JSD6628 6.6mm F2.8 M16 Lens: FOV and Ball Detail on 5MP
JSD6628 6.6mm M16 Lens: Specifying Field of View and Ball Detail on a Wide 5MP Sensor
On an 8.5 mm wide sensor, focal length and horizontal coverage are the same decision. Here is the arithmetic that decides it, and what an M16 lens at F2.8 can actually deliver.
Key takeaway: The JSD6628 is a 6.6 mm fixed-focal M16 lens with an F2.8 aperture, a True 50MP rating and distortion specified below -3 percent. On a 2464 x 2064 sensor with 3.45 µm pixels its calculated coverage is 65.6 x 56.7 degrees, and holding 90 degrees horizontally on the same 8.50 mm active width would take a focal length of 4.28 mm or shorter. On this sensor the two goals are settled by one number: coverage or ball detail, not both.
The design brief behind a wide-angle tracking camera
A camera mounted beside the net has to hold both players, the racket arc and the flight path of the ball inside one frame, and it has to resolve the ball well enough for a detection algorithm to fit a trajectory. Those two requirements are bought with the same millimetres of focal length, which is why the arithmetic in the next section comes before the product table.
The sensor sets the budget for both. A 2464 x 2064 CMOS imager with 3.45 µm pixels presents an active area of about 8.50 x 7.12 mm and a diagonal of about 11.08 mm. Those are not round numbers the lens can be chosen around; they are the aperture through which every degree of field of view and every pixel of ball detail has to pass.
Four failure modes show up on court, and each is cheaper to avoid on paper than in a re-tooled housing:
Coverage shortfall. The angles published for a candidate lens were measured on a different sensor format, so the horizontal field on the real imager is narrower than the design assumed, and the far court edge falls outside the frame.
Undersized image circle. A lens whose image circle is smaller than the sensor diagonal produces corner shading and a soft, unrecoverable corner. On a wide sensor the corners are exactly where the far player sits.
Signal starvation. At 120 fps the exposure budget per frame is limited, and a small aperture that was acceptable at 30 fps turns into a gain and noise problem once the frame rate rises.
Shutter architecture mistaken for an optics problem. On an imager whose readout is progressive rather than global, a fast ball is displaced row by row during readout and smears across a band of pixels that no aperture and no focal length can recover. Shutter type belongs in the sensor selection, before the lens is specified.
None of these is an argument against wide-angle M16 optics. Each is an argument for doing the coverage arithmetic before the lens is ordered, and for choosing the part whose published data lets that arithmetic be finished.
"Field of view is not a property of a lens. It is a property of a lens and a sensor together, measured on one stated axis."
— JSD optical engineering note, wide-format M16 designsThe hard constraint: angle and focal length are locked
For a rectilinear lens focused near infinity, the horizontal field of view follows directly from the focal length and the width of the sensor's active area:
HFOV = 2 x arctan( sensor width / (2 x focal length) )
Two consequences follow, and they run through the whole selection process.
Focal length sets angle. On an 8.50 mm wide sensor, a 6.6 mm lens gives 65.6 degrees horizontal; a 6.0 mm lens gives 70.6 degrees; a 4.28 mm lens gives 88.0 degrees. There is no version of a 6 mm lens that reaches 88 to 90 degrees horizontal on this sensor, because the geometry does not contain that outcome.
The two requirements trade against each other. Angle and detail are bought with the same millimetres of focal length. That trade can be quantified before any lens is ordered. The pixel span of a tennis ball at 15 m is:
pixels = ball diameter x focal length / ( distance x pixel pitch )
With a 67 mm ball, 3.45 µm pixels and a 15 m working distance, the constants give about 1.294 pixels of ball span per millimetre of focal length.
