Flash Measurement System for PCB Inspection

A printed circuit board is a flat part that behaves like a three-dimensional one. The laminate is thin, the copper is mirror-bright, the solder mask is matte and often a different colour from batch to batch, and the dimensions that matter sit at every scale at once: a 0.1 mm trace, a 0.6 mm pad, a 2 mm slot, a 250 mm board outline. Add gold fingers that cannot be scratched, a panel that has been through reflow and no longer lies flat, and the measurement problem stops being about magnification and becomes about how the part is held and how much of it can be seen at once. A flash measurement system answers that problem with a single calibrated exposure: the stage does not travel during the measurement, nothing presses on the copper, and every programmed feature in the field is found by software rather than by an operator moving a crosshair.

This page is written for the person who has to decide whether that approach fits a specific board. It covers the field arithmetic first, because that is what decides fit; then the accuracy budget, because repeatability is routinely mistaken for it; then lighting and fixturing, which are where most PCB measurement projects actually succeed or fail. The boundaries are stated as plainly as the capabilities, including the checks that belong to AOI or to a cross-section rather than to any optical measuring machine.

What a PCB actually asks a measuring instrument to do

The dimension list on a PCB is short but heterogeneous. On the bare board it is conductor width and spacing, pad diameter, annular ring, solder-mask opening, slot and rout width, hole position relative to a datum, and the overall outline. On an assembled board it adds connector position, component offset and coplanarity questions that are no longer 2D. On a flexible circuit it adds coverlay opening, stiffener outline and bend-area features that only exist while the part is held flat.

Three properties of the part shape the instrument choice. First, most of these dimensions are edge-based and measured in the plane, which is exactly what a calibrated field can do well. Second, the tolerance classes are mixed: an outline dimension carries a tolerance an order of magnitude looser than a trace width, so a single uncertainty figure quoted for the whole board is misleading in both directions. Third, the board is a low-contrast object in some places and a mirror in others, so the lighting has to change feature by feature rather than be fixed for the job.

Why single-exposure measurement fits panel and array work

A board is a repeated object. A panel carries an array of identical circuits, and a production lot carries many identical panels. That repetition is what makes a single-field instrument efficient: the measurement program is taught once on one circuit, and every subsequent circuit is measured by the same program in the same calibrated field, with no re-datuming and no operator judgement about which edge to pick.

The mechanical side matters just as much. A probe or a clamp on a finished board is a defect risk: gold fingers, ENIG pads and thin traces do not tolerate a stylus, and a clamp that flattens a warped board changes the dimensions being measured. A flash measurement system places no load beyond the part's own weight - the published stage load for the IMAGE 3 series is 5 kg, far above any panel - images from above, and measures what the field sees. The constraint on a PCB job is therefore never payload. It is field size, calibration and the smallest feature the optics can resolve.

Does the board fit the field? Run this arithmetic first

The IMAGE 3 series reaches a wide-field measuring range of up to 300 x 200 x 75 mm, model dependent, with a precision measuring range from 118 x 118 mm to 218 x 118 mm depending on configuration. Those two numbers define two different jobs, and confusing them is the most common way a PCB project is specified wrongly. The wide field is the panel-level view: it captures many circuits at once at a coarser pixel scale. The precision range is the tight-tolerance view: it captures a smaller region at a finer pixel scale, which is where trace width and annular ring belong.

Panel formats are the first check, and the arithmetic is simple enough to do before any quotation. An 18 x 24 inch panel is 457.2 x 609.6 mm (derived at 25.4 mm per inch), which exceeds both axes of a 300 x 200 mm field, so no rotation brings it inside. Halving it once gives 228.6 x 304.8 mm, which still exceeds the 200 mm short axis in both orientations. A quarter panel, 228.6 x 152.4 mm, fits comfortably. The practical conclusion for a fabricator running 18 x 24 inch panels is that a 300 x 200 mm class instrument measures a quarter-panel at a time, or a sampled set of circuits, not the whole panel in one exposure.

Part or panel format	Footprint (mm)	Fits a 300 x 200 mm field?	Practical approach
Single small circuit	under 100 x 100	Yes, many per exposure	Array measurement: several boards in one field
Quarter of an 18 x 24 in panel	228.6 x 152.4 (derived)	Yes	Four placements per panel
Half of an 18 x 24 in panel	228.6 x 304.8 (derived)	No - 304.8 mm exceeds the 300 mm axis	Two placements per panel
Full 18 x 24 in panel	457.2 x 609.6 (derived)	No	Per-region sampling plan or a larger field
200 x 250 mm test coupon panel	200 x 250	Only in the 200 mm orientation, with no carrier margin	Check fixture outline before committing

The table is arithmetic, not a product limitation to be argued with. What it changes is the recipe: a job that does not fit becomes a sampling plan with a defined set of measured regions, and the inspection report has to say which regions were measured. That is a legitimate and often preferable way to run production - most PCB dimensional checks are about watching a process, not about certifying every square millimetre - but it has to be a decision, not a discovery after installation.

