D810 sharpness and aliasing

The Nikon D810 has no anti-aliasing (AA) filter. The D800E has a tricky now-you-see-it-now-you-don’t (non-AA?) filter. Is there any difference in the sharpness of the two cameras? How about aliasing and sensitivity to moire?

I mounted a Sigma 50mm f/1.4 FG lens on the 810. I set up an ISO 12233 target. so that the vertical extent was 400 pixels. I lit the target with a Paul Buff Einstein electronic flash set to a little over 4 watt-seconds, yielding a flash duration of about 200 microsesonds. I put the camera on an Arca Swiss Cube using a RRS L bracket.  I mounted the head to a set of stiff RRS carbon fiber legs. I set the shutter delay to 3 seconds and the ISO to 64. I focused wide open, using live view, then stopped the lens down to f/5.6. I connected a synch cable to the PC connector on the camera. I made three exposures.

I opened the lens up, refocused, stopped down, and made three more exposures. Then I did that all over again, for a total of nine images.

I mounted the lens on a D800E, set the ISO to 100, turned the strobe down to 2.5 watt-seconds, and repeated the protocol, giving me nine more images.

I updated Lightroom to 5.6, and ACR to 8.6. Now I don’t need to use the Adobe DNG converter. I cropped to the chart in Lr, and exported all 18 images as layers to Ps.

All the D810 images looked about the same. There was some variation in the D800E images due to inconsistent focusing with the D800E’s brain-dead live view.  I picked what looked like the best D800E image (all three in the set with the same focus were really close), and picked an D810 image at random.

Here’s the whole chart:

 

D810

D810

 

D800E

D800E

There’s not a lot to choose between the two.

Let’s take a closer look at the rosette on the upper left:

D810

D810

D800

D800

The bars opposite the number “4″ are at the Nyquist frequency. Tighter bars are aliased. In both images, you can see aliasing up to “8″ which is the sampling frequency.  This is textbook performance, except that the D800E image is a little snappier for the horizontal lines. That could easily be due to sub-pixel shifts at capture time.

Some people have said that the D810 has more false color than the D800E. Let’s look at the area of the chart with the worst false color artifacts:

D810

D810

D800E

D800E

There’s not a lot to choose between the two images. Note that both cameras are showing aliasing to the right of the “8″ in the lower left corner of the image. That bar marks the 1 sample/cycle point, so the aliasing indicates capture — albeit wrongly reproduced — of target information with spatial frequency greater than twice the Nyquist frequency. This indicates the potential of this lens to work well with a sensor of less than half the pixel pitch of the D8x0 sensor. 200 megapixel cameras? Bring ‘em on.

The ISO 12233 target is much more discriminating than a normal photographic subject. I think the new camera’s resolution, sharpness, and susceptibility to artifacts is essentially the same as the old one, under perfect conditions.

Under not-so-perfect conditions, the improved LV and lower vibration susceptibility of the D810 will make a difference.

 

 

Unintended eclipses with the D810 built-in flash

I suppose that the best flash, like the best camera, is the one you have with you. In that case, there’s a lot to be said for that dinky little flash that lives on top of the Nikon D810′s prism housing.

On the other hand, it doesn’t stick up very far when you turn it on. That causes potential red-eye problems, and also is an invitation for the lens to block the beam. This tends to be more of a problem with zooms than primes, at closer distances, with lens hood attached, and at shorter focal lengths. There’s no sure-fire way to say if a suspect lens will suffer from the problem without testing, Here are a few such tests.

