- Ed Dozier
Nikkor 24-70mm f/2.8 AF-S E ED VR Review
This review will concentrate on the lens MTF50 resolution performance and how well the lens auto-focuses. Repeating the Nikon specifications of the lens isn’t why you should be reading this, since that data is readily available in any number of other places.
This lens is often referred to as the "wedding photographer's lens". On FX, it gets you 95% of the focal lengths you'd need at a typical wedding, in addition to the wide-enough aperture for dim venues and great bokeh. Personally, I'd pair this baby with the Nikkor 85mm f/1.4 and you're good to go.
The usual disclaimer: this is looking at a single copy of the lens. Yours will be different, but hopefully ‘similar’.
These tests were done using a Nikon D610 (24 MP, 5.95 micron pixel) with un-sharpened 14-bit compressed RAW format.
Nikkor 24-70 f/2.8 VR with lens hood. Massive.
Here is a link to get pretty good overall information on this lens. They reduce resolution measurement down to a single number for an f/stop setting. It’s not that simple; resolution is a 2-dimensional thing, and there are sagittal and meridional characteristics within those two dimensions.
First things first: focus calibration. My lens copy needs a fine-tune adjustment of about +16 on my camera, which is vaguely alarming for such a high-cost “pro” lens. I say “about +16”, because this lens needs a different focus calibration setting at different focal lengths. I am using the MTF Mapper software to get focus calibration measurements and MTF50 measurements, which I explain at this link
You should really be using tools that give you answers “by the numbers” to get the most out of your equipment. The focus target I use is an A0-sized print, oriented 45 degrees to the camera sensor plane. Since the camera focuses on the vertical edge of a giant trapezoid, there’s no ambiguity about where the camera is focused. See the picture of the target below (the trapezoid looks more like a rectangle, due to the perspective distortion). I use “AF-C” with a back-button focus for this test, since that’s exactly the way I shoot the camera.
Too bad Nikon focus calibration is so primitive, compared to Sigma. This lens (read ‘most lenses’) desperately needs calibration for different focal lengths and distances. Oh, well. I had to split the difference for the fine-tune value, giving more bias to 70mm (since focus errors are more noticeable there).
For a typical factory-calibrated lens, my D610 needs a fine-tune value of +6, so measuring a +16 (or a little higher) doesn’t inspire confidence in Nikon quality control. As long as it stays within +- 20, I suppose I shouldn’t care.
In case it’s not obvious, you never use “Live View” for focus calibration tests. Only use “phase detect” focus. I always use “AF-C” and not “AF-S”, since that’s exactly how I always shoot (in addition to back-button focus).
Fine-tune +16 not quite enough at 70mm. The peak should align with the blue line.
Fine-tune +16 a little too much at 24mm. The blue line is where the focus sensor was placed on the target; the green peak is where actual focus was located.
What the focus target looks like at 70mm. It’s rotated 45 degrees to the camera. The left side is farther from the camera than the right side. I taped it on top of my resolution target, which is why you see faint gray squares (it doesn’t affect the calibration).
Focus Speed and Repeatability
This lens focuses (phase detect) very quickly and was totally repeatable (in decent light), so no complaints there. By the way, you should always conduct focus tests in good light, or else you’re wasting your time with trying to calibrate against a moving target. EV 10 or brighter is a good light level to use while testing.
Even in dim lighting, I didn’t experience any focus hunting.
Vibration Reduction (VR)
This lens has state of the art VR, which is one of the biggest upgrades over the previous version of the 24-70 lens. I got nearly 4 stops improvement. I’m guessing that this is the main reason somebody would opt for this lens versus getting the previous version.
‘E’ Electronic Aperture
Maybe you have the D5 and are shooting at 12 fps. Maybe you can tell the difference in the electronic aperture versus the mechanical aperture. I honestly can’t tell the difference in shooting with a mechanical or electronic aperture.
For those of you interested in shooting infrared, this lens produces a minor hotspot in the middle of the field of view. If you stick to f/5.6 or wider, the hotspot virtually disappears. Since I usually shoot at 24mm and f/8 with infrared, I use Lightroom with its "radial filter" to decrease the exposure in the middle of the shot by a little more than a third of a stop to fix the hotspot. I also have to shift focus to about 8 feet to get infinity in focus with my 850nm infrared filter (again at 24mm).
Centering: out-of-focus rings should look perfectly symmetric
The out-of-focus rings look perfectly symmetric about the center, indicating that the optics are well centered. This also gives you some idea about the bokeh, which I consider to be excellent.
Here’s the meat of this review. Although the data below is in terms of “MTF50 resolution”, keep in mind that camera sensors with larger pixels will actually produce worse results in terms of “MTF50 lp/mm” values for a lens. You need to combine the MTF50 values with the size of the camera sensor to arrive at the overall picture resolution, typically expressed in “line pairs per picture height” (LP/PH). It’s simple to convert, however: LP/PH = MTF50 * SensorHeight_mm.
The D610 sensor height is 24mm, so lp/ph is just (MTF50 * 24.0). For instance, a lens that has an MTF50 lp/mm of 40 gets (40*24) or 960 lp/ph. Like I said, simple.
I use an “A0”-sized resolution target, where all of the small squares have a little slant to them. The MTF Mapper software I use gets answers that are very similar to Imatest. The main difference is that MTF Mapper is designed to provide answers separated into “meridional” and “sagittal” directions, just like the MTF line charts that manufacturers distribute. As a side note, however, manufacturers only give you "theoretical" and not actual MTF curves. Since the software can produce full 2-dimensional results, you actually see how the whole sensor performs. I imagine you expect your camera to produce 2-dimensional photographic results, so why wouldn’t you want measurements done the same way?
What the resolution target looks like. Mine is mounted ‘upside down’.
Very good center results wide open at 24mm. Corners aren’t too bad, either.
Note that the meridional and sagittal measurements aren’t that different, so that means that astigmatism is well-controlled, too. Sagittal-direction corners at this aperture are among the best I have seen; meridional-direction corners aren’t so hot. Center MTF50 lp/mm is 45, so this means 1080 lp/ph.
Note poor meridional (tangential) and good sagittal corner resolution at f/2.8
Stop down to get the sharpest corners; otherwise, just set the aperture to get the desired depth of field. Center performance is stellar, unless you go beyond f/11. It’s slightly less sharp at 70mm, but not enough to really notice.
I have read complaints about this lens for sharpness, but my copy was able to get 1200 lp/ph (MTF50 = 50 lp/mm) in the center at just about every focal length.
I got at least 720 lp/ph (MTF50 = 30 lp/mm) in the corners at all focal lengths for meridional (tangential), except around 35mm, where it topped out at 600 lp/ph. For sagittal direction, the corners were nearly as sharp as the center, which is just brilliant.
The real key to getting sharp pictures is focus calibration. I’m astonished at how many people ignore this fact. It’s a pain to do the calibration, but you’re rewarded every time you take a shot after that. Don’t get lazy.
I’ll continue to harp at Nikon about being so primitive with their focus fine-tune abilities. So is just about everybody else, except Sigma. Mirror-less manufacturers claim their on-sensor phase-detect solves this, but the focus speed just isn’t there (yet).