The camera module passed every bench test. Sharp in daylight, clean at night, good contrast at the far end of the test room. Then it went behind the cover panel of the product, and the night picture turned grey, as if someone had breathed on the lens. The module did not change. The only new part was a few millimetres of plastic in front of it.
Night haze behind a cover window comes from the camera’s own infrared LEDs leaking into the lens through the window instead of reaching the scene. It is a big problem because the real image signal is tiny: in our model of a crib camera, only about 1 in 4 million of the emitted IR comes back from the mattress into the lens. A leak of about 1 in 100 million spread over the frame already matches the signal from the far corner. There are three leak paths: reflection off the inside of the window, light guided sideways inside the window, and scatter from dirt or scratches on the outside. A smaller air gap does not fix the first one, and a sealing gasket does not fix the second. The reliable fix is a light barrier that goes through the full window thickness, or separate windows for lens and LEDs.
This is Part 04 of our series on designing a camera into a crib. The physics applies to any camera with IR LEDs behind a shared cover window, but a crib rail makes it more likely: designers want one flush panel that matches the furniture, and that is the layout most prone to leaks.
§01Why a tiny leak ruins the night picture
In daylight, the IR LEDs are off and the scene is lit by the room, so window leaks do not show. At night, the LEDs are the only light source, and the camera has to collect what bounces back from a matte mattress tens of centimetres away. Very little of it comes back.
The conclusion does not depend on the exact assumptions. The image signal is millions of times weaker than the light the LEDs emit, and the LEDs sit millimetres from the lens. A window that redirects even a minute fraction of that light toward the lens will add a grey veil that lifts black levels, flattens contrast and hides detail at the far corners first, where the signal is weakest.
§02Three ways IR gets from the LED to the lens
FIG. 01 · Schematic cross-section, not to scale. The dashed outline shows where a sealing gasket sits. Path 2 travels inside the window, above the gasket.
Reflection off the inside of the window
Every uncoated surface reflects part of the light that hits it. For common window plastics at normal incidence, that is around 4–5% per surface, and the share rises steeply at shallow angles. Some of the LED light reflects off the inner face of the window, back into the cavity, and lands on the lens.
Stopped bya light barrier between LED and lens that is pressed against the window, with no gap.Light guided inside the window
A clear window is also a light guide. Light that enters it at a steep enough internal angle, through scattering in the material, a textured or decorated surface, or a molded edge, is trapped by total internal reflection and runs sideways inside the plate. Where it meets a scratch, a printed edge or the area over the lens, some of it escapes. Mobile phones face the same problem with proximity sensors behind cover glass, which is why their sensor windows are usually separated by ink or by an opaque break in the glass.
Stopped bya break in the window itself: separate windows, or an opaque divider molded through the full thickness. A gasket on the inner face does not reach it.Scatter from the outside surface
Light that passes cleanly through the window can be scattered straight back by whatever is on the outer surface: fingerprints, dust, residue from cleaning wipes, scratches. Because the outer surface is very close to the lens, scattered light reaches it at high intensity. This path changes over the product’s life, as the window gets touched and cleaned.
Reduced bya scratch-resistant surface, keeping the window recessed rather than proud of the rail, and placing the lens as close behind the window as the design allows. It cannot be removed completely, only kept small.§03Why a smaller gap does not fix reflection
A common first response to path 1 is to bring the window closer to the LEDs and lens, on the reasoning that a smaller air gap leaves less room for reflected light. The geometry says otherwise.
With a smaller gap, a reflected ray has to leave the LED at a shallower angle to reach the lens. The LED emits less light at shallow angles, but the window reflects much more of it there. The two effects largely cancel.
| LED-to-lens spacing · air gap | Ray angle from window normal | LED intensity at that angle | Window reflectance | Path strength |
|---|---|---|---|---|
| 6 mm · 3 mm | 45° | 71% | 6.2% | 4.4% |
| 6 mm · 1 mm | 72° | 32% | 20.7% | 6.6% |
| 6 mm · 0.5 mm | 81° | 16% | 41.9% | 6.9% |
| 10 mm · 2 mm | 68° | 37% | 16.5% | 6.1% |
| 10 mm · 0.5 mm | 84° | 10% | 58.2% | 5.8% |
TABLE 01 · Polycarbonate window (n ≈ 1.59), unpolarised light, LED with a 60° half-intensity angle facing the window. “Path strength” = LED intensity × reflectance, as a share of the LED’s peak intensity.
Across every combination, the reflection path stays at roughly 4–7% of the LED’s peak intensity. Spacing the LED further from the lens does not help much either. Neither change makes a difference compared with the one-in-a-hundred-million figure from §01. The only thing that closes path 1 is an opaque barrier that seals against the window.
§04Window materials
Material choice changes surface reflection only a little. It matters more for durability and for how the window behaves at the IR wavelength, which is not something you can judge by eye.
| Material | Reflection per surface, straight on | Total internal reflection beyond | Practical notes |
|---|---|---|---|
| PMMA (acrylic) | ≈ 3.9% | ≈ 42° | Harder surface than uncoated polycarbonate, more brittle on impact. |
| Polycarbonate | ≈ 5.1% | ≈ 39° | Tough on impact, softer surface that scratches more easily without a hard coat. Some cleaning chemicals can attack it. |
| Glass | ≈ 4.3% | ≈ 41° | Hard and scratch resistant, but breakage is a question most nursery products avoid. |
TABLE 02 · Reflection from refractive index at normal incidence. Values are typical for each material family; specific grades differ.