Nikon 17-35 f/2.8 with hood at 17mm focal length

Nikon 17-35 f/2.8 with hood at 17mm focal length

Nikon 17-35 f/2.8 with hood at 35mm focal length

Nikon 17-35 f/2.8 with hood at 35mm focal length

Nikon 24-70 f/2.8, no hood at 24 mm focal length

Nikon 24-70 f/2.8, no hood at 24 mm focal length

Nikon 24-70 f/2.8, no hood at 35 mm focal length

Nikon 24-70 f/2.8, no hood at 35 mm focal length

 

Nikon 24-70 f/2.8, no hood at 70 mm focal length

Nikon 24-70 f/2.8, no hood at 70 mm focal length

Nikon 35 f/1.4, no hood

Nikon 35 f/1.4, no hood

Nikon 24 mm f/1.4, no hood

Nikon 24 mm f/1.4, no hood

Nikon 14mm f/2.8, built-in hood

Nikon 14mm f/2.8, built-in hood

Autofocus on the Nikon D810

I don’t have equipment to quantitatively test the dynamics of autofocus systems.  It has been my experience that the dynamics are what differentiate one phase detection AF system from another, for the most part. Properly adjusted — and, if necessary, tweaked for each lens and distance range — they all seem to focus about as accurately, providing they reliably focus at all.

So the tool that I have to evaluate the improvements in the D810 AF system is subjective comparison. I took two Sigma 50mm f/1.4 DG lenses, and mounted one on a D810 and one on a D4. In a room dim enough that the exposure meter read f/1.4 at 1/20 second at ISO 125, I set both cameras to AF51. I focused back and forth between a near and a medium-distance subject, alternating between the cameras. I also used different focusing points, keeping them the same between the two cameras. I used moderate and low-contrast targets. The D4 was a hair faster, but neither system hunted.

I turned down the lights so the exposure meter read f/1.4 at /10 second at ISO 2000. Same result.

I took the lens off the D4 and mounted it on a D800E. In the brighter light, the D800E was a little slower than the D810, In the dim light, the D800E could not reliably focus at all on some things.

I’d have to say that, for static subjects, the AF  in the D810 is improved over the D800E, and nearly as good as the D4. If it has anywhere near the D4′s ability to track motion, it will be a big improvement.

EFCS results with a 400mm f/2.8 lens on a D810

Next up, in the series of lenses that I’m using for EFCS testing on the D810: the Nikon AF-S 400mm f/2.8 D II. Not a VR lens; the one just before the Canon patents ran out.

I started at f/5 at 1/500 and worked down in 1/3 stop intervals.

The results, measured for horizontal edges:

400 MTF50

 

400 MTF30

 

From 1/250 to 1/125, there’s not much to choose between the two shutter modes. The divergence of the results at 1/500 are a surprise to me, and could be the result of the interaction of vibration with the shutter opening, or could just be experimental noise (although, from looking at the slightly lower shutter speeds from the non-EFCS mode, there’s probably a pattern there).

One of the issues could be that the 3 second shutter delay that’s the most Nikon provides on the D810 isn’t enough for big iron lenses like the 400/2.8. The only way to test that hypothesis is to dig up a remote release. I may do that.

Note the dip in EFCS MTF at around 1/250. That could be noise, or it could be the acceleration of the second curtain.

The EFCS and non-EFCS MTFs show no sign of converging at slow shutter speeds. The oscillations of such a big lens take a long time to damp out. All the more reason to use EFCS and not let them get started in the first place.

 

 

EFCS results with Nikon 70-200 on the D810

I posted sharpness test results with and without EFCS with the Zeiss 135mm f/2 APO-Sonnar here. EFCS seemed to make a significant difference at shutter speeds from a few tenths of a second to a hundreth or so.

Some people have reported no improvement with common zoom lenses. I thought I’d try the AF-S Nikkor 70-200 f/2.8 G II ED, a well-respected and good-selling lens.

One important difference between the mounting arrangements of the Zeiss 135mm lens and the Nikon zoom is that the Zeiss is (marginally) light enough not to need a rotating lens collar and tripod foot, where the 70-200 needs — and has — one. In landscape mode, mounting a lens to a tripod head with the collar and foot is more susceptible to shutter shock than mounting the camera body to the head. The reason is that the shutter’s motion can be directly resisted by the head and tripod if the camera is clipped in, whereas the shutter’s motion takes place at one end of a torsion pendulum if the lens collar provides the attachment point.