| Focal length | HFOV on an 8.50 mm wide sensor | Ball span at 15 m | What it buys, and what it costs |
|---|---|---|---|
| 4.28 mm | 88.0° | 5.6 px | Meets the coverage target; about 20 percent less ball detail than the 5.17 mm reference |
| 4.58 mm | 85.7° | 5.9 px | Still inside the coverage family; ball detail about 11 percent below a 5.17 mm reference lens |
| 5.17 mm | 78.9° | 6.7 px | The common starting point for this sensor class; neither target is met |
| 6.6 mm (JSD6628) | 65.6° | 8.5 px | Ball detail about 25 percent higher than the reference; horizontal coverage 22 degrees short of the target |
Read as a table rather than as two separate requirements, the brief resolves into a single question: which 10 to 20 percent matters more on court, the far-court edge of the frame or the pixel count on the ball? A system that plans pose estimation and racket tracking across both halves of the court usually keeps the coverage; a system that plans to fit ball trajectories against a known court geometry usually keeps the detail. Both answers lead to a workable camera. What does not work is a brief that asks for both and assumes the lens will supply them.
Two further numbers belong in the same calculation. Vertical coverage of 50 degrees or more is comfortably met by any lens in this table, since the sensor is 7.12 mm high: the 6.6 mm part gives 56.7 degrees vertical, and even a 4.28 mm part gives 79 degrees. Distortion matters more than it looks, because a court is a set of straight lines that the algorithm has to trust; a per-cent-level barrel figure is straightforward to characterise and correct, but an undocumented figure has to be measured by the software team before it can be corrected at all.
JSD6628: the measured specification
The JSD6628 is a 6.6 mm fixed-focal lens in an M16 (S-mount) barrel, rated for True 50MP and specified for 1-inch-class large-format sensors. It is the point in the JSD M16 range where a longer focal length and a wide image circle are available in the same small barrel, which is what makes it the natural candidate when ball detail is the priority and coverage has to be re-planned around it.
For a project whose housing drawing is already moving, the figures that matter are these: 6.6 mm focal length, F2.8 aperture, M16 x P0.5 thread, 102 degrees diagonal field of view at the 1-inch format (91 degrees horizontal as published), a True 50MP resolution rating, and distortion specified below -3 percent. Total track length and the mechanical interface are covered in the drawings below.
What each figure changes in the design
| Specification | Published value | What it lets you decide |
|---|---|---|
| Focal length | 6.6 mm | Covers 65.6° horizontal and 56.7° vertical on an 8.50 x 7.12 mm sensor. Fixes the mounting position for the required court coverage, and fixes the ball footprint at every distance. |
| Aperture | F2.8 | Sets the illumination budget for short exposures. F2.8 passes about a quarter of the light of an F1.4 design, so lighting and gain planning start here rather than at the camera. |
| Mount and thread | M16 x P0.5 | Sizes the lens holder, the panel cut-out and the board keep-out. M16 x P0.5 is not interchangeable with the M12 x P0.5 thread used by most compact vision modules, and an adapter adds length to the optical path that has to be drawn, not assumed. |
| Field of view | 102° diagonal at 1-inch format; 91° horizontal as published | Recalculate on your own sensor before use. Angles published at one format do not transfer to another; the table in the previous section shows the values for an 8.50 mm wide sensor. |
| Resolution rating | True 50MP | Leaves headroom over a 5MP imager, so the lens rather than the sensor sets the practical limit. It also means the part can follow a sensor upgrade within the same barrel. |
| Distortion | Below -3% | Characterised with a single profile. Request the figure with its test conditions so the residual can be verified against a chart at the working distance. |
| Optical construction | 1 molded glass asphere + 3 glass + 5 plastic elements (8 in total) | Accounts for the size and the thermal behaviour of the assembly. Relevant where a sealed outdoor enclosure is planned and operating temperature swings. |
What a resolution requirement expressed in lp/mm is actually asking for
Resolution is often written as a spatial frequency rather than as a sensor class, and on a 3.45 µm pixel pitch the two meet at 145 line pairs per millimetre, the Nyquist limit of 1 / (2 x 3.45 µm). A project that asks for modulation transfer function data at 145 lp/mm is asking the optics to resolve up to the sensor's own limit, not beyond it.
