In my previous post, I worried that there were focus problems with my new Nikon 70-200mm f/2.8 lens. I'd based this on a few test pictures taken while visiting the park with Anthony, and concluded with the realization that I need to do more rigorous tests to discount the possibility that my own errors were causing the problems.

I later posted a short summary of this to a forum at Phil Askey's the most excellent Digital Photography Review (see thread here), and got some comments reiterating that it might be my fault. Modern SLR camera auto-focus works on finding contrast, and while I think that the subjects in my test shots stand out with high contrast from their background, some of them also had highish-contrast aspects within the background, so it's possible that the autofocus latched on to them instead of my intended target.

I've never run into this problem with my 18-200mm f/3.5-5.6, either because it doesn't have focus problems, or even if it were to have one, because its depth-of-field is much deeper. With the zoom at 200m, a subject 5 meters (16 feet) away has a depth of field of about 5.3 inches at the lens's fastest aperture of f/5.6, more than twice the depth as the new lens at its fastest aperture of f/2.8 (a depth of field of only about 2.6 inches). Testing the new lens wide open (at f/2.8) places the most demand on all aspects of focus (camera, lens, and me).

So, today I did more tests.

Focus Test Charts

There are many focus test charts available — I've used this and this previously. I've learned a lot with them, but felt I could do better, so I finally made my own. A very shrunk version is shown at right.

The black bar in the center has very high contrast from the background, surrounded on all sides with very low contrast. The auto-focus can't latch on to the low-contrast surroundings, thereby guaranteeing that if the camera focus spot is pointed anywhere near the center bar, it will focus on it and only it.

The chart is tipped back away from the camera, so that everything below the black focus bar is closer to the camera than the focus bar, and everything above the focus bar is further away.

A properly working camera and lens (and operator ) focusing on the black bar should result in the black bar being in focus, as well as everything to its left and right. That includes the far edges of the test chart, which have various repeating high-detail patterns and text. Things above and below the main focus line, because they are different distances from the lens, should get progressively out of focus.

There are numerous little features about my chart that make it more useful than the others I've seen, but I'll save that discussion for a later writeup. One thing I'll mention now is that the focus bar is surrounded not by blank white, but by text — low-contrast text that is invisible to the auto-focus, but not in the final picture. It helps a lot.

Here's a close-up of the focus bar and the low-contrast text that surrounds it (which doesn't look so low-contrast in this image, but does so when printed, at least on my printer):

When printed at 600 dpi, the black bar is 1.8mm tall (a bit over 1/16th of an inch).

Here's a close up of the detail seen to the left of the focus bar:

That repeats up and down both sides of the chart.

Retesting my Nikon 70-200mm f/2.8 VR

With the new chart in hand, I set it up in the sun by a window. Here's what one shot looked like:

Here's a full-resolution crop with exaggerated contrast to highlight the low-contrast area, showing that the focus bar is in front of the in-focus area. This matches exactly the results I saw before.

Here's a test at the other end of the zoom range: 70mm:

Here's the full-resolution crop.

Compared to the test above, more of the chart is in focus, and the focus changes more gradually. This is because the chart is not at such a steep angle to the lens, and also because at this distance and focal length, the depth of field is more than double. So, it's harder to pinpoint just where the center of focus is, but it seems to me that its at about the 7 to 9 range (black numbers on the left; you can scroll the image), and not the 0 that it should be:

Conclusions

It's still a work in progress, but I'm now more confident of my technique and of the results: I have a bit of back focus with my camera and/or lens.

I recently bought three new lenses for my Nikon D200. In my previous post, I told of the Sigma 30mm f/1.4 that had such horrible focus problems that I had to return and reorder it. Today I write about the Nikon 70-200mm f/2.8G IF AF-S VR that I finally received on Saturday.

I should mention my mindset before receiving this lens. Other than the short stint with the Sigma 30mm, I've had just one lens since getting the D200 in January, the Nikon 18-200 f/3.5-5.6 VR. It's quite a bit thicker and longer than a standard SLR 50mm lens, and with its lens hood it looks all the bigger, so I tended to “feel like a pro” with it. Well, at least I thought that I looked like one.

I knew the 70-200 on order was a bit bigger because at f/2.8, it lets in four times the amount of light at 200mm than the 18-200 at its f/5.6. But nothing prepared me for pulling it out of the box. My first impression was that I was sent a telephone pole by mistake, it was so long and heavy.

The specs say that it's three pounds, which sounds like nothing, but after giving it a try for just one or two minutes, my left arm was really tired. I'm wimpy, but not that much of a wimp. Perhaps they filled it with concrete by mistake?

And the length — it's huge! Not “huge” like the massive lenses that sports and wildlife photographers use, but it's much longer than I expected. I'd bought a camera bag the other day for it, getting a bag larger than I thought I'd need, but it turns out that the bag is too small and I can't use it with this lens. Oops!

