses review is being extended to cover this impressive new addition to the ZF line. I think the 28/2 did an outstanding job on this particular image (after all, if one just has the right camera and lens, the rest is easy*).

* like raising children

Lens review: Canon EF 14mm f/2.8L IIPERMALINK

Looking for an extreme wide angle lens for your Canon EOS body? Check out my latest lens review. The review will be available for download on Saturday Dec 29, 2007, but may be purchased now.

CameraQuest lens adapter—followupPERMALINK

Following up on my Dec 22 entry regarding a gap with a CameraQuest Leica-to-EOS adapter, I’ve received a replacement and it does not manifest the problem—problem solved. Thanks to Stephen Gandy of CameraQuest for responding so quickly to the problem. I’ll be returning the defective adapter, and if anything definitive is found as to the cause, I’ll report on it here.

Nikon D3 viewfinder weirdness—followupPERMALINK

Following up on my Dec 22 entry regarding the “fireball” in the viewfinder, I’ve discovered what causes it: the Nikon anti-fog eyepiece! With the eyepiece in place the extremely bright spot is present; remove the anti-fog eyepiece and it disappears. I had used the as-shipped eyepiece until quite recently, which apparently does not have this issue.

Zeiss ZF 28/2 Distagon—depth of fieldPERMALINK

One literal approach to photography is “f/8 and be there”—be ready, be at the scene, and make everything “sharp”. Strangely enough, this long-standing idiom makes a lot of sense for today’s high resolution digital SLRs, where stopping down past f/8 degrades image contrast and sharpness (see Diffraction).

But assuming one is “there” and the subject matter is placed as desired (the “composition”), what other factors bear on the result? Certainly lighting plays an enormous role, but so does depth of field; it is the visual impact and perception of the composition that changes as a result of these secondary factors. Consider the two images below; one is shot wide open at f/2 and the other is shot at f/8 (handheld, framing is not exactly the same). Which do you prefer?

 
Nikon D3 + Zeiss ZF 28/2 Distagon, f/2 (left), f/8 (right)

If you belong to the “everything sharp front-to-back” camp, you’re going to be annoyed at the f/2 image, and still be annoyed at the f/8 image, since even f/8 cannot provide enough depth of field to render more than a narrow band of sand crisply. A good friend of mine will find the f/2 image distasteful (“it’s blurry”), but I prefer it.

The Zeiss ZF 28/2 Distagon offers great artistic possibilities at f/2 (for several different optical reasons). By comparison, an f/2.8 lens has 1.4X the depth of field of an f/2 lens, and thus loses such expressive potential. This is one reason that I wish Zeiss had made the 25/2.8 Distagon an f/2 design instead, and why a 20/f/1.4 would be appealing.

Zeiss ZF 28/2 Distagon in stock at B&HPERMALINK

If you’ve been lusting over the newly-available Zeiss ZF 28/2 Distagon, B&H Photo has it in stock as of today (though that could change as stock is sold). Read yesterday’s entry for thoughts on this new lens. I’ve also been working steadily to update Zeiss ZF Lenses with Nikon D3 results.

Zeiss ZF 28/2 Distagon

Zeiss ZF 28/2 Distagon PERMALINK

I made a number of nice images with the Nikon D3 + Zeiss ZF 28/2 Distagon today, which I’m adding to Zeiss ZF Lenses. Without a doubt, the ZF 28/2 Distagon has considerable “personality” which must be understood to exploit it fully. The Zeiss designs are clearly “artist’s lenses”—they do not render images generically like the brand-name alternatives. Some of them draw so differently that the usual “by the numbers” comparisons are actually quite misleading. The ZF line offers image rendition unlike any other lens line, together with astonishing consistency in color rendition and flare control.

Here is one (cropped) image taken today in Pescadero, CA. Artistry of another kind...

Nikon D3 + Zeiss ZF 28/2 Distagon, 1/320 @ f/4, handheld

In Zeiss ZF Lenses, example images are presented at larger size (typically 1536-pixels wide), with more and larger actual pixels crops using minimal compression, along with analysis designed to ferret out the personality and characteristics of the lens. In this way you can learn, in advance, what to expect from each ZF lens prior to purchase, as well as useful insights that could otherwise take months to learn.

Nikon D3 viewfinder weirdness PERMALINK

I ran into a really strange phenomenon today—while shooting the Zeiss ZF 28/2 Distagon with the sun positioned in the upper left of the frame, I noticed an small, but intense fireball in the lower right of the viewfinder, clearly a derivative of the sun itself, but much smaller and more focused.

Since the actual image shows outstanding flare control, it’s clear that the issue lies with the D3’s viewfinder optics, or possibly some interaction with certain lens(es). I’ve never before seen a problem like this, and it’s disconcerting. I reproduced the problem again later under somewhat different lighting conditions with a less intense result. I don’t yet know if it’s something peculiar to the D3 and the 28/2 Distagon, or a general problem, and I haven’t yet tried the 28/2 Distagon on the Canon EOS 1Ds Mark III.

Update: Looking carefully at the resulting image, the “fireball” does seem to be in the same place in the frame as seen in the viewfinder, but the ZF 28/2 Distagon’s flare control supresses it superbly. Perhaps there is some focusing screen interaction. Once understood, the problem can be ignored from an intellectual basis, but still the visual feedback remains disconcerting.

Nikon 24-70 FX PERMALINK

B&H Photo in New York has Nikon’s fantastic 24-70 f/2.8 in stock (it might have been hard to find till now). If you get one, please use the B&H link at the top of this blog and help support this site. (Search B&H for “Nikon 24-70”).

Problem with CameraQuest lens adapter— followupPERMALINK

Yesterdays’s entry described an issue with the Leica-to-EOS adapter I use. Stephen Gandy of CameraQuest has said he’ll be sending me a replacement. When it arrives I’ll report further.

