Fit an ordinary night-vision camera module into a crib rail, switch off the lights, and the picture usually looks wrong in the same way. The strip of mattress next to the rail glows white, the middle looks acceptable, and the far corners fade into grey. The camera is working as designed, but it was designed to sit a metre or two away from the mattress, not in the rail.
In a crib rail, three effects stack: distance, the angle at which infrared light strikes the mattress, and the LED beam pattern. In our model of a camera 25 cm above a US full-size mattress, distance alone gives a 17× brightness difference between the near edge and the far corner. Add the angle of incidence and it becomes 67×. Add a standard symmetric LED ring and it reaches about 97×, or 6.6 photographic stops. Narrow-beam LEDs make it far worse. A zoned layout, with a wide low-power group for the near side and higher-power groups aimed across at the far corners, brings it down to about 11–18×. It needs roughly 40–60% more total LED power for the same darkest-point brightness, which is a heat problem inside a closed rail.
This is Part 03 of our series on designing a camera into a crib. Part 02 chose the lens. This part lights what the lens sees.
§01Three effects that stack
Our earlier article on night vision range covers the inverse-square law for a freestanding camera. In a rail, two more effects pile on top of it.
| What is included | Brightest ÷ darkest, 25 cm | Same, 40 cm |
|---|---|---|
| Distance only | 17× · 4.0 stops | 7× · 2.8 stops |
| + angle of incidence on the mattress | 67× · 6.1 stops | 19× · 4.2 stops |
| + symmetric LED ring, 60° half-angle | 97× · 6.6 stops | 27× · 4.7 stops |
| + symmetric LED ring, 30° half-angle | ≈ 4,200× · 12 stops | 208× · 7.7 stops |
TABLE 01 · Camera at mid-rail, LED ring aimed along the lens axis (55° below horizontal at 25 cm, 62° at 40 cm).
Distance. The nearest point of the mattress is about 25 cm from the LEDs, the far corner about 102 cm. Squared, that is 17×.
Angle of incidence. Light arriving straight down onto the near edge spreads over a small area. Light arriving at a shallow angle onto the far corner spreads over a much larger area, so each square centimetre gets less. At 25 cm, that factor alone takes 17× to 67×. This is the part most night-vision explanations leave out, and it is specific to cameras that look across a surface rather than at it.
Beam pattern. An LED is brightest along its axis. A ring aimed at the middle of the mattress sends its strongest light to the near half. With a 60° half-angle, the result is 97×. With a 30° half-angle, the near corners fall almost completely out of the beam, and the difference passes 4,000×.
§02What 6.6 stops does to the picture
The camera has to pick one exposure for the whole frame. With a 6–7 stop spread across the mattress, there is no setting that suits both ends.
The strip next to the rail looks clean. The far half of the mattress drops into dark grey, and the far corners are close to black.
The far side becomes visible, but the near edge and the middle clip to flat white, which hides texture and movement where the camera is closest.
Auto-exposure usually lands between the two, so both ends lose. In our model with a symmetric 60° ring at 25 cm, only about half of the mattress area sits within one stop of the median brightness. The rest is either noticeably too bright or noticeably too dark.
FIG. 01 · Each layout is shown against its own median, so the bars compare spread, not absolute brightness. The zoned layout is the constrained version in §04, with no LED aimed shallower than 20° below horizontal.
§03The narrow-beam trap
When a night image looks too dark at the far side, the common supplier suggestion is narrow-beam IR LEDs, because they are sold for “longer night vision range”. On a shelf camera aimed at a distant crib, that can be reasonable. In a rail it makes things worse.
A narrow beam concentrates light along its axis. Aimed at the middle of the mattress from 25 cm, that axis passes close to the near half. The near edge gets brighter still, and the near corners at both ends of the camera’s own rail fall outside the beam entirely. In our model, a 30° half-angle ring leaves the near corners at about 1% of the median brightness. The overall spread goes from 6.6 stops to about 12.
§04A zoned layout, and what it costs
The fix is to stop treating the IR as one light source. Split it into groups that each handle the part of the mattress they are suited to.
Near group: wide beam, low power, aimed steeply down
The near side is close and receives light at a steep angle, so it needs very little. A wide-beam group aimed almost straight down at modest drive current is enough. In our model, a 60° half-angle aimed 85° below horizontal did this job.
Far groups: narrower beams, high power, aimed across at the far corners
Two groups aimed outward, toward the far corners, at much higher drive current. This is where a narrower beam earns its place, because these groups have a specific distant target. In our model, two groups at 30° half-angle, turned 45° to each side and aimed 20° below horizontal, running at many times the near group’s power, gave the best result within that aiming limit.
| Layout | Spread at 25 cm | Spread at 40 cm | Mattress within ±1 stop of median, 40 cm |
|---|---|---|---|
| Symmetric ring, 60° half-angle | 97× · 6.6 stops | 27× · 4.7 stops | 60% |
| Zoned, far groups 20° below horizontal | 18× · 4.2 stops | 6× · 2.5 stops | 96% |
| Zoned, far groups nearly horizontal | 11× · 3.5 stops | 4× · 2.1 stops | ≈ 98% |
TABLE 02 · Best layouts found by search within the stated limits. “Nearly horizontal” aims the far groups about 4° below horizontal, which also sends light onto the far rail and into the room. That is why we show the 20° version as the practical case.
Two things stand out. First, zoning removes most of the problem but not all of it. At 25 cm, even the best layout leaves 3.5–4 stops of spread. Second, zoning and mounting height work together. At 40 cm, the zoned layout puts 96% of the mattress within one stop of the median. That is a picture a parent would describe as evenly lit.