This lens forced me to put the strap back on my camera. Generally, I find that a camera strap just gets in the way — the 18-200 is so light that I don't mind just carrying the camera most of the time, or putting it into a small camera bag for longer periods of disuse. I didn't have a camera bag that could hold the new lens, and I had no intention of carrying it the whole time, so I put the strap back on so that I could sling the camera/lens across my back while walking.

(Is it just me, or do others wish that the strap attach points were on the back plane of the camera, rather than the front? If they were on the back plane, the camera would lie flat against you rather than tipped at an odd angle.)

Testing The Lens

Anthony wanted to go to the playground, so I thought it'd be a great chance to test the lens. For the most part, I kept the aperture at f/2.8 so that I could play with the shallow depth of field, and test the focus.

This next shot is a lady getting into a taxi near the playground. It's not much of a picture, but somehow the smoothness of it really looks nice to me:

Looking at the full size image (click on the pic above, or here) shows a lot of detail in her kimono and purse, and the out of focus green in the background just seems really nice.

The next image reminded me that even with a fast lens and fast shutter speed, you have to watch out for subject movement. The shutter speed was a fairly zippy 1/_800^th of a second, yet there's still quite a bit of movement in Anthony's fingers as he slaps the sand mold over. The tips of his fingers look to move about a centimeter during the exposure, which would place their speed at 8 meters/sec, or just a bit short of 18 miles per hour. Fast fingers!

A Touch of Rear Focus

Unfortunately, the lens is not without its problems. It has rear focus, which means that the actual point in focus is a few inches behind where it thinks it's setting the focus. It's not a problem for shots of distant subjects, nor for shots with relatively small (high “f” number) apertures, but it manifested itself many times during our trip to the playground.

Here's a perfect example:

It's a nice picture, showing his inside-out shirt (he dressed himself) and a somewhat serious face, but unfortunatly, it's not in focus.

The focus point was his left eye, so the eyes should certainly be in focus. With these lens parameters, the field of acceptable focus should extend about 1.6 inches in front and 1.6 inches behind, for a total depth of field of 3.2 inches encompassing his nose and sideburns. However, a 100% crop shows the eyes and nose out of focus, with his sideburns and parts of his ear in focus.

By the way, the “3.2 inches” number I cite for the depth of field is necessarily an approximation. The focus deteriorates any distance from the focus point — the entire concept of “depth of field” is about how much out-of-focus blur is considered acceptable for common uses, such as in viewing an 8x10 at arm's length. Blowing up the same image to a wall-sized print would place a higher demand on sharpness, thereby substantially shortening what is considered the depth of field.

For those interested, Wikipedia has a lot of info about depth of field.

---

Here's another example. The truck is the focus point:

The depth of field in this case is about 10.8 inches (from 5.3 inches in front of the focus point to 5.5 inches behind). The truck seems to be just shy of the near edge, with the sand just behind it starting to look really sharp.

---

And one more example, with the leaf in the center being the focus point:

The depth of field here is about 12.7 inches (6.2 inches in front; 6.5 inches behind), and the leaf is clearly not quite in it.

I need to do some more tests to make sure that the problem is not something as simple as the camera latching onto something different from what I think it's latching onto, or movement on my part between the focus and the shot, but at this point it seems that I'll have to send this in to Nikon for adjustment.

Update: In trying to understand the focus issues, I designed an Autofocus Test Chart that fixed the deficiencies I found in the other test charts I tried. In the end, I did send the lens back to Nikon. It took two tries, but Nikon finally fixed it.

Wishing for Focus Problems

On the way home, we came across this, er, style abomination while waiting to cross the street. Like passing roadkill on the highway, I couldn't help myself but to look.

I found myself wishing for focus problems

Unfortunately, not even the fuzzy hat is a focus problem — her sense of style is really that bad.

I don't want to end with that picture, so I'll end with another picture of the beefy, hunky lens:

As I recently wrote, I love my Nikon 18-200 zoom except that it's a bit “slow”, optically speaking: at f/3.5-5.6, the lens diameter is too small to allow enough light for reasonable shutter speeds in low-light situations. So, two weeks ago, I finally picked up a Sigma 30mm f/1.4 prime lens.

At f/1.4, it's 3 stops “faster” (optically speaking) than my current lens, meaning it can allow 8 times more light, in turn allowing a shutter speed 8 times faster. That makes it great for low-light situations.

A byproduct of a large aperture (low “f” number like 1.4) is that it can produce an extremely shallow depth of field, meaning that most things are out of focus except for objects more or less exactly such-and-such a distance from the lens. This can be used creatively in a variety of ways.

Because of the shallow depth of field, focusing must be spot on or what you intend to be sharp won't be. Unfortunately, this lens has a reputation for having auto-focusing problems: if you're lucky, the one you get is perfect, but if you get a bad one, the focus point can be a few inches to the front or rear of where you told it to focus. If you get one of these, you have to either return it to the store and try your luck with another, or send it to Sigma to have them fix it.