Problem with CameraQuest lens adapterPERMALINK

In the past I’ve recommended the lens adapters sold by Stephen Gandy at CameraQuest. These adapters allow (for example) mounting Nikon or Leica or Olympus lenses on Canon EOS bodies. Most of my experience has been with the Nikon-to-EOS adapters.

While shooting the Leica R 180/2.8 APO today on my Canon EOS 1Ds Mark III using a CameraQuest adapter, I discovered a problem: with the camera turned vertical, a slight gap would appear between the lens adapter and the lens (perhaps 0.2 mm or so). This gap was enough to see focus to pop in and out when using Live View (pushing on the lens)—a serious problem indicating misalignment of lens and sensor. Oddly, the gap does not seem to occur with the camera positioned horizontally, so I suspect a design problem.

Perhaps this is a problem specific to my particular Leica-to-EOS adapter rather than a general weakness in the CameraQuest adapter design for Leica lenses. The Leica 180/2.8 APO is a medium-weight lens at 2.2 pounds / 1kg. It might be that the design simply cannot sustain a lens of that weight, but since the 180/2.8 has no tripod collar, mounting the lens is not an option (and hanging a 2.7 pound camera off the lens isn't likely to fare any better!).

I’ve contacted Stephen Gandy at CameraQuest for an explanation, and I’ve requested that a replacement be sent in order to verify whether the problem is with my particular adapter. It remains to be see how Mr. Gandy will handle this situation. Expect a follow-up when I know more.

Also worth noting is that so far I have not observed this problem with the CameraQuest Nikon-to-EOS adapters, but I also have not mounted any lens over about about 1.5 pounds (the Zeiss ZF 100/2 Makro-Planar, which is also shorter). Nor did I observe a problem with the 1.7-pound Nikon 180/2.8 ED. However, I’ll have to now take a closer look.

Nikon lenses have a 46.5mm flange focal mount distance; Leica lenses have a 47.0mm distance. Canon lenses have a 44mm distance. So in theory the Leica adapter should be able to be 0.5mm (20%) thicker eg 3.0mm vs 2.5mm (and hence stronger) than the Nikon adapters.

Zeiss ZF 28mm f/2 Distagon in handPERMALINK

FedEx is apparently working overtime, because the expected Monday delivery showed up early. I had only a little time to shoot the Zeiss ZF 28mm f/2 Distagon this evening, but my initial impressions of its image quality are highly favorable, though with a strong “personality” along the lines of the ZF 25/2.8 Distagon. I expect to release an interrim update to my Zeiss ZF Lenses review in the next few weeks.

The ZF 28/2 I’m using is an evaluation copy (which I’ll almost certainly purchase). Brand-new, it arrived with a lens hood that would not attach (I experienced this problem once before with the lens hood for the 35/2 Distagon). I was able to disassemble the lens hood, jostle the problematic internal ring, and reassemble the hood so that it functioned properly.

Zeiss ZF 28mm f/2 DistagonPERMALINK

I expect to receive an evaluation copy of the Zeiss ZF 28mm f/2 Distagon within a few days. I’ll be reporting on it here, but detailed evaluation will go into Zeiss ZF Lenses.

Those who purchase Zeiss ZF Lenses by the end of 2007 will receive the updated version with the 28/2 Distagon (and Nikon D3 results) free of charge. On Jan 1, the price will increase, so be sure to purchase the review before the end of 2007—no exceptions. The new price will include the 28/2 Distagon and Nikon D3 results, and (possibly) some Canon EOS 1Ds Mark III results.

Zeiss ZF 28/2 Distagon

Getting detailed information about digital images (EXIF)PERMALINK

I’ve advertised my Nikon D200 for sale, and a buyer asked how many shutter actuations it had (less than 1900, barely used). But I was stymied at first in determining the answer to this basic question; the explanations I found on the web were vague and when I tried GraphicConverter v5.6.2 (one recommended solution), I could not locate the information at all.

So I turned to Phil Harvey’s ExifTool. It is a command line tool (like my own DiskTester and diglloydTools), but don’t be scared off by that. For detailed information and/or repeated use, a command line tool is much more useful than a GUI.

Read ExifTool and Digital Image Information (nerds only).

Was it a good 2007 blog?PERMALINK

This blog and the free articles are provided as a Service to photographers everywhere, and I hope you have enjoyed it. The best way to encourage more of the free stuff is to buy one of the paid reviews—a win/win.

Photo practicePERMALINK

Shooting mundane subjects is a good way to hone one’s eye, and five minutes with the banale is a challenge—make the best of it on a regular basis and something really nice will eventually result at an unexpected time. Here is my morning’s 5-minute effort.

Canon EOS 1Ds Mark III + 14/2.8L II @ f/11

Canon EOS 1Ds Mark III + 14/2.8L II @ f/11

Persimmon bountyPERMALINK

Persimmons produce their stunning bounty every year, but bear especially heavily in alternate years. This year (2007) was unusually prodigious; the tree seen below lost a large limb due to the weight of the persimmons tearing it off in a windstorm. Several hundred persimmons were already harvested prior to this photo being taken. Production was likely well over 1000 persimmons on this one small tree. Amazing really.

Canon EOS 1Ds Mark III + EF 14/2.8L II @ f/11

The zipper effectPERMALINK

I first reported on what I called the “zipper effect” way back in 2005 in my D2x vs EOS review (still well worth reading). Call it the “zipper effect”, the “jaggies” or the “staircase effect”—it isn’t pleasing—but might not matter much for prints.

Mouse over the example below. The original was converted in Canon’s Digital Photo Professional with sharpening level 2. Mouse over to see what happens after applying Photoshop Smart Sharpen {200, 0.5}, a level of sharpening that is not at all unreasonable for a large print. Observe especially the shadows at left with the striped appearance—rather odd, probably an artifact of today’s sensor technology—scanned film won’t show such striping. Yet another reason a non-Bayer (non-interpolated) sensor would be a Very Good Thing.

