Micro-contrast has no standardized definition, and the most careful published one is perceptual rather than numerical: Zeiss’s own Hubert Nasse describes it as the contrast in structures you can just about see, or just cannot quite see, with the naked eye. The number that stands in for it on a data sheet is contrast at 40 line pairs per millimeter, which is a proxy for that band rather than a definition of it. Everything else the term is asked to carry, from glow to bite to three-dimensionality, either reduces to that number, reduces to a different number nobody publishes, or reduces to nothing you can test. Sorting those three piles takes an afternoon and a box lined in flat black.
The Property That Has a Number
Contrast, to an optical engineer, is a ratio. Photograph a pattern of black and white bars, then measure the brightest and the darkest values that come back out. The difference between them divided by their sum is the modulation, and it runs from zero for a uniform gray to one for true black against true white.
Modulation transfer function is that ratio measured in the image, divided by the same ratio in the target, computed one spatial frequency at a time. Spatial frequency is how tightly the bars are packed, counted in line pairs per millimeter at the sensor. Coarse bars at 10 lp/mm come through nearly intact. Fine bars at 40 lp/mm come through degraded. Every lens has some frequency above which the bars arrive as flat gray.
Manufacturers who publish MTF pick a few frequencies and plot contrast against distance from the center of the frame. Canon plots 10 and 30 lines/mm, and since late 2018, its charts show the lens wide open only. Nikon plots 10 and 30. Zeiss plots 10, 20, and 40. Leica publishes the most of anyone, and the data sheet for the Leica Summilux-M 35mm f/1.4 ASPH. says exactly what it is showing: contrast in percentage for 5, 10, 20, and 40 lp/mm across the height of the 35mm format, tangential and sagittal, in white light. Canon and Nikon label their frequencies lines per millimeter while Zeiss and Leica label theirs line pairs per millimeter, though both labels mean the same quantity: cycles per millimeter. The reasons charts from different makers do not compare cleanly lie elsewhere, in the frequencies each one chooses, in curves computed from the design rather than measured off a finished lens, and in different white-light weightings.
Lloyd Chambers, in an article Zeiss hosts on its Lenspire site, gave the term the cleanest definition it has ever received. “Micro contrast refers to contrast of very fine details, such as 40 line pairs per millimeter (lp/mm),” he wrote, adding that overall contrast is the 10 lp/mm figure and that 20 lp/mm covers the middle. Read that way, micro-contrast is neither mysterious nor unmeasurable. It is the right-hand end of a curve you can download as a PDF.
H. H. Nasse, who wrote many of the Carl Zeiss camera lens division’s technical papers, used the word the same way and drew a sharper line with it. In “How to Read MTF Curves,” published in December 2008, he pointed out that when photographers say contrast, they always mean micro contrast, meaning structures at the edge of what the naked eye can pick out. The black-versus-white difference of a chessboard filling the frame is something else, which he called macro contrast, and which depends on veiling glare rather than on how well the lens is corrected.
How Fine 40 Line Pairs per Millimeter Really Is
A Sony a7R V lays 9,504 pixels across 35.7 mm of sensor. That is a pitch of 3.76 µm and a Nyquist limit of 133 lp/mm, which is the finest pattern the sensor can sample at all. The 30 and 40 lp/mm curves the manufacturers publish therefore sit at 23% and 30% of that sampling limit. Those frequencies were chosen in the film era, and they describe fine detail rather than the finest detail a modern body records.
Put a file on the wall, and the arithmetic gets friendlier. An 8-by-12-inch print from a full frame file is an 8.5x enlargement, and at the classic 25 cm viewing distance, one degree of your visual field covers 4.36 mm of paper. Detail recorded at 10 lp/mm reaches your eye at about 5 cycles per degree, which is almost exactly where human contrast sensitivity peaks. Detail recorded at 40 lp/mm reaches your eye at about 21 cycles per degree, well down the sensitivity curve though nowhere near the roughly 60 cycles per degree where vision gives out entirely.
Micro-contrast is real, then, and you can see it. You are simply several times less sensitive to a given percentage of contrast loss at 40 lp/mm than at 10 lp/mm, at normal print sizes and normal viewing distances. Double the print, and the balance shifts back. That relationship, not any property of the glass, is why the same two lenses can look identical in a web-sized image and different in a 24-inch print.
The Ceiling Your Aperture Already Set
Before comparing two lenses at 40 lp/mm, check how much contrast your aperture has already spent. Diffraction alone, in a lens with no aberrations whatsoever, caps modulation transfer at a value that depends only on f-number and wavelength.