The MTF50 results with the 70-200 set to a focal length of 135mm:

Nikon Zoom 135 MTF50

The EFCS helps quite a bit.

The MTF30 graph:

Nikon Zoom 135 MTF30

Pretty much the same story.

Why the gradual improvement in the EFCS results as teh shutter duration gets longer? My guess is that we’re looking at the effect of the acceleration of the trailing curtain.

Why does the mechanical shutter look better than than the EFCS at 1/250? I think that’s just statistical variation.

As usual, don’t try to compare the MTF50 and MTF30 values from one post to another. Focusing errors, lighting differences, camera to target distances, etc, make those comparisons invalid.

EFCS on the Nikon D810

Before I get to how the electronic first curtain shutter (EFCS) on the D810 works, it will be instructive to look at how the all-mechanical focal plane shutter operates.

Here’s a photograph of an analog oscilloscope with the time base set to 0,5 milliseconds/division. EFCS was disabled, and the shutter speed was 1/1000 second:

 

_8100843

 

The apparent advancing of the exposure at the bottom and the top of the image is due to the interaction of finite phosphor decay and the sinusoidal input signal, which excites the phosphors more at the upper and lower parts of the wave, where it is moving more slowly.

You can see that the shutter moves in the opposite direction from the shutter in a Sony alpha 7 camera: from the bottom to the top of the image, or from the top to the bottom of the camera. The fastest strobe synch speed for the D810 is 1/250 second, or 4 milliseconds (ms), so you’d think that the curtains would take about that long to travel across the sensor.

You’d be wrong.

How far wrong? Let’s put some construction lines on the image:

_8100843-Edit

 

It looks like the shutter takes about 5 divisions, or 2.5 milliseconds, to complete its trip across the chip. Nikon must need the other 1.5 milliseconds for something. To allow for unknown delay in the strobe firing circuitry? To allow for radio or optical triggers? Who knows, but it’s there.

Note the time the shutter is open at any point on the image. Two divisions or a bit less, or one millisecond.

Another thing to notice is both curtains of the shutter appears to move at a fairly constant speed: the lines are straight. That will make it easier to design an electronic first curtain.

Now let’s look at the same setup with EFCS on:

_8100865-3

Looks pretty similar, doesn’t it? Here it is with the construction lines:

_8100865-2

We can see that the EFCS faithfully reproduces the effect on the sensor of the mechanical first curtain, as the reset signal that creates the electronic first curtain travels from the bottom to the top of the image in about the same amount of time as its mechanical cousin.

 

NIKON D810 EFCS VIBRATION with a 135MM LENS

I mounted a Zeiss 135mm f/2 APO-Sonnar on the D810 and repeated the tests of the previous two posts.

Here’s what happened, first in landscape orientation, measuring the edges perpendicular to the shutter motion:

Zeiss 135 MTF50

Zeiss 135 MTF30

And in portrait orientation, also measuring the edges perpendicular to the shutter motion:

Zeiss 135 MTF50p

Zeiss 135 MTF30p

EFCS really earns its keep here.

I would expect an even wider divergence with longer lenses.

Nikon D810 EFCS vibration w/ 50mm lens in portrait orientation

[This post, like the previous one, was extensively revised on 7/28/2014, and for the same reason.]

I repeated the tests of the previous post with the camera in portrait orientation. This is a more severe test, since the shutter moves in the left/right direction when the camera is oriented with the long side up and down, and the tripod is not as stiff in that direction.

I measured the MTF of the vertical slanted edges. In the previous post, I measured the horizontal slanted edges. The reason is that the shutter vibration is orthogonal to those edges in both orientations, and therefore those edges are more sensitive to shutter vibration.

Here’s what resulted:

Sigma 50 MTF50p

Sigma 50 MTF30p

In portrait mode, EFCS helps quite a bit over a broad range of shutter speeds.