Accuracy on a board: repeatability is not the number you quote

The current IMAGE 3 model sheet gives repeatability of up to ±0.25 µm, model dependent, and measuring accuracy from ±2 µm in precision mode without a connected sensor to ±(1.5 + 0.02L) µm with one, where L is the measured length. Those are two different statements and they apply to two different questions. Repeatability describes how tightly the same feature repeats when it is measured again. Accuracy describes how close the reported value is to the true value, and on a board it is dominated by the length term whenever the dimension spans more than a few millimetres.

Evaluated from the published formula, the length-dependent allowance behaves as follows (all values derived):

Measured length L	0.02L term (derived)	Total in connected precision mode (derived)	Typical PCB dimension
5 mm	0.1 µm	±1.6 µm	pad pitch, small slot width
50 mm	1.0 µm	±2.5 µm	connector row, local outline
200 mm	4.0 µm	±5.5 µm	board outline, datum-to-datum across a panel

Read the table against the drawing tolerance, not against the repeatability figure. A ±5.5 µm budget on a 200 mm outline is generous for a board outline and hopeless for a 50 µm trace; a ±1.6 µm budget on a 5 mm feature is workable for pad geometry but still not a sub-micron CD claim. The honest method on a PCB job is to split the dimension list into two classes, measure each class in the field where its tolerance is achievable, and state the uncertainty class on the report rather than a single number for the whole board.

Two consequences follow for recipe design. Long dimensions should be measured in the wide field, where the whole span is inside one calibrated exposure and there is no stitching error to add. Tight dimensions should be measured in the precision field, where the length term is small and the pixel scale is fine enough to place an edge repeatably. Trying to serve both with one magnification produces a recipe that is imprecise on the tight features and unnecessarily slow on the loose ones.

Lighting a board: copper, mask and bare laminate behave differently

A PCB is the hardest lighting problem in dimensional metrology because the same field contains three surfaces with different reflectance behaviour. Bare copper and gold fingers are mirrors: a ring light alone produces a specular band that the edge detector reads as a displaced edge, and the measurement then tracks the highlight instead of the feature. Coaxial illumination delivered along the optical axis keeps the specular return centred and makes the true edge the strongest transition. Solder mask is the opposite: matte, often green or black, and it absorbs enough light that a coaxial-only setup gives a soft edge with poor contrast, so multi-angle surface light is the better choice. Through-holes, slots and routed outlines need the part silhouetted, which is what the parallel backlight is for.

The IMAGE 3 configuration reflects that reality: adjustable multi-angle surface light plus coaxial light for surface features, and a parallel backlight for contour work, switched per feature inside the program rather than set once for the job. The reason this matters more on a PCB than on a machined part is that a single board can require all three within one measurement program - coaxial on the gold fingers, angled surface light on the mask-defined opening, backlight on the slot.

There is a second-order effect worth knowing before a recipe is frozen. Solder mask colour and gloss vary between batches and between fabricators, and HASL finishes are deliberately uneven. A recipe tuned on one batch can drift out of contrast on the next, which shows up as an edge-detection failure rather than as a wrong number. The robust practice is to keep the lighting parameters in the program per feature, and to re-run a golden sample at each batch change so the failure appears as a failed check rather than as silent measurement drift.

Which PCB dimensions a flash system holds, and which it does not

Inside a calibrated field, the reliable list is the 2D edge-based set: conductor width and spacing, pad diameter and position, annular ring, solder-mask opening, slot and rout width, board outline and hole-to-datum position. These are measured from one exposure with the part anywhere in the field at any rotation, because the pixel-to-millimetre coefficient belongs to the calibrated field and not to the part position. The software side is a taught program with one-click measurement and Excel report output, so a repeated panel check is a fixture load and a button, not a metrology session.

Outside that list, three categories need to be named before a purchase order. Height and coplanarity are not part of the 2D chain: they require the 8055 point-spectral confocal sensor, which is standard on the MAX configuration and optional on MAX2, and its published range is ±3 mm with a 55 mm working distance, a 17 µm spot and 0.02 %F.S. resolution - a height sensor, not a full-board 3D scanner. Hole-wall copper thickness and plating integrity are cross-section measurements and cannot be inferred from a top view. And sub-micron critical dimension work is outside the instrument's class altogether: the 0.1 µm minimum display unit is a display resolution, not a statement that a 0.1 µm line can be measured with usable uncertainty.

Where the flash system stops, and which head takes over

The boundary on a PCB line is usually drawn in three places, and each one has a different neighbour.

Defect detection is AOI work, not metrology. AOI asks whether a joint is present, whether a trace is broken, whether a component is rotated; it optimises for coverage and speed at a pixel scale set by the smallest defect, and it does not produce a calibrated dimension. Choosing the optics for that job is a separate exercise, covered in PCB AOI optics: pixel size, lens type and working distance. A flash measurement system complements AOI by quantifying the dimensions AOI only flags.

High-magnification review of a small region is a lens problem, not a field problem. When the question is a 20 µm feature, a scratch on a gold finger or the edge quality of a drilled hole, the answer is pixels concentrated on a small area: a telecentric lens for PCB and FPC inspection on a machine-vision camera, or a stereo microscope for hands-on rework, as compared in PCB repair station optics.

