Why the Edge of a Micro Hole Blurs in Backlighting: Checks in Order

A micro hole in backlight should produce the cleanest edge in machine vision: a disc against a contrasting field with a step transition. When it arrives as a ramp several pixels wide, the station is locating an edge that does not exist yet, and every downstream number inherits the uncertainty.

Four mechanisms produce the ramp. Diffraction sets a floor no lens can beat. Defocus adds to it whenever the part is not exactly at the working distance. The illumination's angular spread smears the edge independently of the optics. And insufficient pixel sampling turns a sharp edge into a two-pixel interpolation problem.

The reason to be disciplined is that the fix differs in each case, and two of the four fixes are free. Changing the illumination costs part of an afternoon; buying a higher-NA lens costs a procurement cycle and may make the depth of field worse.

What the edge of a hole actually is

A hole in backlight is imaged as the boundary between light that passes through the aperture and light that does not. Its sharpness depends on how collimated the light is, how wide a cone the lens accepts, and whether the aperture plane is the one in focus — not on surface finish, material or colour. That is why backlight gauging is robust, and why a blurred edge reliably signals that one of those three is wrong.

The six checks, in the order worth running

1. Confirm the backlight is collimated and actually covers the field

A diffuse backlight emits at every angle, so light reaches the part up to and slightly past the geometric edge, producing a penumbra rather than a step. This is an illumination numerical aperture effect, independent of the lens.

The POMEAS VP-GLTD telecentric parallel LED backlight series is designed for this, with beam diameters of 30 mm, 60 mm, 100 mm, 145 mm, 189 mm and 265 mm, working distances covering 45–500 mm, and red, green, blue and white options. The product page states that parallel illumination reduces the boundary blur caused by light diffracting or spreading around the edge.

Coverage is the second half of this check. The beam diameter has to cover the whole field with margin, because brightness falls near the beam boundary and an edge sitting there produces a different grey level than the same edge at the centre.

2. Confirm which aperture you are measuring: the hole, or its entrance

In a part of finite thickness a through-hole has an entrance aperture, an exit aperture and a bore wall between them; which one the lens sees depends on the focus plane and on how collimated the illumination is. If the two differ, the measured diameter depends on which is imaged.

The check is to move focus deliberately through the thickness and watch the reported diameter. If it changes, the station is measuring different apertures at different times, and the blur is partly a focus-plane ambiguity rather than an optical limit.

3. Compare the part thickness against the published depth of field

If the entrance and exit faces are separated by more than the depth of field, one of them is defocused, and a defocused edge is a wide, low-gradient ramp. Published POMEAS values at the magnifications used for micro-hole work:

At 0.017 mm, PMS-LGX8-95 has essentially no tolerance for part thickness or Z variation; the entrance plane has to be the focus plane by design. At 3.03 mm, PMS-LMC04-150 will hold both faces of a thicker part, at the cost of a lower magnification.

4. Check the diffraction limit against the edge transition you need

The lens sets a floor. The diffraction-limited resolution is commonly estimated as:

resolution ≈ 0.61 × λ / NA

Taking λ = 550 nm as the assumed wavelength — a test assumption, not a POMEAS published value — against published NA figures:

If the required edge transition is finer than the lens's published resolution figure, no change to illumination, sampling or algorithm will produce it. The only routes are a higher-NA model, which costs depth of field, or a shorter wavelength, which is why the VP-GLTD series offers four wavelength options.

This figure is a diffraction limit, not a measurement accuracy. A station that resolves a hole edge does not thereby measure the hole diameter to that figure; accuracy additionally depends on pixel sampling, contrast, edge definition, illumination stability and the algorithm.

5. Check the sampling before blaming the optics

A sharp edge spanning one pixel is not measurable to sub-pixel precision. The pixels across the transition are set by the pixel pitch divided by the magnification against the optical transition width from check 4: if the optical edge is 2 µm and the projected footprint 5 µm, the camera is the limit.

POMEAS publishes object-side resolution figures from 1.1 µm (PMS-LGX8-95) through 10 µm (PMS-LTC1-65) and 13.42 µm (PMS-LTC1-110C) on models relevant to micro-feature work. Matching one to a camera is arithmetic the buyer has to do with the camera's own datasheet.

6. Confirm the algorithm is not adding the blur

Averaging filters, a threshold set mid-ramp rather than at a defined intensity, and a normalisation that changes with ambient light all move the reported edge position. The check is to plot the intensity profile across the hole edge in a raw frame: if the profile shows a step and the result a ramp, the blur was added after the optics.

