Lensrentals once put ten copies of the same Nikon 70-200mm f/4 VR on an optical bench, focused each one at 70mm, then zoomed to 200mm without touching anything else. Best focus moved by an average of 0.374 mm, and the identical test on a cinema zoom of the same range gave 0.008 mm. The space between those two figures is the whole difference between a lens that zooms and a lens that only changes its focal length.
What Parfocal Actually Means
A parfocal lens holds focus while its focal length changes. Focus on a face at the long end, zoom back to wide, and the face is still sharp. A varifocal lens does not behave that way. Change its focal length, and the plane it was focused on slides forward or backward, so you have to focus again before the shutter is any use to you.
In optical design, that distinction separates a zoom lens from a lens that merely varies focal length, and the vocabulary is older than anything in your bag. Lenses with movable elements that changed magnification existed in the 1890s, and every one of them made the operator refocus after every change. A zoom lens, in the strict sense, is that idea with the refocusing engineered out.
The record credits Clile C. Allen with the first patent for such a design in 1902, and the first built in quantity was the Bell & Howell Cooke Varo 40-120mm for 35mm motion picture cameras in 1932. Roger Cuvillier's Pan-Cinor arrived around 1950 using an optically compensated zoom group, and Pierre Angénieux's landmark 17-68mm of 1956 set the pattern for the modern mechanically compensated zoom. Mechanical compensation itself was not his invention. The Varo was already moving its groups on cams in 1932, and Frank Back's optically compensated Zoomar is remembered precisely because it took the other route. Still photography came late and came in through the side door. The 36-82mm f/2.8 Zoomar of 1959, designed by Back, built by Kilfitt and sold under the Voigtländer name, is generally described as the first production zoom for 35mm stills. It was optically compensated, so it held focus exactly at only a handful of points in its range and drifted between them. The first zoom you ever mounted was not a zoom by the textbook definition, and neither is the one on your camera right now.
Parfocal is not a yes-or-no property either. No zoom holds perfect focus at every point in its range, so the real specification is a tolerance: how far focus is allowed to drift before the picture stops being acceptably sharp. Roger Cicala, who ran the bench test above, puts it plainly. "Parfocal is not an absolute definition," he wrote. "What is acceptable as parfocal to one person in one situation is unacceptable in another."
The Adjustable Compensator Your Photo Zoom Does Not Have
Inside any zoom, one group of elements does the zooming. Optical designers call it the variator, and sliding it along the axis changes magnification and therefore focal length. Left to itself, that motion also drags the image plane back and forth, which is exactly the behavior that makes a lens varifocal. Holding the image still takes a second moving group, the compensator, whose only job is to cancel the image-plane travel the variator creates.
There are two ways to move those groups. In an optically compensated zoom, the variator and compensator are locked together and travel identically, which is cheap to build and starves the designer of the variables needed to correct aberrations. Those designs faded by the 1960s and are now effectively extinct. Nearly everything modern is mechanically compensated instead, with cams or slot-and-pin mechanisms guiding each group along its own curved, non-linear path as you turn the barrel. A zoom lens design report from the University of Arizona's College of Optical Sciences walks through both forms, and it makes the point that even the good version only holds the image still to within a tolerance, never perfectly.
Cinema zooms spend real money on that tolerance. "Cinema zooms are made with a moving compensating group that has the specific purpose of maintaining focus while the lens is zoomed," Cicala wrote. "This group is adjustable to make certain it compensates properly. Photography zooms do not have anything like this compensating group." That adjustability is where the money goes. It buys more elements, tighter cam tolerances, and a technician who can trim the group's position when a lens comes back from a shoot.
Broadcast glass goes further and hands part of that calibration to you. Fujinon puts a back-focus ring on its broadcast and cinema barrels to set flange focal distance, because the same lens sits at a slightly different distance from the sensor on different bodies, and because a large swing in ambient temperature moves it too. Fujinon's back-focus procedure is the parfocal test written out in steps: lock the camera down, put a Siemens star chart about 10 feet away, open the aperture wide, zoom to the long end and focus with the front ring, then zoom wide and recover the lost sharpness using the back-focus ring rather than the focus ring. Repeat until zooming changes nothing. Your photo zoom has no such ring, because there is nothing on it you were ever meant to calibrate.
What an Optical Bench Says About It
The numbers come from a March 2016 bench test by Cicala and Aaron Closz. Across ten copies of the Nikon 70-200mm f/4 VR, all of them already tested and rated optically excellent, the average change in best-focus position from 70mm to 200mm was 0.374 mm. The extremes ran from 0.742 mm in one direction to 0.273 mm in the other, a spread of just over a millimeter across copies of a single lens model. One copy drifted forward and another drifted backward, and both came off the same production line. The Zeiss 70-200mm T2.9 cine zoom averaged 0.008 mm through the same test with a total range of 0.05 mm, and Lensrentals treats anything past that as a lens due for service.
Millimeters at the sensor are hard to feel, so convert them into something you can set on a dial. The slab of space at the image plane inside which a point still records as a point is the depth of focus, and it depends on two things only: the f-number and how large a blur circle you are willing to call sharp.
