JSD2628 2.8mm M12 Lens for OCR Scanners – <−0.35% Distortion, 5MP, 1/4-Inch
JSD2628 2.8mm M12 Lens for Document Scanning: Why Geometry Decides OCR Accuracy
A compact, ultra-low-distortion M12 lens designed to keep document geometry straight and OCR software simple — backed by the optical-engineering support to confirm the details that matter.
Key takeaway: The JSD2628 lens is a 2.8 mm, F/2.8 M12 imaging lens with a 1/4" image circle and 5 MP resolution, specified around geometric accuracy rather than pixel count. Its TV distortion stays below −0.35%, so page geometry is largely correctable with fixed scaling instead of a per-page warp model. Three specifications — IR-cut configuration, optical materials and back focal length — are worth settling before a design freeze.
Why OCR Accuracy Starts with the Lens
A document scanner can output a page that looks clean and still fail recognition. The cause usually sits upstream of the OCR engine: text baselines that curve, page borders that bow, character strokes that blur into their background. Fixing those defects in software costs compute, resolution and calibration time; designing them out of the optics costs none of the three.
The JSD2628 lens is a 2.8 mm, F/2.8 M12 imaging lens with a 1/4" image circle and 5 MP resolution, specified around geometric accuracy rather than pixel count. What follows is the criteria document-scanning optics are judged on, where the JSD2628 sits against them, and the three specifications — IR-cut configuration, optical materials and back focal length — worth settling before a design freeze.
Recognition engines assume a page behaves like a page: text lines run horizontally, table rules meet at right angles, margins stay parallel. Skew — a uniform rotation — is cheap to correct. Distortion is not. Barrel or pincushion error varies with distance from the optical axis, so compensation needs a non-linear warp map, and applying one means resampling pixels.
Three costs follow. The warp map runs on every page and competes with recognition for the same processor. Resampling shifts pixel values by fractional amounts near the field edge, softening the boundaries the engine uses to separate strokes. And the map is unit-specific — a correction calibrated on one scanner does not automatically hold across a production batch.
“Distortion control in the lens removes those costs at the source, once, passively, with no processing overhead per page.”
— JSD OpticsEdge contrast is the other half of the problem, and it depends on how the lens and sensor handle infrared; that is covered in the engineering section below.
What Document-Scanning Optics Must Get Right
Six parameters decide whether a document lens survives production. They are not equally recoverable downstream, which is what makes the ranking matter.
| Parameter | Why it decides OCR quality | How it is checked during selection |
|---|---|---|
| TV distortion | Non-linear geometry error that software must model and resample | Measured across the full field, not only on axis; compared against the dewarping budget |
| Image circle vs. sensor diagonal | An under-covering circle darkens page corners; an oversized one wastes resolving power | Image circle matched to sensor diagonal before any other comparison |
| FOV and working distance | Fixes the capture footprint at a given platen gap or feed height | Derived from EFL and active sensor height, then validated against enclosure dimensions |
| IR handling | Near-infrared reaching the sensor lifts the noise floor and flattens stroke contrast | Confirm whether the sensor filters IR on-package; if not, plan for an IR-cut element |
| Thermal focus behaviour | Focus shift during warm-up blurs small print across a batch | MTF evaluated after a warm-up soak rather than cold |
| Mechanical fit | Barrel diameter, thread and total track set the minimum module height | Mount stack and TTL checked against the enclosure before layout begins |
Ranked by irreversibility, distortion and IR handling sit at the top: both are set at the optical level and only partly recoverable in software. Working distance, resolution and mechanical fit can be traded against cost and footprint.
The JSD2628 Lens: Specifications Translated into Scanner Benefits
| Parameter | JSD2628 |
|---|---|
| EFL | 2.80 mm |
| F# | 2.8 |
| FOV (D/H/V) | 84° / 75° / 58° |
| TV Distortion | < −0.35% (barrel) |
| Optical Construction | 2G3P (2 glass + 3 plastic aspherical elements) |
| Mount | M12 |
| Barrel Diameter | φ14 mm |
| TTL | 20.0 mm |
| Resolution | 5 MP |
| Image Circle | 1/4" (Ø5.0 mm) |
Distortion below −0.35%. This is the figure that changes the software architecture. At that level, page geometry is largely correctable with a fixed scale, and in some pipelines no warp model is needed at all — which removes a per-page processing step and a per-unit calibration stage from production.