That distinction decides what to request from a supplier. At 145 lp/mm the useful deliverable is a modulation transfer function curve plotted at that frequency for the centre, the 0.7 field point and the corner, at the aperture the system will actually run. If the intent is a design margin of twice Nyquist, the request becomes a 290 lp/mm curve, and the optical design budget changes accordingly. Stating which one is meant costs one sentence and saves a design cycle.
Why an F1.6 brief costs more than an F2.8 part
Aperture is the one line in the brief that cannot be met by choosing a different focal length, because it is set by the optical design rather than by the sensor. In terms of exposure the difference is two stops: F2.8 delivers roughly a quarter of the light of an F1.4 design and about half of an F2.0 design, which is the difference between a lighting plan that a standard court installation can carry and one that needs dedicated illumination. The fastest standard aperture in the M16 range is F2.0, so a request at F1.6 or faster points to a custom optical design at this image circle. The lighting plan, not the catalogue, decides whether that step is worth taking.
Mechanical envelope and signal budget
A lens is only specified once the drawing and the electronics agree with the datasheet. The mechanical drawing fixes where the optics sit relative to the sensor plane, which fixes the housing depth.
Two interface values deserve an explicit line in the specification table rather than a footnote. The first is total track length: the product page publishes 34.49 mm, and the drawing below shows the barrel envelope the designer has to clear. The second is back focal length, which is the distance from the last optical surface to the image plane and therefore the number that fixes the sensor offset in the holder. Where a value is listed as available on request, the right move is to request it with the drawing set and confirm it against the current revision before the holder is released.
Coverage geometry for a 6.6 mm lens on an 8.50 x 7.12 mm active area: the 5.54 mm corner radius is the minimum image-circle radius the lens has to fill, and the two angles are what that focal length can deliver on this sensor. The published 91 degrees horizontal belongs to the 1-inch format and does not transfer here.
Fixing focus for a 5 to 20 m working range
A fixed-focus design is the right default for a court camera, and the depth of field that a short focal length and a small aperture produce is the reason. On this sensor the hyperfocal distance works out at roughly 4.5 m, so a focus set once holds the whole 5 to 20 m range inside the tolerance of a 3.45 µm pixel, and the residual geometric error at the centre stays a small fraction of that pixel.
Two cases still need checking rather than assuming. Focus position should be set at the working distance the detection algorithm leans on, not at the middle of the range, because the near end of a 5 to 20 m span is where defocus appears first; and in an outdoor installation the setting has to hold across the temperature range of the housing without a re-collimation step in the field.
Where this lens earns its place
Four application patterns map cleanly onto a 6.6 mm M16 wide-format lens, and each one starts from the same engineering position: the sensor is wide, the target is small or distant, and coverage has been planned around the focal length rather than demanded from it.
Sports and broadcast tracking cameras
Challenge: a ball or puck a few centimetres across, tracked from 5 to 20 m, at frame rates where exposure is measured in milliseconds.
Value: a long focal length for the format puts 8.5 pixels on a 67 mm ball at 15 m, and the 102-degree diagonal at 1-inch format gives a single camera a usable field from a fixed position. The 50MP rating keeps the optics ahead of the sensor, so the remaining limitation is set by the detection algorithm rather than by glass.
Large-format embedded vision modules
Challenge: a board-level camera with a 1-inch-class sensor needs an image circle that covers the diagonal, in a barrel small enough not to define the enclosure.
Value: an M16 barrel with a native large-format image circle removes the need for a C-mount interface and its flange distance, which is often what makes a compact sealed module impossible.
Industrial inspection with geometric output
Challenge: the pipeline derives dimensions from pixels, so distortion has to be low and characterised rather than known only by reputation.
Value: distortion specified below -3 percent is a workable single-profile figure. Request it with its test conditions, and verify the residual against a chart at the production working distance.
Semi-outdoor infrastructure vision
Challenge: a camera beside a court, a gantry or a roadside pole sees rain, dust and temperature swing.
