Digital Q&A

Note: these are historically old and haven't been updated or revisited in some time. A different Q&A setup is being worked on, at which time I'll massively update this page.

Send Thom a question by clicking here: QandA
last update: Aug 1, 2003

Question: I've seen some posts in the forums that say you should do no in-camera sharpening because you'll enhance the noise and others saying you should do some sharpening in the camera since the camera sharpens before jpeg compression and you may avoid sharpening the edges of artifacts. Any opinions?

Answer: I'm of the opinion that sharpening should never be performed unless you know the final size at which the image will be used. Once the halo artifacts caused by sharpening are added to a file, it's nearly impossible to remove them. See my sharpening article if you're unfamiliar with what sharpening does.

While most cameras sharpen prior to JPEG compression, that's a two-edged sword. Indeed, it does prevent sharpening from being applied to the artifacts, but it doesn't prevent artifacts from being added to the sharpening! (Read that previous sentence as many times as necessary to make sense--it really is the core of the dilemma.) As detail increases in an image, the JPEG compression engine produces two byproducts: (1) the file size becomes larger due to the added detail, which doesn't compress as well; and (2) JPEG-caused artifacts (primarily "mosquito" and "interline" noise) increase.

If you capture an image digitally with as little post processing as possible (i.e., RAW file with no sharpening, no up sampling, no hue/contrast/tone adjustment, etc.), the primary problem you have to deal with is the anti-aliasing effect digital sampling produces (and its relative, moire). You'll certainly need to sharpen the image prior to printing it, but if you know the print method and size, you can hold the halo artifacts produced by sharpening to levels that aren't readily seen in the final image. For example, the Epson inkjet printers all have a dot-gain of about 30%, and it's relatively easy to hide sharpening. If you sharpen in camera, then discover that you need to sharpen again after resizing an image, you'll end up sharpening the original sharpening artifacts, and at some point that build-up of artifacting becomes readily visible.

The D1x is a unique digital camera, in that it's the only one I know of that currently down samples photosite information. When you down sample one axis and up sample another, as the D1x does in the large JPEG sizes and Nikon Capture does with NEF images, photosite related noise is generally reduced. It also masks some anti aliasing effects, generally meaning you need less sharpening. As you'll see when you get my book, NO and LOW sharpening in JPEG large sizes are remarkably similar in artifacts, so I tend to leave my D1x on LOW sharpening when shooting in JPEG. The Coolpix 995 also does a remarkably good job at LOW sharpening. With earlier Coolpix models, I'd tend to turn sharpening OFF. The D100 absolutely requires sharpening when shooting JPEGs (but again, I tend to do that post shoot). Both the Fujifilm S2 Pro and Kodak Pro 14n have a tendency towards more moire when you sharpen in camera.

top | home

Question: I've got a D1 (or D1X, D1H, Coolpix 990, Coolpix 995). Does the Nikon Flash Guide cover flash use with these cameras?

Answer: The Nikon Flash Guide was written prior to my having much experience with digital cameras, thus only touches on some of the differences. Check the Nikon Flash Guide corrections page for more information about digital flash use. At some point later in 2003 I'll have a Second Edition available that includes the digital stuff, but at this point I'm waiting for an SB-800 before finishing up that version of the book.

top | home

Question: I've tried all the different white balance settings, and even used a white card to set white balance manually, but I still can't quite seem to get neutral results. Any suggestions?

Answer: First, make sure that the monitor you're viewing has been calibrated and set correctly (I discovered recently that my monitor had drifted a bit, which was influencing my color balance results; I now use Colorvision's OptiCal with the USB Spyder monitor to regular tune my monitor). Next, stop using a white card! D1 users especially have discovered that using a neutral gray card works better, and I've found this to be the case on virtually every digital camera I've encountered. Why? Because most whites aren't exactly white, plus any over- or underexposure while setting white balance skews the results. Kodak's 18% gray cards don't trigger exposure problems and are neutral in color, solving both problems.

top | home

Question: I've taken quite a few photos with my Coolpix 990 using a single Sunpak433D. These turned out predictably. I added a second flash using a wein slave, again with predictable results. I then added a 500w quartz halogen work flood deflected from a large white polysterene surface above and behind the camera. Many of these photos turned out incredibly blue. I tried setting the white balance using an identical piece of polystyrene board in the subject position. This worsened the problem. What gives?

Answer: According to The Manual of Photography, most tungsten-halogen lamps have a color temperature of between 2700 and 3400K. Many also have a tendency to change color temperature with time. So, if you set your white balance to that needed by the halogen lamp, when you add flash, whatever they light will be remarkably blue, more than 149 MIREDs different (MIREDs are the measurement to assess color temperature differences and the difference between two MIRED numbers is used to decide filtration). Generally, if you're working with lights that have two different color temperatures, you have to filter one of the light sources to get acceptable color. You could, I suppose, bounce the halogen off a slightly blue card and set the white balance to that, but you'd be operating by trial and error until you found the correct color blue.

top | home

Question: My photo dealer says that Nikon Coolpix models produce inferior results to Olympus models with the same resolution. Is he right?

