Multi-FOV M12 Lens Family for 1/2.9" 5MP Sensors
M12 Lens Selection for 1/2.9" 5MP Sensors: How to Plan a Multi-FOV Camera Family
Same sensor, same M12 mount, three field-of-view tiers: wide, standard and ultra-wide.
Key takeaway: JSD3024 (110° DFOV, 3.17mm EFL), JSD3422 (84° FOV, 3.45mm EFL) and JSD2120 (136.2° DFOV, 1.88mm EFL) all mount an M12 thread and are specified for a 1/2.9" 5MP sensor. They work as three tiers of one optical platform rather than three interchangeable parts: JSD3422 holds the geometric accuracy budget at <-0.73% distortion, JSD3024 carries the shortest mechanical envelope at 9.94mm total track length, and JSD2120 extends coverage to 136.2°. Read as a set — image circle and distortion checked per tier — the three lenses let one camera module be offered in wide, standard and ultra-wide variants.
1. Why One Lens Is Not a Lens Family
Embedded vision projects rarely buy a single lens. They buy a field-of-view roadmap: one base camera module, then two or three lens options so the same product line can be offered with a wide and a narrow version. The lens is the least expensive item in that roadmap, and the one whose choice sets calibration effort, mechanical stack and detection behaviour all at once.
A familiar pattern: an 84° lens for the measurement variant, a 110° lens for the general-purpose variant, a 136° lens for the wide variant. Each pick is defensible on its own. Two design cycles later the calibration pipeline is maintained three times over, the detection model behaves differently at each field of view, and the mechanical team is supporting three housings.
The cause is not optical. It is a platform decision, taken at the selection stage.
The three traps of tier-by-tier selection
Calibration does not transfer. A distortion model is fitted at one focal length and holds only there. Changing tier discards it and forces a refit, and tiers whose distortion differs by an order of magnitude cannot share a single correction path.
Distortion budgets drift apart. An 84° lens and a 136° lens do not sit in the same distortion class. Treating them as equivalents spends geometric accuracy that the measurement variant needed, without anyone deciding to spend it.
The mechanical envelope moves with the optics. Total track length does not stay put across tiers, so the family carries a mechanical spread that the enclosure has to absorb as one budget rather than as three separate lens specifications.
A lens family is defined before the focal length is fixed: the sensor target, the mount and the mechanical envelope are set first, and the individual tiers are selected against them.
Lens family selection orderThe sequence matters because of where the cost lands. Re-cutting a family after the module has been tooled means re-qualifying one sensor against three lenses instead of one, and re-running a mechanical stack that was already signed off. The case worked through below is a 1/2.9" 5MP platform offered at 110°, 84° and 136°, matched to three M12 lenses from a single range; the same sequence applies to any multi-FOV roadmap.
2. What a 1/2.9" 5MP Sensor Demands from an M12 Lens
A 1/2.9" sensor at 5MP is a compact, high-density target, and its active diagonal — about 6.28mm — governs most of the selection. A lens whose projected circle is smaller than that diagonal loses the corners; one that projects a good deal more leaves image circle unused.
Image circle clearance comes first
Image circle is read before any FOV figure. A lens offered for a 1/2.9" target is expected to project a circle larger than 6.28mm, because that surplus is what absorbs sensor placement tolerance, barrel tilt and cover-glass refraction. Corner darkening that survives an illumination fix usually traces back to this row.
JSD3024 is specified with a Φ6.8mm image circle on the 1/2.9" target — the widest margin of the three.
JSD3422 is specified with a Φ6.6mm image circle, the middle of the set.
JSD2120 publishes a 6.3mm image circle on the same target — the tightest of the three, and the figure to verify against the exact sensor part number before tooling is committed.
Spare circle and field of view pull against each other. In this set the widest tier carries the least margin, which is why the ultra-wide member is the one whose clearance is checked against the sensor part number first.
