Zoom Lens for Machined Parts: One Camera, Several Magnifications

A machine shop rarely inspects one part number. The same station is asked to look at a turned sleeve on Monday, a milled bracket on Tuesday and a batch of small spacers on Wednesday, and the features that matter change with the part: a chamfer here, a thread relief there, a slot that has to be seen to the bottom. A fixed-focal lens answers that with a drawer of lenses and a setup sheet, and every change of part number becomes a change of optics. A motorized zoom lens answers it with a magnification command, which is why the zoom body has become the default head on mixed-part inspection stations - provided the magnification range, the stand-off and the depth of field are settled before the station is drawn.

This page is about the optics of that decision for machined work specifically: how to size the magnification range from a part family, what the stand-off has to clear inside an enclosure, where the depth of field runs out as magnification rises, and the point at which a machined part should leave the zoom body and be measured on a telecentric head instead. Every field of view quoted here is derived from the sensor size and the magnification rather than read from a separate datasheet table, and the derivation is shown so it can be repeated for a part the table does not cover.

Why machined work arrives as a family, not as a part number

The parts that share a machining centre also share a tolerance band, a material and usually a customer, but they do not share a size. A family of turned parts can run from a 6 mm spacer to a 45 mm flange, and the features to be judged scale with them: a 0.2 mm chamfer on the spacer and a 1 mm thread relief on the flange are both edge details, and both want roughly the same number of pixels across the feature rather than the same field of view in millimetres. That is the argument for variable magnification in one sentence - the requirement is constant pixel coverage of a feature whose physical size changes, and only a zoom can hold that constant without swapping hardware.

The second reason is presentation. Machined parts arrive in fixtures that were designed for the machine tool, not for the camera: a vice, a collet, a pallet with three parts on it. Moving the camera to a second position to get a second field of view means an axis, a re-homing step and a cycle-time cost that is easy to underestimate. Changing magnification instead keeps the optical axis where it is, and the fixture design does not have to reserve travel that the machine does not have.

The cost of that choice is that magnification is now a variable. A fixed lens has one number that can be checked once and engraved on a setup sheet; a zoom body has a range, and the range only stays trustworthy if the recipe records where inside it the axis was sent and the axis reports back where it arrived. Everything later on this page assumes that discipline, because without it a zoom head is a manual lens with extra parts.

Sizing the magnification range from the part envelope

Work from the largest part in the family, not from the smallest. The largest part sets the lowest magnification the body must reach, and the finest feature on the smallest part sets the highest. For each part, write down the width that has to be visible in one image - the part plus the fixture margin that part-position variation demands - and call it W. The magnification implied by W is:

magnification = sensor width / W

On a 2/3-inch sensor the active area is 8.8 by 6.6 mm, so the arithmetic is short enough to do on paper. The table below applies it across both zoom ranges POMEAS builds for this kind of station. Every field of view in it is derived from the sensor size and the magnification; none of them is a separate datasheet figure.

Magnification	Field of view on 2/3" sensor (derived, H x V)	6.5X series (0.7X-4.5X)	12.5X series (0.58X-7.5X)
0.58X	15.17 x 11.38 mm	below range	lowest magnification in range
0.7X	12.57 x 9.43 mm	lowest magnification in range	in range
1.0X	8.80 x 6.60 mm	in range	in range
2.0X	4.40 x 3.30 mm	in range	in range
4.5X	1.96 x 1.47 mm	highest magnification in range	in range
7.5X	1.17 x 0.88 mm	above range	highest magnification in range

Two results from the table decide most projects. First, the widest view either series can give on a 2/3-inch sensor is about 15 mm across, at 0.58X. A 60 mm bracket needs 8.8 / 60 = 0.147X, which is outside both ranges, so a family that reaches that size cannot be covered in one image by either body - the honest options are a second station, a two-position camera, or accepting two images stitched by the software rather than one. Second, the highest magnification either body reaches is 7.5X, which gives about 1.17 mm across; if the finest feature needs more than that, the requirement has moved into microscope territory and should be treated as a different instrument, not as a zoom.

Between those two limits the ranges overlap generously, and the choice between them is decided by the low end. If the family needs more than roughly 12.5 mm of visible width, the 6.5X body has already run out and the 12.5X body is the one to quote, because it reaches 15.17 mm at 0.58X. If the family tops out below 12 mm and the detail work stays under 4.5X, the 6.5X body does the same job with a longer stand-off, which is usually the more valuable of the two margins on a machine-tool installation.

