Water spray pattern inspection with a backlit pass circle

The manual check is simple: hold a plastic template with a drawn circle at the height the spray must reach, and look whether the wetted area fills the circle. This article turns that check into a camera measurement for fixture and test engineers. You will place a transparent plate at the template height, backlight it, and pass or fail each head on a coverage number — at least 95% of the reference circle — with the lens, exposure and calibration arithmetic worked out on the page.

Can a camera replace the plastic template for spray pattern testing?

Yes. A transparent plate at the template height plus a backlight behind it reproduces the manual test exactly, and software passes a head when the wetted area covers at least 95% of the reference circle after a morphological close.

The template makes one judgement: the filled fraction of a fixed circle, at a fixed height, from a fixed head. Keep that geometry and replace the eye with a measured area ratio. Mounting the plate at the required height folds the height requirement into the same image — a head that falls short wets nothing and coverage collapses toward zero — so one number tests both conditions. And because the circle lives in software as a region of interest, nothing mechanical wears, sags or gets lost in a drawer.

Why can't the vision tool find individual water droplets?

Clear water barely contrasts under front light, and droplets move, merge and split between frames, so any measure built on finding individual droplets flickers from test to test. Detect the union of everything wet inside the circle instead.

A transparent droplet only bends light at its rim; under front light an edge detector fires on sparkle, not on water. Behind a bright field the same droplets become dark silhouettes with real area, and whether the pattern arrives as 200 separate blobs or one connected film stops mattering: the coverage of the union is the same number. That invariance keeps the measurement stable across cycles. Counting droplets, or fitting an ellipse around scattered blobs, drifts every time the nozzle spits slightly differently.

Backlight or front light for imaging a water spray?

Backlight, always. Water is transparent: front light returns glare dots and mid-gray noise, while a panel behind the plate turns every wetted patch into a dark shape on a bright, even field that a single threshold can split.

Use a flat diffuse panel, not a collimated one. Coverage needs the same brightness at every point inside the circle; the collimated versus diffuse backlight trade-off matters for edge gauging, not for area fill, and the illumination geometry overview shows where backlight sits among the options. The backlights category is the place to compare panels and wavelengths.

Drive the panel so the dry plate reads 200-220 of 255 gray levels and the wet film drops below 80; set the threshold halfway, near 140. White or red both work on a mono sensor; move to 850 nm infrared when room light or walk-by shadows leak into the cell. Grab one dry frame every cycle and threshold the difference image rather than the raw frame — that cancels panel ageing and ambient drift.

Which camera and lens cover the circle at your working distance?

A 12 mm C-mount lens at 150 mm working distance frames an 85 × 64 mm field — a Ø60 mm circle with a visible ring around it — and any camera from about 2 MP upward resolves the fill decision far finer than the decision needs.

Work the field from the circle outward:

1. Field needed: a Ø60 mm circle plus a few millimetres of visible ring gives 85 mm.
2. A 1/1.7-inch sensor is about 7.4 mm wide, so m = 7.4 / 85 = 0.087.
3. f = 150 mm × 0.087 / 1.087 = 12.0 mm — a standard 12 mm C-mount lens.
4. Check: m = 12 / (150 - 12) = 0.087, so the field is 85 × 64 mm and the scale is 85.1 mm / 4000 px = 21 µm per pixel.
5. Detectability: a 2% shortfall of the circle is 57 mm², about 20,000 pixels even at a 2.5 MP scale. Resolution is never the constraint here; the threshold and the closing filter are.

The LS-CMR-0166 delivers 4000 × 3000 pixels at 30 fps over USB3 from a 1/1.7-inch sensor, a comfortable fit for the plate method. For the strobed in-flight variant below, the MV-CU020-90GC pairs a 2.5 MP global shutter with GigE in C-mount, and the area scan cameras page holds the wider GigE, USB3 and CoaXPress spread.

What exposure freezes the spray?

On the plate nothing moves, so 5-20 ms at constant backlight is enough. In flight at an assumed 2 m/s, hold blur under two pixels: at 21 µm per pixel that means a 20 µs strobe pulse.

The arithmetic: blur = speed × time = 2,000 mm/s × 0.000020 s = 0.04 mm, which is 1.9 pixels at 21 µm. Double the jet speed and you halve the pulse. A 20 µs pulse is routine for a strobe-rated LED panel driven by a light controller, while constant-on brightness collapses at that width — one more reason the plate method is the default. A short flash also wants a global shutter: a rolling shutter reads a brief flash as bright and dark bands across the frame.

How does the wetted image become a pass or fail number?

Threshold the difference image, close it with a 10-pixel disk, intersect with the circle region, and divide the wet area inside the circle by the circle's area: the head passes when the ratio reaches 0.95.

The worked-example circle is Ø60 mm, so its area is π × 30² = 2,827 mm². The 5% fail margin is 141 mm² — a hole you could cover with a Ø13.4 mm disk, and the patterns the manual test rejects leave gaps far larger than that. The noise floor is far smaller: a two-pixel threshold band along the pattern edge covers about π × 60 mm × 0.043 mm = 8 mm², or 0.3% of the circle, and closing adds roughly 1%. With the floor near 1-2%, the 95% limit clears it by three points.

The closing disk does the bridging: 10 pixels at 21 µm is a 0.21 mm radius, which fills holes up to about 0.42 mm. Work on a hydrophilic surface — cleaned glass or treated polycarbonate — so the film connects instead of beading.

Each cycle runs the same five steps:

1. Grab the dry reference frame.
2. Fire the spray and wait a fixed settle delay.
3. Grab the wet frame and subtract the reference.
4. Threshold at the mid-level and close with the disk.
5. Intersect with the circle region and apply the 0.95 rule.

How do you set the 95% limit so good and bad heads separate?

Run the bench across a golden sample: known-good heads should land at 97% coverage or above and seeded-bad heads below 92%, leaving the limit at 95% in the middle of the gap. If the groups overlap, fix the lighting before touching software.

Build the sample before the fixture ships. Run 15-20 proven-good heads — expect medians near 98-99% — then 5-10 heads seeded with real defects such as a partly blocked nozzle, a tilted head or reduced supply pressure, which should land between 80 and 91%. The three-point guard band on each side absorbs fixture drift. Locate the plate with hard stops so the reference circle sits in the same pixels every cycle, and calibrate millimetres per pixel from a backlit dot plate — the backlit dot calibration plate guide shows the set-up. Structure the shake-down like the lens and lighting test plan: one variable at a time, coverage logged per cycle. When a known-good head lands at 94%, suspect the threshold, the plate wetting or splash on the light window — not the limit.

Which parts build the bench?

Three choices decide the build: a flat backlight sized to the plate, a C-mount camera with the 12 mm lens, and — for a single station — a smart camera that runs the coverage check on board instead of a PC.

Panel size follows the plate: a 150 mm square panel suits the single-circle fixture, a 300 × 200 mm panel a wider plate or a two-head station, and the high-power panel is the strobed in-flight option.

The LS-BKL-1397 adds the centre hole for heads that must pass through the light plane on axis; the LS-BKL-0865 covers 400 × 300 mm at 92.2 W and is the panel for strobed in-flight duty. For a single station with no PC, the LS-SMC-0063 runs the imaging, thresholding and pass/fail on board with its fixed 16 mm lens. Mount it so the fixed lens frames the 85 mm field from the worked example — the same geometry, one device — and browse the smart cameras page for the rest of the family.

Two numbers size the whole bench: the template circle diameter and its height above the head. Send them through the lighting configurator or the camera configurator, or contact the applications team for a bench drawing review.