FAQ

Getting started with 3D

What depth sensing technology does Zivid use?

Zivid uses Temporal Structured Light, which integrates a sequence of spatially encoded patterns to achieve high accuracy depth even on structure-less surfaces.

What is structured-light 3D, and how is it different from time-of-flight and stereo?

Zivid cameras use Temporal Structured Light: the projector casts a sequence of encoded light patterns, and the camera reconstructs an accurate 3D point cloud from how they deform on the surfaces. This gives high accuracy and dense data even on structure-less surfaces, where stereo (which relies on visible texture to match between two views) struggles. Time-of-flight (ToF) measures the travel time of light per pixel. Structured light typically achieves much higher accuracy and finer detail, which is what most robotics and inspection tasks need.

What is 3D machine vision?

3D machine vision captures the shape and position of objects as a point cloud, where every pixel has a real X, Y, and Z coordinate in millimeters. This lets a robot or inspection system reason about geometry, such as where an object is, how it is oriented, and its true size. Zivid provides this with high-accuracy Temporal Structured Light cameras and an SDK for production integration.

What is the difference between 2D and 3D machine vision?

A 2D camera produces a flat image of color and brightness, but no depth, so it cannot directly measure distance, shape, or the orientation of an object. A 3D camera produces a point cloud where every point has a real X, Y, and Z position, so it measures geometry and not just appearance. 2D suits tasks such as barcode reading, presence checks, and surface inspection, while 3D enables bin picking, robot guidance, assembly, and dimensional measurement. Zivid cameras also provide a calibrated 2D color image aligned to the 3D point cloud, so you can use both together.

How does an industrial 3D camera differ from a consumer depth sensor?

Consumer depth sensors are designed for low cost and human-scale sensing, with relatively low accuracy and limited robustness to ambient light and difficult materials. Industrial 3D cameras like Zivid are built for repeatable sub-millimeter accuracy, robustness in factory conditions, calibrated and traceable measurements, and an SDK intended for production integration. For applications such as bin picking, inspection, and assembly, that accuracy and repeatability is what makes a reliable robot cell possible.

Cameras, field of view, and accuracy

What is the Field of View of the Zivid cameras?

Zivid 3

XL250

Distance [mm]	FOV [mm]
1300	938 x 938
2500	1750 x 1750
5000	3537 x 3537

Zivid 2+

MR130/M130

Distance [mm]	FOV [mm]
800	481 x 406
1300	793 x 658
2000	1220 x 1012

LR110/L110

Distance [mm]	FOV [mm]
700	694 x 547
1100	1096 x 858
1700	1692 x 1324

MR60/M60

Distance [mm]	FOV [mm]
300	234 x 236
600	579 x 469
1100	1096 x 858

Zivid 2

M70

Distance [mm]	FOV [mm]
500	514 x 320
700	754 x 449
1500	1519 x 959

L100

Distance [mm]	FOV [mm]
600	644 x 372
1000	1147 x 680
1600	1618 x 988

To calculate the FOV for different distances, click here.

What accuracy, trueness, and precision can the camera achieve?

Accuracy is described as a combination of point precision (the repeatability of individual XYZ measurements) and dimension trueness (how correct measured distances are). The table below gives the typical point precision and dimension trueness error at each model’s focus distance.

Model	Focus distance	Point precision	Dimension trueness error
Zivid 3 XL250	2500 mm	265 µm	0.24 %
Zivid 2+ MR130 / M130	1300 mm	208 µm	0.33 %
Zivid 2+ LR110 / L110	1100 mm	255 µm	0.33 %
Zivid 2+ MR60 / M60	600 mm	71 µm	0.16 %
Zivid 2 M70	700 mm	85 µm	0.17 %
Zivid 2 L100	1000 mm	244 µm	0.18 %

See Zivid Specs Terminology for the definitions, and the datasheet for your model for the full numbers across the working range. To reach and keep the best accuracy, run Infield Correction, allow the camera to warm up, and keep Thermal Stabilization enabled.

What kind of sensor does Zivid use?