Value: a short M16 barrel is easier to seal at the module level than a larger interface. Where IP protection is a requirement, specify the rating for the assembly rather than assuming it from the lens alone, and add the gasket and window to the same drawing.
M12 and M16 wide-angle options side by side
Most compact vision modules in the 5MP class are built on M12 wide-angle lenses, and the first instinct is to reach for one here as well. The comparison is worth making explicitly, because the difference is not only the thread.
| Consideration | Typical M12 wide-angle lens | JSD6628 M16 6.6 mm |
|---|---|---|
| Image circle | Commonly sized for 1/2.5" to 1/1.8" sensors, so an 11.08 mm diagonal is not covered | Designed for 1-inch-class large-format sensors; confirm the published circle against your diagonal before the holder is released |
| Coverage planning | Often the reason for choosing M12: a very wide angle from a compact barrel | Deliberately narrower on a wide sensor, so mounting position is planned around 65.6° horizontal |
| Ball detail at 15 m | Short focal lengths for the format give 5 to 7 pixels on a 67 mm ball | 8.5 pixels at 6.6 mm, about 25 percent more than a 5.17 mm reference |
| Mechanical interface | M12 x P0.5, smallest holder and the widest choice of off-the-shelf holders | M16 x P0.5, larger holder and panel cut-out, more room for a sealed front window |
| Resolution headroom | Frequently rated 5MP to 8MP, matching the sensor rather than exceeding it | True 50MP rating, so the lens does not become the limit if the sensor is upgraded |
| Corner behaviour on a wide sensor | Corners are the first place an undersized circle shows up, as shading and softness | Built for the format, so corners are inside the design's working field rather than outside it |
| Illumination at 120 fps | Fast M12 designs at F1.6 to F2.0 are available off the shelf | F2.8 standard, F2.0 at the fast end of the M16 range |
The two routes solve different problems, so the choice follows from which constraint is fixed. Where the widest possible angle from the smallest possible barrel is the priority, an M12 design is the starting point and the coverage arithmetic will show whether it fits the sensor. Where pixel count on a small, fast target matters more, the M16 route in a 6 to 7 mm focal length is the one that keeps the corners and the resolution headroom. Where the brief sits between the two, JSD5042, a 50MP M16 lens at 126 degrees diagonal for the 1/1.55-inch class, is the adjacent part worth evaluating against the same coverage table, and the 1-inch large-format lens range shows what else exists at this image circle.
When to specify a different lens
A 6.6 mm M16 lens answers a specific combination of requirements. Two others call for a different part.
Horizontal coverage of 88 to 90 degrees on a wide sensor is a hard requirement. That is a focal-length decision rather than a lens decision: on an 8.50 mm wide active area the ceiling is about 4.28 mm. Wide-angle M16 entries in the JSD range sit closer to that figure, and the closest of them reaches the coverage target at the cost of a substantially larger distortion figure that the software pipeline has to absorb. Which of the two is acceptable is an algorithm question, and it is worth resolving with the real detection thresholds in hand.
A 6 to 6.5 mm focal length is required together with 88 to 90 degrees of horizontal coverage on the same sensor. Those two requirements cannot both hold on an 8.50 mm wide active area, at any aperture and in any barrel. The options are to re-plan coverage around a longer focal length, as the table in this article does, or to move to a sensor whose active width is at least about 12.3 mm so that a 6 mm lens produces 88 degrees. The second option changes the camera platform and the lens then has to cover the correspondingly larger image circle. A custom design cannot resolve a geometric conflict, and establishing that before the housing is committed is cheaper than discovering it on court.
In each case the answer exists inside the M12, M16 and M18 ranges: the ranges are organised around sensors and their image circles, so a requirement can be mapped to the part that fits it rather than a part being stretched to cover it.
Working with JSD Optics on a wide-format camera
JSD Optics develops M12, M16 and M18 imaging lenses around image sensors rather than around a fixed list of focal lengths. For a wide-format project that order of work matters, because the sensor decides first: active area, pixel pitch, image circle and mechanical keep-out define what the lens has to be before the focal length can be discussed at all.