Working distance inside a machine enclosure

On a stand-alone inspection bench, working distance is a preference. Inside a machining centre it is a constraint set by everything else in the cell: the tool changer, the coolant nozzles, the chip conveyor and the door interlock all have a claim on the space in front of the part. The 6.5X motorized body (PMS-LZ-650104) is specified at 82 plus or minus 2 mm working distance as a bare lens, and the 12.5X family (PMS-LZL-12100 and its variants) at 77.4 plus or minus 2 mm. Both take a 2/3-inch sensor on a C-mount, so the camera side is not what limits the layout - the stand-off is.

Two things consume that stand-off before the part is reached. The first is any protective window or air-purge housing used against coolant mist; a plane-parallel window sits in the optical path, so the focus it produces is not the focus the bare lens produces at the same mechanical distance, and the shift grows with the thickness of the window. The second is any front attachment or video tube: a 1X video tube and a 0.5X front attachment change working distance and magnification together, which is useful when a wider view or a longer stand-off is required, but it means the figure to design around is the one for the complete stack rather than for the bare body. Both are reasons to quote from a drawing of the actual stack and confirm focus on a real part before the mount is machined.

Depth of field against feature height

Machined parts are not flat, and the features that matter usually stand up or cut down. A thread relief is a groove, a chamfer is a slope, a milled pocket has a floor and a wall, and a burr is a ridge above the nominal surface. Depth of field has to be compared with the height range of the features being judged, not with the flatness of the fixture.

In the 12.5X family the datasheet gives 2.75 mm of depth of field at 0.58X with a numerical aperture of 0.025, falling as magnification rises. The 6.5X body is specified from 1.9 mm at 0.7X down to 0.1 mm at 4.5X. At the low end of either range there is room to see a whole small part in focus; at the high end the depth is thinner than a thread relief is deep, and the focus has to be set on the feature being judged rather than on the surrounding face. That is not a defect of the lens, it is the usual consequence of trading depth for resolution, but it does change the integration: a station that inspects a groove at 4.5X needs a focus axis, or a recipe that moves focus between features, and it needs the focus position written down next to the magnification.

Where a zoom body stops and a telecentric lens has to take over

The distinction is not image quality, it is what the measurement depends on. A zoom body forms an image whose magnification changes with object distance: move the part 0.2 mm closer and the image grows. For judging a feature - is the chamfer there, is the slot clear, is the thread relief complete - that is irrelevant, and the flexibility of variable magnification is worth far more. For certifying a dimension through the thickness of a part it is decisive, and that is where a telecentric head belongs.

The three calls that normally move to a telecentric head are outer diameter on a turned shaft, edge position on a part whose height varies across the batch, and any bore whose diameter is specified at a depth. On a machined part with real thickness variation, an object-side telecentric lens holds magnification constant as the surface moves, which is exactly what a diameter measurement needs and exactly what a zoom body cannot promise. The silhouette side of that argument, including which machined features genuinely leave the measurable set, is set out in the guide to telecentric lenses for machined parts.

In practice the two heads coexist rather than compete. A mixed-part cell often carries a motorized zoom head for appearance and feature checks and a telecentric head for the dimensions that have to be reported, sharing a motion platform instead of sharing one lens. Deciding which head owns which characteristic before the station is drawn is cheaper than adding the second head afterwards.

Resolution is a curve, not a number

Resolution is the specification most often quoted as a single figure and most often misapplied, because on a zoom body it changes across the range. In the 12.5X family the datasheet gives 13 micrometres at 0.58X; at higher magnification the figure improves, and the improvement does not follow a rule that can be extrapolated from the low end. For the 6.5X body the resolution has to be read from the datasheet of the specific model at the magnification the station will actually run, and it should not be carried over from another model in the same series.

The practical consequence is that resolution has to be checked at the two ends of the range the recipe uses, not at the middle. If the detail pass runs at 3X and the overview at 0.8X, the numbers that matter are the ones at 3X and 0.8X, and a quote made from a single catalogue figure will be optimistic at one end and pessimistic at the other. Where a model's figure at a given magnification is not published, treat it as unknown and ask for it rather than interpolating - a resolution number that was guessed is worse than a blank, because it will be used to accept or reject parts.

Lighting a turned or milled surface

A machined surface is a set of directional scratches and facets, and it scatters light differently depending on how it was cut. On a turned part the tool marks run around the circumference, so a ring light produces a bright band and a dark band that follow the feed; on a milled face the marks are directional and a ring light can make one wall of a pocket disappear. The lighting decision is therefore separate from the lens decision, and it has to be made per feature rather than per station.

Three arrangements cover most machined work. A ring light is adequate for through-holes and for locating a part in the fixture, where the requirement is contrast against the background rather than surface detail. A diffuse dome evens out directional scatter and is the usual choice for judging a chamfer or a radius on a turned face, because it removes the feed-mark banding that a ring light amplifies. Coaxial illumination sends the light back along the optical axis and is what makes the bottom of a blind hole or the flank of a thread relief visible at all; POMEAS builds the 6.5X body in both a plain motorized version and a coaxial version (PMS-LZ-650105), so that choice can be made per station instead of by adding an external coaxial unit in front of the lens, which would change the stand-off again.

