You ran the firmware update the camera software offered on first launch, the progress bar reached the end, and now the depth camera is gone: nothing under its name in Device Manager, an empty list in the vendor's enumeration utility, and the same result on a second computer. This is the most common way a 3D camera fails without actually failing, and the repair is documented and free. You will see what recovery mode is, why the fix works only on the computer where the update ran, how to carry the recovery update out step by step, and how to tell a stuck update from a genuinely dead sensor. It is written for engineers integrating 3D sensors — depth cameras, laser profilers, structured-light systems — on Windows hosts.
Why did my 3D camera disappear right after a firmware update?
The camera is almost certainly not dead: a failed or interrupted update leaves it sitting in recovery mode, a minimal bootloader state that enumerates over USB with different identifiers, so the operating system lists an unknown device where your camera used to be.
The update tool writes the new image, commands the device to reboot, and waits for it to come back as a camera. If the write completes but the reboot handshake stalls, or the new image fails its signature check, the device stays in the bootloader. That bootloader is a separate, minimal program with its own USB identity, so the operating system has nothing to match against the camera driver it installed earlier the same day.
The signature is consistent:
- the update progress reached the end, then a message reported that the device was slower than expected to reconnect;
- the camera is absent from the enumeration utility on every computer you try;
- Device Manager shows either nothing at all or a single generic entry, never the camera model;
- disconnecting and reconnecting changes nothing, because the bootloader state persists by design.
If the camera vanished with no update in the picture, work the ordinary link checks first: power budget basics for a depth camera with no image and USB link checks after a cable swap. Recovery mode is specifically the after-update case.
Is a camera stuck in recovery mode bricked for good?
No — recovery mode exists precisely so a failed update can be repaired in the field: the bootloader that the update tool talks to is deliberately kept intact, so rewriting the main firmware image brings the camera back with no hardware change.
The bootloader is written at the factory and the update process does not overwrite it. That is the design: whatever happens to the main image, a small known-good program remains that can accept a new one. Rewriting the main image is a field procedure with the vendor's update tool, not a factory repair, and a camera that comes back through it is identical to one that updated cleanly.
One consequence matters: power-cycling cannot clear the state. A bootloader with no valid main image has nothing else to boot, so the twentieth replug behaves exactly like the first. Stop replugging, leave the camera on one port, and move to the recovery procedure. The generic DFU sequence is written up in DFU recovery steps for a camera that looks bricked; the part most people miss — which host must run it — is the next section.
Why does the recovery fix work only on the computer that saw the failure?
Because the working installation is the only host with the recovery driver bound: when the camera re-enumerated in recovery mode, that PC cached the firmware-update class driver for it, while every new computer sees an unrecognised interface and never exposes the device to the update tool.
When the camera dropped into the bootloader, the first host enumerated the new USB identity and bound a firmware-update class driver to it, and that binding is cached in the host's driver store. The vendor's update tool therefore has a device it can claim on that machine. On a fresh computer the same identity meets a driver store with no matching binding: the interface stays unrecognised, nothing useful installs, and the update tool's device query returns an empty list — which reads exactly like a camera that is not detected.
This is why the instinctive next step, trying another PC, always makes things look worse: the second computer is the one host guaranteed not to see the device. Move the camera back to the original host before you conclude the hardware has died. Update-survivor driver bindings are a Windows-wide pattern rather than something specific to depth cameras; the same mechanics sit behind USB3 camera driver conflicts after an update.
Which USB port and cable should a recovery update use?
Use a USB 3 port on the host machine itself with the camera's original cable: a USB 3 port supplies up to 900 mA and a USB 2 port only 500 mA, so a power-hungry depth camera on an under-powered port drops off the bus mid-write. Keep passive USB 3 cable runs under about 3 m, the length the USB 3 specification allows a standard cable: longer runs train a link only marginally and fail exactly when the tool starts moving data.
Root port means a port soldered to the host's own board. Hubs and docks share and reshape the link: a hub's downstream ports do not each get the full 900 mA, and a camera that browns out during a write fails again for a fresh reason. Front-panel case headers are wiring rather than silicon, and they are the next most common culprit after hubs.
Link fallback is the quieter failure. A USB3 camera that cannot train a full-speed link drops when the tool starts moving data, which looks like a flaky update rather than a port problem. The same behaviour on embedded hosts is covered in why USB3 cameras fall back on some hosts.
How do you run the recovery update, step by step?
Run the vendor's firmware update tool on the original host, invoke it with its recovery flag, and point it at the signed firmware image you downloaded from the manufacturer, because that combination is what claims a device that no longer presents its normal identity.
- Return the camera to the original host. The recovery binding lives in that computer's driver store; no other machine can see the device until the main image is written back.
- Confirm the state. Device Manager shows an entry the camera never had before, typically a firmware-update or unknown device. Reading the USB IDs tells you which state the device sits in: see the USB ID check for a camera stuck in firmware update mode.
