No depth stream? Your depth camera is in passive IR mode

Your own capture code shows a clean RGB image, an infrared feed so dark it looks dead, and no depth frames at all — while the manufacturer's viewer, on the same camera and the same PC, shows all three streams perfectly. The camera is almost certainly fine: the stream example you started from sets the depth mode to passive IR, and passive IR computes no depth by design. This guide gives you the diagnosis in order of probability, the exposure and coverage arithmetic for the mode you switch to, and the checks that separate a mode setting from a real hardware failure.

The symptom pattern and the one-line cause

The pattern is distinctive. RGB streams clearly because the color camera runs on its own sensor. The IR feed is not absent — it is nearly black, a very dark grey with faint scene structure. Depth is not degraded or noisy; there are simply no depth frames to display.

One setting explains all three observations: the depth mode. The vendor's viewer opens the camera in an active IR mode, so it shows depth. Example stream code ships with passive IR selected, so your application gets exactly what passive IR gives: a single wide-baseline infrared image lit only by ambient IR, and no depth image at all. Zero frames is the mode working as intended, not an error state.

Why passive IR computes no depth

Active IR modes drive the camera's infrared projector. The sensor reads the projected pattern returning from the scene and computes one depth value per pixel from it — encoded-light triangulation in structured-light cameras, phase or direct time-of-flight returns in ToF cameras. No projection, no depth.

Passive IR leaves the projector off and uses the depth sensor as a plain infrared camera on a wide baseline. There is no pattern to decode, so the camera returns one infrared image and nothing else. That is why the symptom is total: depth does not degrade in passive IR, it is simply not computed.

If you are weighing active IR depth against stereo depth for a station, the trade-offs run wider than one mode setting — the active infrared depth versus stereo comparison walks them.

Why the IR feed is nearly black

The projector is also the IR camera's main light source. In passive IR the sensor sees only what the scene reflects from ambient infrared — room lighting, warm equipment, daylight through a window — so an exposure tuned for the projected pattern lands several times short. In an active mode the IR feed is bright, and you can see the projected texture in it; that texture is the depth signal.

You can recover a usable passive IR image, but not depth:

At 30 fps the frame period is 1/30 s = 33 ms, and exposure cannot exceed it. An exposure near 10 ms — comfortable with the projector running — is about 3x short in passive IR; raise it toward 33 ms or switch auto-exposure on. An added IR flood light brightens the passive image further. None of this produces depth: depth needs the pattern, not brightness.

Test and fix, in order of probability

1. Read back the depth mode in your application. Passive IR is the answer in most cases. Set it to NFOV unbinned — the mode the viewer runs — and restart the stream.
2. Confirm your code subscribes to the depth stream, not only to IR and color. A capture graph that never opens depth cannot display depth; if frames stop somewhere in the chain, find where depth frames stop shows the chain to check on any OS.
3. Check the pixel format on the depth stream. An unhandled format can drop frames before display — unknown pixel format diagnosis covers the readback.
4. Check which device your API opened. On many depth cameras color and depth appear as separate device nodes, and a wrapper can hand you the color node only; RGB depth stream interface selection covers the choice.
5. If depth arrives but looks shifted against color, that is registration, not a missing stream: depth to RGB registration offsets in pixels is that fix.

Choosing the active mode: the coverage arithmetic

NFOV unbinned is the right starting point: full depth resolution over a narrow field, roughly 0.5–3.9 m of working range in this camera class. Binned narrow modes trade pixel count for light and reach, and wide-field modes trade resolution for span — roughly 0.25–5.5 m and a far wider view. Size the field before you commit:

Take a 75° horizontal field at a 1.5 m working distance: covered width = 2 × 1.5 × tan(37.5°) = 2 × 1.5 × 0.767 = 2.3 m. A 640-pixel-wide depth image over that span gives 2300 / 640 = 3.6 mm per depth sample. A 120° wide field at the same distance covers 2 × 1.5 × tan(60°) = 2 × 1.5 × 1.732 = 5.2 m — 2.3x the span at the same pixel count, so about 8 mm per sample unless the wide mode carries more pixels.

Cameras that cannot fall into passive mode

If a station keeps tripping over mode settings, a camera whose depth always depends on its own projected light removes that failure path. A structured-light stereo camera such as the MV-DPS200P-02, a laser galvanometer stereo camera like the MV-DLS300P-04, and the ToF MV-DT01WMU all carry their own illumination, so depth never silently stops behind a passive setting:

Values as printed in each product's data sheet; open a part for its drawing and full specifications.

For depth that reads distance on every pixel regardless of scene light, time-of-flight is the direct route:

Browse the full 3D sensors category or the camera finder to narrow by type; the configurator walks camera, lens and light in one pass.

When the mode is right and depth is still empty

• Working distance outside the mode. Narrow unbinned modes in this class start near 0.5 m; a target at 0.3 m returns nothing. Step back to 0.8 m and retest.
• Ambient IR saturation. Direct sun on the target floods the projected pattern and punches holes in depth. Shade the target — too much ambient light is the enemy here, not too little.
• Two projectors in one room. Cross-talk between cameras corrupts both patterns; trigger them alternately or mask one.
• Bandwidth. Depth plus IR plus color over one link can exceed it, and streams drop silently — camera interface bandwidth, cables and distance sizes the link.
• Stereo rather than ToF. On a stereo depth camera, RGB working with no depth has a different cause — stereo camera works for RGB but no depth covers the missing-mono-sensor case.

Checklist before you call it hardware failure

Before a warranty claim or a replacement order, tick these:

• Depth mode read back as an active mode (NFOV unbinned), not passive IR.
• Depth stream subscribed, with a pixel format your pipeline handles.
• IR feed bright and showing projected texture — proof the projector runs.
• Target inside the mode's working range and out of direct sun.
• One projector active at a time.
• The vendor's viewer shows depth on the same PC and port — proving the chain end to end.

The decision, case by case

• Viewer shows depth, your application does not: your depth mode is passive IR. Set NFOV unbinned and restart the stream. This is the case almost every time.
• IR feed dim but depth present: exposure. Raise it toward the 33 ms frame-period ceiling at 30 fps, or enable auto exposure.
• Depth empty in every application, viewer included: subscription, pixel format, working range or saturation — work the list above before suspecting the camera.
• Coverage matters more than detail: move from a narrow mode to a wide-field active mode; at 1.5 m that buys 5.2 m of span instead of 2.3 m.

Still stuck on a camera you bought from us? Send the depth-mode readback and one IR frame through the contact page and we will work it with you.