Commanding magnification from the recipe

Once the optics are settled, magnification has to become a value the control can command. The motorized zoom bodies use a closed-loop DC servo driven over a 9-pin RS-232 interface, single or dual axis, which means the magnification can be sent as part of a recipe and read back afterwards. The read-back is the more valuable half of the pair: the check that a station has stayed in calibration is a comparison of commanded and reported magnification after a few hundred cycles, and that comparison is only possible if the recipe stores both.

A machined-part recipe normally ends up with four fields that belong together: the magnification for the overview pass, the magnification for each detail pass, the focus position for each of those magnifications, and the lighting state that goes with each. Writing them as one record matters because they drift together - a change of magnification alters depth of field, which alters the focus position that is acceptable, and a focus position restored without its magnification is the most common cause of a station that passes its first article and then rejects a batch.

Integration checklist

• Family envelope measured on real parts, largest and smallest, with fixture and part-position margin added, before any body is quoted.
• Lowest and highest magnification derived from that envelope and compared with the range of the body on the actual sensor, not on a nominal one.
• Stand-off budgeted for the complete stack - window, front attachment, video tube - and checked against the cell layout, not against the bare-lens figure.
• Depth of field at the detail magnification compared with the height range of the features being judged, and a focus axis allowed for if it does not cover them.
• Resolution read at both ends of the magnification range the recipe will use, with unpublished figures marked unknown rather than interpolated.
• Lighting chosen per feature: ring for location, dome for chamfers and radii, coaxial for blind holes and thread reliefs.
• Magnification, focus and lighting stored as one recipe record and read back on a schedule, so drift appears as data instead of as a rejected batch.

What this page does not cover

Stamped and pressed parts are a different optical problem, because the feature of interest is often a burr or an edge in a thin section rather than a machined face, and the throughput requirement is usually higher; that case is covered in the guide to flash measurement for stamped parts. Plastic and moulded parts bring their own lighting difficulties, including transparency and sink marks, which are treated in the note on flash measurement for moulded and medical parts.

If the open question is whether a zoom body is the right class of lens at all - against a fixed-focal lens, a telecentric head or a longer focal length - the six specifications worth comparing are listed in alternatives to a 12X zoom lens. If the arithmetic on field of view, magnification and working distance needs the general form rather than the machined-part version, the FOV and magnification calculation sets it out with worked examples.

Frequently asked questions

Can one zoom lens cover a whole family of machined parts?

Yes, within the magnification range of the body. A 6.5X series body spans 0.7X to 4.5X and a 12.5X series body spans 0.58X to 7.5X, so the question is whether the smallest and the largest part in the family both fall inside one range on your sensor. Divide the sensor width by the field of view each part needs; if both results sit inside the range, one body covers the family and the part change becomes a recipe change rather than a lens change.

When should a machined part go to a telecentric lens instead?

When the measurement is a dimension taken through the thickness of the part: an outer diameter on a turned shaft, an edge position on a part whose height varies, or a bore whose diameter is specified at a depth. Those calls depend on the magnification staying constant as the object distance changes, which is what object-side telecentricity provides and a zoom body does not. A zoom body is the better fit when the part family varies in size and the job is to see features clearly rather than to certify a dimension.

How do I choose between the 6.5X and 12.5X zoom series?

Start from the widest field of view the family needs. On a 2/3-inch sensor the 6.5X series reaches about 12.57 by 9.43 mm at 0.7X, while the 12.5X series reaches about 15.17 by 11.38 mm at 0.58X, so the 12.5X body is the one for larger envelopes and for detail work that has to go beyond 4.5X. The 6.5X body keeps a longer stand-off at 82 plus or minus 2 mm against 77.4 plus or minus 2 mm for the 12.5X family, which matters when a fixture or a tool holder eats into the space above the part.

Does a protective window change the working distance?

Yes. A plane-parallel window in front of the lens is part of the optical stack, so the focus it produces is not the focus the bare lens produces at the same mechanical stand-off, and the shift grows with window thickness. Treat the window, any front attachment and any video tube as one stack, set the stand-off for the complete stack, and confirm focus on a real part rather than on the bare-lens datasheet figure.

Next step

Send the largest and smallest part in the family, the finest feature that has to be judged, the stand-off your cell allows and the sensor you intend to use through the contact page, and the magnification range can be checked against the datasheet of the specific body instead of against a general table. The two zoom bodies referred to throughout this page are the PMS-LZ-650104 6.5X motorized zoom lens and the 12.5X motorized zoom lens; the rest of the range is in the zoom lens catalogue.