Camera	Sensor
Zivid 3	8.07 MP CMOS
Zivid 2+	5.10 MP CMOS
Zivid 2	2.86 MP CMOS

The sensor resolution equals the number of points in the point cloud; see Point Cloud Structure and Output Formats for the resolution per model and how to read it from the SDK. For how sensor resolution is defined, see Zivid Specs Terminology.

How much does the Zivid camera weigh?

Camera	Weight [Kg]
Zivid 3	2.700
Zivid 2+	0.920
Zivid 2	0.945

What material is the Zivid camera casing made of?

The Zivid cameras are constructed of both aluminum and magnesium parts to create a lightweight and robust casing.

How do I choose an industrial 3D camera for my application?

Match the camera to your scene: the size of the objects, your working (standoff) distance, and the field of view you need to cover at that distance. Then check that the resulting spatial resolution is fine enough to capture your smallest important feature, and that the accuracy meets your tolerance. For Zivid specifically, the number in a model name is its focus distance in centimeters (for example, the Zivid 2+ MR130 is focused at 130 cm). Use the field-of-view and imaging-distance calculator to match a model to your scene.

What accuracy can an industrial 3D camera achieve, and how much do I need?

Zivid cameras reach sub-millimeter point precision and dimension trueness on the order of 0.2 % at a given imaging distance. For example, a Zivid 2+ MR130 at 1000 mm measures a 100 mm distance to within about ±0.2 mm. The exact numbers depend on the model and the imaging distance. The accuracy you need depends on your task: bin picking, robot guidance, and assembly typically needs sub-millimeter, while picking larger parts or boxes can tolerate more. See Zivid Specs Terminology for how precision and trueness are defined, and the datasheet for your camera model for exact numbers at your working distance.

Capture quality and difficult materials

Which 3D camera works for shiny, transparent, or dark parts?

All Zivid models handle these difficult materials well, so the choice is driven by your object size and working distance (see the camera table above), not by the material. For highly specular metal parts and bins, the Sage Engine gives the best results and is available on the Zivid 2+ MR130, MR60, and LR110. The Omni Engine, for transparent and reflective scenes, is available on all Zivid 2+ models. Zivid 3 is especially robust to ambient light and reflections thanks to its significantly stronger projector.

How well does the camera capture shiny, reflective, or specular metal parts?

Shiny and specular metal is challenging for any 3D camera, but Zivid is designed to produce good data on these surfaces. Use the Manufacturing Specular or Semi-Specular presets, which provide the high dynamic range needed for specular surfaces and which are good for strong inter-reflections. A dark, absorptive background reduces inter-reflections on cylindrical or spherical parts. See Dealing with Highlights and Shiny Objects and Optical Properties of Materials.

Can the camera capture transparent or translucent objects?

Transparent and translucent objects are challenging because light passes through them, but they can be captured with the right setup. Use the Omni Engine (Zivid 2+) with the Consumer Goods presets. Keep the object close to the focus distance with a uniform gap to the background. See Dealing with Transparent Objects.

How well does the camera capture dark, black, or matte surfaces?

Dark and matte surfaces reflect little light and are challenging, but Zivid is designed to produce usable data on them. Use Specular-family presets with longer exposure times with HDR. See Dealing with Highlights and Shiny Objects and Optical Properties of Materials.

Which presets and settings should I use?

Start from the preset closest to your application (the Manufacturing, Consumer Goods, and Inspection families), then fine-tune the acquisitions, aperture, and exposure. See Capturing High Quality Point Clouds for a full guide.

Can Zivid cameras be used outdoors or under direct sunlight?

The upper limit that we use when testing is approximately 1500-3000 lux. We measure by positioning the LUX meter across different areas of the camera’s FOV at the imaging distance, and take the maximum value as the ambient light level. Direct sunlight (32,000-100,000+ lux) is not feasible for most cases; physical shading of the scene is required. Under some outdoor conditions, where ambient light stays within these limits, the Zivid 3 can be used outdoors thanks to its significantly stronger projector.

Performance and speed

What capture and cycle time can the camera achieve?