Canon 1Ds Mark III, actual pixels

Does it matter? Probably not at all in most size prints of most subjects, but there are always exceptions. I printed a section equivalent to a 36" X 24" print at 156dpi on my Epson R2400. The staircasing effect was visible upon close inspection, in fact unreasonably close inspection (and even more noticeable in the roof tiles). Such digital artifacts do have a visual impact even if they are not directly perceived. With luck, future sensors will move away from the path of ever-higher resolution, and instead move towards elimination of such artifacts, which will heighten the sense of “alive” in photographs.

Hard drive failure—emails lostPERMALINK

A hard drive failure might have resulted in losing some emails late Sunday December 16th. If you sent email to diglloyd.com between 15:00 and 23:00 please resend it if you received no response.

Digital SLR resolution — the futurePERMALINK

Digital photographers have never stopped asking for more resolution. The 21-megapixel Canon EOS 1Ds Mark III has now reached a point where that need has largely been satisfied. Improving upon 21 megapixels means eliminating the sharpness-robbing anti-aliasing (AA) filter that both Canon and Nikon insist on using (an AA filter is either not present or is optional with medium format backs, one reason for their high resolving power). Eliminating color filter array interpolation would also result in a leap forward in image quality, but doing so remains an elusive goal. Finally, some responsibility remains with the photographer—absolutely spot-on focus and optimal aperture (f/5.6 - f/8) are required—nothing but perfect technique with top lenses will do.

Having examined the 21 megapixel images produced by the Canon 1DsM3, my conclusion is that lens performance is a limiting factor (including diffraction beyond f/5.6), but that there do exist high performance lenses which still have performance in reserve of and in excess of 21 megapixels, even across the entire frame. Such lenses include the Zeiss ZF 100/2 Makro-Planar, the Leica R 180/2.8 APO-Elmarit-R and certainly some Canon and Nikon optics as well.

A “prime” lens means a fixed focal length—not a zoom. Zoom lens performance cannot be optimal at every focal length; a prime lens is optimized for a single focal length and focusing distance (assuming top-notch design and assembly) .

Assuming the elimination of limiting factors like the AA filter, improved sensor technology, and evolution of lens technology, digital SLRs in the neighborhood of 36 megapixels might provide satisfying image quality. However, such a camera will require a world-class lens that must not be stopped down beyond f/5.6 (or maybe even f/4), making real improvements in detail elusive for practical photography. See for yourself in Diffraction—Canon EOS 1Ds Mark III.

Click to read article

Digital file sizes — be careful what you wish forPERMALINK

For years, the complaint that we’ve had as digital photographers has been: “more resolution”. For most photographers, 21 megapixels is the end game; higher pixel quality is more important.

Pixel quality can continue to improve, and the ever-present anti-aliasing filter remains an annoyance for many types of photography—one that the Leica M8 has dispensed with and one reason its images fare so well in comparisons. Maxmax.com will modify certain cameras, such as the Nikon D200 and Canon 5D to remove the AA filter, their so-called “hot rod” modification.

The 21.1MP Canon EOS 1Ds Mark III delivers RAW 14-bit (CR2) files ranging in size from 20MB to a whopping 37MB. By comparison, 14-bit lossless-compressed Nikon D3 NEF files generally range from 14-16MB, but can be as large as 20MB. Larger files result from images that contain a great deal of fine detail; images that contain large chunks of blue sky or other smooth or blurred areas will be much smaller.

It’s not just storage size either—Canon EOS 1Ds Mark III files take substantially longer to view and process in Canon’s Digital Photo Professional. And one need only open a 6MP Nikon D70 image to realize how much faster RAW-file processing is; a 12MP camera demands twice the processing power of a 6MP camera. So consider carefully whether you really need something like a 21MP camera—there is a pronounced lag in flushing files to the storage card on the 1DsM3, even with the premium SanDisk Extreme IV cards.

Both the Canon EOS 1D Mark III and 1Ds Mark III offer an “sRAW” option (“small RAW”) which is 1/2 the resolution (1/4 the size). That reduction is too large compromise for many types of shooting; an additional choice that yielded a 71% (1/2 the size) image would be more useful. The Nikon D3 offers 12MP, 6.8MP and 3.0MP options, but only for JPEG files, failing to address the issue for RAW. Update: a reader suggests that the sRAW feature might offer approximately 1 stop additional dynamic range. I haven’t tested or verified that claim, but a 5MP resolution and additional dynamic range might be quite useful for certain types of photography eg photojournalism or weddings, where large numbers of images are taken.

Canon EOS 1Ds Mark III vs Nikon D3—resolutionPERMALINK

I’ve now made three different A/B field comparisons between the 12.1MP Nikon D3 and 21.1MP Canon EOS 1Ds Mark III. In all three cases, the results seen are highly consistent, with the 1DsM3 showing higher resolving power. The image shown below is the full frame image whose crops are shown below.

Canon EOS 1Ds Mark III + Zeiss ZF 50/2 Makro-Planar @ f/8

The entire aperture series from f/2 through f/16 was shot on both the D3 and the 1DM3 using the Zeiss ZF 50/2 Makro-Planar, a lens of impeccable sharpness. The shallow depth of field at aperture f/2 was used to verify matched focus, critical to making a valid comparison (both cameras were focused with their respective Live View features). Two other A/B comparisons were also made at a focal length of 14mm (different lenses); one of these is shown further below.

Aperture f/8 is optimal on both cameras—at f/11, diffraction begins to degrade the image in a small but noticeable way; therefore the f/8 images were used. Both images were processed at Neutral with all sharpening disabled (Digital Photo Professional, Capture NX). The Nikon D3 image was up-rezed to the same pixel width as the 1DsM3 image (5616). A Photoshop CS3 Smart Sharpen of {100%, 0.5} was then used on the L channels in Lab mode. Note especially the detail in the tree bark and moss.