Pixel pitch sets the MTF question
A 1/2.9" sensor at 5MP in a 4:3 aspect ratio works out to roughly 2,560 × 1,920 pixels and a pixel pitch near 1.94µm — an estimate taken from the format diagonal, with the sensor vendor's specification governing. At that density the lens, not the sensor, is usually the limiting element off-axis. Centre performance on a competently built M12 lens holds up; the corner of a 136° field is where designs separate. A "5MP" label states the sensor resolution a lens is specified to support, not uniform sharpness across the whole frame.
CRA and the cover glass already in the stack
Chief ray angle matching fails more quietly than distortion. A sensor stack carries a cover glass and a microlens array optimised for a particular CRA curve; light arriving at the corner pixels from the wrong angle loses sensitivity there, and the result reads as shading. It gets treated as an illumination problem and is in fact a matching problem. The CRA curve and the back focal length are two inputs worth settling at evaluation stage, before a holder drawing is frozen.
3. The Three Tiers, Side by Side
All three lenses share an M12 thread and a 1/2.9" 5MP specification. Past that common ground each tier delivers a different property, so the table reads as three answers rather than as a ranking.
| Parameter | JSD3024 | JSD3422 | JSD2120 |
|---|---|---|---|
| FOV tier | Wide | Standard / low distortion | Ultra-wide |
| DFOV (1/2.9") | 110° | 84° | 136.2° |
| HFOV / VFOV | By sensor format — confirm with JSD | 75° / 47° | 122.2° / 78.6° |
| Effective focal length | 3.17mm | 3.45mm | 1.88mm |
| Aperture | F2.4 | F2.2 | F2.0 |
| TV distortion | <-10.82% (barrel) | <-0.73% (barrel) | <-11.02% (barrel) |
| Optical construction | 3G3P | 1G4P | 2G4P |
| Image circle | Φ6.8mm | Φ6.6mm | 6.3mm |
| Total track length | 9.94mm | 14.51mm | 16.88mm |
| Resolution rating | 5MP | 5MP | 5MP |
| Mount | M12 × P0.5 | M12 × P0.5 | M12 × P0.5 |
| Sealing | Not sealed | Not sealed | Not sealed |
Only two rows genuinely separate the tiers: distortion and total track length. Image circle and mount are pass/fail conditions shared by the set, so they belong to the platform rather than to an individual part.
JSD3422 — the tier that buys geometric accuracy
At <-0.73% TV distortion across an 84° field, the JSD3422 sits in a different class from the wide tiers — around 15× lower than either. The consequence is arithmetic: a corner point 3.1mm from the image centre in an 84° field shifts by roughly 12 pixels at the estimated 1.94µm pitch under a -0.73% figure, and by roughly 170 pixels under a -10.8% one. The first case is absorbed by a single radial term in most calibration pipelines. The second needs a full distortion model and still leaves residual error at the extreme corners.
Where image geometry maps to a physical distance, that difference is the measurement error floor. The 1G4P construction reaches the low-distortion target with one glass element, which is a design decision rather than an incidental result.
JSD3024 — the tier that buys depth
At 9.94mm total track length, the JSD3024 is 4.6mm shorter than JSD3422 and 6.9mm shorter than JSD2120 — the difference between fitting behind a slim front panel and forcing a deeper housing. Its <-10.82% distortion is characteristic of a compact wide-angle design and places it in the software-corrected group rather than the metrology group. For a 110° field on a 1/2.9" sensor that is an ordinary trade, and it is the one that carries the shortest stack in the set.
JSD2120 — the tier that buys coverage
The JSD2120 reaches 136.2° DFOV at 1.88mm EFL with the widest aperture of the three at F2.0 — about half a stop more light than JSD3024, which shortens the exposure time a given scene needs. The published HFOV is 122.2° and the published VFOV is 78.6°; wide-angle FOV figures are quoted against different definitions across the industry, so JSD confirms the measurement convention behind these on request. Coverage at that width changes the image geometry by design, and <-11.02% barrel distortion comes with it.
4. Six Selection Criteria That Actually Decide the Tier
Worked in order, the six checks below eliminate candidates before sampling budget is spent on them.
Image circle margin. Confirm the projected circle against the sensor's active diagonal with placement tolerance included. It comes first because no calibration step can recover it: a lens that misses the diagonal loses corner light that no software restores.