With the right settings and good hardware, capture times below 100 ms are achievable, but only for certain applications and distances. Other, more challenging applications at long distances may require captures of 1 s or longer. Capture time depends on the number of acquisitions, exposure, and processing, and on the host hardware (the network card and the GPU). Reduce time with fewer acquisitions, subsampling, and a region of interest. Use the Calculate 3D Capture Speed article for representative benchmarked numbers, and see Capture Speed for measured capture-speed benchmarks.

Can the camera capture moving objects?

Zivid cameras can capture slowly moving objects, but feasibility depends on the speed, the field of view, and the required quality. Motion requires a shorter capture, which you get with a fast, single acquisition and 4x4 subsampling, as used in the Parcels Fast settings. See Calculate 3D Capture Speed for representative numbers, and contact customersuccess@zivid.com to discuss feasibility for your application.

Can I increase the frame rate while sacrificing on quality and/or resolution?

By applying sub-/downsampling, see Sampling (3D), it is possible to decrease the acquisition and capture time by reducing the resolution. Despite the reduced resolution, it is noteworthy that the quality of the point cloud may actually be enhanced.

How much bandwidth does the camera at maximum occupy?

Camera	Bandwidth [Gbps]
Zivid 3	10.0
Zivid 2+	7.07
Zivid 2	5.37

Check out Calculate 3D Capture Speed for more representative numbers.

What computer and GPU are required to run the camera?

A dedicated GPU is strongly recommended; integrated graphics works but is significantly slower. See GPU Requirements for requirements and Recommended Industrial PCs for tested PCs.

Which industrial PC and network card do you recommend?

For best performance, use a dedicated GPU together with a 10 GigE network card. Some industrial PCs ship with 2.5 GigE networking, so confirm a 10 GigE add-in card option with the supplier when you need full bandwidth. See Recommended Industrial PCs for tested industrial PCs, GPUs, and 10 GigE network cards.

Software, API, and integration

What APIs does Zivid Support?

• C++

• C#/.NET

• Python

• GenICam/HALCON (Experimental)

• ROS

• NVIDIA Isaac Sim extension (Experimental, for simulation)

Does Zivid support GenICam?

Zivid Software includes a GenICam GenTL producer (Experimental)

Using GenICam with Zivid cameras is tested in following HALCON versions:

• 19.05 Progress, 20.05 Progress, 21.11 Progress, 24.05 Progress, 24.11 Progress-Steady, 25.05 Progress

For other GenICam GenTL consumers, we provide a technical preview. This may allow use of a Zivid camera in other software packages that enable interfacing GenICam GenTL compliant cameras.

What Operating Systems are supported by Zivid?

Windows: 10 / 11

Linux: Ubuntu 20.04 / 22.04 / 24.04

Note

Support for older operating system versions has been removed over time; see the SDK Changelog for which versions a given SDK release supports.

Will the point cloud data differ when using Zivid Studio and the API?

No, for the same settings the data is identical between Zivid Studio and the API. The reason captures often look different is the starting point, not the data. Zivid Studio guides you to pick a tuned preset, while a minimal API capture uses the default settings. To reproduce a Studio capture exactly, export the settings as a YML file from Zivid Studio and load that file through the SDK. This is the recommended way to keep Studio and your application in sync.

How can I process the point cloud data from Zivid?

Zivid does not provide point cloud processing software. HALCON and EyeVision are examples of paid processing software while OpenCV, Open3D, and PCL are free alternatives to use depending on the target application.

In what format can files be saved?

The point cloud files can be saved as .zdf, which is the format that Zivid Studio uses. You can also export the file as a PLY file (*.ply as ordered or unordered) / ASCII points file (*xyz) / Point cloud data file (*.pcd). See Point Cloud Structure and Output Formats for the point cloud structure and outputs.

Does Zivid work with ROS, my robot, and HALCON?

Zivid is robot-agnostic and integrates through standard interfaces. The SDK is available for C++, C#/.NET, and Python, with an official ROS / ROS2 wrapper (zivid-ros) and an Experimental GenICam GenTL producer tested with HALCON. The camera outputs point clouds in its coordinate frame; relating that to a specific robot or PLC is done through hand-eye calibration and your integration code.

Does Zivid include picking, detection, or inspection software?