At f/4 with the classic full frame circle of confusion of 0.030 mm, the total depth of focus is 0.240 mm, and because you set focus before you zoom, the sensor starts in the middle of that slab with 0.120 mm of room in either direction. The average copy in that test drifted 0.374 mm, three times the room it had. The worst copy drifted six times the room it had. To bury the average drift inside that tolerance, you would have to stop the lens down to roughly f/13. Judging sharpness at the pixel level of a 45-megapixel body rather than at print size shrinks the acceptable blur circle to about 0.009 mm, which pushes the requirement past f/40. That is off the end of the barrel, and it is well inside the range where diffraction alone draws a circle several pixels wide. The cine zoom's 0.008 mm sits at about one eleventh of its own allowance wide open, which is why nobody who rents one ever thinks about this.
The standard has also moved under everyone's feet. A lens that looked parfocal in standard definition can be obviously varifocal in 4K, without one element having changed. The pixels got smaller and you started judging at 100 percent, so the blur circle you are willing to call a point shrank with them. Cicala's summary of the whole exercise was a single line: "The 'parfocalness' of a photography zoom has significant copy-to-copy variation." That is why a forum thread declaring some model parfocal gives you nothing you can act on. It tells you that one copy, on one body, satisfied one person.
Why Autofocus Hides the Problem
You have almost certainly never watched this happen, because your camera fixes it faster than you can notice. You zoom, you half-press, the motor shifts the focus group a fraction of a millimeter, and the frame is sharp before your eye has finished settling on the subject. The drift is there on every shot. It just never survives long enough to reach the file.
Lens designers know that, and they spend the freed constraint on everything else you shop for. A designer released from holding the image plane still across the zoom range can trade that away for more range, a faster maximum aperture, closer focusing, less weight, better aberration correction, and a lower price. A 24-105mm f/4 that weighs 700 grams exists partly because nobody asked it to be parfocal.
Canon's own product page for the RF 24-105mm F2.8 L IS USM Z lists "Electronic Parfocal Performance" among the lens' special features and credits it to the pair of Nano USM motors, which deliver "a professional electronic parfocal focus capability while zooming." The lens is not optically parfocal. It reads where the zoom ring is and drives the focus group to compensate in real time, which is a servo loop imitating a compensator group. Mirrorless focus groups are light and their motors are quick, so the imitation is convincing, and it is still a different thing from glass that never let the image plane move at all.
Nikon drew the line even more carefully. The NIKKOR Z 28-135mm f/4 PZ, announced on February 12, 2025, at a suggested $2,599.95, is a power zoom built for video work. Its announcement claims a design that suppresses focus breathing and an internal zoom that keeps the center of gravity steady, and it says nothing at all about parfocal behavior across the optical 28-135mm range. The single place the word appears is the Hi-Res Zoom feature on the Z8 and Z9, where "from 135mm-270mm, the zoom is effectively parfocal." That stretch is a sensor crop rather than an optical zoom. No glass moves, so nothing can drift, and the most reliably parfocal zoom on a modern camera turns out to be the one made entirely of pixels.
Where the Trick Stops Working
Autofocus covers for a varifocal zoom right up to the moment you ask the lens to do something the autofocus system is not watching.
- A push-in during a take, shot on manual focus, starts sharp and ends soft, and the audience reads it as a mistake rather than a move.
- The old camcorder habit of zooming in to nail focus and zooming back out to compose is precisely the operation a varifocal lens punishes, and it is the habit most people bring with them into stills bodies shooting video.
- Night work sets you up for it. You focus a 24-70mm on a bright star at 70mm because that is where focus is easiest to judge, then zoom to 24mm for the composition, and autofocus is not coming to help you at f/2.8 under a dark sky.
- A focus stack assumes the only thing changing between frames is focus distance. Nudging the zoom ring partway through changes considerably more than that.
- Adapted stills glass on a cinema body inherits the drift with nothing to mask it, since there is often no phase-detect coverage and frequently no autofocus running at all.
- Tethered studio work hides it best and hurts most. Set focus on the eye, reach over to tighten the crop with the zoom, fire, then spend ten minutes wondering why this frame looks softer than the one before it.
Those situations are what cine zooms are actually sold to solve, and the price gap looks absurd until one of them costs you a job. A Fujinon MK18-55mm T2.9 covers only a Super 35 sensor and gives up a great deal of the versatility of a full frame standard zoom, and what the money buys is a compensator group, a technician's ability to adjust it, and a back-focus ring for trimming the result against your specific camera. If you are moving from stills into motion work and trying to sort which cinema lens features are marketing and which ones change what you can actually shoot, the Introduction to Video tutorial covers the ground where the two disciplines collide.
How to Test the Copy in Your Hands
Since the answer belongs to your copy rather than to the model, measure your copy. The test takes about ten minutes and needs a tripod, a target with fine detail, and live view magnification.
Lock the camera down at your normal working distance for that lens rather than at infinity, because a zoom can be nearly parfocal at one distance and badly varifocal at another. Open the aperture all the way, since stopping down is the one thing guaranteed to hide the error you are hunting. Zoom to the long end, punch in to maximum magnification, and set focus manually as precisely as you can manage. Then zoom to the wide end without touching the focus ring, punch in again, and look. Reverse the direction and repeat, because the mechanism has slop and the result is not always symmetric. Finish by shooting a frame at each end and comparing them at 100 percent, since a rear screen at arm's length forgives a drift your files will not.
Then run it again at a second distance, and if you shoot video, run it a third time at your working aperture instead of wide open. Write the verdict on a strip of gaffer tape and stick it inside the lens cap. When a lens holds at 3 meters and falls apart at 15, you want that written down before you are on set with a client watching the monitor.
Lead image by Angry-Monkey Photography, CC BY-SA 2.0. Source.

4 hours ago
10







English (US) ·