Sensor identity, not sensor size. Two sensors both sold as 1/2.9" can differ in pixel pitch, chief ray angle and active area, and each difference moves the answer.
Distortion budget in output units. State the allowable distortion as a corner displacement in millimetres or pixels rather than as a percentage, because that displacement is what the calibration step has to absorb; a percentage figure hides the pixel pitch that decides it.
Corner MTF, not centre MTF. Measured MTF at centre, 0.7 field and full field is the evidence that settles an image-quality gate; a resolution label is not.
Mechanical envelope. Total track length and barrel diameter set the module stack, and the tiers here span 9.94mm to 16.88mm. Either the module is sized for its longest member, or the depth-constrained variant takes the shortest lens.
Environmental sealing. Where the site involves washdown, dust or humidity cycling, filter the candidate list on IP rating before anything else: none of the three tiers here carries a rating, and section 6 lists the sealed models that follow the same family logic.
The order carries as much weight as the list. The first two checks are pass/fail on fit, the middle two set image quality, and the last two decide whether the module can be built and where it can be installed. Sample availability is confirmed model by model at enquiry rather than taken from a catalogue date.
5. Application Mapping: Which Tier Goes Where
Tier selection starts from the dominant error source in the task, not from field of view: a task that measures geometry and a task that detects presence are buying different properties from the same sensor.
JSD3422 — measurement, code reading, alignment
Where the image has to map to physical dimensions, geometric error converts straight into measurement error. A low-distortion model keeps correction to a radial term, which shortens the calibration routine, steadies first-pass yield at the frame edge, and reduces how often a station has to be re-checked. The 84° field then fixes the standoff at which the subject fills the frame.
JSD3024 — compact modules with general-purpose coverage
Where depth behind the front panel is contested, the 9.94mm stack is what allows a slim sensing head while 110° still covers a workstation or a doorway. Distortion correction moves into the pipeline so the mechanics do not have to grow — an explicit trade rather than a hidden one.
JSD2120 — full-scene awareness and navigation
Where the system has to hold as much of the environment as one frame can, including the near field that a narrower tier leaves outside it, 136.2° at F2.0 keeps exposure time workable in dim light. The output of such a task is presence or position in scene, which is what allows a wide projection to be handled by a distortion model instead of a metrology chain.
| If the function is… | Leading criterion | Suggested tier |
|---|---|---|
| Dimensioning, metrology, optical alignment | Distortion | JSD3422 (84°) |
| Label, barcode and OCR reading | Distortion + working distance | JSD3422 (84°) |
| Compact embedded camera, tight enclosure | Total track length | JSD3024 (110°, TTL 9.94mm) |
| Robot end-effector, close-range pickup | TTL + FOV | JSD3024 (110°) |
| Presence and occupancy detection | FOV | JSD2120 (136.2°) |
| AGV / AMR navigation, obstacle awareness | FOV + aperture | JSD2120 (136.2°, F2.0) |
| Indoor monitoring, one camera per zone | FOV | JSD2120 (136.2°) |
The table describes one product platform with three market variants: one qualified sensor, one mount, one supply chain, and the lens as the only variable.
6. Sealing and Environment Decisions
JSD3024, JSD3422 and JSD2120 are not sealed models. They carry no IP rating, which suits clean, dry, indoor enclosures — the usual setting for embedded vision and indoor monitoring.
Washdown, dust, humidity cycling or outdoor exposure changes the requirement. The seal is a property of the lens barrel and has to be designed in from the start; it cannot be added at module level once the barrel has been chosen.
Sealed alternatives in the same family logic
Two IP-rated models cover the same sensor target and mount convention:
JSD3023 — 1/2.7" / 1/2.9", 5MP, F1.8, TTL 16.2mm. A sealed wide-angle option for a wet or dusty environment.
JSD2157 — 1/2.7" / 1/2.9", 5MP, 6G full-glass, F2.0. A sealed option in the ultra-wide coverage class.
Both keep the sensor target and the M12 mount convention, so the change stays confined to the optics; the trade is a larger barrel and a longer stack. The full sealed selection sits in the 1/2.9" sensor lens range, and a lens holder is available where the module needs a defined focus and locking method.
