Leer en español: El carácter de un objetivo se puede medir
Experiments with M-mount lenses
Why do two 50 mm lenses with the same aperture draw differently? What makes a 1961 Summilux produce that background that dissolves, while a modern aspheric leaves the outlines of the blur clearly marked? I have been turning this over for a while, and the short answer is that a lens’s character has a precise physical definition: it is the set of aberrations the designer decided not to correct. And if it can be defined, it can be measured, compared and put on a map.
That is what the M-mount optical character chart does. It is an interactive catalog of 129 lenses, from 1925 to 2026, that reconstructs the defocus disc of each one from its coefficients. It also defines a distance between lenses to answer the question behind almost all the others, namely which one resembles which and in what exactly. It runs entirely in the browser, with no server and no account, and comes with an illustrated manual with all the detail. Both are in Spanish.
1. The idea
A perfect lens has no character. Limited only by diffraction, it would form the same image as any other perfect lens of its focal length and aperture. Everything we recognize as a lens’s personality (the creamy bokeh, the halo over highlights, the nervousness of out-of-focus highlights) comes from the residuals: uncompensated spherical aberration, longitudinal chromatic aberration, field curvature, mechanical vignetting.
The tool takes eleven of those residuals as coordinates and places each lens in a space whose origin is the ideal lens. The distance to the origin tells you how much character there is; the direction tells you what kind. The 1933 Summar sits 3.93 from the ideal, and the APO-Telyt 135 at 0.74. The whole history of optical design fits between those two extremes.
2. The defocus disc, calculated rather than drawn
The centerpiece is the disc simulator. Every bokeh ball it shows comes from a geometric ray trace over the lens pupil, with a surprisingly compact formula behind it. The landing radius of each ray is
r(ρ) = aρ + bρ³
where a is the defocus, which changes sign between background and foreground, and b is the lens’s residual spherical aberration, which never does. From that asymmetry comes a result anyone who has used a classic lens knows without knowing why it happens: the same lens that gives a creamy background gives a foreground with hard outlines. When the two terms share a sign, the disc has a bright core and a fading edge. When they oppose each other, a caustic appears, and with it the bright ring of nervous bokeh. The simulator has a background/foreground switch, and pressing it with a pre-aspheric Summilux loaded is enough to watch the inversion live.
The model also includes the aperture polygon according to the real blades, off-axis pupil clipping (cat’s eye, calibrated in transmittance: the Noctilux f/1 passes 26% of its pupil’s light in the corner, almost two stops) and longitudinal chromatic aberration as a per-channel focus shift, from which the green fringe behind and the magenta one in front, as seen in real photographs, emerge on their own.
3. Three use cases
The substitute from another brand.
An APO-Summicron 50 costs what it costs. The similarity tab, with the other-manufacturer filter on, returns the Voigtländer APO-Lanthar 50 f/2 at a distance of 0.12, percentile 0: it is the closest cross-brand pair in the entire catalog. The table also says which generator separates each pair, so you can see at a glance whether the remaining difference is chromatic aberration, pupil geometry or veiling flare.
The false twin. The Voigtländer Nokton 50 f/1 is constantly compared with the Noctiluxes, and the tool lets you check how much truth there is in that. With the aspheric Noctilux 0.95 the distance is 0.27, second nearest neighbor in the whole catalog, and the breakdown shows that 68% of that distance is a single residual, the ripple from aspheric molding: in everything else they are almost the same lens. With the classic 1976 Noctilux f/1 the distance rises to 1.15 and is spread across six different generators. The two share the cat’s-eye geometry in the corners, which is what feeds the bar-room comparison, and differ in everything that governs the transition into blur. In practice, the Nokton substitutes for the 0.95, and resembles the classic f/1 only from a distance.
The one that needs calibration. The Zeiss C-Sonnar 50 f/1.5 shifts its focal plane when stopped down, and on a rangefinder that means it focuses where it is not pointing. Its sheet overlays that shift on the depth of focus: when the curve leaves the band, recalibration stops being optional. Filtering the catalog by the focus-shift trait returns the 25 lenses in the mount that deserve that scrutiny before you buy them for a rangefinder body.
4. The metric and its test
Defining a distance is easy; the hard part is proving it measures something. The test the tool uses is clean: each lens’s optical family (Sonnar, Biogon, double Gauss, apochromat) does not enter the distance calculation. If the metric captures something real, lenses of the same family should land close together on their own.
In the chart, the nearest neighbor shares the optical family in 73% of the 128 cases, compared with the 18% that chance would give. The groups the metric forms unaided are the four Biogons, the apochromats from three different manufacturers, and every reissue next to its original: the nearest neighbor of the 2025 Summilux Classic is the pre-aspheric E46, at 0.15, with the 1961 E43 in the 4th percentile of all pairs. That the reissue lands by itself on the series it reproduces is exactly what Leica intended when making it.
One validation is still pending, and the document says so in its own method: that the numerical distance corresponds to the difference an observer perceives. That requires a forced-choice experiment with pairs of discs, and until then the metric’s weights are a reasonable convention.
5. What is data and what is judgment
Every number in the chart belongs to one of three levels, and the tables always keep them apart. Cataloged values come from the manufacturer’s sheets: focal length, elements, blades, weight. Derived values are exact calculations based on those: entrance pupil, diffraction cutoff, the distance itself. And the ordinal indices (creaminess, veiling flare, microcontrast) are judgments on a 0 to 100 scale built from published MTF curves and sustained agreement between observers. That third level is the weak link, which is why the only lens in the catalog with a factory MTF, a Light Lens Lab 50 f/1.5, has the published points drawn over the modeled curve, with the discrepancies in plain view: the model does well at the center and fails at the edge at full aperture because it has no astigmatism term. I prefer a model with its failures disclosed to one that looks perfect because nobody has checked it.
6. Where to find it
The chart and its manual are open, each in a single HTML file that can be saved and used offline (both in Spanish):
The ordinal indices are interpretive. Before buying anything, check against your own measurements; the tool is for knowing what to measure and where to look.
Deja una respuesta