Looking at one of the first photos I took, I knew I'd gotten a lemon:

Clicking on the picture above to see a larger version shows pretty much nothing is in focus, although you can see a stripe across the width of the paper Anthony's looking at is more in focus than the rest. The focus point was his left eyelash (the one on our right), but the lens missed it wildly.

Anyway, I didn't know about that until later when I looked at them on the computer. When I first got back from the store, I'd had time to take just a few test shots of Anthony before we headed out to eat with Fumie's folks. It was dark out and I wanted to see what the large aperture could do, so I left it wide open most of the time. As I said, I didn't know about the focus problem until I got back, so many of the pictures below have focus issues.

As with most pictures I put in my posts, clicking on them yields a larger version.

We went to a restaurant that's about a five-minute walk away. It specialized in Fugu (pufferfish) and Matsutake mushrooms.

As you can see in the upper-left of the large version of the next picture, my copy of the lens suffers from severe chromatic aberration. The black letters on the white sign show a strong blue fringe, so strong that I wondered if it was actually part of the sign. (I checked the next day; it's not.)

To take the picture above, I had my back pressed deep into the corner of the room. At 30mm, this lens is just a bit too long for a confined space like this. I really would have preferred a 24mm over a 30mm, but it seems no one makes such a Nikon-mount 24mm f/1.4 lens.

Despite the picture above having nothing in focus except the edge of the tray in front of Anthony, it's a good example of how a shallow depth of field can be used to isolate a subject. If the focus had been correct, Anthony would be crisp and sharp, but the background and stuff on the back edge of the table would all remain out of focus. With the smaller aperture of a common lens, the background would be more in focus, more distracting.

In the next picture, I have Fumie and the foreground out of focus, with the waitress the main point of focus.

The shot above really highlights the shallow depth of field.

The shot below doesn't say much about the lens; it's just a nice shot of Fumie's dad, despite being out of focus.

I'd played around a lot with the shallow depth-of-field the lens could produce, but I wanted to test its light-gathering abilities more. I thought I'd take a picture of the river next to our building, but Anthony was fussy and would have none of it. In the end, I was able to take one shot hand-held with just one hand while holding a fussy, wriggling Anthony in the other:

It's clearly out of focus, and not the lens's fault this time, but considering the situation I think it's pretty darn good. My other lens would have needed an almost two-second exposure, which would have yielded nothing but a blur. (However, under better conditions my 18-200 can produce nice night photos, such as this 30-second exposure from a bit further down the river, looking towards where the above picture was taken.)

Back at home, once I looked at the first batch of photos, I realized that I had a lemon, and kept it to mostly manual focus for the next few days. (This had me pining for a split-focus screen for my D200, like film SLRs of days past. Does anyone in Kansai install them?) The photos from my previous three posts about Anthony's birthday presents (one, two, three) were all taken with this lens.

In particular, this one photo was worth the cost of the lens:

This picture just couldn't have been taken with an f/2.8 lens (f/2.8 being a common maximum aperture on many pro-level lenses), as such a lens would have needed a 1/_15^th-second exposure, and no four-year-old stays still for that long. Well, I could have pushed the ISO way up to recover some of the shutter speed, but the result just wouldn't have been the same.

Update: the ghosting problem originally reported in this paragraph is likely not the lens' fault, but probably due to the fact that I had a filter on the lens (a clear-glass “protector” filter). I've not yet re-tested with my recently-acquired good copy of this lens, but I wanted to at least put a note here. Originally, I wrote... As much as I cherish this picture, it shows yet another problem with this lens: ghosting. The version above is almost just as it came out of the camera. Looking at (a reduced-size copy of) the original, you'll see that I used Photoshop to remove ghost reflections of the candle flames that appeared on and next to his head (as well as a light outside showing through the curtain).

I knew that this lens really sucked in the light, but with focus problems, chromatic aberration, and ghosting, the copy I had just sucked. I brought it back to the store, and while there I was able to test another copy (using a focus test chart among other methods [Update: here's the better focus-test chart] that I've used since) and its focus was clearly better than the copy I had. In the end I returned it and got my money back instead of exchanging, mostly because I'd since learned of a place I could get it for 20% less.

Today, I reordered it at a different store (for $100 less) along with two Nikon lenses, a 17-55mm f/2.8G IF AF-S DX and a 70-200mm f/2.8G IF AF-S VR. I expect to have them in time for fall foliage pictures. I'll use the Sigma f/1.4 lens only in special low-light situations, or for its shallow depth of field. Otherwise, I expect the two new Nikon zooms to get a lot of use, leaving my current f/3.5-5.6 18-200 zoom as an all-in-one walkaround/travel lens.

We'll see.

Update: finally got a good one.