A Meaningless Experiment

I was setting up a few lights at home to shoot pictures of my daughters, and I took the picture you see above while testing the hair light, you can see it is quite hot (I ended up reducing it by one stop), and what you can't see, is that it is causing lens flare, because it was not gridded.

While importing the pictures to lightroom, I decided to see if I could get any shadow detail out of the file, so I pushed the exposure by 4 stops, and pulled the highlights down, and I got the picture you see below.

Impressive? I thought so, but of course the colors look weird (the background is supposed to be gray, not blue-ish), and I lost some highlight details, but nonetheless I was impressed by what I could get out of the OMD's small sensor, despite the 4-stop under-exposure, and the complete lack of any light except for the hair light.

And before I get attacked by anyone, I do not suggest that you should mess up your lighting and try and fix it in post, it will never come close to a well lit picture, I just did it for fun.

And for more fun, I decided to turn the picture to B&W to get rid of the strange color cast, and apply some strong noise reduction. Below is the B&W version and a 100% crop.

Read More

Nikon D600 + Nikkor 24-70: A Day with My Precious Ones (and my kids too)

Last weekend I took our kids to Disneyland - just the three of us.  It was my first time to bring them there by myself, so instead of my usual two camera combo, I just brought one camera and one lens - the Nikon D600 and the Nikkor 24-70 2.8G.

In this post I will discuss how the combination performed in terms of:
- versatility
- image quality
- autofocus speed and accuracy
- low light performance (autofocus and image quality)
- dynamic range

(Click on the pictures for higher-resolution versions.  FYI no additional sharpening was used - only LR defaults.  All shots were handheld with no flash.  Except as otherwise noted, I did not apply noise reduction, clarity, vibrance or saturation.)
VERSATILITY
I was initially concerned that I didn't have a telephoto lens or ultrawide angle lens with me, but the Nikkor 24-70 2.8G covered many kinds of shots, from scene setters to close-up portraits.  To me, the 24mm had a similar feel to my Tokina 11-16 and the 70mm portraits felt tight enough to isolate the subject.

Traffic Jam.  24mm

36mm. Clarity +15

45mm

56mm

70mm

IMAGE QUALITY
For controlled shots, I found the images were sharp, even wide open at f/2.8.

The 24-70's bokeh is reasonably good in my opinion, though not in the same league as the Sigma 50 1.4:

The one thing that I was not impressed with was the amount of flare I was getting.  Please note that I did have a UV filter on (reviewed here), so I will look into this again in the future, without a UV filter.  UPDATE 10/12/12: I found that the rear element of my lens had a smudge in it.  I have since cleared the smudge, but I suspect the flare that I saw was from the smudge.  Will retest.

AUTOFOCUS SPEED AND ACCURACY
Consistent with previous tests, I found the D600's autofocus to be very good, speedily acquiring focus on active subjects.  Although the focus point was not always perfect, a high percentage of the shots had more than acceptable focus.  As expected, it was a lot easier to get good focus with an aperture of f/4.0 or f/5.6 than f/2.8.

I was surprised that the D600 and 24-70 were able to focus quickly enough to capture fast-moving subjects, even when I had no time to select the AF point.  The ride was at full speed in the shot below and IIRC the camera was on Auto Area AF with AF-C:

LOW LIGHT PERFORMANCE
Although we visited Disneyland in the daytime, we did drift into more dimly lit areas and it also got dimmer toward the end of the day.  On previous trips, I probably would have used flash under the same circumstances but on this occasion, I wanted to see what the Nikon D600 could do.  In my opinion, using high ISO in dim lighting is a more realistic test of a camera's high ISO performance than using a well-lit test scene.  (Note: The scenes below might look brighter than they actually were because I adjusted the contrast.)

From prior testing, I found that the D600's high ISO performance was exceptional for the viewing sizes that I normally use.  I therefore allowed the Auto ISO to choose any ISO, all the way up to 25600.  Sometimes, the camera would choose weird ISO settings like 8063.  Not sure why but it wasn't an issue for me.  Anyway, here are some high ISO samples.  No noise reduction used except as noted.

2800 ISO

8063 ISO

I can play too! 10,159 ISO

One thing that I like about the D600's high ISO is that it can still retain detail.  I prefer a noisier but more detailed image than one that has no noise but also no detail.  Another benefit of the D600 is that because of its high resolution, noise reduction is more effective if and when applied.

20,000 ISO. Noise reduction +30. Saturation -22.

25,600 ISO. Noise reduction +50. Clarity +31. Saturation -20.

I also want to note how well the D600's autofocus performed in these low light conditions.  The last couple of shots above were in a very dim room lit by a low-wattage incandescent bulb.  The last shot was at 25600 ISO, f/2.8, 1/80, or about an Exposure Value of 1.  It was dim enough that it was hard to see into the shadows with my naked eyes.  IIRC the AF assist did not activate because the AF point wasn't in the middle.  Even in such extreme circumstances the D600 and 24-70 focused with no hesitation at all.