Zivid provides the camera and the SDK, which deliver a calibrated, true-color point cloud plus tools such as region of interest, normals, downsampling, and hand-eye calibration. Zivid also provides motion planning software. Object detection, segmentation, pose estimation, and CAD comparison are done with your own software or third-party packages. See How can I process the point cloud data from Zivid? for examples of compatible software.

SDK development

How do I get the camera intrinsics?

The SDK exposes camera intrinsics in both OpenCV and HALCON models, and you can read them from a connected camera, from a frame, or from a saved file. Each camera exposes two sets: fixed intrinsics for the camera and estimated intrinsics that account for the settings used in a capture. See Camera Intrinsics for how to obtain them.

How do I set capture settings in code, and how do I find the valid range for a parameter?

Settings such as exposure time, aperture, and brightness are set through the SDK just as in Zivid Studio. The valid range for each can differ between camera models. A value outside the valid range is rejected when the camera captures, so query the camera’s settings information to read the supported ranges for your model rather than hard-coding them. The ExploreSettingsMetaData sample shows how to read the valid range of each setting. The simplest way to reproduce a known-good capture is to export the settings as a YML file from Zivid Studio and load it through the SDK.

Can I update the camera firmware from code?

Yes. Firmware can be updated programmatically instead of through Zivid Studio; see the FirmwareUpdater sample in C++, Python, and C#.

Which ROS version does Zivid support?

Zivid provides an official wrapper, zivid-ros, for both ROS and ROS 2. Match the wrapper release to your SDK version, as noted in the wrapper’s documentation.

Does Zivid run on ARM platforms such as NVIDIA Jetson?

The SDK supports Ubuntu on ARM, including NVIDIA Jetson, and can run inside a Docker container. See Install Zivid on Jetson Linux for Jetson setup and Docker for running in a container.

Connectivity and hardware

What kind of Ethernet cable may I use for Zivid cameras?

The Zivid cameras come with a 5 m CAT6 Ethernet cable. Zivid also provides Ethernet cables in 5 m, 10 m, and 25 m options. See Data Cables for more info. Zivid strongly recommends that the cables are carefully checked before use or if run time errors occur. It is also strongly recommended to tighten the screw on the camera side when setting up the camera.

What power supply does the camera need?

Zivid cameras are powered from 24 V DC. We strongly recommend using the Zivid Power Supply and Power Extension Cables to ensure compliance with Safety and EMC emission and immunity standards. If you must use a third-party supply, a 24 V industrial supply, or robot internal wiring, see that page for the requirements the power must meet.

Can I connect to a camera by its IP address in the SDK?

Yes. You can connect by serial number, or directly by IP address or hostname using a CameraAddress. See the connect-by-IP section of the Capture Tutorial.

Mounting, calibration, and multiple cameras

What kind of calibration is performed by Zivid?

Zivid cameras come precalibrated from the factory and are ready to use. Our floating calibration takes temperature and iris (aperture) into account. The point cloud provided by the camera is calibrated, which means that X, Y and Z are correct according to real scenery. There is a pixel to pixel correspondence for the X, Y and Z matrices and the color image.

Zivid also offers Infield Correction tool designed to verify and correct the dimension trueness of Zivid cameras in the field.

Hand-eye calibration is used to calibrate the camera to a robot. It refers to finding the transformation between the coordinate systems. Zivid has an official tool for helping customers with the Hand-Eye Calibration calibration.

What is in-field correction, and which values are acceptable?

In-field correction verifies and corrects the camera’s dimension trueness in the field. A corrected dimension trueness within roughly 0.1 % is expected, but it depends on the camera model. It is useful if you suspect accuracy has drifted, for example after a robot collision. It also helps tighten accuracy to the best achievable for your field of view and imaging range. See Infield Correction for an overview and Guidelines for Performing Infield Correction for board distances, repeated measurements, and how to judge the values; if results stay far outside the expected range, contact customersuccess@zivid.com.

Which calibration board should I use?

The correct board depends on your camera model and working distance, and also on your robot mounting and application. See Infield Correction for the board-to-model recommendations (for example ZVDA-CB01 vs ZVDA-CB02), and Zivid Calibration Object for the calibration objects used in hand-eye.

Can the camera be mounted on a robot arm as well as stationary?