Canon EOS 1Ds Mark III <=> Nikon D3
(Mouse-over the image to see it switch between cameras)

Even in the out-of-focus background areas, the difference shows, though it’s less important:

Canon EOS 1Ds Mark III <=> Nikon D3
(Mouse-over the image to see it switch between cameras)

The Canon EOS 1Ds Mark III is the resolution champ. Below is one more example, cropped from the image shown yesterday. Focus matching was done as before to ensure a valid test. The Canon 14/2.8L II was used on the 1DM3 and the Nikon 14-24/2.8FX was used on the D3. Processing was as with the example above, except that the D3 image was given Smart Sharpen of {200, 0.5} and the 1DM3 image was given only {100, 0.5}; the D3 images stand up much better to sharpening, but also requires more of it. (The per-pixel quality of the D3 seems especially high).

Canon EOS 1Ds Mark III <=> Nikon D3
(Mouse-over the image to see it switch between cameras)

There are other approaches to upsampling and sharpening; these examples were done with the simplest approach. Upsampling using Genuine Fractals was also tried, showing the same gap in resolution. Photographers looking for a substantial resolution boost have a clear choice in the Canon EOS 1Ds Mark III, but what else could be expected given 75% more pixels? Other imaging qualities (noise, color, etc) might make the Nikon D3 more appealing for some applications. Those factors take more time to evaluate.

Canon EOS 1Ds Mark III vs Nikon D3PERMALINK

Does anyone really expect that a 21.1-megapixel camera would not handily outperform a 12.1-megapixel camera in terms of resolution? Field shooting today in multiple trials leaves no doubt, but there is considerably more to the story than just resolution. Stay tuned for a body-to-body comparison of the Nikon D3 with the Canon EOS 1Ds Mark III (the resolution aspect).

Nikon D3 + 14-24/2.8FX @ 14mm

Some will cry “foul!” that comparing the two cameras is wrong because of the mismatched resolution, but it’s just as reasonable as comparing a DX sensor camera to a full-frame camera as I did in D2x vs EOS (still highly instructive even if those cameras aren’t on your shopping list). For anyone considering a full-frame digital camera, The D3 and 1DsM3 are the two contenders, and there are numerous aspects to “performance” beyond resolution. Anyway, one is free to not read anything that disrupts cherished cognitive commitments.

Nikon D3 low light performancePERMALINK

Shooting the Nikon D3 in low light is more feasible than with any camera I’ve used before, and mating it to the Zeiss ZF 35/2 Distagon works great as an available-light combination. Sharpness is amazingly high considering the 1/40 sec shutter speed (handheld), and f/4 aperture.

Nikon D3 + Zeiss ZF 35/2 Distagon, 1/40 @ f/4 handheld, ISO 1600

Canon EOS 1D Mark III sub-mirror fixPERMALINK

In my Nov 1, 2007 commentary on the Canon EOS 1D Mark III sub mirror fix, I wrote “...customers deserve a substantial and tangible acknowledgement of the inconvenience and expense. Doubling the warranty to two years instead of one year would seem appropriate”. Canon has apparently following through to some degree, apparently extending the warranty to one year beyond the repair, at least in some cases—whether this is a general policy is unclear, but I’ve asked Chuck Westfall of Canon to clarify.

Update: Chuck Westfall of Canon USA confirmed today the warranty extension as follows: “Canon USA is extending the warranties of EOS-1D Mark III cameras repaired for AF mirror adjustment”.

Please support free contentPERMALINK

Countless hours go into shooting, analysis and writing for this blog and the free articles—please support this site by using the B&H Photo and/or Amazon links at the top of this page (and elsewhere)—make it a habit. When you buy something (anything) using those links, this site gets a commission at no cost to you. It’s a win-win situation that might eventually make it possible to provide the in-depth formal reviews free of charge.

21 megapixelsPERMALINK

It seems a shame to shoot 21 megapixels and then leave 19 of them blurred, but megapixels do not a photo make. The new Canon EOS 1Ds Mark III offers the highest resolution yet seen in a digital SLR, and it will satisfy even a few megapixel-crazed photographers. This photographer expects to be happy also, but more equipment sales are now required to fund this new acquisition.

Canon EOS 1Ds Mark III + Zeiss ZF 100/2 Makro-Planar

Canon EOS 1Ds Mark III + Zeiss ZF 100/2 Makro-Planar

Nikon D3 noise behaviorPERMALINK

The noise levels of the Nikon D3 are extraordinary by digital SLR standards. But higher ISOs such as ISO 25,600 can be usable but less than pleasing. See for yourself in Nikon D3: Noise (part 1).

Nikon D3 at ISO 25600 (click for more)

I’ve collected the Nikon D3 information this site into a Nikon D3 Table of Contents page. Future D3 comments/articles will be found referenced from there.

The Balclutha—sailing cargo ship extraordinairePERMALINK

If you visit San Francisco, be sure to visit the historic Balclutha, one of the last cargo ships to use the wind for propulsion (wind technology might experience a resurgence soon as a supplement to bunker-fuel). Triple-masted, and built for trips around Cape Horn through 70-foot waves, she was a ship that did not suffer fools or cowards.

Keeping in mind that precise focus is uncertain, and there is atmospheric haze, examine the actual pixels extreme corner crops—they’re awfully good considering the range of focus and that it’s a full-frame camera. Click the image below to see them.

Nikon D3 + 24-70/2.8G, 1/2500 @ f/8, ISO 400

Moored permanently at the Hyde Street Pier in San Francisco (Hyde Street and Jefferson near Ghiradelli Square), she bobs idly while awaiting visitors who would never consider eating salt pork or hardtack.

Nikon D3 + 24-70/2.8G @ 27mm, f/5.6 @ 1/1250 sec handheld, ISO 400

Nikon D3 virtual horizon featurePERMALINK

The Nikon D3 “virtual horizon” feature apparently has limited accuracy—pay close attention and don’t throw away that bubble level just yet!