Dark to the eye is not dark to the camera
Designers often want the window to look black so the camera disappears into the rail. There are plastics made for exactly that: dark in visible light, transmissive in the near infrared. There are also dark plastics that look identical and block most IR. The only way to know is the material’s transmission curve at the LED wavelength, around 850 nm or 940 nm, measured at the actual window thickness.
The same applies to the gasket. Black foam or rubber is not necessarily black in the infrared, because some black pigments reflect IR quite well. A gasket that looks black but reflects at 850 nm adds a reflection path of its own. Ask for IR reflectance data, or test it in the night setup below.
Coatings
An anti-reflection coating on the window reduces path 1 at the source, and a hard coat protects against the scratches that feed path 3. Both add cost and need to survive the cleaning products the end user will actually use. Neither replaces a light barrier.
§05What is different about a crib
The same window in a ceiling security camera would sit out of reach for years. In a crib rail, the window faces a child at close range, and that changes what happens to path 3 over time.
- It will be touched. Hand prints and saliva on the window scatter IR straight back into the lens.
- It will be cleaned often, with whatever wipes the household uses. Some cleaning chemicals can haze or crack certain plastics.
- Crib rails get chewed. That is why rail covers are sold for teething children. A window flush with a chewable surface will collect scratches, and each scratch is a scatter point.
The design response is to recess the window behind the rail surface rather than leave it flush, to specify a hard-coated surface rated against common cleaners, and to test night image quality after a scratch and cleaning cycle, not only on a fresh sample. Whether a recessed window and its retention meet the crib’s structural and small-parts requirements is a separate check, covered in Part 05.
§06Layouts that work
Lens and LEDs each have their own opening in the rail, with solid rail material in between. This closes paths 1 and 2 completely. It is the most reliable layout, but it shows the LEDs as separate features on the rail.
A two-material panel with an opaque section molded through the full thickness between lens and LEDs. It looks like one flush window from the outside, and it blocks paths 1 and 2. Tooling is more complex.
Blocks path 1 if the gasket seals properly and is IR-absorbing. It leaves path 2 open. Acceptable only if a night test on the real panel shows the haze is under control.
The default when nobody looks at the night image until late sampling. It works in daylight, but the night image shows haze.
None of these choices can be made by the camera module supplier alone. They belong to the rail and enclosure design, which is why the window needs to be part of the camera specification from the first drawing, not added by the industrial designer after the module is chosen.
§07How to catch it in sampling
Cover window night test
Run in a dark room with the camera in its real enclosure. Compare each step against the bare module to see what the window adds.
- Black target test. Point the camera at black velvet or into an open dark space, IR on. A perfect setup shows near-black. Any grey level is stray light from the window and enclosure.
- Window on, window off. Repeat step 1 with the window removed. The difference in black level is what the window costs you.
- Cover the LEDs one group at a time with opaque tape, to see which LEDs feed the haze.
- Far-corner contrast. Repeat the seven-card mattress test from Part 03 behind the window and compare far-corner contrast with the bare module.
- Fingerprint test. Put one fingerprint on the window over the lens and repeat step 1.
- Wear test. Run a scratch and cleaning cycle on a sample window, then repeat steps 1 and 4.
- Gasket material check. Image the gasket material under the IR LEDs. If it shows up bright, it reflects IR.
Where the numbers come from. The signal and leak figures are order-of-magnitude estimates from the model in the method note. The reflectance figures follow from the refractive index of each material family. Neither is a measurement of a specific product. The test above is how a specific design should be judged.
If you are planning a camera behind a flush panel, in a crib or in any other enclosure, send us the panel drawing early. The window layout is easier to change than the tooling built around it. Our Camera & Video Monitoring page covers embedded camera projects, and the development process shows where enclosure sampling sits.
§08Questions buyers ask
Why does my camera look hazy at night behind a cover window but fine in daylight?
In daylight the IR LEDs are off, so the window has nothing to leak. At night the LEDs are the only light, and some of their output reaches the lens through the window instead of through the scene. Because the real image signal is millions of times weaker than the LED output, even a very small leak shows as a grey veil.
Will reducing the gap between the window and the LEDs stop the reflection?
Not by itself. A smaller gap forces the reflected light to leave the LED at a shallower angle, where the LED is dimmer but the window reflects more. In our calculation the reflection path stays at roughly 4 to 7 percent of the LED’s peak intensity across gaps from 0.5 to 3 mm. An opaque barrier sealed against the window is what stops it.
Is a gasket between the lens and the LEDs enough?
It stops reflection off the inner surface if it seals and absorbs IR. It does not stop light travelling sideways inside the window, because that light is above the gasket, within the window’s thickness. That path needs separate windows or an opaque divider through the full thickness of the panel.
Can I use a black window over an IR camera?
Yes, if the material is made to transmit near-infrared. Some dark plastics transmit well at 850 or 940 nm and some block it, and they can look the same. Check the transmission curve at your LED wavelength and window thickness before choosing.
Why does the window matter more in a crib than in other cameras?
Because it faces a child at close range. It will be touched, cleaned often and possibly chewed. Fingerprints, cleaning residue and scratches scatter IR back into the lens, so night image quality should be tested after wear, not only on a new sample.
Camera going behind a panel?
Send us the panel or rail drawing with the window, lens and LED positions. We will flag the leak paths and suggest a layout before the panel is tooled.
Send Your Panel Drawing → Window layout · Gasket and divider design · Night test · PCBA