Especially with high-resolution digital cameras, full-image sharpness is limited to a narrow zone even when stopped well down, demanding careful attention to detail for best results.
My recent PHOTO Techniques articles have explored how diffraction, focus shift, and field curvature all can lead to blur. The focus shift and field curvature issues can be mitigated by increasing the depth of field, so long as diffraction is held at bay by not stopping down too far.
Depth of field means “the zone of reasonably sharp focus,” with the term “sharp” being both arbitrary and ambiguous. The word “zone” is truer to reality, as “depth” implies a fixed-thickness layer of sharpness at a fixed distance, which is generally not the case once f ield curvature and other aberrations are taken into account. Furthermore, lenses are complex physical objects that have build variances from design specifications that can at best only approach the theoretically best performance. Lenses also vary in rendition, so that two lenses at the same aperture can show strikingly different real-world sharpness.
Circle of confusion
The circle of confusion refers to the blur produced by an out-of-focus point of light. Wide open, there is lens-barrel vignetting causing a partial eclipse of the blur circle (the “cat’s eye” effect). It’s one reason why off-center detail can be sharper than expected—the blur “circle” is smaller than it is at the center.
shaped more or less in a planar fashion; a tall tree sticking too far out of that plane would be blurred due to the tilt. The distant trees in Figure 6 are close to the tilted plane of focus, which extends from the foreground to Yosemite’s Lempert Dome in the distance. They remain sharp, and the whole image is sharp at ƒ/2.8. That’s the power of tilt. Figure 5 shows the test scene at ƒ/8; stopping down would never match the results with tilt.
Focus stacking employs computing power to merge multiple images taken at different focus distances into one composite image.
Focus stacking extends depth of field from the close focus limit of the lens to inf inity! However, the process is rarely perfect; there are often merging artifacts that require retouching, especially along edges or areas of movement. Photoshop CS4’s focus-stacking feature is fun to try, but turn your attention to Helicon Focus software, which offers fewer artifacts (www.heliconsoft.com).
Practical in-the-field focus stacking might mean stacking only two or three images; after all, the wind blows and the sun and clouds move; it will depend on the subject matter and conditions. In the studio, many more stacked images are practical. Since focus stacking requires careful incremental focus adjustments, a top-quality manual- focus lens is your best choice, so choose the very best optics available, ignoring the automation issues.
Figure 8 shows an image made with a 280mm lens at ƒ/5.6; there is simply no hope of getting the wagon in sharp focus throughout, not even at ƒ/22. Focus was carefully bracketed in nine frames from the leading edge of the wagon to the rear. The stacked image is sharp across the entire wagon left-to-right and front-to-back. But there’s a catch—odd-looking blur along the board-edges of the wagon. Helicon Focus offers a retouching mode where such artifacts can be cloned out, but it’s tedious to f ix a large area. Cleaning up the rightmost wheel, with its dangling chain, was really easy and quick. Other areas looked great.
Depth of field is exceedingly shallow in terms of resolving to sensor resolution. Zero in on the most important subject matter to distract the viewer into believing that everything is sharp. When feasible, use a tilt/shift lens to move the plane of sharp focus as needed, and consider focus stacking for situations where time and lighting allow it.
Stopping down one stop changes the diameter of the lens diaphragm by a factor of 1.4, which means that out-of-focus blur circles are reduced by a factor of 1.4 linearly, improving resolved detail by a factor of 1.4.
But for detail to double in linear resolution, then two stops are required. For example, ƒ/5.6 is two stops more than ƒ/2.8, yielding a 2× improvement in resolving power. Aperture ƒ/11 is again two stops more than ƒ/5.6, or 4× the linear resolving power over ƒ/2.8. It should be clear that even a quadrupling of resolution doesn’t amount to much when detail is strongly out of focus to begin with.
Figure 1 shows actual-pixels crops of a single star, taken with the Zeiss ZF 85mm ƒ/1.4 Planar at ƒ/1.4, using the Nikon D3. The series was progressively defocused to show just how the blur circles look in reality (and near optical center). We can see various optical aberrations at play, probably axial chromatic aberration as well as spherical aberration and spherochromaticism. It should be obvious that there is more going on that just a simple blur circle.
