· 7 min read
Bayer Demosaic Conversion Versus a True Monochrome Sensor
Why removing the color filter array raises a digital sensor's resolution and sensitivity compared with desaturating a Bayer color file to grayscale.
Written in by Simon Lehmann Editor
A fixed-body camera holds the lens parallel to the film and centred on its axis, which forces two compromises. The plane of sharp focus stays parallel to the film, so a receding subject can only be held by stopping down to f/45 or f/64 and paying for it in diffraction and exposure time; and tilting the whole camera up to include a tall building makes its verticals converge. A view camera removes both constraints by letting the lens standard and the film standard pivot and slide independently. The whole craft rests on two geometric rules and one division of labour, and once you can see how they produce numbers you can lay focus where you want it and keep your verticals plumb.
A view camera is two movable frames joined by a bellows: the front standard carries the lens, the rear standard carries the ground glass and film. Each performs the same six movements. Rise and fall translate a standard vertically, parallel to the film plane; shift translates it horizontally in the same parallel manner. These three are pure linear displacements that change no angle between lens and film. Tilt rotates a standard about a horizontal axis, swing about a vertical one; these change the relative angle of the two planes.
The division of labour follows from a simple mechanical fact. A front tilt or swing only re-aims the lens: it moves the plane of sharp focus while the film stays put, so the magnification across the frame and therefore the subject’s shape are unchanged. A rear tilt or swing changes the angle of the film to the subject, which alters the magnification from one edge of the frame to the other, and that differential magnification is exactly what makes parallel lines converge or diverge. So the working rule is: front for focus, back for perspective. Ansel Adams gives the canonical photographer’s treatment of all six in The Camera (1980), and it is the best single reference to keep on the shelf.
With lens and film parallel, the plane of sharp focus is parallel to both. Tilt the lens and that plane swings away, but not arbitrarily. The Scheimpflug principle states that the subject plane, the lens plane and the film plane must all meet along one common line for the whole subject plane to be sharp; that line is the Scheimpflug line. The relationship is geometric, not optical. Theodor Scheimpflug, an Austrian army captain, set it out in British Patent No. 1196, filed 16 January 1904 and accepted 12 May 1904, “Improved Method and Apparatus for the Systematic Alteration or Distortion of Plane Pictures and Images.” He credited the earlier patent of the French engineer Jules Carpentier, British Patent No. 1139, filed 17 January 1901 and issued 2 November 1901, a perspective-correcting enlarger.
The Scheimpflug line tells you the focus plane passes through a line, but not which plane, since infinitely many planes can pass through one line. The hinge rule supplies the missing constraint. Harold M. Merklinger formulated it and coined the term hinge line in Focusing the View Camera: the plane of sharp focus, the front focal plane of the lens (the plane one focal length in front of the lens, parallel to the lens plane) and the parallel-to-film plane through the lens centre all converge along the hinge line. The distance J from the lens to that line depends only on focal length and tilt angle, alpha = arcsin(f / J), i.e. J = f / sin(alpha). For small tilts Merklinger gives a field approximation: alpha in degrees is about f / (5J), with f in millimetres and J in feet.
Put numbers in. You have a Schneider Symmar-S 210mm f/5.6 on 4x5, shooting a table-top that recedes from just below the camera, the relevant ground plane sitting about 10 ft below the lens axis. Then alpha is roughly 210 / (5 x 10) = 4.2 degrees of forward tilt. That is the whole point made concrete: about four degrees swings the plane of sharp focus from parallel-with-the-film all the way down onto the receding surface. A few degrees really is a large swing, because J is short.
Once tilted, the hinge line is fixed in space. Refocusing with the back can then do only one thing, Merklinger’s image: the plane of sharp focus “rock[s] like a teeter-totter on the Hinge line.” This justifies the standard ground-glass procedure: focus the far detail with the back, apply front tilt, refocus, and iterate. Each refocus rocks the whole plane about the fixed hinge, so a few passes converge on a setting that holds near and far together.
Stopping down still buys depth of field, but with a tilted lens that depth is not a slab parallel to the film. The near and far limits are themselves planes, pivoting about lines parallel to the hinge line and offset from it on either side. The in-focus region between them is therefore a wedge, with its thin end at the hinge near the camera and opening out with distance. The practical consequence is blunt: depth of field is meanest in the foreground close to the hinge and generous far off. When you place the focus plane, lay it so the shallow near-camera end falls where you have least subject depth.
This is where the aperture payoff lands. The receding plane that would have demanded f/45 or f/64 by brute depth of field alone can be held at f/22, or near wide open, once the focus plane is laid into the subject by tilt. You trade a couple of degrees of front tilt for two or three stops, escaping the diffraction softening and long exposures that the small apertures cost.
Every movement spends the lens’s image circle, the disc of usable image it projects. Coverage must exceed the film diagonal before any movement is possible; the surplus is what is available for rise, fall and shift. The 4x5 diagonal is about 153mm, 5x7 about 210mm, 8x10 about 312mm. A Symmar-S 210mm covers roughly 294mm at f/22 (about 70 degrees), which is generous on 4x5, just covers 5x7 as a normal lens, and will not cover 8x10.
For architecture, choose a wide-angle lens not only for its view but for its surplus coverage. A Nikkor-SW 90mm f/8 projects about 235mm at f/22, an angle of coverage near 105 degrees, against the 153mm 4x5 diagonal, leaving roughly 80mm of surplus. To photograph a tall building you keep the back plumb (no rear tilt) so the verticals stay parallel, then apply front rise into that surplus to bring in the top of the building. You are spending the wide image circle on rise, not on a wider view, which is why the wide lens is the right tool even when you don’t need the angle. On 4x5 the 90mm SW will use nearly all the rise it has.
Run past the edge of the image circle and you get hard mechanical cut-off, an abrupt dark corner where the format has run off the disc. That is distinct from the gradual fall-off toward the edge of coverage, which follows the cos^4(theta) law of natural illumination and dims smoothly rather than cutting. Wide lenses like the 90mm SW show enough cos^4 fall-off that they are commonly used with a centre filter to even the exposure across the frame. The available movement, then, is as much a property of the lens as of the camera.
· 7 min read
Why removing the color filter array raises a digital sensor's resolution and sensitivity compared with desaturating a Bayer color file to grayscale.
· 6 min read
How weighting red, green and blue channels in conversion reproduces the effect of physical filters, and where sensor color response sets the limits.
· 7 min read
Silver-halide grain is a clumped, developed structure; sensor noise is photon shot noise plus read noise. Why each looks distinct in a monochrome print.
The grainmag companion app
Meter and place your tones without a signal. No account, no internet required — just you, the light, and the grain.