At f/2.8, a perfect lens returns 92% modulation at 40 lp/mm. That falls to 78% at f/8, 70% at f/11, 56% at f/16, and 41% at f/22. The same calculation run at 10 lp/mm gives 98, 94, 92, 89, and 85%. Stopping down from f/5.6 to f/16 therefore costs a perfect lens 28 points at the fine end and 7 points at the coarse end.
Twenty-eight points is a larger swing than you will find between most pairs of good modern lenses at the same aperture. Shooting two 24mm lenses at f/16 does not invalidate a comparison between them, since diffraction charges both the same toll. It does mean the quantity being argued over is a sliver sitting on top of a much larger loss you chose. Micro-contrast is not a fixed property of a lens. It is a curve that moves every time you turn the aperture ring.
Processing moves it too. Nasse described edge enhancement as making the bright side of an edge brighter and the dark side darker, and noted that this “improves the micro contrast and the edge steepness, and the subjective impression of sharpness is significantly improved, without significantly increasing the resolution of detail.” A small-radius unsharp mask lifts the transfer function in precisely the band where micro-contrast lives, and Lightroom’s develop module applies one to every raw file you open before you touch a slider, at an amount of 40 and a radius of 1.0 pixel. Clarity works at the other end of the scale, a large-radius midtone control that shifts coarse contrast rather than fine, and it sits at zero until you drag it. Deconvolution goes further, modeling the blur of a specific lens and camera pairing and inverting part of it, which is the trick behind tools like DxO PureRAW. Neither can restore detail that fell below the noise floor, and both leave halos when pushed, but both add exactly the quantity the forums insist cannot be added in post.
The Part the Charts Genuinely Miss
Nasse spelled out the real limitation in the second part of the same paper, published in March 2009. “The contrast at a spatial frequency of zero is always set equal to 100%,” he wrote. “Therefore the MTF curve describes the change relative to this reference value; it does not measure the absolute image contrast. For the same reason, the MTF does not take into account any deterioration of the contrast due to stray light.”
Stray light behaves like a second, very broad point spread function lying underneath the sharp one. Some fraction of the light that should have landed on a point instead scatters off barrel walls or bounces between glass surfaces and smears across a large part of the frame. That broad component carries no information above a few line pairs per millimeter, so at every frequency you care about it does nothing but scale the sharp core down by a constant factor of one minus g. Normalizing the curve to 100% at zero frequency does not divide that factor back out. It survives, at every frequency, which means a measurement that genuinely captured the glare would show it. The reason a published chart does not is simpler and more damning: the chart never contained it. Nasse says so outright: “MTF measurements say nothing about this macro contrast. They gauge only the correction of the lens, i.e. the small deviations of the light beams, while the macro contrast depends on the veiling glare of the lens, i.e. on the large deviations.” A design-stage curve is a sequential ray trace with no scatter or double-bounce paths in it at all, and a conventional MTF bench is configured to characterize the core imaging response rather than the faint, wide halo around it. Recording that halo takes angular coverage and dynamic range the ordinary setup is not asked to provide, which is why stray light gets a measurement of its own rather than falling out of this one.
A lens that dumps 3% of every highlight into a uniform haze and a lens that dumps 0.3% can publish the same MTF curve. On paper, they are twins. In a backlit frame, one has a black point well above the other. A veil that is genuinely uniform is largely recoverable, because adding the same value to two tones leaves the difference between them intact and a black-point move puts it back where it was. What does not come back is the part of the veil that is not uniform, along with whatever it pushed up into clipping or down into the noise. Shooting into the light is where the difference becomes obvious, and it is the single most common situation people describe when they say a lens has pop.
The property has a standard. ISO 9358, first published in 1994 and reconfirmed in 2019, defines both the veiling glare index and the glare spread function and specifies how to measure them, using either an integral black-patch method or an analytical one. A revision that adds point source transmission is in development. No camera lens manufacturer publishes a veiling glare index for a photographic lens.
Other gaps are narrower but real. Canon’s own documentation notes that its charts cover one aperture and one focus distance and say nothing about distortion, vignetting, chromatic aberration, coma, or flare. Curves computed at a single wavelength also flatter a lens by hiding longitudinal chromatic aberration, the defect apochromatic designs exist to remove, and makers do not all use the same convention. Zeiss and Leica both state that theirs are made in white light. Compare an apo lens against a non-apo lens using charts drawn under different conventions, and you are comparing the conventions.
Where the Folklore Runs Out
The most repeated mechanical claim in the micro-contrast literature is that modern lenses have too many elements, that every air-to-glass surface bleeds contrast, and that a simple old design therefore holds tone better. The first half of that is physics.