The goal in stopping down is to reduce the size of the circle of confusion. Not only does stopping down help eliminate some optical aberrations (e.g., the ring of purple haze), it reduces the size of the blur circle. A lens diaphragm whose diameter is reduced by half (two stops) cuts the blur diameter in half also, yielding a blurred dot (circle of confusion) of half the diameter and a quarter of the area.
Smaller blur circles look sharper than larger ones. Eventually, the circles are small enough that we perceive the detail as “reasonably sharp” and therefore within the depth of f ield. But until the circle of confusion approaches and exceeds the sensor resolution, the image fails to utilize the full sensor resolution. That is why the term “sharp focus” in the context of depth of f ield is oxymoronic—only a narrow zone of focus actually resolves to sensor resolution. This is especially noticeable at closer ranges, where depth of f ield is shallow.
Depth of field and depth of focus
Depth of field and depth of focus have an inverse relationship. Depth of f ield refers to the subject matter (real objects), and depth of focus refers to its inverse: the area of sharp focus at the sensor. The two ideas are really the same concept, except that one is an object space, and one is related to the image being projected onto the sensor.
We can get depth of field just as we want it, but it’s depth of focus that determines whether the image is faithfully recorded. Assuming perfect alignment of components (unlikely), a digital sensor will faithfully record the image. Not so with film, which not only has substantial thickness, but can warp as much as 200 microns.
With film, one layer/area might have perfect focus, and thus resolve detail at critical sharpness, but other layers or areas might be substantially blurred because the film does not lie f lat. For example, medium-format 220-roll film is generally considered sharper than 120-roll film, for film- flatness reasons. (See Zeiss Camera Lens News #10). Film flatness also changes over time, as the film starts to curl slightly. Sharpness can be subtly or greatly degraded because the f ilm is warped out of the zone of sharp focus (“depth of focus”).
Digital sensors are perfectly flat, but alignment of the sensor to the lens mount is critical. The chances of a DSLR having perfect lens-mount-to- sensor alignment are slim, and this shows up with asymmetric sharpness with ultra-wide lenses (especially 24mm and wider). Focus accuracy is difficult even with Live View, which is generally accurate to no more than 40 microns; a 60 micron difference at the sensor is equivalent to 10 feet versus 12.5 feet on a 21mm lens. As a practical matter with wide-angle lenses, signif icant depth of f ield is required for any hope of consistently sharp imaging across the frame.
The term “sharp” by itself is fairly meaningless; it’s an arbitrary judgment based on print size or viewing distance. View a magnified image, and what appeared sharp at a smaller size might be blurry. But reproduced on a postage stamp, the image might contain more than adequate detail.
A rational way of judging acceptably sharp (e.g., adequate depth of f ield) has been to reference a print size and a viewing distance, taking into account the acuity of the human eye.
But that approach has its own issues with human perception. Why not make the sharpest possible image so that its future potential is not limited to any particular print size?
Lens depth-of-field markings assume a circle of confusion of 30 microns, which is about 30 times larger (in area) than the photosites on today’s high-resolution DSLRs. Thus, detail in a 24-megapixel camera is allegedly sharp when equivalent to 0.8 megapixels. Consider such things when upgrading to a higher-resolution camera.
As a practical matter, consider a 24-megapixel versus 12-megapixel camera, both with the same size sensor. A primary reason to move from 12 to 24 megapixels is to increase the maximum size of sharp prints.
Yet at the same aperture, the amount of blur hasn’t changed, but it is now simply spread over more pixels. Nothing has been gained except in a narrow zone of focus where resolved detail approaches the sensor resolution. Stopping down one more stop will improve detail rendition overall, so long as diffraction doesn’t reduce the resolution in the critically focused areas. The appropriate approach to depth of f ield is: how can sensor resolution be fully utilized? The answer to that involves depth of f ield and also diffraction; there is no free lunch.
More depth of field is often the goal, but photographers sometimes spend considerable sums to get less of it (e.g., a Canon 85mm ƒ/1.2L). Higher shutter speeds might be the goal, but it’s often also about minimizing depth of field. But consider that the poppy image in Figure 2 was taken at ƒ/8. Less depth of field can make an image appear sharper because the subject matter in critical focus will be clearly separated from the background and comparatively sharp.