Two more engineering considerations worth deciding early
Infrared performance. None of the three models carries a published IR specification. IR behaviour depends on the coating and on the IR-cut decision, and IR-cut filtering and IR-corrected optical design are two separate decisions, neither substituting for the other. Applications that need NIR sensitivity are best raised at enquiry stage, where measured data for a specific band, IR-cut filter options and IR-corrected designs can be put against the application at hand.
Focus and mounting convention. M12 lenses are focused by thread rotation, so the module needs a defined focus procedure and a locking method, whether adhesive, set screw or a dedicated holder. An assembly line that cannot absorb a manual focus step is usually better served by a lens holder than by a tighter thread tolerance.
7. Why JSD Optics as the Lens Partner
JSD Optics develops and manufactures imaging lenses for camera module manufacturers, and the design work starts from the image sensor rather than from a catalogue page.
Sensor-first matching. Given the sensor, resolution, FOV, working distance, distortion budget and mechanical constraints, the outcome is either a standard lens from the catalogue or a custom development.
Standard plus custom. The catalogue covers 5MP through 50MP across 1/3" to 1/1.3" formats, and the same design team handles custom lenses when the standard tiers do not fit.
Engineering data on request. Measured MTF data, optical construction details and mechanical drawings are supplied so validation can happen before tooling is committed.
Manufacturing and quality. ISO 9001 and IATF 16949 certified, with in-house design and mass production.
Application support. Custom EFL, CRA matching and mechanical modifications are quoted on request, and integration questions that come up after the lens is chosen sit inside the same scope.
Where none of the three tiers fits the envelope or the sensor, the same team develops a lens against the sensor — so catalogue part and custom design are two outcomes of one matching step rather than two separate tracks.
Related ranges: low-distortion M12 lenses, the super wide-angle lens series and the machine vision lens range.
❓ Frequently Asked Questions
The questions that come up most often when a 1/2.9" 5MP module is planned across three field-of-view tiers.
Are JSD3024, JSD3422 and JSD2120 interchangeable on the same camera module?
They share an M12 thread and the same 1/2.9" 5MP sensor target, so they can be fitted to one module footprint in principle. They are not mechanically identical: total track length runs 9.94mm (JSD3024), 14.51mm (JSD3422) and 16.88mm (JSD2120). A module designed around the longest lens accepts all three; a module designed around the shortest does not.
Which of the three is the low-distortion option?
JSD3422. Its TV distortion is <-0.73% across an 84° field, against <-10.82% for JSD3024 and <-11.02% for JSD2120. Where an application measures geometry rather than detecting presence, JSD3422 is the tier that should carry it.
Can the 136.2° JSD2120 be used for measurement tasks?
It is optimised for coverage rather than geometry. At <-11.02% barrel distortion a full distortion model and careful calibration are required, and residual error at the extreme corners stays larger than on the 84° tier. JSD2120 fits decisions about presence, position in scene or navigation; JSD3422 fits cases where a pixel maps to a distance.
Do these lenses work with a sensor other than 1/2.9"?
JSD3024 is also specified for 1/2.8" and covers a Φ6.8mm image circle; its 110° DFOV figure applies to the 1/2.9" target, and the field widens on a larger format. JSD3422 publishes a Φ6.6mm image circle for 1/2.9", and JSD2120 publishes 6.3mm, the tightest of the three. Fitting a lens to a sensor larger than its image circle allows loses corner light, so the exact sensor part number is what confirms the usable margin.
What information is needed to recommend a tier?
Five inputs: the image sensor part number, the required resolution, the target field of view, the working distance or measurement range, and the mechanical constraint of available depth and barrel diameter. Adding a distortion budget and an environmental requirement brings back a matched tier plus alternative options.
Planning a Multi-FOV Camera Family?
Send the sensor part number, the target FOV tiers and the mechanical envelope. JSD will match the lens set and confirm the engineering data — image circle margin, distortion and measured MTF — so the design can be validated before tooling.
Or email us at info@jsdoptical.com
Consult now