The cost: power and heat
Sending light across the mattress at a shallow angle is inefficient, because the light lands on a large area. In our model, a zoned layout needs about 40–60% more total LED power than a symmetric ring to bring the darkest point to the same brightness. In an open shelf camera that is a battery or adapter question. Inside a closed crib rail, it is heat that has nowhere to go. Thermal design is covered in Part 05.
The LEDs do not have to sit next to the lens
Our model puts every LED at the camera position because that is what most modules do. In a crib, the rail is long. Far-corner LEDs placed further along the near rail, toward the ends, would shorten their distance to the far corners and steepen their angle of incidence. That costs wiring and makes the enclosure more complex, but it attacks the geometry directly rather than spending more power on it.
§05Wavelength at close range
Night-vision LEDs commonly come in two wavelengths, around 850 nm and around 940 nm. The trade-off is well known, but it reads differently when the LEDs are 25 cm from a child’s face rather than across a room.
- 850 nm LEDs usually show a faint red glow when you look straight at them. Camera sensors are more sensitive at this wavelength, so less power is needed. In a rail, that glow sits right next to the sleeping child, inside the field of view.
- 940 nm LEDs are close to invisible, but silicon sensors are noticeably less sensitive there. Matching the brightness of an 850 nm design takes more drive current, which adds to the heat problem from §04.
Whichever wavelength is chosen, raising drive current on LEDs mounted at short distance from a child should be reviewed under a photobiological safety assessment such as IEC 62471, which classifies lamps and LED products by the optical radiation they emit. We mention it because a zoned design changes how much IR is emitted and in which directions. It is a check item for the project, not a statement about any specific design.
§06What image processing can and cannot add
Firmware and the image processor have a role, but only after the hardware has done most of the work.
- Per-zone drive current can be tuned in firmware, and even adjusted per mattress position, but only if the board has separately driven LED groups. A single ring on one driver has nothing to tune.
- Wide dynamic range modes in the image pipeline can recover some of the remaining spread. They do much better on a 3–4 stop scene than on a 6–7 stop scene.
- Exposure metering can be weighted toward the mattress area instead of the whole frame, so bright rails and dark room do not pull the exposure off.
None of these can add light to a corner that the LEDs do not reach.
§07How to night-test a crib camera sample
Night test protocol · crib-embedded camera
Run this in the real rail, behind the real cover window, in a dark room. Testing a bare module on a bench will look better than the final product.
- Place matte grey test cards at seven positions: near edge middle, centre, far edge middle, both far corners, both near corners.
- Repeat at every mattress position the crib allows.
- Capture raw or lightly processed frames if the module allows, and read the pixel values of each card. Do not judge from the parent display alone.
- Record the spread in stops between the brightest and darkest card, and agree a target with the product team before testing, not after.
- Check clipping at the near edge and noise at the far corners in the same frame.
- Measure the enclosure temperature after the IR has run for a full night cycle, with the rail closed.
- Look at the LEDs from the mattress position to judge any visible glow.
Step 1 has to happen behind the real cover window. A window that leaks IR sideways into the lens adds a grey haze over the whole image and reduces contrast at the far corners. That problem is the subject of Part 04.
Where the numbers come from. The figures are our own calculations from the model in the method note, on a standard mattress size. They describe what the geometry and LED physics require. They are not measurements of a specific LED, module or product. Real LEDs have beam patterns that only approximate the cosⁿ model, and real mattress covers are not perfectly matte. Confirm any layout with the night test above.
If you have a crib drawing, we can run this model on your rail position and mattress heights before LEDs are chosen. Our Camera & Video Monitoring page covers embedded camera projects, and the development process shows where night testing sits.
§08Questions buyers ask
Why is the night image uneven on a camera built into a crib rail?
Three effects stack. The far corner is about four times farther away than the nearest point at a 25 cm mounting height, light reaches it at a shallow angle so it spreads over more area, and a standard LED ring aims its strongest light at the near half. In our model the combined difference is about 97 times, or 6.6 stops.
Do narrow-beam IR LEDs help a crib camera see the far side?
Not if they are in a ring aimed at the middle of the mattress. A narrow beam makes the near side brighter and leaves the near corners nearly dark. In our model a 30° half-angle ring raises the spread to about 12 stops. Narrower beams work only as dedicated far-corner groups in a zoned layout.
What is a zoned IR layout?
Splitting the infrared LEDs into separately driven groups: a wide, low-power group aimed steeply at the near side, and higher-power groups aimed across at the far corners. In our model it reduces the brightness spread at 25 cm from about 97 times to about 11 to 18 times, depending on how shallow the far groups are aimed.
Does zoned IR use more power?
Yes. In our model it needs about 40 to 60 percent more total LED power to bring the darkest point to the same brightness as a symmetric ring, because light aimed across the mattress lands on a larger area. Inside a closed rail that becomes a heat management question.
Should a crib camera use 850 nm or 940 nm infrared?
850 nm needs less power but usually shows a faint red glow, which in a rail is right next to the child. 940 nm is close to invisible but needs more drive current for the same image brightness, which adds heat. Either way, the emission at close range should be reviewed under a photobiological safety assessment such as IEC 62471.
Night image uneven on your crib sample?
Send the rail drawing, mattress positions and your current LED layout. We will model the spread across the mattress and suggest a zoning approach before you change hardware.
Send Your Crib Drawing → IR layout model · LED zoning · Night test · PCBA