Rotating the image below by 1° counter-clockwise makes it perfectly level. While 1° doesn’t sound like much, it is obviously “way off” to the human eye. To be fair, in other shots it was only off by 0.1° and 0.4°. The 1° figure might be a worst case; Nikon should tighten up the tolerances for what is considered level.

Nikon D3 leveled using “Virtual horizon” feature

Nikon D3 leveled using photographer’s eyeballs

Nikon D3 examplesPERMALINK

I’m starting to really like the D3 images which are extremely smooth and delicate, indeed stunning now that I’ve dialed in appropriate sharpening. Color can be astonishingly pleasing with some images, and not as pleasing with others—I’m still trying to figure that out. And it’s very clear that the D3 images have incredibly high pixel quality very receptive to further processing.

I shot some personal images (family) this morning which I won’t be posting, but I’m extremely impressed with the “poise” of the images, their depth and contrast and realism. Part of it might be the astonishing Nikon 24-70FX zoom, which I’m almost prepared to call the best zoom I’ve ever used, even though my copy appears to be “off” slightly along the left edge.

We had a beautiful white frost this morning in the San Francisco Bay Area. Below is one of my favorite fruits, a Fuyu Persimmon in the blue light of morning. All of these snapshots were taken handheld, perhaps a tripod could have extracted better detail, but I’m not going to restrict myself to posting tripod-only examples, and heck I think they look darn good. No post processing was done other than modest sharpening; additional effort could make them really come alive.

Some photographers sharpen their images for an at-first-glance “zing”; I don’t like such sharpening because subtleties of tone are lost. Evaluating images is best done with minimal processing and sharpening.

Nikon D3 + 24-70FX @ 70mm, f/6.3 @ 1/200 sec, ISO 800, handheld

Nikon D3 + 24-70FX @ 52mm, f/5.6 @ 1/320 sec, ISO 800, handheld

Nikon D3 + 24-70FX @ 24mm, f/8 @ 1/80 sec, ISO 800, handheld

Gigapxl projectPERMALINK

Some folks are really serious about resolution. Have fun reading about the Gigapxl project. A custom camera, custom-designed lens, and huge film record unbelievably detailed images, though the claims are somewhat outdated (“600 prints from a digital SLR”). Have fun with some of the Google Earth images too.

The new 21MP Canon EOS 1Ds Mark IIII should allow about 100 images to be stitched together for a gigapixel image. The Zeiss ZF 100/2 Makro-Planar would likely be an excellent lens for such an application. I think I’d be perfectly happy with something more manageable though, perhaps 100 megapixels.

Multi-core computersPERMALINK

Even low-end computers have dual CPU cores these days, and higher end units like Apple’s Mac Pro have four or even eight CPU cores. Why don’t we see huge speedups with these machines? Read my Multi-core Computers article to understand why.

Poor CPU utilization in Photoshop CS3

Nikon D3 image qualityPERMALINK

I’ve now had time to shoot some objective tests with the Nikon D3 and Canon EOS 1D Mark III. It will take many hours to analyze all the differences (color, dynamic range, detail, noise, etc) but my view now is that Canon 1D Mark III color rendition can be more or less pleasing than the Nikon D3, depending on subject matter and white balance and RAW-file processing—the two cameras are like two different films. Accumulating more experience will clarify the color issue.

I was in for a real treat today around lunchtime—just look at this stuff. This image is stunning with even a little sharpening, and could certainly make a first-rate 24 X 16" print, and it probably would hold up extremely well at 36 X 24". I scaled it up 200% using Genuine Fractals 5.0 and was extremely impressed with the integrity of the resulting file.

Nikon D3 + Zeiss ZF 50/2 Makro-Planar @ f/11

The way the D3 NEF images are processed results in widely varying contrast and color rendition (this is also true of the 1DM3, but the variants are so obviously awful in Canon’s Digital Photo Professional that I stick to Neutral or Standard). Nikon’s “D2x mode” Picture Control settings produce quite different results as well. I’ve now settled on very simple settings in Nikon Capture NX, and I’m liking the results a lot more: Picture Control=Neutral, White Balance=Daylight (or gray patch), Auto Color Aberration, everything else turned off.

I’m going to go out on a limb and make some statements here, some of which are not yet proven to my satisfaction, but which I suspect will prove to be true, or at least not very false (image quality is subjective and multivariate, not a binary thing).

1) Superb image quality on a per-pixel basis. Initial disappointments (due to a complex mix of lens/color/sharpening/focus issues) have given way to the view that the Nikon D3 offers superb per-pixel image quality, possibly the best yet seen in a digital SLR or at least as good as anything else. This includes astonishing quality when scaled up, probably due to its 14-bit pixels with exceptionally low noise and smoothness.

2) Beautiful results for black and white shooters. I’ve only done a few conversions, but the results suggest unusually smooth and tonally-beautiful results (color rendering issues are of course avoided in B&W). Confidence level: high.

3) Focus matters. Getting top results from the D3 means using its Live View at maximum magnification and nailing the focus precisely, and choosing the focus point intelligently so as to render the scene as a whole optimally. This presents a challenge in many situations. Best results demand technical perfection, just as with a 4X5 view camera. The results are commensurate with the care taken.

4) Lenses matter. I was astonished at how much better the new 24-70FX is over the 28-70 AF-S (outside the central 2/3 area), even though my 24-70 seems a bit blurred on the left side (see Brand-new Blur). The 14-24 is also impressive.

5) Useful features. The Canon 1D Mark III Live View looks lousy and feels like yesterday’s technology compared to the D3 implementation. Nifty goodies like the Virtual Horizon feature in the D3 are very useful, and Nikon’s intelligent mirror lockup implementation is useful constantly. On the flip side, I prefer the Canon menu layout; the D3’s layout hides too many items, forcing too much scrolling. There is much more to be said on ergonomics, but those are a few highlights.

Followup on Nikon D3 aliasingPERMALINK

Following up on Monday’s entry on aliasing, it appears that the ugly horizontal bars seen are the result of Capture NX demosaicing algorithm; processing the same image with Adobe Camera RAW does not result in the ugly horizontal green stripes, but it does with RAW Developer.