2.8 mm EFL with 84° diagonal coverage. Capture width at a fixed platen distance follows from focal length and the sensor's active area, not from focal length alone. A design targeting a given page width at a given module height should derive the required geometry from those two numbers and check it against the 84° diagonal figure rather than assuming a 2.8 mm lens frames a full page at any distance.
1/4" image circle (Ø5.0 mm) with 5 MP resolution. These describe two different limits. The image circle is a coverage limit — fall short of the sensor diagonal and page corners darken. Resolution is a sampling limit — exceed what the sensor resolves and the extra optical performance is paid for but unused.
M12 mount, φ14 mm barrel, 20.0 mm TTL. Together these define the space claim. TTL sets the distance from sensor plane to the front of the barrel, so it constrains enclosure height more directly than focal length does, and it is the number to put in the mechanical layout first.
F/2.8 aperture. In a scanner with fixed focus and controlled LED or lamp illumination, aperture mainly sets exposure time and depth-of-field margin. F/2.8 keeps light throughput workable while leaving depth of field to absorb minor tolerances in the platen-to-lens path.
Application Scenarios
Flatbed and platen scanners. The lens-to-glass distance is fixed for the life of the machine, so any distortion appears identically in every scan — as curved page borders and bowed rules. Fixed focus also means the focus position is set at assembly and must hold through the production run.
Sheet-fed check, form and ID readers. Subjects arrive at slightly different positions and heights within the feed path. What matters here is repeatability: identical geometry item to item keeps recognition thresholds stable across a batch, and small character heights make stroke contrast the decisive factor rather than resolution.
Compact embedded scan modules. Size is the constraint. A φ14 mm barrel and 20 mm total track fit into slim housings, and the threaded M12 interface lets focus be set during assembly and locked with a set screw or adhesive — a simpler service path than a glued fixed-focus module.
Comparison and Selection Guidance
| Consideration | Typical general-purpose 2.8 mm M12 lens | JSD2628 |
|---|---|---|
| TV distortion | Commonly 1–3%; a warp model is routine | Specified below −0.35% |
| Downstream effect | Per-page resampling, per-unit calibration | Geometry largely handled by fixed scaling |
| Sensor match | Image circle often resolved after the sensor is chosen | Matched to 1/4" / Ø5.0 mm circle |
| Mechanical integration | Barrel and TTL vary by supplier and model | φ14 mm barrel, 20.0 mm TTL, standard M12 thread |
A practical order for narrowing candidates: sensor format and image circle first, because a mismatch cannot be corrected later; geometry second, because it sets the scope of the software; mechanical fit third, because it constrains the enclosure; aperture and resolution last, since both can usually be traded.
1/4" is one point in a wider range, not the limit of what is available. JSD Optics manufactures imaging lenses from 5 MP to 50 MP for formats between 1/3" and 1/1.3", so a scanner migrating to a larger sensor is not tied to this single model.
Engineering Considerations Before You Commit
IR-cut configuration. The standard JSD2628 is supplied without an integrated IR-cut filter, so in default form it should be treated as IR-transparent; filtering is a configuration option rather than a standard feature. The decision follows from the sensor, not the lens. If the sensor package already carries an IR-cut filter, the standard unit is the simpler path. If it does not, unfiltered near-infrared reaches the silicon and raises the noise floor at fine detail — exactly the spatial frequencies where character strokes live. One mechanical consequence is worth planning for early: inserting a filter changes the glass stack between lens and sensor, which shifts effective back focus and total track.