Continuous measurement of a moving strip is a different geometry again. If the product is a flexible circuit on a roll or an etched strip coming out of a process, the instrument is an inline dimensional measurement head, not a stage-mounted flash system.

The constructive arrangement on a PCB line is therefore not one instrument but a division of labour: flash measurement for the dimensional set that repeats on every circuit, AOI for presence and defect, and a magnified head for the small number of features that need resolution rather than field. All three read into the same QC record, and none is asked to do another's job.

Making a panel recipe survive production

Four things decide whether a PCB measurement program still passes a golden sample three months after installation.

Fixture flatness. A board that has been through reflow is not flat, and a measurement taken on a warped board measures the warp as well as the feature. Telecentric optics hold magnification constant when the object moves in and out of focus, which removes the size error but not the focus error, so a vacuum plate or a kinematic support at the datum holes is part of the measurement, not an accessory. When warp is the question itself, it needs the confocal height sensor and a defined sampling grid, and the result is a height map rather than a 2D dimension.

Field calibration as a scheduled item. Because the calibrated field is the length reference, the calibration artefact and its re-check interval are the most important maintenance items on the instrument - more important than any lens cleaning routine. A practical schedule re-checks the field calibration at the start of each shift and after any change of magnification recipe, and logs the residual so that slow drift appears as a trend rather than as a batch of boards that fail on one side of the panel only.

Program discipline. Feature definitions, lighting per feature, magnification and datum choice belong in the stored program, and the program version belongs on the inspection report. A recipe edited on the floor without a version stamp turns a process trend into an unexplained step change, which is exactly the kind of signal a dimensional check exists to provide.

Sample handling. Boards are thin, flexible and easily marked. Support the part so it cannot deflect under its own weight, keep the measurement side free of contact, and load the same way every time - the same datum holes, the same orientation. The instrument can measure a part anywhere in the field at any rotation, but a consistent load is what makes consecutive measurements comparable at the micron level.

Integration checklist for a PCB measurement job

• Panel format checked against the 300 x 200 mm wide field and the 118 x 118 to 218 x 118 mm precision field before any optics discussion; if the panel does not fit, define the measured regions explicitly.
• Dimension list split into long dimensions (wide field, length term applies) and tight dimensions (precision field, small length term), with a stated uncertainty class for each.
• Smallest feature of interest compared with the field calibration and the sensor, not with the 0.1 µm display unit.
• Height, coplanarity or warp requirement identified up front, so the 8055 confocal option is priced with the instrument rather than after it.
• Lighting decided per feature on real boards: coaxial for copper and gold, angled surface light for mask-defined edges, backlight for slots and through-holes.
• Fixture and datum scheme agreed with production - vacuum plate or datum-pin support, consistent load orientation.
• Calibration artefact, re-check interval and residual log written into acceptance, before the instrument is released to the line.

Frequently asked questions

Can a flash measurement system measure a full PCB panel in one shot?

Only if the panel fits the field. An 18 x 24 inch panel is 457.2 x 609.6 mm, which exceeds both axes of a 300 x 200 mm field, and even a half panel at 228.6 x 304.8 mm exceeds the short axis in both orientations. A quarter panel at 228.6 x 152.4 mm fits. For full panels the practical answer is a sampling plan with defined measured regions, or an array measurement of several small circuits in one exposure.

How do you measure gold fingers and ENIG pads without marking them?

By not touching them. Measurement is a single exposure from above with no probe and no clamp, so there is no contact on the finish at all. The optical issue is the opposite of a mechanical one: gold and bare copper are mirrors, so coaxial illumination is used to keep the specular return centred and make the true edge the strongest transition. Ring or angled light alone tends to measure the highlight rather than the pad edge.

Is ±0.25 µm repeatability the accuracy on a 200 mm board dimension?

No. That figure describes how tightly the same feature repeats. Accuracy on a long dimension is dominated by the length term: with the published ±(1.5 + 0.02L) µm figure in connected precision mode, a 200 mm dimension carries a 4 µm allowance from the length term and about ±5.5 µm in total, both derived from the formula. Quote the two classes of dimension separately on the inspection report.

Can it check board warpage or hole-wall copper thickness?

Hole-wall copper thickness is a cross-section measurement and cannot be obtained from a top view. Warpage is a height question: the standard 2D chain does not measure it, and the 8055 point-spectral confocal sensor, standard on the MAX configuration and optional on MAX2, provides a ±3 mm height range with a 55 mm working distance and a 17 µm spot for a sampled height map. A full-board 3D scan is a different instrument category.

Next step

Send the panel format, the fixture outline, the dimension list with tolerances and the smallest feature that has to be resolved through the contact page. The field check comes first and it is arithmetic, so a panel that does not fit is settled in one reply rather than in a quotation round. For the instrument itself, the IMAGE 3 image dimension measurement guide covers the series and its configurations, the configuration page covers what changes between models, and the flash measurement versus VMM comparison covers the case where a travelling stage is the better answer. The product page for the IMAGE 3 instant vision measuring system lists the current model data. For a small stamped or connector-level part on the same line, the flash measurement for connectors and flash measurement for stamping pages apply the same method to smaller parts.