Same Lens, New Sensor: The Minimum Requalification Before You Accept the Combination

The approved combination is not the lens alone. When the sensor behind it is discontinued and the replacement has a different pixel size, the Nyquist frequency of the system moves, and every lens specification judged at the old pixel scale must be re-read at the new one. The requalification scope is a short, named list: MTF at the new Nyquist, chief ray angle and image circle, back focal length against the new cover glass, distortion, and a side-by-side read against the original batch data. Most of that evidence can come from the lens maker’s batch documentation, re-issued at the new conditions, before you touch the drawing.

If this is close to your build but not exact, custom optics design at Lensmind is how we turn a spec sheet into a production design.

Why a Pixel Pitch Change Is a Lens Change

A lens does not know how many pixels sit behind it. What changes is where on the lens contrast curve your system now operates. The Nyquist frequency of a sensor is set by pixel pitch: a smaller pitch pushes Nyquist higher, into a region of the modulation transfer function where the lens delivers less contrast than it did at the old pitch. A larger pitch does the opposite, which sounds like free margin but is not, because a coarser pixel can also hide real edges that your measurement algorithm was tuned to see.

This is why the question “is the lens still good?” is incomplete. The correct question is “does the lens still deliver the contrast and geometry that our inspection, measurement, or imaging task needs, at the new sampling scale, with the same locked focus?” That question can be answered with evidence, and the evidence already exists in a form you can request: per-batch test documentation covering the agreed family parameters, typically MTF, distortion, BFL, and depth of field, with batch number and test conditions attached.

Symptoms, Causes, and Dispositions

After a sensor swap, the complaints that reach engineering look like image problems. Each one traces back to a specific optical item, and each item has a defined check. Use this table to route the symptom before anyone starts turning focus rings or adjusting gain to compensate. If you need a refresher on reading the curves involved, start with how to read an MTF chart for a machine vision lens, then come back to the table.

Symptom after the swap	Most likely optical cause	Check that settles it	Disposition
Fine edges soft, especially toward the corners	Lens contrast has dropped at the new, higher Nyquist frequency	Batch MTF re-read at the new spatial frequency, compared against the limit your original acceptance used	Request re-issued batch data; if margin is gone, negotiate a spec variant or a lens revision
Corner shading or color cast near the field edge	Chief ray angle of the lens no longer matches the new sensor’s microlens design	CRA curve of the lens vs the new sensor’s CRA specification, plus relative illumination data	Request the lens CRA curve from the maker; confirm match before first article
Focus will not reach, or best focus sits at the end of the adjustment range	New cover glass thickness or package height has shifted the image plane	BFL stack-up calculation against the new sensor’s cover glass specification	Re-shim or specify a mount variant; document the spacer in the assembly record
Measurement values shifted with no visible blur	Pixel scale change has altered calibration coefficients and algorithm thresholds	Recalibration on a known target, then a gauge R and R comparison against the old record	Recalibrate and re-issue inspection thresholds under change control
Dark corners on a physically larger active area	The lens image circle never covered the new format	Image circle diameter vs the new sensor diagonal	Do not proceed; select a lens rated for the larger format before any other test

Item by Item: What to Re-check and Who Does It

The shortest defensible scope has three blocks. Each block has an owner, and the lens maker owns most of the evidence.

Contrast at the new Nyquist

Take the new pixel pitch, compute the Nyquist spatial frequency, and ask the maker for the batch MTF data at that frequency and at half of it, across the field positions your original qualification used. If the original drawing called for a minimum contrast at a named image height, the same callout is re-judged at the new frequency. This is a reading exercise on data the maker already produces for every batch, not a new test campaign. The one case that genuinely needs new bench work is when the new pitch sits beyond the frequencies on the original test report; then ask the maker to measure, on a sample from current production, at your named frequencies and conditions.

Coverage, chief ray angle, and relative illumination

Confirm three geometric facts in one pass. First, the image circle still covers the new active area, with a stated margin, not a “should be fine.” Second, the lens exit chief ray angle profile matches the microlens design of the replacement sensor; a mismatch shows up as corner shading and color cast that no firmware fix removes cleanly. Third, relative illumination across the field stays inside the budget your software assumes. All three are optical design outputs the maker can supply as curves with conditions. If the new sensor is the same optical format, this block often closes in an afternoon; if the format grew, it is the block most likely to stop the program, so do it first.

