D-SLR State Of The Art, Part II
What to know about the latest digital sensors, ISO and image quality
Has a full-frame sensor measuring 35.9x24.0mm with a maximum resolution of 6048 x 4032 pixels, translating to a pixel size of 5.94 microns. That’s larger than the D300’s, but much smaller than the D3’s. We haven’t had a chance to use this new camera yet, but based on pixel size alone, we’d expect its image quality to fall between that of the D3 and the D300.
But wait! The D3X has twice as many pixels as the D3 (albeit smaller ones). To produce a given-size print, a D3X image can be downsampled to match the D3’s resolution, and doing so results in reduced image noise. It turns out that the D3X’s much higher pixel count trumps the D3’s bigger pixels, and the D3X images will produce better-looking prints of any given size, all else being equal.
The same holds true for Canon’s 21.1-megapixel, full-frame EOS-1Ds Mark III and 10.1-megapixel, APS-H-format EOS-1D Mark III. The 10-megapixel model has bigger pixels and better per-pixel performance, but the 21.1-megapixel model delivers better-looking prints of any given size (and especially at sizes beyond the 10.1-megapixel image’s range).
The Plot Thickens
To further complicate things, all else isn’t always equal. Consider Canon’s new EOS 5D Mark II. It uses a 21.1-megapixel sensor quite similar to the one in the EOS-1Ds Mark III, with the same 6.4-micron pixel size, yet it has a top normal-range ISO setting of 6400 and a top expanded ISO setting of 25,600 vs. the EOS-1Ds Mark III’s top normal-range setting of 1600 and top expanded ISO of 3200. And Canon states that the EOS 5D Mark II produces the best image quality of any EOS D-SLR ever. What happened? Well, in part, Canon’s new DIGIC 4 image processor; in part, a new RGB color filter for the sensor; and, in part, proprietary improvements throughout to raise sensitivity and lower noise.
While microlenses increase effective sensitivity, they don’t help much with dynamic range. That’s because a given-size pixel can hold only so many photons.
If you have a one-gallon bucket with a diameter of six inches and a one-pint bucket with a diameter of two inches, and stick them both out in the rain, the large bucket will collect more water faster. If you put a funnel with a six-inch diameter over the one-pint bucket, it will collect more rainwater faster, but it can still hold only one pint.
When a pixel fills with photons, saturation—“blow-out” of highlights—results, and photons can even spill over into adjoining pixels. This is sometimes referred to as “blooming.”
Here again, sensors with bigger pixels tend to deliver greater dynamic range than sensors with smaller pixels. However, today’s sensors and image processors can deliver a wide dynamic range even with multimegapixel cameras and their relatively small pixels.
|How Accurate Are Digital ISO Speeds?|
Practitioners of the Zone System made famous by Ansel Adams discovered that exposing film at their ISO speeds (ASA speeds, in those days) didn’t give them the desired density on Zone I; the shadow areas were underexposed. Their film-speed testing generally revealed a need to feed a lower ASA speed into their exposure meters.
If you use a handheld meter, or the Sunny 16 rule, you’ll likely find that digital ISO speeds also are a bit optimistic. The Sunny 16 rule states that for a frontlit subject in bright sun, the proper exposure is a shutter speed of 1/ISO at ƒ/16, or its equivalent. For ISO 200, that would mean an exposure of 1⁄200 sec. at ƒ/16, or 1⁄400 sec. at ƒ/11, or 1⁄800 sec. at ƒ/8, etc. I’ve found that with most D-SLRs (I’ve tested 52 over the years), the Sunny 16 rule results in underexposure. (The cameras’ built-in TTL metering pretty well compensates for this, but I’ve found some cameras require +0.3 or +0.6 exposure compensation for most scenes.)