Focal length, sensor size, and depth of field
For the same angle of view, smaller- sensor cameras require shorter focal-length lenses. Shorter focal lengths offer more depth of field at any f-stop. One of the major drawbacks of small-sensor cameras is the effective impossibility of smoothly blurred backgrounds; no point-and- shoot camera offers an ƒ/1.0 or even ƒ/1.4 lens.
Figures 3 and 4 compare details of an image shot with a full-frame Nikon D3x; the camera also offers a “DX crop” mode used for some detail. The 24–70 zoom was set at 70mm for the full-frame shot and approximately 46mm in “DX mode” (cropped sensor). At ƒ/5.6, the DX-crop frame shows notably better front-to-back sharpness than the FX frame. The equivalent full-frame f-stop for equivalent depth of f ield should be ƒ/5.6 * 1.5 = ƒ/8.5. Aperture ƒ/8.5 cannot be set on the camera, but it falls between ƒ/8 and ƒ/9, and ƒ/9 is only 1/3 stop more than ƒ/8, so the test is within roughly 1/6 stop. Aperture ƒ/9 is used here in the crop for comparison purposes.
Examination of the image front- to-back reveals that 70mm FX-frame sharpness somewhere between ƒ/8 and ƒ/9 is indeed a match for the ƒ/5.6 image on the 46mm DX crop, with the ƒ/9 crop indeed a smidgen sharper. There is a small margin of error, of course (exact focal length and lens performance and exact focus matching), but the comparison proves out the theory: on the DX frame at ~46mm, depth of field at ƒ/5.6 matches that of the FX frame at 70mm at ƒ/9.
Practical tips for field work
Focus carefully, always!
For large prints, critically accurate focus is essential. Stopping down helps, but overcomes only a modest amount of error; ƒ/2.8 versus ƒ/11 is only a factor of four in resolved (linear) detail, hardly enough to compensate for a focus error of even a few feet at 10 feet with a 21mm lens. Try it yourself on an outdoor subject and critically assess your images; it’s demanding work to nail it. This problem will only worsen with future DSLRs in the 30–40 megapixel range.
While variations in the zone of sharpness can be acceptable with three-dimensional subject matter, focusing at 3 meters instead of 5 meters will visibly degrade the sharpness at infinity, even with an ultra-wide 21mm lens.
Bias focus to the most important subject matter, or infinity
Focus on the most important area of the subject, or at least bias the focus toward it as a means of ensuring crisp rendition. If distant subject matter is important, bias strongly toward infinity; the image will generally fail to appear sharp otherwise, even at the hyperfocal distance (the old “1/3 in” rule, roughly speaking). You are not wasting depth of field by doing so.
Depth-of-field markings on the lens can be helpful, though with experience they’re simply not needed. Add two stops to obtain acceptable sharpness. For example, if the depth of field indicated for ƒ/4 is desired, use the ƒ/4 markings as a guide, but expose at ƒ/8 (see Figure 5). Avoid ƒ/22—it’s a losing proposition due to diffraction.
To ensure mediocre sharpness, use the hyperfocal distance, the old “1/3 into the scene” rule. You’ll get good average sharpness, but sharpness at infinity and close-up distances will suffer. That might be fine for some subjects, but for important subject matter at infinity, the image will disappoint.
Using the so-called hyperfocal distance can work reasonably well, simply because most photographers using this rule stop down to ƒ/16, or even ƒ/22. But this sacrif ices contrast and detail, especially with 20+ megapixel DSLRs. Especially for scenes where detail at inf inity is important, bias toward infinity.
Tilt lenses to the rescue
A “tilt” lens allows adjustment of the focal plane relative to the film or sensor; the plane of sharp focus is skewed for better alignment to the sensor (it can also be deliberately misaligned to throw things out of focus). Tilt has long been used in large-format view cameras, where the front and/or rear standards can be “swung” or “tilted.”
The tilt effect is so powerful that at ƒ/2.8, a lens tilted appropriately can easily be sharper from foreground to inf inity than at ƒ/16 without the tilt. With cooperative subject matter, ƒ/5.6 or ƒ/8 can yield not only superb lens performance, but also ample depth of field. The subject matter must be