DYNAMIC RANGE
One thing that I was looking for in the D600 was its dynamic range.  I have a Fuji S5, which I love, and I wanted a camera that would have a dynamic range that could match that of the S5.

When we visited Disneyland, the sun was blazing, providing a good opportunity to test the D600's dynamic range.  Here are some examples that amply demonstrate the D600's DR:

In this shot, Belle was wearing a satin-like glove which fell under direct sunlight.  Nonetheless, none of the details of the glove were lost.

This following scene was extremely backlit.  It was hard for me to see much more than silhouettes against the sun's glare.  It was the kind of scene that is the specialty of the S5.  The D600 performed just as well, retaining all details with the exception of specular highlights (note that the delicate diffuse highlights, such as the glossy surfaces of Dumbo were not blown out).

Now I have to come clean about a couple of things.  For these DR shots, the straight-out-of-the-camera didn't look like the final images that you see here.  Second, I had to adjust the automatic exposure of some of these shots.  You'll see what I mean when I post in more detail about the D600's dynamic range.  In the meantime, here are more samples.

MORE SAMPLES

Smile for daddy!

Read More

Nikon D600 vs. Fuji S5 Pro highlight dynamic range

Many recent DSLRs from Nikon have been tested to show a very wide highlight range.  For example, DPReview reports that the Nikon D800 has as much as 5 stops highlight range with Active D-Lighting at Extra High.

The D600 looks to have promising highlight range.  The D600 has been tested by DXO to have a dynamic range similar to that of the D800.  In my own testing, the D600 also appears to have a fairly wide highlight range, based on the latitude I have seen for recovering highlights.

In this post, I compare the highlight dynamic range of the Nikon D600 with that of the Fuji S5 Pro, a unique DSLR that has unparalleled highlight range.  I wanted to see whether the Nikon D600 has met or even surpassed the S5's highlight dynamic range.

TEST

I setup several objects of varying degrees of specularity (reflectiveness) indoors.  I prepared three hotshoe flashes - one SB-800 on-camera acting as a master, another SB-800 as a slave, and a YN-565 as a second slave.  All flashes were firing at the same manual power level, bounced toward the ceiling.  I then took shots of the objects at these settings:

- flash only at 1/4 power

- flash only at 1/2 power

- flash only at full power

- full power flash + ambient normally exposed

- full power flash + ambient +1 exposure

All shots were taken at base ISO (100 for both the D600 and the S5) in raw, converted to DNG for processing in Lightroom 4.

NOTE:  In the case of the D600, there is no raw support yet from Adobe so I used the workaround of tricking Lightroom into thinking that the D600 files were from a Nikon D4.  Blogged here.  The resulting files are cropped.  Not only that, the crop is from the top left corner therefore the cameras' perspectives are slightly different.  Moreover, no one knows whether the D600 files with true raw support will be the same or better than these "fake" D600 DNG files.

FUJI S5 RESULTS

First, I took a baseline shot that was deliberately underexposed so that it would fall within the dynamic range of both cameras.  I removed the ambient (100 ISO, 1/200, f/5.6), and fired all three flashes at 1/4 power.

As you can see, detail is visible everywhere.  For example, even the specular highlights of the ribbons are not clipped - they have color.  The metal disc held by the bear has visible detail even with the specular highlight.

Having established a baseline, I then took the progressively brighter shots.  I first examined the brightest shot:

I applied a correction of -3EV to make it easier to observe whether the detail was retained:

It appears that the S5 was able to retain detail everywhere.  I was not able to exceed the S5's highlight range with this test.

D600 RESULTS
As with the S5, I took 5 test shots.  In the first two shots, the D600 clearly was able to retain all detail.

I took a look at the middle shot (flash only at full power):

I applied a -3EV correction to make the details in the specular highlights more observable:

It appears all detail was retained, though the specular highlight of the red ribbon seems almost ready to lose color.

I took a look a the fourth shot.

With a -4EV correction:

It looks like a little detail was lost in the metal disc, in the red ribbon's specular highlights, and above the bear's nose and left paw (camera right):

I also looked at the fifth shot.  As expected, a little more detail was lost:

ACTIVE D-LIGHTING
I also tested whether Active D-Lighting would improve the D600's result.  Note: the slave flash positions got moved slightly by accident.  Anyway, here is the 5th shot with Active D-Lighting set to Extra High:

With the same correction applied above:

It appears to have more detail than the 5th shot above.  Though not as good as the 3rd shot, the ADL might even be better than the 4th shot, though there is a lot of lost detail above the bear's nose.  Because the flash positions got moved I can't be absolutely certain.  My son was pulling me away to play with me before I could re-run the tests for ADL.

CONCLUSION:
Clearly, the aging Fuji S5 Pro still has substantially more highlight range than the Nikon D600.  Because I was not able to reach the S5's limits, I can't say exactly how much better the S5 Pro's highlight range is compared to the D600.  Without Active D-Lighting, the S5's highlight range advantage over the D600 appears to be well over 1 stop.  With Active D-Lighting at Extra-High, the S5's highlight advantage is reduced but I couldn't say for sure by how much.