- Collect the right files. The signed firmware image for your exact model, from the manufacturer's download area, plus the update tool from the same SDK generation. A frequent reason recovery seems to fail is a tool from one generation pointed at an image from another — see firmware and SDK version mismatches that report no device.
- Connect through a root USB 3 port with the camera's own cable, nothing in between.
- Run the update tool with its recovery option and the image file as input. The exact command line is documented per camera family; for the 3D sensors we stock it is the DFU procedure in the DFU firmware update guide.
- Let the write finish. Stay plugged in until the tool reports completion; re-enumeration afterwards takes 5 to 15 seconds.
- Verify. The enumeration utility lists the camera under its normal name with the new firmware version, and the SDK opens the depth stream. If the image is back but streams are missing, the driver stack needs a rebuild: rebuild it after a firmware update.
What if the recovery update fails too?
Escalate in this order: one more recovery attempt on the original host with a different port and cable, a known-good camera swap to separate host from device, then an RMA, because a device that will not claim even under the recovery tool has a hardware fault.
Run that second attempt through a different root port with a different cable, because a brown-out during the first write can imitate every other failure. Check the pair again: tool from the SDK generation the release notes name, image built for your exact model. If the tool still cannot claim the device, run the swap test.
The swap test takes one minute. Put a healthy 3D camera on the same port of the same host: if it enumerates, the host is fine and the fault follows the suspect camera; if it does not, the port or the driver store needs attention and the camera may be innocent. Keep a known-good sensor on the bench for exactly this job.
- Type
- Depth / ToF camera
- Type
- Stereo / structured light
- Type
- Stereo / structured light
A camera that will not claim even under the recovery tool, with model and tool generation confirmed, has a flash or controller fault that field procedure cannot reach. Send it to the manufacturer's service channel with the timeline — versions, host, error text — and line up the replacement from the 3D sensors range.
Laser profilers and stereo 3D cameras
170 models · from $699.00How do you keep this from happening on the next update?
Treat firmware updates as a maintenance window, not a first-run prompt: update only when you can spare the camera for 10 to 15 minutes, always on the host that will run it, and only with the update tool and SDK version that the release notes pair together.
- Commission on the production host. Accept first-run firmware prompts during set-up only, never mid-run: a camera that misbehaves on reboot can be out of service for the better part of an hour.
- Pair tool and image before you start. Read the release note that names which SDK generation and firmware version belong together, and record both against the camera.
- Archive the signed image per version. A copy in the project folder turns a future recovery from a search into a five-minute job.
- Update one camera at a time, through a root port, and stay at the bench until the tool reports completion.
- Bookmark the recovery pages before you need them. Getting a board into bootloader mode and the DFU update done safely are the two that matter on a line-down afternoon.
Nothing here needs new hardware: the camera that vanished after an update is on your bench, in recovery, waiting for the host it met first. Run the recovery write there, confirm the depth stream opens, and file the signed image where you will find it next time. If the swap test has already sentenced the camera, start with the camera finder, walk the system configurator, or contact the applications team to have a recovery procedure checked for your exact model. The 3D topic page collects the rest of the depth-camera fixes on this site.
Questions
Does a recovery update wipe the camera's calibration?
No. Recovery and normal updates rewrite the main firmware image only; calibration and user parameters live in separate flash regions and survive on supported sensors. If measurements drift after a version jump, run a quick trueness check against a known part and repeat the camera's correction routine if the numbers moved — not a recalibration from scratch.
How do you know the update has failed rather than just being slow?
On a healthy link the device enumerates within seconds of a replug — call it 5 to 15 seconds — and a depth-camera image write finishes in one to three minutes. Call the update failed when the progress bar has not moved for two minutes, or when the tool reports completion and then errors on reconnection. One full minute of patience after a completion message still beats an unplug mid-write.
Will unplugging and replugging, or a full power cycle, clear recovery mode?
No. Recovery state is designed to persist: it survives power cycles, replugs and weeks unpowered without degrading. Replugging only re-runs the same enumeration each time. Give one retry on a different root port for a clean link, then go straight to the recovery write — it is the only thing that changes the device's state.
Which firmware image do you download for the recovery?
The signed image for your exact model, from the manufacturer's own download area, matching the SDK generation of the update tool. Record the version against the camera. Skip beta-marked images on production hardware and never take an image file attached to a forum thread — a mismatched or unsigned image fails in ways that look like hardware faults.
How is recovery mode different from an ordinary driver clash?
Timing and identity. A driver clash shows the camera under its own name in Device Manager while the SDK fails to open it, usually with no update involved. Recovery mode appears immediately after an update, shows a generic or firmware-update entry instead of the model, and looks identical on every computer, including ones that never had the driver installed.


