The twin-lens reflex solved a problem that vexed early reflex design: how to see the image up to the instant of exposure without a mirror that swings out of the light path. Its answer was to split the two jobs between two separate lenses, stacked vertically. The lower lens forms the image on film; the upper lens, of matching focal length, sends its image up through a fixed 45-degree mirror to a ground glass viewed from above. Because the viewing path never crosses the taking path, the screen stays bright through the moment the shutter fires. Franke & Heidecke fixed the template with the original Rolleiflex in 1929, and three decades later the interchangeable-lens Mamiya C series of the 1960s pushed the design to its mechanical limit. The cost of that elegance is built into the geometry, and it shaped how the cameras were used.

Two lenses, two viewpoints

Stacking the lenses means they observe the scene from positions separated by their vertical spacing. On the Mamiya C series the lens-axis separation is exactly 50 mm. At infinity that displacement is negligible, but as focus moves closer the two fields of view diverge: the taking lens, sitting lower, records a frame shifted downward relative to what the ground glass shows. This is parallax error, and it grows as subject distance shrinks, which makes it most acute in close portraiture and copy work.

The size of the error follows from similar triangles. The vertical shift of the frame at the subject plane is just the lens baseline b; expressed as a fraction of the captured frame it scales with magnification, so it climbs steeply as you focus closer. Take the Mamiya’s b = 50 mm. At a 1 m subject distance the taking lens sees a field offset downward by 50 mm; against a vertical field of view of roughly 0.5 to 0.6 m at that distance, that offset is about 8 to 10 percent of the frame height. Halve the distance to 0.5 m and the same 50 mm becomes roughly 15 to 20 percent of the frame. That is the difference between a comfortably composed head-and-shoulders and a portrait with the top of the skull guillotined off the negative.

Correcting the offset

Manufacturers attacked the problem in stages. The crudest aid was a set of engraved correction marks on the ground glass showing how the frame would migrate at close focus. The Rolleiflex Automat, introduced by Franke & Heidecke in 1937, did better: a moving frame beneath the focusing screen, coupled to the focusing mechanism, tracked the taking lens automatically across the whole range from infinity down to 0.9 m, so the indicated field always matched what fell on film. For tripod work the Mamiya Paramender takes the most literal approach of all. It raises the whole camera by exactly 5 cm, equal to the C-series 50 mm lens spacing, so after you compose and lock focus the taking lens rises into the precise position the viewing lens had occupied. The offset is not estimated but physically cancelled.

The optical close-up solution is the Rolleinar, a Rollei attachment set supplied in Bay I, II and III mounts and strengths 1, 2 and 3, with the Bay I kits covering roughly 40 in down to about 10 in. It is a paired attachment. The element over the taking lens is a plain close-up dioptre; the element over the viewing lens carries an offset wedge prism, the Rolleiparkeil, fitted with a red alignment dot that must sit uppermost. That prism tilts the finder image downward by just enough to realign it with the taking lens, correcting parallax optically at the moment of viewing rather than after the fact.

The waist-level view and its reversal

The single 45-degree mirror corrects the image vertically but not horizontally, because the number of reflections governs handedness. One reflection flips the image left-to-right while leaving up and down intact, so the ground glass presents a laterally reversed picture: a subject stepping left in reality slides right on the screen. A pentaprism restores a fully correct image precisely because it adds further reflections to make an even total, undoing the flip. The TLR keeps the single mirror, so held at waist level and viewed from above, that reversal is the standard working condition and a persistent difficulty with moving subjects.

The viewing lens carries no diaphragm. On a Rolleiflex it is a Heidosmat, typically f/2.8 or f/3.2 depending on the model, chosen purely to throw maximum brightness onto the screen, while the taking lens is a separate and stoppable Tessar, Xenar, Planar or Xenotar. Because the viewing optic never stops down, most TLRs cannot preview depth of field at all. The rule is not absolute: certain Rolleiflex models and the Mamiya 105 DS lens, which carries a stoppable diaphragm in the viewing optic, provide depth-of-field preview as documented exceptions.

Why the negative is square

Most TLRs record on 120 roll film, which Kodak introduced in 1901 for the Brownie No. 2. The film is nominally about 61 mm wide and paper-backed, and the standard medium-format TLR frame is a nominal 6x6 cm that actually measures about 56x56 mm, yielding twelve exposures per roll. The square is not an arbitrary choice. A non-square frame would require rotating the body to switch between portrait and landscape, but a waist-level finder cannot be turned on its side without becoming unusable, and rotating the camera would change the vertical relationship of the two lenses. The square sidesteps the problem entirely: every exposure has the same orientation.

That geometry pushes the decisive choice to the printing stage. You compose on the 56 mm square ground glass and frame loosely, then decide at the enlarger whether to print the full square or crop to 645 or a tighter rectangle, dodging the burden of orientation out of the moment of exposure. The discipline is concrete rather than mystical: a roll of FP4 Plus developed in ID-11 gives a negative dense enough to crop hard without the grain falling apart, so the deferred decision costs you nothing in print quality. The square is what you commit to at capture; the rectangle, if any, is what you earn in the darkroom.