3D camera range drops when hot: thermal derating fixes

The camera passed commissioning at full range. Weeks later the farthest target sits beyond the depth cutoff after the line has run since dawn, yet the same target reads cleanly again the next morning. That pattern — range that shrinks over hours and returns after a cool-down — is thermal derating, and it is the most common late-onset range failure in active depth cameras. This page gives you the differential: the two settings that mimic the symptom, a timed range test that confirms heat with numbers, the airflow and duty-cycle fixes that move the range back, and the margin rule that keeps it from returning. It is written for anyone running ToF or structured-light depth cameras on production duty.

Rule out the two settings that mimic heat

Two settings shorten range without any hardware fault, so clear both first — they take a minute each.

First, the depth preset. Depth camera SDKs ship named scene profiles, and a profile written for bright scenes or short range cuts emitter power and exposure. If the camera booted into the wrong profile, far targets vanish from the first frame — instantly, not gradually. Select the profile that matches your lighting and reach before anything else.

Second, the viewer's display clipping. Depth viewers carry a display-only maximum-distance setting, and one left at a small value hides everything beyond it on screen while the stream itself stays intact. Return it to its default maximum and retest.

The discriminator is time. A settings fault is constant from boot; derating grows hour by hour and reverses on cooling. Exposure choices matter here too: long range needs exposure headroom, the depth camera frame rate and exposure interaction shows why high frame rates cap it, and the API exposure range query tells you what headroom the driver allows before you commit a value. Depth-threshold and maximum-depth settings behave the same way; the stereo point cloud noise fixes cover them setting by setting.

Why heat shortens range

An active depth camera projects infrared light, watches for the return, and converts the delay or shift into depth. Three things in that chain drift with temperature:

• Emitter output falls as the laser junction heats. A typical VCSEL emitter drifts about 0.06 nm per kelvin in wavelength, so a 17 K warm-up moves the emission roughly 1 nm — enough to slip against the receiver's narrow bandpass filter and cut the signal the sensor sees.
• Timing references drift in time-of-flight designs, so depth noise grows before the range number visibly collapses.
• Past an internal limit the camera protects itself: it clamps emitter drive, exposure or frame rate. That protective clamp is the range loss you are seeing.

Range tracks light. For a target the camera resolves, returned signal falls with the square of distance, so to first order maximum range scales with the square root of emitted optical power:

Data sheets state an ambient window for the camera — many active depth cameras specify 0-35 °C — but the same table carries a condition: the housing's case-temperature limits must hold everywhere in that window. A camera in free air at 24 °C meets the rating; the same camera bolted inside a sealed housing can pass its case limit in a 22 °C room. For how published performance shifts with ambient, see depth camera performance vs operational temperature range.

Confirm it with a timed range test

Half an hour of setup buys a definitive answer. You need a K-type thermocouple meter, a matte white paper-faced target and the farthest working distance your cell truly needs.

1. Power the camera down for at least an hour at stable room temperature so it starts cold, and record the air temperature at the camera.
2. Place the white target square to the camera at that farthest working distance — 4.0 m in the example below.
3. Power on, then every 30 minutes for four hours record two numbers: the farthest distance where depth stays valid and confident, and the housing temperature at the same instant.
4. Plot both against time.
5. Repeat once with a small fan moving air across the housing as a control.

Read the plot against this rule: range loss that grows with housing temperature and recovers under the fan is thermal derating, confirmed. Loss that sits flat from boot regardless of temperature is settings, optics or ambient light — go back to the settings section. Treat a reversible swing of 10% or more across a shift as derating worth fixing, and anything under 3% as ordinary warm-up drift to design around.

Worked example: the ten-hour shift that took 4.0 m down to 3.5 m

Inputs, from a documented night-shift failure: the camera read its target at 4.0 m at cold start; after roughly ten hours of continuous running the farthest valid depth sat at 3.5 m; no settings had changed, and the range returned by the next cool-down.

The rule in words: retention is warm range divided by cold range, and because range scales with the square root of emitted power, the implied power drop is one minus retention squared. The margin rule converts the result into an orderable spec.

retention = 3.5 m / 4.0 m = 0.875, so 87.5% of the cold range survives the warm shift — a 12.5% loss.

implied power drop = 1 − 0.875² = 1 − 0.766 = 23.4%.

Cross-check the other direction: a 20% optical drop predicts 4.0 m × √0.8 = 4.0 m × 0.894 = 3.58 m, matching the observed 3.5 m within measurement scatter — the numbers are self-consistent.