Hybrid 2G3P construction. The lens uses two glass and three plastic aspherical elements. Molded aspheres are how compact M12 lenses achieve tight geometry control inside a short barrel at manageable cost; the trade-off is thermal. Plastic elements change refractive index and shape with temperature faster than glass, so focus position drifts more across a wide ambient range. Room-temperature scanning equipment usually tolerates this. A device with internal heat sources, or one specified across a wide operating window, should have focus behaviour evaluated under soak conditions, and JSD Optics can review whether an alternative material configuration suits the requirement.
Back focal length. A discrete BFL figure is not published in the standard specification table; the 20.0 mm total track is the mechanical reference. BFL in a mounted design is not a single fixed number — it depends on the sensor's cover-glass thickness and on whether an IR-cut filter is present — which is why it is confirmed against a specific sensor stack rather than quoted generically. For design review, the JSD Optics engineering team can provide the exact BFL together with CRA data, MTF curves and mechanical drawings.
Ingress protection. The JSD2628 is not a sealed design. IP67 is available on selected models in the range; for indoor document-scanning equipment, sealing is rarely a functional requirement.
Why JSD Optics Is Built for This Kind of Project
JSD Optics has developed and manufactured imaging lenses since 2009 and works with camera module manufacturers, embedded-vision developers and OEM/ODM teams rather than end users. That distinction matters when a specification has to be matched to a particular image sensor: the company is an imaging lens developer, not a general optics supplier. Credentials include ISO 9001 and IATF 16949 certification, roughly 200 staff, and production from 5 MP to 50 MP across 1/3" to 1/1.3" formats, primarily M12 and M16.
For a document-scanning programme, three capabilities are relevant:
Sensor-first selection. Projects start from the image sensor and the application requirements around it. An existing catalogue lens is recommended when one fits; when none does, a custom design is developed.
Design data on request. BFL, CRA data, MTF curves and mechanical drawings are supplied for design review, so the optics can be checked against the enclosure before tooling.
Custom development. Custom EFL, CRA matching, IR-cut filter integration and mechanical modification — from optical design through prototype, testing and mass production.
Confirming the JSD2628 Lens Specification Against Your Design
The JSD2628 addresses document scanning and OCR on the criteria with the least downstream flexibility: geometry corrected in the optics, a clean 1/4" image circle, and M12 integration in a compact barrel. Two items are worth settling before a design freeze — whether the sensor filters infrared on-package, and how much temperature swing the scanner sees in normal operation — while the back focal length is best confirmed against the actual sensor stack rather than a catalogue value.
Three ways forward:
Ready to build: request an evaluation sample for in-house testing.
Still validating: request MTF curves, CRA and BFL data, and mechanical drawings for a design review.
Planning a custom configuration: book an engineering consultation covering IR-cut options, CRA matching or mechanical modification.
Supplying the sensor model and target working distance allows JSD Optics to cross-check compatibility — including alternative formats or material configurations — before any commitment is made.
Contact JSD Optics to request a JSD2628 evaluation sample or engineering data.
❓ Frequently Asked Questions
The questions design engineers ask most often before specifying the JSD2628 lens for a document scanner.
What makes the JSD2628 suitable for OCR?
Its TV distortion specification, shown in the table above, is several times tighter than the 1–3% common to general-purpose M12 lenses. That keeps page geometry within a fraction of a percent and reduces how much correction the recognition stage has to perform.
Does the JSD2628 include an IR-cut filter?
Not in the standard configuration. IR-cut filtering is available as an option, and the right choice depends on whether the image sensor already filters infrared on-package.
Are the optical elements glass?
No. The construction is hybrid 2G3P — two glass elements and three plastic aspherical elements.
What is the back focal length?
A discrete BFL value is not published in the standard specification table; the total track is 20.0 mm. The exact BFL is supplied against a specific sensor stack, together with CRA and MTF data.
What sensor size does it match?
A 1/4" sensor with a Ø5.0 mm image circle at 5 MP resolution.
Ready to Specify the JSD2628 Lens?
Send your sensor model and target working distance — JSD Optics will cross-check compatibility before any commitment.
Or email us at info@jsdoptical.com
Consult now