Where POMEAS telecentric lenses land on those six

Reading the table against the six checks:

- PMS-LGX8-95 is the highest-NA option at 0.3, publishing 1.1 µm resolution at 9.5 mm with a field of 0.4 mm × 0.3 mm on a 1/4-inch sensor. It sets the finest diffraction floor here and the smallest depth of field, which makes check 3 decisive. - PMS-LTCM2-65-S and PMS-LTC2-110 are both 2.0× and differ in the two numbers that matter: resolution 2.2 µm against 8 µm, working distance 65 mm against 110 mm. For a small hole on a flat part, the 65 mm model's resolution figure is the reason to choose it. - PMS-LTCM12-110 publishes 6.71 µm at 1.2× with 0.67 mm depth of field, a Φ30 mm body and a field of 7.3 mm × 5.5 mm on a 2/3-inch sensor, with object-side telecentricity <0.02° and TV distortion 0.01% — the middle option. - PMS-LMC04-150 publishes 3.03 mm of depth of field at 0.4× and 148 mm, the configuration for a hole in a part with real thickness. It publishes no resolution and no NA.

Each row lists only what that model publishes. Blank entries are gaps in published data, not zeros, and no value should be carried across from another row.

What those numbers mean on the line

Scenario 1 — The ramp disappears when a collimated backlight replaces the diffuse panel

A station gauging sub-millimetre holes reports an edge ramp about four pixels wide, and the backlight is a diffuse LED panel. Replacing it with a VP-GLTD collimated source — a 30 mm beam at 45–90 mm, or a 60 mm beam at 90–180 mm — collapses the ramp. The lens was never the problem.

Scenario 2 — Two apertures, one measurement

A drilled hole in a 0.8 mm plate reports different diameters depending on where the operator focused: PMS-LTC2-110 publishes 0.5 mm of depth of field, so both faces are not in focus at once. Defining the measurement on the entrance face, or moving to PMS-LMC04-150 at 3.03 mm, resolves it.

Scenario 3 — The floor is diffraction, and illumination cannot fix it

The station needs an edge transition finer than 2.2 µm. PMS-LTCM2-65-S publishes exactly that at NA 0.152, and PMS-LGX8-95 publishes 1.1 µm at NA 0.3. Below those there is no POMEAS model in this group that will do it; the realistic levers are a shorter wavelength and accepting a coarser transition.

Scenario 4 — The edge is sharp in the raw frame and blurry in the result

The intensity profile across the hole edge shows a step, but the reported diameter still wanders. This is check 6: the algorithm is smoothing, or the threshold is defined against a normalisation that moves with ambient light. Fixing it costs nothing in hardware, and it is worth confirming before any optical change is specified.

What this article does not tell you

This is a technical reference, not a diagnosis of your station. No claim is made about what is blurring the edge on a specific installation. Identifying which mechanism dominates requires the intensity profile from your own camera.

Optical resolution is not measurement accuracy. The diffraction figure limits how sharp the edge can be; accuracy additionally depends on pixel sampling, contrast, edge definition, illumination stability, and the algorithm.

Every figure here is the published value for the specific POMEAS model named beside it, not measured under a single common condition. The 550 nm wavelength is an assumed test wavelength, not a POMEAS published value, and the results of that calculation are calculations rather than specifications. Where a field is absent — depth of field for PMS-LTCM2-65-S, NA for three of the five, resolution for PMS-LMC04-150 — read it as "not published".

The telecentric content already on this site covers numerical aperture, sensor matching and working distance selection as separate pages. This article is about the micro-hole edge case.

FAQ

What is the most common cause of a blurred micro-hole edge in backlight?

A diffuse backlight. Light arriving at every angle illuminates past the geometric edge and produces a penumbra. A collimated source such as the VP-GLTD series is the usual remedy.

Can a telecentric lens overcome diffraction blur?

No. Diffraction sets a floor of roughly 0.61 × λ / NA. At NA 0.3 that is about 1.12 µm at 550 nm, which matches the 1.1 µm published for PMS-LGX8-95.

Why does the measured hole diameter change when I refocus?

Because a through-hole has an entrance and an exit aperture. If both are not inside the depth of field, the station is measuring whichever one is sharper at the time.

Does a higher-NA lens always give a sharper hole edge?

It gives a finer diffraction limit and costs depth of field. PMS-LGX8-95 at NA 0.3 publishes 0.017 mm of depth of field, so the part has to be positioned accordingly.

How many pixels do I need across the edge?

Enough that sub-pixel interpolation has something to work with. Compare the camera's pixel pitch divided by the magnification against the optical transition width from the lens's published resolution.

Should I change the wavelength?

It is worth trying, because edge contrast and diffraction both depend on wavelength. The VP-GLTD series offers red, green, blue and white so the choice can be made rather than assumed.

Next Step

If you are specifying a replacement and need measured values for your own conditions — your sensor size, your working distance, your feature size — send us those four numbers and we will run them against the current catalogue and tell you what actually fits, including when nothing does.