Three things carry more weight than a brochure when a lens becomes part of a product. The specification database is organised for sensor-first matching, so a project can start from the imager and work forward. JSD Optics implements ISO9001:2015 and IATF16949 quality management systems and operates production and laboratory facilities for optical measurement, which is where modulation transfer function, distortion and relative-illumination data come from when a design review needs them. And lens assemblies are available for evaluation before a design is frozen, which turns a coverage or illumination argument into measurements taken on the actual court rather than estimates from a table.
Current specifications and ordering data: JSD6628 6.6 mm F2.8 M16 low-distortion lens for 1-inch-class 50MP sensors.
❓ Frequently Asked Questions
The questions that come up most often when a wide-format M16 lens is evaluated for a high-speed vision system.
What field of view does a 6.6 mm M16 lens give on an 8.5 mm wide sensor?
About 65.6 degrees horizontal, 56.7 degrees vertical and 80 degrees diagonal. Those angles are calculated from the active area, not taken from the published 91 degrees horizontal, which applies to the 1-inch format at a 16 mm diagonal. Always recalculate field of view on the sensor you are actually using.
Can a 6 mm lens give 88 degrees horizontal coverage on this sensor?
No. On an 8.50 mm wide active area, 88 degrees horizontal requires a focal length of about 4.28 mm or shorter. A 6 mm lens gives 70.6 degrees and a 6.6 mm lens gives 65.6 degrees, regardless of aperture, barrel or supplier. The two requirements have to be traded against each other before a part is selected.
How large does the image circle need to be for an 8.50 x 7.12 mm sensor?
The diagonal is 11.08 mm, and covering the four corners of the rectangle needs a radius of about 5.54 mm, so plan for an image circle of at least 11.1 mm. Anything smaller shades the corners, and on a court camera the corners are where the far side of the play sits. Confirm the published image circle against the drawing before the holder is released.
What does a 145 lp/mm modulation transfer function requirement mean for a 3.45 µm sensor?
145 lp/mm is the sensor's own Nyquist limit, calculated as 1 / (2 x 3.45 µm). A requirement at that frequency means the lens should resolve at the sensor's limit, not beyond it. Ask for the modulation transfer function curve at that frequency across the centre, the 0.7 field point and the corner, at the aperture the system will run. If the intent is a design margin of twice Nyquist, the request becomes a 290 lp/mm curve.
Why does the aperture matter at 120 fps?
At 120 fps the exposure window is 8.3 ms or less, so the lens has to deliver enough light inside that window. F2.8 passes roughly half the light of an F2.0 design and a quarter of an F1.4 design. In controlled lighting that is workable; in a dim hall it shows up as gain and noise. The fastest standard M16 aperture in the JSD range is F2.0, and faster apertures at this image circle are a custom discussion.
Is a 6 to 6.5 mm M16 lens possible for a wide sensor?
A 6 to 6.5 mm lens exists and is available in M16. What is not possible is that focal length together with 88 to 90 degrees of horizontal coverage on an 8.50 mm wide active area, because the geometry rules it out at any aperture. The workable paths are to plan coverage around 65 to 71 degrees with a 6 to 6.5 mm lens, or to keep the coverage target and move to a sensor whose active width is at least about 12.3 mm.
Can samples be evaluated before the design is frozen?
Yes. Lens assemblies are available for evaluation, and engineering data such as detailed drawings and measurement results can be prepared for a design review. Share the sensor, the required coverage axis, the working distance range and the mechanical keep-out, and the evaluation set can be matched to the project rather than to a catalogue page.
Size a wide-format M16 lens against your real sensor
Send the sensor, the active area, the working distance range and the mechanical keep-out. JSD optical engineers will run the coverage and detail arithmetic with you, confirm the image circle, and prepare an evaluation set.
Or email us at info@jsdoptical.com
Consult now
