VR

VR Tests, Pass 3 Part 2: Not Impressed

I’m looking for expert opinion, but my current conclusion is that my lens is broken. I’m supposed to get a 3-stop increase in handholdability from VR on this lens, and I don’t get anything like that. I still want comments on whether I’m being unreasonable about what I categorize as “critically sharp”. I think I’m being somewhat lax, myself; if it looks good at 1:1 I’ll call it critically sharp, and not inquire further.

So, here’s a set of carefully-shot tests, at increasing shutter speeds. Each setting I shot 10 times; then I rested for at least 5 minutes before resuming (I don’t think my arms and hands are really that shaky, but I was trying to give the lens every chance). I was using aperture priority, and adjusting the ISO, so I didn’t really get 10 shots the same each time; apparently I was too close to a boundary, so I got some at one shutter speed and some at another. They’re named and categorized based on the shutter speed recorded in the EXIF data. All test shots are at 200mm and f/5.6. The camera is the Nikon D700. The lens is the Nikkor 70-200/2.8 VR AF-S G. Click through to see the 100% crop of the center of the frame for each test. See the tripod results (showing what really sharp results look like, and what the full frame looks like before cropping down to just the center) here.

At 1/15 VR off: Good 0, marginal 1 (#1), bad 0.

1/15 VR on: Good 1 (#2), marginal 2 (#3,4), bad 1 (#1).

1/20 VR off: Good 2 (#2,7), marginal 0, bad 6 (#1,3,4,5,6,8).

1/20 VR on: Good 1 (#6), marginal 3(#1,3,5), bad 2(#2,4).

1/30 VR on: Good 0, marginal 3(#1,2,3), bad 0.

1/40 VR off: Good 1 (#4), marginal 0, bad 9 (#1,2,3,5,6,7,8,9,10).

1/40 VR on: Good 0, marginal 5(#1,2,4,6,7), bad 2(#3,5).

1/60 VR off: Good 1(#2), marginal 3(#6,7,9), bad 5(#1,3,4,5,8).

1/60 VR on: Good 0, marginal 0, bad 3(#1,2,3).

1/80 VR off: Good 0, marginal 0, bad 1(#1).

1/80 VR on: Good 2(#2,5), marginal 1(#3), bad 3(#1,4,6,7).

1/125 VR off: Good 1(#1), marginal 0, bad 0.

1/160 VR off: Good 1(#6), marginal 4(#1,7,8,9), bad 4(#2,3,4,5).

1/160 VR on: Good 8(#2,4,5,6,7,8,9,10), marginal 1(#3), bad 1(#1).

Once again, I get a surprising proportion (i.e. greater than 0) of usable shots at 1/20 and 1/40 seconds without VR. I wasn’t leaning on anything, sitting, resting my arms on anything; I was standing up all by myself, and holding the camera in my own two hands. And I get very little improvement in sharp shots, few critically sharp shots, when VR is turned on.

Anybody else with this lens, do you get better VR results than this? Is mine broken?

Notes for future testing: Target is good.Â Procedure (with rest) is good. Need to use manual exposure. Need to have more light, so I can try some higher shutter speeds.

Continue reading VR Tests, Pass 3 Part 2: Not Impressed

VR Tests, Pass 3, Part 1: The Manual is Right

Yes, this topic has been annoying me quite a bit; why do you ask?

I was doing a set of tripod-based tests tonight, under the conditions I’ve established for the planned hand-held series, to produce example of just how sharp this target should look through this lens. (One of the questions last time was whether I was expecting too much sharpness.)

I was shooting very carefully, doing nearly everything I could to eliminate sources of vibration: tripod, LiveView mode (so the mirror was up), remote release (so my touching the body didn’t introduce any vibrations), contrast-detection focus (so any alignment issues in the other AF system won’t cause focus errors).

I shot 5 copies of each test case; each group of 5 looks substantially the same, indicating that my procedures were at least consistent from shot to shot.

Here’s one of the full frames, reduced to a sensible jpeg size, to give you an idea of what the setup was. All the images after this will be 100% crops from the center when you click through the thumbnail.

First, here are 100% crops from a few shots without VR at various ISOs and shutter speeds:

Now, here are 100% crops from a few shots with VR at the one ISO (Lo.1) and shutter speed (1.30sec) I tested:

That, as the saying goes, is just not right.

The manual says to NOT use VR on a tripod. Based on this step of the experiment, I have to say that the manual is definitely right on this one! (The manuals for various generations of Nikon VR give different answers to whether it should be used on a tripod. I checked the manual for this specific lens, and it definitely says not to use VR on a tripod. They’re right!)

So, some day soon, the rest of the test, actual hand-held shots with VR on and off, using this same target.

VR Test Part 2

Having checked out the results of the first test, I decided to run tests at a higher shutter speed, and with longer test series.

As before, these are 100% size crops out of the center, including the focus point. For this series, I had to go to ISO 800 to get the shutter speed I wanted.

VR off, 1/30:

Scarily, 1, 6, and 9 are sharp, hand-held free-standing at 1/30 sec. with a 200mm lens. That’s 3 out of 10, a 30% success rate. This is not supposed to happen.

What I see in the VR on group is a much higher percentage of not fully sharp images. Only two and three are fully sharp, about the same as without VR! But 5, 7, 9, and 10 are quite close to sharp.

This exactly matches my subjective impressionâ€”VR rarely produces a fully satisfactory image for me, but often helps to get a usable one. It’s certainly no substitute for a tripod!

Either that or the VR in my Nikkor 70-200mm f/2.8 isn’t working right.

VR Test

“Vibration Reduction”; Nikon’s tradename for optical image stabilization. The camera and lens sense the degree of camera motion, and deflect elements in the lens to cause counter-vailing motions, resulting (if it all works right) in a sharper image. They claim about a 3-stop improvement (in terms of lower shutter speeds usable hand-held).

The rule of thumb is that you can safely hand-hold the camera down to a shutter speed of 1/(focal length).Â This is a 35mm rule of thumb, and it’s the 35mm-equivalent focal length that matters here.Â So for a 200mm lens on a full-frame DSLR, the safe shutter speed (by rule of thumb) is 1/200 sec. Or, with VR, about 1/30.

The following test photos are small crops from the center of the frame, containing the focus point. They were all shot hand-held, free-standing (I wasn’t leaning against anything).

I score that as 1 and 2 acceptably sharp, the rest not.

And now some examples shot with VR off.

None of the VR off examples are acceptably sharp.

So; the VR off case certainly works as expected, no hope. The VR case produced two acceptably sharp photos a full 4 stops below where it should have been okay by rule of thumb. And a lot of failures, but I was seriously pushing the limits here.

Next post will be another run, a bit more careful, with 10 shots of the same test subject for each series. But this is getting long enough and ugly enough as it is.