ACR, RAW Developer and Capture NX also have radically different ideas about color and contrast, all images were processed at 5500°K with minimal settings and sharpening off (Smart Sharpen done later in CS3), no attempt was made to match them further in terms of contrast or color.

Click to see larger images/crops. The Capture NX and RAW Developer results look troublesome for prints of even moderate size; such artificial-looking results are easily detected by the eye. Of course there are techniques for reducing the impact of such nasties, but extra work is required. For this image the Adobe Camera Raw results show the least amount of undesirable artifacts.

Nikon D3 sample image, crops taken at upper left

      
Capture NX, Adobe Camera RAW, Iridient RAW Developer 1.7.1
Observe horizontal green stripes in Capture NX image at left

Nikon D3 with 24-70 and 14-24 and morePERMALINK

I’ll post some examples soon taken with Nikon’s new 24-70/2.8 and 14-24/2.8. Both lenses are impressive performers, marked improvements over previous efforts in that zoom range, at least in the f/5.6 - f/8 range, where I made several hundred images today. If you’re shooting a Nikon D3, these should be considered must-haves. The 24-70 is far better than the older 28-70 AF-S on full-frame images; both lenses seem to have been designed from the ground up to “sing” on full frame. These lenses are going to make a lot of D3 users very, very happy. It’s a shame that both of them lack aperture rings, rendering them useless on Canon bodies. Both lenses are also very substantial items—they look serious and you’re not going to do any surreptitious photography with them.

D3 image quality—what I’m finding is that being much more aggressive on sharpening (level 4 or 5 in Picture Control) is helping mitigate the lower-contrast edge detail when the images are enlarged somewhat. This is really different behavior from the Canon EOS 1D Mark III, which manages to produce high-contrast on very fine details with minimal sharpening (along with extremely pleasing color and je ne sais quoi). Two very different sensors and electronic designs no doubt. I prefer doing the least amount of sharpening necessary, but the D3 characteristics might mean only that I have to readjust my expectations, not that something is “wrong” with the images. It might also mean that D3 images can withstand more manipulation and greater enlargement, but that should be understood as speculation.

I’ve tentatively concluded that the D3 can produce significantly more detail than the Canon 1D Mark III on real images (ignoring resolution charts), but the amount and type of sharpening will determine whether images are persuasive to that argument or not. Technical issues (lens, focus, etc) can dwarf any differences in detail rendition—exercise skepticism with all A/B comparisons, especially those that do not establish how focus was matched.

Human visual perception is a complex area which depends on a multitude of factors, including macro and micro contrast, color rendition and accuracy, smoothness, etc. My response to Canon EOS 1D Mark III images is one of delight; my response to Canon EOS 5D images is quite the opposite. The Nikon D3 is going to delight a lot of people, but my personal visual perception places it between those two Canon models; the D3 images do not please me as readily as the Canon 1D Mark III images do (the opposite could be true for you). I think my reaction is due both to the way in which finer details are rendered (micro contrast) as well as color rendition. Perhaps over time RAW-file processing parameters will be found that make the D3 produce results more to my liking, but my current assessment is that the style of the D3 images is simply different, and inherent to the camera/sensor/electronics. (For that matter the style of the 21MP Canon EOS 1Ds Mark III might turn out to be significantly different from the 10.1MP 1D Mark III!).

Making an analogy to an historical relic (celluloid film)—if you prefer Fuji Velvia over Fuji Provia or Fuji Astia, or slide film over negative film, realize that cameras have differences just as film did/does. It would be a boring situation if all digital cameras produced the same style of images, and it would be a mistake to assume that your personal response is necessarily the same as mine, or anyone else’s.

Nikon Capture NX sharpening bug!PERMALINK

I’ve found a blatant bug which explains why some Nikon D3 images processed in Nikon Capture NX 1.3.0) go soft (Mac OS X 10.4.11, MacPro). Let your own eyes see the story—mouse over the image below; the only difference in processing is that the blurry one was processed with exposure compensation of +1/2 stop. Apparently, using exposure compensation in Capture NX disables any sharpening set in Picture Control (set like this no sharpening will happen). Perhaps there are other such combinations that defeat sharpening.

Nikon D3, actual pixels. Mouse over: f/2 <=> f/2 +1/2 stop

Nikon D3 —California Buckeye nutsPERMALINK

I had only 20 minutes to shoot some backyard snaps today as Life has decided that the week I received my D3 would be the most constrained time in months in terms of free time to shoot!

These images of Aesculus Californica nuts (Sept 23) were taken with the Zeiss ZF 100mm f/2 Makro-Planar, one of the very finest lenses I’ve had the pleasure of shooting (see Zeiss ZF Lenses and also Sept 28 blog entry). The 100 Makro has oodles of reserve resolving power, well beyond what the D3 is capable of.

Depth of field with the D3 is extremely shallow, yet stopping down beyond f/8 shows reduced contrast due to diffraction. Fortunately, shooting wide open with a lens like the 100/2 Makro Planar can be a purely artistic choice in terms of lens performance; depth of field is part of that artistic choice (and the Makro-Planar is unique among macro lenses in offering an f/2 maximum aperture).

Images were processed in Nikon Capture NX, using Picture Control Sharpening = 5, fairly strong, with a modest sharpening for the full-frame downsized samples (no further sharpening done on the actual-pixels crops). D3 images need strong sharpening.

Mouse over the image below to see f/2 vs v/2.8. Then click on the image to see more, including actual-pixels crops. Ask yourself if the f/2 aperture is worth it—I think it yields a distinctly different possibility than f/2.8, something not available with Nikon or Canon 100mm macro offerings. If the Zeiss ZF lenses interest you, then see Zeiss ZF Lenses.

