Start with the number, because every other claim in this piece answers to it. The total depth of focus at the film plane is 2 x N x c, where N is the f-number and c the circle of confusion. For 35mm, c is conventionally 0.03mm. So a lens at f/2 gives 0.12mm of total depth of focus, or +/-0.06mm; wide open at f/1.4 the budget collapses to about 0.084mm, +/-0.042mm. That is roughly half the diameter of a human hair, and it is the target both a rangefinder and an SLR must hit on every frame. Everything that follows is about who hits it and how each system fails.

The Coincident-Image Mechanism

A rangefinder triangulates. Two windows separated by the mechanical base length view the subject from slightly different angles. A cam coupled to the lens helix rotates a beamsplitter so that a superimposed image in a central patch shifts horizontally; when the lens is focused at the subject distance, the two images coincide. The principle goes back to the coupled coincident-image rangefinder Leitz built into the M3 in 1954.

The geometry is a long, narrow triangle, and its limiting term is the eye. The human eye resolves about one arcminute, roughly 0.0003 radian, so that is the smallest angular misalignment between the two images you can judge. That fixed angular error, projected back through the optics, becomes a distance error, which the lens then converts into a defocus at the film plane. The wider the base and the more the patch is magnified, the smaller the distance error for a given angular slop.

That triangulation is independent of the taking lens. A rangefinder focuses a 21mm and a 90mm with identical mechanical precision, because the patch knows nothing about the lens in front of it. The catch is that the required precision is wildly different between those two lenses, and the rangefinder has no way to know that either.

Effective Base Length, and Why Magnification Matters

Raw base length understates accuracy, because the patch is viewed through a magnifying eyepiece. The figure that governs real performance is the effective base length (EBL): mechanical base length multiplied by viewfinder magnification. The Leica M-A (Typ 127) datasheet states all three figures directly: a 69.25mm mechanical base, a 0.72x finder, and a 49.9mm EBL.

Magnification and base trade against each other, which is why EBL, not raw base, is the number to watch. Comparing bodies built on the same 69.25mm mechanical base:

• Leica M3, 0.91x finder: ~63mm EBL
• Leica M6/MP, 0.85x finder: ~59mm EBL
• Standard M6/M-A/MP, 0.72x finder: ~49.9mm EBL
• Leica CL: ~18.9mm EBL

The Voigtländer R3A makes the point in reverse: a 1.0x finder, but only a 37mm mechanical base, so its EBL is just 37mm despite the life-size view. A small base behind a strong finder loses to a long base behind a modest one. The M6 alone proves it: swap its 0.72x finder for the 0.85x and the EBL climbs from 49.9mm to roughly 59mm on identical hardware.

The Long-Lens Ceiling, Derived

The minimum EBL needed for an accurate focus is b' = (e x f^2) / (k x z), where e is visual acuity in radians (~0.0003), f is focal length, k is f-number, and z is the circle of confusion (0.03mm). Two terms do the work: required EBL rises with the square of focal length and falls with f-number. Long fast lenses are punishing on both counts.

Walk it through. A 50mm f/1.4 wants roughly (0.0003 x 50^2) / (1.4 x 0.03) = about 18mm of EBL, well inside the 49.9mm of a 0.72x M body. A 90mm f/2 wants (0.0003 x 90^2) / (2 x 0.03) = about 40mm, still under 49.9mm but with little margin left once you allow for a slightly tired eye or a patch that is not perfectly aligned. Push to 90mm f/1.4 and the requirement jumps past 57mm, beyond what a 0.72x finder delivers; you would need the M3’s 63mm.

This is the real reason the longest native rangefinder-coupled M lens is a 135mm, and never faster than f/2.8 rather than f/2. The fastest one, the 135mm f/2.8 Elmarit-M, even shipped with a permanent 1.5x magnifier over the finder, boosting the effective base specifically to keep that aperture honest. The optics for a fast 135 are not the obstacle; the rangefinder is. The f^2 term means a 135mm needs more than seven times the EBL of a 50mm at the same aperture, and no 35mm rangefinder base is long enough to keep a fast one honest.

A Frame You Would Actually Shoot

Take a head-and-shoulders portrait on HP5 Plus rated at EI 400, a 90mm f/2 wide open, focusing on the near eye. Two separate things have to go right. Depth of field in front of the lens decides how much of the face sits acceptably sharp, eyelashes versus eyebrow; that is a property of the optics and the distance, the same on either system. But whether the eye lands sharp at all depends on whether the focusing system’s error stays inside the +/-0.06mm budget at the film plane. On the rangefinder, that is the angular slop of the patch projected through a 49.9mm EBL, near its limit at 90mm f/2. On an SLR, you are looking at that exact plane on a bright f/2 screen and confirming it directly. Same negative, two different ways of being wrong.

Through-the-Lens SLR Focusing and Its Failure Modes

An SLR sidesteps the triangulation entirely by focusing on a screen whose matte surface lies at a path length optically equal to the film rails, folded there by the mirror box. That screen is rarely plain ground glass; it pairs a matte surface with a Fresnel field lens that evens out brightness across the frame so the corners do not go dim. Because you judge the actual projected image, accuracy scales with the lens: a faster, longer lens throws a steeper cone and snaps in and out of focus more obviously, exactly the regime where the rangefinder runs out of base.

The aids have a design trade built in. A split-image or microprism wedge is cut to a particular cone angle, and the steeper the wedge the stronger the focus snap, but the wider the aperture at which it darkens. The common split+microprism screen, the de facto standard on manual-focus 35mm SLRs through the 1980s and descended from the Nikon F of 1959, is built around roughly an f/4 cone. At f/5.6 the eye must be precisely centred or one half of the split darkens; by about f/8 one half is always black and you are forced back onto the plain matte collar. Designers cannot have both a hard snap and a screen that works stopped down on a slow lens; they choose.

Parallax and the Shared Hidden Variable

The rangefinder’s other native weakness is the one in this article’s title. Because the finder sees from beside the lens rather than through it, the bright-line frames shift as you focus, the camera’s attempt to correct parallax, and even corrected they are worst at minimum focus, around 0.7m on most M bodies. The finder shows you neither the true field of view nor the actual depth of field; you frame by an approximation. The SLR, sharing one optical path, frames exactly at any distance, macro included.

Underneath both systems sits the same hidden variable: a mechanical reference held to about 0.04mm. On the rangefinder it is the cam, roller, and vertical alignment of the patch; on the SLR it is the mirror rest, the screen seat, and the flange-to-film distance. Get it wrong and focus shifts invisibly, which is a measurable claim, not a rhetorical one. On a 90mm f/2 Summicron a shim error of about 0.04mm consumes essentially the whole depth-of-focus budget; the Hexar RF, with its wide flange tolerances and short 0.6x finder, is well documented to drift past infinity when that adjustment screw is even slightly out, exactly the failure a fast 90 punishes hardest. The rangefinder’s limit is its fixed base; the SLR’s is its dependence on a bright, accurately seated screen. Both live or die inside that same +/-0.04mm.