Please recall the flaws I mentioned at the beginning.  I plan to retest this when the D600 gains real raw support, and this time I will try harder to find the S5's limit, and will test ADL more rigorously.

Read More

Choosing the Exposure for Dynamic Range

Over the weekend, we visited Knott's Berry Farm, one of the first theme parks in the U.S.  What I like about Knott's is that the lines are much shorter than those of Disneyland therefore we get to experience more rides within the same amount of time.  I also think Knott's has a more interesting variety of rides for young children like our 5-year old, and has more thrill rides (though not as many as Magic Mountain).  Although the production value is not as high as that of Disneyland, it is still quite high.

During our visit the heat was intense, reaching 95°F (35°C), just one degree below the all-time record of 96 for that area, and 10 degrees hotter than the average for this time of the year.  With the sun blazing overhead, the dynamic range was very wide.  I used two different exposure strategies to deal with the dynamic range.

PROTECTING HIGHLIGHTS
When the ambient is very bright, it is very easy to exceed the highlight range of a digital camera and blow the highlights.  I used to not care so much about it myself.  In the shot below, you can see that the highlights from the sunlight-as-kicker have no detail at all.

However, when I started learning about shooting film, I became more attentive to highlights. With print film (negative film) I have tremendous latitude for exposing highlights, and it's almost impossible to blow highlights.  With digital, once a certain threshold of light is reached, the sensor can't record any more details, and all I suddenly get is white (i.e. blown highlights).  Blown highlights are therefore one indicator that the image was shot digitally.

(BTW there's nothing wrong per se with looking digital.  It's just a personal preference.  I grew up during a time when digital video was used mostly for lower-budget movies and shows.)

On this occasion I was using two cameras, the Fuji S5 Pro and the Nikon D90.  They have different amounts of shadow range and highlight range therefore I used different exposure strategies.  In the shots below, I took some test shots to show the effect of the exposure strategies.  For the S5, I took a baseline shot using its 'standard' dynamic range mode, and another shot at the same exposure but this time activating its full 400% dynamic range (you can ignore the absence of a highlight on the neck of the baseline shot -- the sun was going down and a shadow fell on the cheek and neck of the mannequin head).  I also took a pair of shots with the Nikon D90: first a baseline shot with a normal exposure, and a second shot that was deliberately underexposed by two stops.  BTW these were at ISO 400, the widest dynamic range for the D90.

In postprocessing I adjusted the shots to recover the highlights and in the case of the underexposed D90 shot, to normalize the exposure as well.  Here is the result:

With the S5's standard mode, there was almost no highlight recovered.  Conversely, with the 400% dynamic range mode, all highlights were fully recovered.  There was not a single pixel that was blown.

With the D90's normal exposure, I was able to recover a few of the highlights but you can see large patches that were blown (the temple, cheekbone and the entire side of the neck).  However, with the deliberately underexposed shot, most of the highlights could be recovered.  There is only a patch of lost highlight near the base of the neck.  

Full resolution S5 400% DR with adjustment (click to download)

Full resolution D90 -2EV with adjustment (click to download)

The downside of this deliberate underexposure is the greater amount of noise in the midtone and shadow areas (after all, I needed to increase exposure 2 stops in postprocessing to compensate for the underexposure).  For me, I usually prefer having extra noise (and using noise reduction) than losing relevant highlights.

---

Back to Knott's Berry Farm.  The sky was cloudless for most of the afternoon, so with the S5, I used manual exposure and set the exposure for the shadow.  Here's a sample shot, straight-out-of-the-cam:

Looking only at the LCD screen, the bright patch of pavement might seem worrying (IIRC it was blinking), but I was confident that most, if not all, of the highlights would be recoverable with the S5, so I concerned myself with getting an adequate exposure of shadow areas to minimize noise.

And here's the adjusted shot (not a single pixel was lost).

As for the Nikon D90, I set the exposure to avoid losing highlights as much as possible.  First, when judging the exposure I paid attention to the blinkies and histogram, dialing down exposure as needed.  Second, I used the 'Extra High' setting of Active D-Lighting, which meant that the camera would underexpose slightly to avoid losing highlight detail.  For JPEGs, it would also boost the shadow and midtone areas to compensate for the underexposure.   However, I was shooting raw, so the boosting would be done in postprocessing.  Here's a sample shot, SOOC:

If I looked only at the LCD screen, it would look very underexposed.  However, I knew that for my purposes, this was the 'correct' exposure.  Here's how the shot looked after adjustment in post-processing.

Most of the highlights were preserved.  There is only a small patch lost, near my wife's elbow.

Here are a few more shots from our visit.  I used two lenses, the Tokina 10-17 fisheye (paired with the S5) and the Nikkor 28-105 3.5-4.5 AF-D (with the D90).

For when my kids are a little older

Knott's has a Peanuts-themed area for younger children, Camp Snoopy

A junior version of the elevator drop

We didn't spend as much time as we wanted because of the scorching heat, but we still had fun.

Read More