Nikon D3 + Zeiss ZF 100/2 Makro-Planar, ISO 200, f/2 <=> f/2.8

Nikon D3 thoughtsPERMALINK

There are objective tests, which take time and study, especially with the D3, which seems to be a wholly different animal than previous Nikon (or Canon DSLRs). Objective tests are tedious and boring to shoot, often precluding interesting work because cameras and lenses need to be coordinated. On the other hand, they provide a reliable way of highlighting real differences between cameras, and are thus a necessary effort.

There are also gut reactions, which I mentioned in my first entry on the D3. I can’t say I’m “wrong” about a gut reaction any more than I’m wrong about liking chocolate ice cream over vanilla, and I’m still trying to figure out (objectively) what I’m reacting to in the D3 images. It is accurate to state that I haven’t experienced the “wow” that I did with my first Canon EOS 1D Mark III images (see June 17 and May 24 entries, among others). Ferreting out why this should be so is no easy task.

Making a purchasing decision based on my gut reaction might not be the best plan, so please don’t cancel your D3 on my account (as one reader did). We all see images differently, and one person’s yuk might be another one’s yum.

Now to some subjective impressions, not subject to the rigors of tests (yet):

• depth of field seems perilously shallow on the D3, curiously so. It might be related to the way the D3 “draws” more so than an actual difference, a perceptual effect (I’ve previously owned the EOS 5D and 1Ds Mark II, so I’m not a stranger to full-frame cameras);
• noise control might be overhyped; from what I’ve seen so far, I don’t really want to to shoot beyond ISO 6400. Usable yes, desirable no.
• The D3 LCD screen is phenomenal. It has to be seen to be understood; images displayed on it look incredible;
• The D3 “electronic rangefinder” focus assistance doesn’t deliver the accuracy needed for critical work; I can do better with my eyes. This is no change from previous models;
• Top image quality outside the central 2/3 area requires stopping down to f/5.6 or f/8 with most lenses. No surprise here, it just means that lens selection is critical for exploiting the sensor resolution, especially at wider apertures.
• Perversely, could some lenses be “too sharp” for the D3? Delivering high resolution at high contrast beyond the Nyquist frequency of the D3 might induce aliasing;

Shape and colorPERMALINK

Depth of field is perilously thin on the full frame Nikon D3. This scene grabbed my attention for its unusual colors and geometric shapes. Click to see a large version and actual pixels.

Nikon D3 + Zeiss ZF 50/2 Makro-Planar, 1/80 @ f/2, ISO 3200

Nikon D3 aliasingPERMALINK

The D3 images I’ve taken so far made me suspect a too-strong anti-aliasing (AA) filter. But that seems inconsistent with the strong aliasing in images like this one, similar to what I’ve seen with medium-format digital backs that lack an AA filter. The pattern seen in this actual pixels crop occurs over large areas of the image, though it comes and goes according to the detail level, ranging from no effect at all to prominent aliasing (horizontal short green stripes and colored speckles throught). The Nikon Capture NX “Color Moiré Reduction” has no effect on it.

Nikon D3 + Zeiss ZF 50/2 Makro-Planar, actual pixels, ISO 800

Nikon D3 and 180/2.8 IF-EDPERMALINK

Bjørn Rørslett of naturfotograf.com reports seeing fewer chromatic errors than expected when using many lenses on the D3. I believe he’s talking primarily about red/cyan color fringing, seen in lenses that are not fully corrected for color errors. This type of color fringing is quite common, and is technically known as lateral chromatic aberration, manifesting itself as a red fringe on one side, and a cyan fringe on the other. There also exists longitudinal (axial) chromatic aberration (generally observed as a green or magenta halo around bright areas); it tends to improve upon stopping down.

The Nikon 180 f/2.8 IF-ED on the D3 shows both types of color errors. The most prominent in the example below is what might be written off as “sensor blooming”, but it is most likely longitudinal (axial) chromatic aberration; it diminishes from f/2.8 to f/8. No evidence of such effects was seen using the Voigtlander 180mm f/4 APO—so it seems fair to say that the Nikon 180/2.8 ED is definitely not apochromatic, a finding I previously confirmed in comparing it to the Leica 180/2.8 APO-Elmarit-R.

Mouse over each image to see it as processed using the Nikon Capture NX “Auto Color Aberration” feature. The first two images are closer to the optical center; the 3rd one is nearer the edge of the frame. (JavaScript must be enabled in your browser for mouse-overs to work).

  
Nikon D3 + 180/2.8 IF-ED, mouse over for ACA

Nikon D3 + 180/2.8 IF-ED, mouse over for ACA

Also seen is lateral chromatic aberration; mouse over to see how nicely the “Auto Color Aberration” feature works—color fringing is eliminated.

with <=> without ACA
(Nikon D3 + 180/2.8 IF-ED)

Nikon D3 focus problemsPERMALINK

Today I encountered a focusing problem that could be reproduced at will with either the Nikon 105mm f/2 DC or the 180/f2.8D. The D3 thinks the frames below are in focus; the focus confirmation light inside the viewfinder goes on (solid), and the focus confirmation beep is heard. (Both of these lenses are the klunky old style focusing ones, not AF-S; I haven’t observed any problems with AF-S lenses.)

Nikon D3 + 105mm f/2 DC. Autofocus, full frame.

Nikon D3 + 180mm f/2.8 ED-IF. Autofocus, full frame.

Turning the D3 on/off does not fix the problem, and as far as the D3 is concerned the image is in perfect focus—forever. Only by switching to manual focus, then refocusing farther away, then switching to autofocus again does the D3 come out of its stupor. Maybe this is a general problem, or maybe it’s just my particular D3.

I also observed focus accuracy errors with both lenses; forcing refocus repeatedly on a tripod-mounted camera resulted in variable sharpness ranging from spot-on to noticeably blurred. Sorry to burst anyone’s bubble, but neither lens is acceptable for work where reliable sharp focus is required (see also The Sharpest Image). The 180/2.8 is brand-new; the 105/2 DC is a few years old.