Nikon EFCS vibration w/ a 50 mm lens in landscape orientation

[Note: this post was extensively rewritten on 7/28. I made the measurements in the original post based on an erroneous understanding of how the EFCS controls on the D810 work., I me-shot all the images to produce what you now see here.]

One of the new features on the D810 is the electronic first curtain shutter (EFCS), which should in conjunction with raising the mirror well before the exposure, reduce vibration.

As I now realize, the EFCS controls on the S810 work quite differently from those on the Sony a7 and a7S. Like the Sonys, there’s a menu item to turn EFCS on and off on the D810. Unlike the Sonys, you don’t actually get EFCS until you set the drive mode to a particular setting. In the D810, that setting is “Mirror up”. (Having no mirror, the Sonys have so such setting.)

In the D810, mirror up drive mode works this way with live view off: pressing the shutter release for the first time raises the mirror. If EFCS has been invoked, the shutter opens. The second time you press the shutter release, the shutter fires — immediately if you have not set a shutter delay, and after 1, 2, or 3 seconds if you have.

With live view on, mirror up works a little differently. Pressing the shutter release does nothing but blank the LCD screen if EFCS has been selected — the mirror is already up and the shutter is open. If it hasn’t, the first press of the shutter release closes the shutter.

The worst part of the system is understanding it or explaining it. In practice, there is an advantage: EFCS can effectively the controlled without going into the menu system. I wish there were a way to control shutter delay the same way. Having an option to have shutter delay onl operate in mirror up mode would do the trick with no more buttons or dials on the camera.

I thought I’d see how much improvement EFCS made with a “normal” lens. I picked the Sigma 50mm f/1.4 DG. I mounted the camera in landscape orientation to an Arca Swiss Cube with a RRS L-bracket, and mounted the head to a sturdy set of RRS legs. Since the shutter in the D810 moves up and down, and the tripod is stiffer in the vertical direction, the landscape orientation is the least vibration-prone way to orient the camera.

I aimed the camera at a slanted edge target illuminated with a Fotodiox 5500K LED flood. I turned EFCS on. I set the shutter delay to 3 seconds. I focused wide open, then set the aperture to f/5.6, the shutter speed to 1/250, the ISO to 125, and made an exposure. Then I turned down the flood (which doesn’t change color temperature as you reduce the light output) 1/3 of a stop, I made exposures with about the same mean sensor level in 1/3 stop intervals at shutter speeds down to 1/4 second.

Then I turned EFCS off and did the whole thing again.

I converted the raw files to DNG with Adobe DNG Converter 8.6.0.250 beta, brought them into Lightroom 5.5, set the white balance to Daylight, cropped to the target, and exported them as TIFF’s. I analysed the modulation transfer function (MTF) of all the images in Imatest.

Here is MTF50:

Sigma 50 MTF50

There’s not much improvement with the EFCS except at shutter speeds between 1/80 and 1/50. Even there the improvement is slight.

Here’s the MTF30 data:

Sigma 50 MTF30

Pretty much the same story.

 

Nikon D810 manual focusing

In making the bookcase pictures for the two previous posts, I got to try out the Nikon D810′s new-and-improved live view.

Wow!

It’s an amazing improvement, not just because it’s so good, but because the D800/D800E live view was so bad. The D810 is right up there with the Sony a7R in the live view department, and that’s a good place to be. I would like to have the Sony’s articulated LCD display, though, not only to be able to conveniently get lower angles, but also so that I didn’t have to extend the tripod so far. If I had a younger back, the latter advantage wouldn’t be as significant as it is.

I also got to try out the manual focusing on the Sigma 50 mm f/1.4 DG lens. I don’t expect much of manual focusing on an AF lens, but the Sigma’s feel is surprisingly good, although the throw is short, like most AF lenses. It’s not up to the standard of the Otus 55mm lens, but what is?

Good job, Sigma.