Answer: The dirty secret of the digital camera business is that virtually all consumer digital cameras use the same CCD sensors, and many even use the same exact Sparc-based chip set internally (true of many of the older Nikon and Olympus models, for example). Sony is the primary provider of CCDs, Fujifilm the primary provider of control chips. What this means is that the differences between most state-of-the-art consumer products breaks down to:

  1. Manufacturer tolerance levels for incoming parts.
  2. Software changes to the fundamental interpolation routines (rare, though getting increasingly more frequent).
  3. Lens quality (and also the low-pass and UV filters mounted over the CCD).
  4. Ease of use factor of the camera's UI (user controls).
  5. Storage media used.

Only 1, 2, and 3 affect image quality, with 3 being the most obvious difference between most cameras (even within a manufacturer lineup--the Nikon Coolpix 990 and 880 have different lenses, and there is an observable difference in quality between the images they produce, despite the fact that virtually every other internal component is the same). If you're not printing 8x10" prints or larger, you may not even notice the differences. Even more to the point: after-shot manipulation with Photoshop or similar software can generally wipe out all the observable differences I've seen in cameras using the same CCD.

So, digital camera choice currently boils down to 4, and to a lessor degree 5. In short, your dealer is wrong. (Note: the Coolpix 995 does differ slightly. Nikon is using a CYMG filter array on top of the CCD instead of a RGGB [Bayer pattern] array. It appears that Nikon has also made a few minor changes to the software that runs on the Sparc set.)

top | home

Question: I've looked everywhere, but can't find a color profile for the Coolpix 990 for Mac users. What can I do?

Answer: First, some background for the rest of the readers: a color profile is a set of translation instructions for computer software to use with digital cameras, scanners, monitors, and printers. A perfect set of profiles makes it so that the red you see through the Coolpix viewfinder (or in the slide you're about to scan) is displayed as the same red on your monitor and prints as the same red on your printer. Without color profiles installed for all your equipment, colors drift from device to device, and that red may end up with too much yellow or cyan or magenta in the final print. Serious photographers care about the colors in their work, so they use color profiles for all their computer equipment. Photoshop, for example, has tools that allows the program to automatically apply profiles for scanners, cameras, monitors, and printers, ensuring that what you see on your screen is what you'll end up with in your final print.

Unfortunately, not all digital products have existing color profiles that you can download or purchase. But that's okay, since you can make your own with a product like MonacoEZcolor. You shoot a picture of the provided IT8 target, then use the program to do the analysis and create your color profile. MonacoEZcolor also can create scanner and printer profiles. You'll need to buy the accessory 35mm IT8 target slide if you want to profile your Coolscan.

Added 3/15/01: The folks that created Qimage Pro now have ICC profiles for the Coolpix 990.

top | home

Question: I've got a lot of older manual focus lenses that I'd like to use with my new D1 (D1X/D1H). Why can't the D1 use matrix metering with them?

Answer: Curiously, the F4 can matrix meter with older MF lenses, but the F5 and the D1 models can't. Obviously, Nikon changed something in how the metering system worked. However, if you really want matrix metering, you can have your older AI and AI-S lenses "chipped." You add to an AI or AI-S lens the chip that indicates the maximum aperture (apparently Nikon's current matrix metering system needs to know exact EV, not just the difference in EV values in each segment; and to get that, the camera needs to know the maximum aperture of the lens being used). These "CPU" chips are actually quite simple, and can be retrofit into virtually any Nikkor or Nikon-mount lens (only those that have maximum apertures of f/1.2, f/2.5, f/8, and f/11 can't be converted, as Nikon never made any AF lenses with those apertures; variable aperture zooms shouldn't be converted, either). Note, however, that focal length information transmitted to the flash may be incorrect, forcing you to manually zoom the flash head. Unfortunately, there's no longer any reliable source to chip a lens.

top | home

Question: How much resolution do I need in a digital camera?

Answer: I like to answer this question a bit differently than some other experts. Here's how I put it: if you can't get excellent 8x10 prints with a 3- or 4-megapixel camera, you're doing something wrong. If you can't get excellent 11x14 prints with a 5- or 6-megapixel camera, then you're doing something wrong. Note that you can produce larger images than what I just suggested, but since most people never stray beyond one of those two sizes, they should start their examination there. A good rule of thumb is that you should be able to resize up to 2x the pixels, so a 6-megapixel camera that produces 3000 x 2000 pixels should be able to generate images up to about 6000 x 4000 before things start to break down. Note further that you only need 240 dpi on most inkjet printers, so we're talking about images larger than most of the consumer printers can produce. Finally, I'll point out that Sports Illustrated and other publications have produced two page spreads from 4-megapixel cameras (and do so with some regularity). In general, resolution alone is not the arbiter of final image quality. Lack of noise and artifacts is much more important to me than extra resolution.

top | home

To get your question answered, send email to: QandA


bythom.com | Nikon | Gadgets | Writing | imho | Travel | Privacy statement | 2001 Thom Hogan. All rights reserved.