Nikon D3 looks goodPERMALINK

Initial concerns about the D3 were apparently misplaced. There is a lot going on with it as a camera—lens performance is critical, depth of field is frightfully shallow, and the files are subtle and delicate, requiring more attention to sharpening than the Canon EOS 1D Mark III. Confusing the issue is the truly awful interface of Nikon’s Capture NX, sure to put off anyone who wants a quick and easy workflow, though RAW-file processing quality is excellent. Even Canon’s Digital Photo Professional interface seems like nirvana compared to Capture NX.

The most striking thing I’m finding is that the D3 reminds me more of the Hasselblad 503CWD than any DSLR I’ve seen before (that is “reminds”, not “equals”). Its 14-bit pixels are almost identical in size, so that theory is at least plausible. Still, the sharpness doesn’t seem quite “there” to me. Perhaps it’s lens-dependent in a oddball way.

The image below is consistent in sharpness with the other images I’ve taken. Click it to see an actual-pixels crop. The image is cropped at left and bottom; the chain-link fence through which it was shot intruded at the sides a bit. Processed through Nikon Capture NX using Picture Control Sharpness = 4. No further processing was done. Saved for web at quality 70 from Adobe Photoshop CS3.

Nikon D3 + Zeiss ZF 50mm f/2 Makro-Planar, 1/500 @ f/8 handheld, ISO 400
(image above is cropped, not full frame)

The overall contrast and depth/believability and color of the D3 images shows great promise, so let us not forget that absolute detail is only one factor of many that make a quality image.

The image below was shot handheld at 1/60 sec @ f/4, so depth of field and sharpness are limited. But the color rendition should be of interest.

Nikon D3 + 50mm f/1.4D, 1/60 @ f/4 handheld, ISO 200

Focus lock and recompose—unlearn it nowPERMALINK

The so-called “focus lock and recompose” technique should be unlearned in the digital age, especially on a full-frame camera like the Nikon D3. The technique consists of focusing on the desired subject matter, then recomposing the image (shifting the view) somewhere else. This technique is a sure recipe for making the desired subject matter go blurry—not just because the plane of sharp focus shifts away from the subject, but also because field curvature might come into play. Field curvature is more common than generally realized, and it’s present in one of my favorite lenses, the Zeiss ZF 25/2.8 Distagon. Stopping down helps, but stopping down ruins any photo which makes intelligent use of zones of sharpness and unsharpness.

When life gives you lemons, maybe there’s a chocolate chip cookie waiting for you!

A soft cookie

Of course, if one is shooting a camera like the Leica M8 with its single center-focusing area, there is no choice—meaning time and experience are required to avoid soft photos shot at wider apertures and/or non-optimal placement of the zone of sharp detail.

Nikon D3 meteringPERMALINK

The Nikon D3 suffers from the same problem that every other DSLR I’ve used does: the inability to meter correctly even for dominant highlights, particularly the red channel in certain flowers. The image below has enough of the bright yellow daisies that one might reasonably expect their exposure to be dealt with appropriately. Why isn’t there a “don’t blow more than 2% of the highlights” mode? Someday there will be, or new technology will eliminate the issue.

In this case, the red channel could be rescued in Capture NX using -2/3 stop exposure compensation (and one stop in ACR, not really worth bothering with for this image, but it makes for a good example). Mouse over to see the Capture NX versions at -2/3 stop compensation. Note also how Capture NX has eliminated the red fringing along the edge of the petals.

in-camera JPEG <=> Capture NX @ -2/3 exp comp
(Nikon D3 + 50/1.4D @ f/8, ISO 200, actual pixels)

in-camera JPEG <=> Capture NX @ -2/3 exp comp
(Nikon D3 + 50/1.4D @ f/8, ISO 200, full-frame)

Nikon D3 with 10.5mm f/2.8 DX NikkorPERMALINK

The movie Fargo has a hilarious scene in which involves a certain disabled person and his pants. I’ve got to keep that in mind when shooting the Nikon 10.5mm DX (2/3 frame) on the full-frame Nikon D3, which can easily record whether or not...well, you get the picture! Not wishing to incriminate myself, I’ll just show this picture instead.

Nikon D3 with 10.5mm f/2.8 DX (full frame)

The 10.5mm DX is designed to cover a 2/3-frame sensor camera like the Nikon D200, D40/x, etc. On a full frame camera, the lens shade intrudes into the frame. Bjorn Rorslett (naturfotograf.com) reports having sawed off the lens shade and finding that the resulting (unsaleable) lens can actually see behind itself (about 200°). Now where’s that hacksaw...!

So—you can either spend US$5000 or so on a superb (but huge and heavy) 7.45mm fisheye from Coastal Optics (see Nov 8 entry), or you can get a diminutive Nikon 10.5/2.8 DX and saw off the lens shade for a most-of-the-frame image.

Nikon D3 updatePERMALINK

I made some A/B comparisons this morning with the 10.1-megapixel Canon EOS 1D Mark III, and feel better than I did yesterday about it (ignoring the JPEGs for now). It looks like yesterday’s statements were related in part to lens selection (with some as yet unexplained results), amount of sharpening used, etc. Additionally, the full-frame sensor is definitely a major influence on across-the-frame image quality (eg the lens really matters), as is focus accuracy and DOF and f-stop used—that much was made very clear in one full-frame comparison. It will be much easier to make faithful comparisons with another full-frame camera like the 5D or 1Ds Mark III.

This is a whole new camera from Nikon, with very different behavior—I should have set my expectations more realistically: it’s going to take considerable time to learn and understand how the D3 behaves on its own and in combination with various lenses. Starting with lenses seems a wise approach.

Family matters take me away for the day, but tonight I expect to post more comments and some samples. Regarding my “gut reaction” yesterday, well, that’s why objective A/B comparisons are always better than shooting just one camera by itself. Too many variables with a new camera, but then again gut reactions are fair game if so-stated. I’ll revert to form with purely objective findings now.