Reflection at an uncoated boundary between air and a common crown glass runs about 4%, which falls out of the Fresnel equations for a refractive index of 1.52. A quarter-wave magnesium fluoride layer, the standard single coating from the 1940s onward, drops it to roughly 1.3% at its design wavelength. Modern broadband stacks hold under 0.5% across the visible band. Veiling glare from internal reflection needs two bounces, one off a later surface and one back off an earlier one, so every pair of surfaces in the lens is a candidate path.
Run it. A 6-element, 4-group double Gauss has 8 air-to-glass surfaces and 28 pairs. Uncoated, it throws about 4.5% of the light into ghosts. Single-coated at 1.3% per surface, it throws 0.47%. A modern 15-element, 11-group design has 22 surfaces and 231 pairs, more than eight times as many paths, and at 0.5% per surface it throws 0.58%. At 0.3% per surface, it throws 0.21%, less than the single-coated classic with fewer than half as many pieces of glass.
The folklore has the mechanism right and the arithmetic backwards. Element count hurts, and it hurts quadratically. Coatings help quadratically as well, and they have improved by more than element counts have grown. The lenses most often praised for glow are the uncoated and single-coated designs of the 1930s through the 1960s, and those are the ones actually flooding the shadows. Some of that glow is the glare, but not all of it, and the difference is visible in a histogram. Nasse shows both: a lens with poor micro-contrast broadens the white peak while leaving the gap between black and white where it was, and a lens with heavy veiling glare lifts the black peak bodily. The first is aberration, usually residual spherical, hanging a halo on the point spread function. The second is stray light landing a long way from where it was aimed. Old glass tends to have both, and only the second one comes off with a black-point slider.
In July 2017, Nasim Mansurov published a satirical piece on Photography Life arguing that modern lenses are over-corrected junk, that lead-free glass ruined them, that nothing with more than nine elements is worth keeping, and that the incompatibility of quantum mechanics and general relativity dissolves once you accept the luminiferous ether, a conclusion he claims to have reached by thinking about glass-to-air surfaces. It is filed under the tag “Satire” and closes with a note saying so: “I hope our readers realize that this article is a satirical piece, aimed at poking fun at those individuals and websites that post nonsense information about lenses and their ‘unique’ qualities.” The comment thread has since drawn hundreds of replies, including readers who could not tell.
Speaking in the voice of the position he was mocking, Mansurov wrote that “this is obviously not quantifiable or measurable, which is why you won’t see discussions on ‘3D pop’ in various reviews, but they are easily seen in images.” Written straight, that sentence appears in a hundred forum threads a year. A property defined as unmeasurable, and visible only to people who already accept it, is not a property of a lens. It is a description of the observer.
The objection Nasse himself kept making survives the parody. Published MTF describes one focus distance, usually infinity, at one or two apertures, at two or three frequencies, on a flat target, with the zero-frequency contrast thrown away. It says nothing about how a lens draws what is out of focus, nothing about its behavior at a meter, nothing about flare. “If I know at which frequency the loudspeakers will have their maximum transmission or how loud 440Hz are, I still don’t know how music sounds,” Nasse wrote. A thin description is not the same thing as a wrong description, and the honest complaint about MTF charts is that they are thin.
How to Settle It in an Afternoon
Two lenses, one body, a tripod, the same aperture, the same light, raw files, and every sharpening and lens-correction default switched off. Shoot the same frame with each, and take the filters and the hoods out of the variables by using both or neither.
Then run the test that actually discriminates. Cut a fist-sized hole in a white board and back it with a shoebox lined in flat black, open face toward the camera, so the lens looks into a cavity instead of at a surface. Light the board hard and evenly, frame so the opening sits in the middle of a large bright field, and expose for the board. Whatever brightness comes back inside the opening is stray light. Black velvet taped to a wall will not do, because even good velvet returns a fraction of a percent of the light hitting it, which is the size of the thing you are trying to read; ISO 9358 realizes its black patch as a light trap for that reason. Read the result as a comparison between your two lenses rather than as an absolute veiling glare index, since one flat board lights a smaller share of the field than the uniform surround the standard calls for. A lens reading 2% of the board’s value and a lens reading half a percent will not render a backlit forest the same way, no matter how closely their charts agree.
Then equalize. Set the same black point and the same white point on both files and look again. Most of the pop that survives a matched aperture does not survive a matched black point, because most of it was macro contrast wearing a technical-sounding name. What remains after that is worth arguing about at 100%: the color fringing on a backlit branch, the halo shape around a specular highlight, the way a flat field holds the corners of a brick wall at two meters. Longitudinal chromatic aberration, residual spherical aberration, field curvature. Each one has a test you can run this weekend, and not one of them needs a new word.