Back focal length against the new cover glass

Image plane position is set by the stack-up of flange distance, mount tolerances, and cover glass thickness. A replacement sensor with a different cover glass moves best focus by an amount your adjustment range may or may not absorb. The physics of this shift, and a worked qualification approach, are covered in our article on cover glass thickness and lens matching. Ask the maker for the nominal BFL and its tolerance from the lens drawing, take the cover glass specification from the new sensor datasheet, and calculate the shift. If the shift is larger than the remaining adjustment travel, the fix is mechanical, a spacer or a mount variant, and it belongs on the drawing as a revision, not as a technician’s habit.

What the Maker’s Batch Data Can and Cannot Replace

A practical requalification budget depends on drawing a hard line between evidence the maker supplies and testing you must run yourself.

The maker can supply, per batch and per serial number where agreed: MTF at named field positions and frequencies, distortion, BFL, and depth of field, each with the test conditions and a pass or fail judgment against the frozen specification. This documentation is a default deliverable on disciplined production lines, not a special favor. For a sensor swap, request that the relevant parameters be re-read or re-measured at the frequencies and conditions the new combination needs, and request the data with batch numbers so it ties to the units you will receive.

The maker cannot supply your application verdict. Whether the new combination still detects your smallest defect, measures your width within tolerance, or survives your thermal profile is a system question. That block stays yours: a calibration rerun on a known artifact, a side-by-side image set against a retained sample of the old combination, and a short production-representative trial before release. If your qualification plan includes environmental re-checks because the new sensor package behaves differently with temperature, IEC 60068-1 is the standard framework for selecting and tailoring those tests rather than inventing a severity table. If your quality system tracks a golden unit or reference lens, this is the moment it earns its keep.

A useful split to write into the plan: the maker proves the optics still meet the drawing’s optical callouts at the new conditions; you prove the system still meets its inspection or measurement obligations with those optics installed.

The Requalification Package, as a Specification Table

When you send the request to the maker, send a table, not a paragraph. This one can be adapted directly into your order documents.

Item	Condition to state	Acceptance basis	Supplied by
MTF at new Nyquist and half Nyquist	Named field heights, working distance, f-number, wavelength band	At or above the original drawing’s contrast limit at those heights	Maker, from batch data or sample measurement
Image circle vs sensor diagonal	New sensor active area and diagonal	Circle diameter exceeds diagonal by the stated margin	Maker, from lens design record
Chief ray angle match	New sensor microlens CRA specification	Lens CRA curve within sensor’s acceptance band across the field	Maker, curve plus conditions
Relative illumination	Same field positions and aperture	No corner drop beyond the software’s assumed budget	Maker, from design or measurement
BFL and cover glass stack-up	New sensor cover glass thickness and tolerance	Focus shift within remaining adjustment travel, or spacer defined	Joint: maker supplies BFL, you supply sensor stack
Distortion	Named image heights, same as original callout	Within the original limit at the new image height	Maker, from batch data
System verification	Locked production configuration, representative parts	Inspection or measurement performance equal to the old record	You, with retained artifacts and the golden reference

Drawing Revision and Concession Handling

Resist the temptation to keep the drawing number untouched and manage the change by email. The drawing that said “lens per specification, approved on sensor X” no longer describes reality. Two clean paths exist, and either is acceptable if it is written down.

The first path is a drawing revision: the bill of materials names sensor Y, and a note lists the requalification items that any released combination must pass. The second path is a controlled specification document referenced by the drawing, listing approved lens and sensor combinations with the per-combination test requirements. The specification route ages better when further sensor changes are likely, because the drawing stays stable while the approved-combination table gains rows.

Whichever route you choose, the release record should contain four things: the evidence package from the table above, the calibration and algorithm re-verification on your side, an explicit statement of which lots are released under the new combination, and the concession terms if any parameter passed with reduced margin. A concession without a written margin and a review date is how a “temporary” waiver becomes permanent.

Conclusion

A discontinued sensor forces one honest question: does the lens still meet its callouts at the new sampling scale? Answer it with a short, owned list rather than a blanket retest. Get the maker to re-issue batch evidence at the new Nyquist, the new CRA band, and the new cover glass stack-up; run your own calibration and application check on top; and put the result into the drawing or its referenced specification as a named, released combination. Teams that do this turn an EOL notice from a panic into a change-control entry, and the next discontinuation notice will find a process waiting for it.