Decision, with its threshold: a reversible 12.5% loss that tracks housing temperature confirms thermal derating, and at this depth the fix is mandatory rather than cosmetic — the application needs 4.0 m and the warm camera holds 3.5 m. Two ways out. Restore the warm range with the fixes below until the step 3 log holds 4.0 m at end of shift. Or re-specify: a 25% margin on a 4.0 m target demands 4.0 m × 1.25 = 5.0 m of cold range, and at the measured 87.5% retention that camera still holds 5.0 m × 0.875 = 4.4 m warm — the target covered with margin to spare.

Fixes that hold range through a full shift

Work in this order — airflow first, because it costs almost nothing and the fan-control run has already told you whether it works.

• Move air across the housing. A small continuous 40 mm axial fan over the camera housing usually recovers the whole loss on its own.
• Vent the enclosure high and low so convection can actually move air, and keep cable entries tight.
• Keep computing heat out of the camera housing. Mount the capture computer where its waste heat cannot wrap the sensor — a fanless industrial mini PC on the frame away from the camera beats a tower warming both.
• Cut duty where the task allows. A cell that inspects once per second can idle the emitter between frames; the trade-offs live in depth camera emitter on/off cycling. Never bypass the emitter drive limits to buy range back — they keep the projector inside its eye-safety class.
• Trim the preset to the true need. Running a maximum-range profile to inspect at half that distance burns emitter power all shift; select the shortest profile that covers the real working distance.

For continuous presence and volume checks at a fixed range, a dedicated ToF sensing camera such as the MV-DT01WMU is the straightforward fit — mount it where the fan's air actually crosses the housing.

Choosing a camera for round-the-clock depth

If the cell runs three shifts, qualify sensors warm, not at bench boot: the camera that matters is the one holding range at end of shift inside its housing. Projected structured light suits close-range detail work where the camera sits well inside its range envelope, and the MV-DPS200P-02 is that type. For logistics-style volume measurement, choose a speckle-pattern structured-light camera built for that duty. Browse the 3D sensors category or the camera finder and filter by the type that matches the task.

When the answer is not heat

The same symptom has other causes, and each has a shape that separates it from derating.

• Fouled window. Dust or condensation on a protective window absorbs infrared strongly; the loss grows over days, not hours, and never recovers on a cool-down. Clean the optics and retest cold and warm.
• Ambient infrared. Direct sun through a roof light or halogen process lighting floods the receiver and cuts range constantly from boot, worst while those lights are on.
• Dark targets. A matte black rubber or foam target can return half the signal of white paper; under the square-root rule that costs about 29% of range (√0.5 ≈ 0.71). Bring the target closer or add fill light rather than re-rating the camera.
• The near-field trap. If the complaint is garbage depth on close objects, that is the minimum-range floor, not derating — why close objects look like noise explains where it starts.

Prevention: qualify warm and spec with margin

Accept the camera at end-of-shift temperature, inside the final housing, not on a bare bench at boot.

• Run the timed range test as part of acceptance, with the enclosure closed and the cell warm.
• Hold the air at the camera inside the data-sheet ambient window — many active depth cameras quote 0-35 °C with case limits met — and measure the air at the camera rather than trusting the room thermostat.
• Carry the margin: farthest warm target × 1.25 sets the cold-range figure you order against.
• Favor steady temperature over swings. Thermal-cycle aging studies on camera modules track sharpness drift across thousands of cycles; a camera living at a steady warm temperature outlasts one cycling cold to hot every shift.

The 3D topic page collects the depth-camera guides on this site, including capture timing and set-up fixes, for the acceptance paperwork.

Checklist: before you blame the sensor

• Confirm the depth preset matches the scene and reach — a bright-scene profile reads as lost range from the first frame.
• Reset the viewer's display maximum distance to its default — it clips the picture, not the data.
• Run the four-hour cold-start log with housing temperature — growth plus recovery is heat, flat is not.
• Repeat the log under a fan — if airflow restores the range, the diagnosis is closed.
• Clean any protective window before the log — fouling mimics derating and never recovers on cooling.
• Order against cold range = farthest warm target × 1.25 — margin survives aging, dark targets and summer.

Build the camera side of the stack with the configurator, started at the camera step:

If the log points at heat, add the fan and the vents when you order — in-stock items ship within one business day, and every order includes free expert support for mount and cooling questions. If the numbers refuse to settle, send the range-versus-time log through the contact page and we will read it with you.