Secrets of Digital Photography



 Jump back to Breaking News...

A point often overlooked in the digital camera field is the behavior of optics. Not lens to lens, but zone to zone within a lens.

Sharpness Challenged?

Digital cameras have lenses, just like every other camera. While most 35mm photographers are used to the idea of performance numbers by area of the image, the general concept of zone-by-zone sharpness variation within an optic system is not a well illuminated subject for digital photographers.

Here is some help to get you through the night.

A lot of recent press about the relative merits of the Nikon 990 and 995 cameras raises questions. The answers are often surprising.

Q: Is the 995 lens sharper than the 990 lens?

Q: Do digital camera lenses perform better at particular places in the image?

Q: Do digital camera lenses perform better at particular f-stops?

Q: Do digital camera lenses perform better at particular zoom settings?

Q: Can a soft lens outperform a sharp lens?

Q: Do images focus flat?

Lenses are mass produced. All lenses are made of elements that are NOT ground specifically to mate with individual samples of corresponding optics. Each and every lens is slightly different from its same-specification replicants. So when complex lens systems are assembled, they are MATCHED to each other by sorting individual elements into tolerance priorities and combining them in sub assembly modules that are themselves prioritized and matched to other modules.

The zoom lens in your camera looks something like this. Some elements move, others stay stationary. Variables are always changing as iris, zoom and focus are adjusted. Computerized tests of each subassembly allows "best combinations" of modules to grow right up to the final item.

The tolerances here are microscopic and in general, a shortfall of one parameter in one module is mated to another that just happens to compensate for it. Many elements, sub assembly modules, clusters of modules and even finally assembled lenses are discarded when they don't match specifications.

The lens is rotationally symmetrical. Rather like it had been designed on a lathe. The optical axis runs down the center (green line) and in theory it is precisely centered over the imager. Any element that wanders a micro-hair off center --as if its own individual optical center were slid slightly to the side-- can introduce optical phenomena that affect one corner or edge.

Lens designers have goals. One is to make the entire lens as sharp as can be over the entire picture, but the magic word is "compromise". How much could you achieve if you allowed a certain tolerance in the general performance of a particular lens design?

Another goal is to create a 3:1 or 4:1 zoom . When the design starts out, the first tool is the computer. It can simulate the performance of the zoom lens well enough to guide the designer through inspired refinements. Just like any creative process, sometimes the designer must back up and start over, but with the long history of designs in his or her computer, the problem-solving precedents are helpful to solve problems along the way.

A perfect lens would be sharp center, edge and corner of every image. That silicon imaging chip is about a dozen times flatter, proportionally, than 35, 2-1/4 or 4x5 film is in a camera, so the target for a digital lens is exemplary.

The imager in the 990 and 995 is the Sony 3.34 megapixel chip. It is built to sub-micron tolerances and as manufacturing of this device "ages" the performance of the chip gets better and the yield of good chips to underachieving chips goes up. Generally, any manufacturing line will get better, more efficient and produce a more "tweaked" product over time. Until some new variable is introduced. If Sony's engineers introduced a new idea, such as a new dye for the magenta filter, it would probably cause a hiccup in the yield for a while.

When the final lens/imager passes inspection, the camera is only partly finished. Motors, control systems, the optical viewfinder (not nearly as critically manufactured--nor does it needed to be) are all added as modules and the camera goes together assembly by assembly, test by test, until a working unit leaves the manufacturing line with its new serial number intact.

How does it perform? Any two samples will likely not perform exactly identically. This is true in all optics. Those items are strictly analog. Digital items can "behave" the same, but they do it by logic.

Serial number xxxxxxx1 and xxxxxx2 will not share every single measurable characteristic.

Lenses almost universally deteriorate in contrast and sharpness starting from the optical center and moving farther off-axis. The corners and edges of an image are the first to show this effect.

But--and this is a big item--most unsharpness in the corners of a frame will quickly evaporate with f-stop. As the iris gets smaller, less of the total surface area of many of the optical elements will actually participate in forming the image. Any curves in the surface of the glass at those places which do not contribute to an optimally sharp image are removed from consideration.

With a half of a stop a lens can go from so-so to great at edges and in corners. At very small apertures light tends to scatter as it scrapes past the physical edge of the iris and this fogs the sharply focused image a bit--a phenomenon known as diffraction.

A model change is dramatic. The same issues come back in new pants. Especially when the central optic undergoes a major revision as has occurred from the Coolpix 990 to 995.

Here is an image from the very upper left corner of a 995 frame showing the difference that aperture can make. The small amount of unsharpness in the corner isn't much, but if you roll your mouse over the image, you will see how it looks at wide open, f/2.6 and at optimum, f/6.0.

A similar image image from a Coolpix 990 that I have been shooting for the past 14 months showed a surprising phenomenon. Great amounts of unsharpness were revealed in the upper left corner. And even the smallest f-stop didn't produce as sharp an image as my new 995 was achieving up there.

The middle of the image was fine and until I did the test, all I had been experiencing was wonderful pictures.

It just goes to show:

  • Lens tests don't make images. You do.
  • And if you think people are zeroing in on the sharpness of the corners of your image as the defining quality that separates the "good" photographs from the "bad" ones, guess again.

What none of these samples show are the effects of different zoom settings (all are full wide), different focus settings (all were essentially at infinity) and different camera sample settings (only one of each camera model was tested).

Your mileage will probably vary. On each of these cameras some corners showed different effects from others. Not even each corner behaved like its immediate kin.

So the answers are:

Q: Is the 995 lens sharper than the 990 lens?

A: In some places, yes. In others, no.

Q: Do digital camera lenses perform better at particular places in the image?

A: You bet.

Q: Do digital camera lenses perform better at particular f-stops?

A: All lenses tend to do this.

Q: Do digital camera lenses perform better at particular zoom settings?

A: Aha! A project for you to find out on your own!

Q: Can a soft lens outperform a sharp lens?

A: Given that the 995 lens is about 1/4 of a single pixel softer than the lens of a 990 (with these two samples) the answer is that this 995 camera performs generally better overall, but a tiny bit less at the smallest resolvable object.

Q: Do images focus flat?

A: Almost never. Some of the soft-corner focus comes from that. Most typically, the zone of sharp focus is a parabola that intercepts the image plane. Depending on the optical design, the curve can be like the one shown here or even the reverse. In some designs, it is a wiggle line that is a more complex curve.

The glow represents the area of very good focus.

Notice how that effect, as shown here, covers MOST of the chip surface. Beyond that would lie an area of okay sharpness, and so on. As the iris closes to a smaller opening, the converging rays get closer together. The area of tight focus expands in a miniature version of depth-of-field.

I would hazard a guess that the recent revelations around the rather large change to the 995 zoom lens will prompt a lot of discussion and perhaps test procedures for evaluating digital cameras that more closely mirror those often applied to larger format cameras.

© 2001 Peter iNova. All rights reserved. Do not reprint. Simply add a link to this page.

Special thanks to Steve of for the use of the 995 images.