Yes. The camera can be mounted on a robot arm (on-arm / eye-in-hand), stationary (eye-to-hand), or on a tripod. On-arm mounting lets a single camera cover several positions and keeps it at the optimal working distance. Stationary mounting is simpler and keeps the camera out of the robot motion path. See Mounting and Mechanical Considerations for Robot Mounting.

Can multiple Zivid cameras be used together?

Yes. Multiple cameras can be combined into one coordinate system with Multi-Camera Calibration, and their captures can be combined with Stitching. Because the cameras project light, capturing simultaneously can add noise where the projections overlap, so sequential capture may be needed. See Multiple Zivid Cameras and Performance Considerations for Multiple Zivid Cameras.

Applications

Is Zivid suitable for inspection and CAD comparison?

Zivid delivers an accurate, calibrated point cloud that you can compare against a nominal CAD model or use for dimensional and surface inspection. The camera and SDK provide the 3D data; the comparison and measurement logic itself is done in your own or third-party software. See How can I process the point cloud data from Zivid? for software options.

Is Zivid used for robotic welding?

Yes, Zivid cameras are used in robotic welding applications, both before and after welding. Before welding, the camera locates the parts and the area to be welded, so the robot can find the seams and edges to follow. After welding, the same 3D data can be used to inspect the weld. The camera provides the 3D data; the weld-path planning and robot control are handled by your welding software. Capturing weld edges and features well relies on the same settings as other reflective-metal scenes, see Dealing with Highlights and Shiny Objects.

Can a 3D camera guide a robot to pick and place parts?

Yes. A 3D camera gives the robot the shape and position of objects so it can compute where and how to grasp them. This is the basis of 3D-guided robotics such as bin picking, machine tending, piece picking in logistics, parcel induction, and assembly. The camera is aligned to the robot through hand-eye calibration, which relates the camera’s coordinate system to the robot’s so a detected point becomes a robot motion.

What is 3D bin picking, and what camera do I need for it?

Bin picking is using a 3D camera and a robot to locate and pick parts out of a bin or container, including randomly placed or mixed parts. It needs a camera with enough field of view to see the whole bin, and enough resolution and accuracy to localize the parts and plan a safe grasp. Good performance on the part material (for example shiny or dark surfaces) also matters. See the Bin Picking & Machine Tending Tutorial for camera selection, field-of-view and clearance considerations, and a full walkthrough.

CAD files and FOV models

Where can I find CAD/STEP files and FOV models?

CAD/STEP files for the cameras (with field-of-view) and mounts are available on the Zivid website.

Maintenance and safety

What temperatures can I safely run Zivid in?

Zivid cameras are designed to work within 0˚C and 45˚C.

We strongly recommend operating the camera within its specified temperature range.

Caution

Using the camera outside its temperature range can damage the camera and void your warranty. Use the camera out of spec at own risk.

Tip

Zivid factory calibration takes temperature into account. However, to further reduce the impact of temperature dependent performance factors, keep Thermal Stabilization enabled.

What are the camera’s environmental ratings (IP rating, temperature)?

All Zivid cameras are rated IP65 for ingress protection and are designed to operate within 0-45 ˚C. Refer to the datasheet for your camera model for the full environmental specifications. For harsh environments outside these ratings, see the protective housing options.

Can the camera image through a protective glass window?

Yes, the camera can image through a window when it is mounted inside a protective housing. Use a flat, optically clear window and keep it clean and free of condensation; an anti-reflection coating reduces glare and stray reflections. See Using Zivid Inside Protective Housing With Glass Window for guidance on housing and window choice.

Can I clean dust off the camera’s fan with compressed air?

Yes, but do not aim compressed air at the fan in a way that spins it faster than its normal running speed. The fan has no built-in protection against being spun too fast, and spinning it too fast with compressed air can damage it. See Service and Maintenance for general cleaning guidance.

What is the expected service life of Zivid camera?

Zivid cameras are designed to capture 25 million point clouds in its lifetime.

This is not to be mistaken for expected Mean Time To Failure (MTTF/MTBF).

Is the structured light that Zivid uses in the visual spectrum?

Yes, the Zivid camera uses visible light for vivid colors.

Is the structured light that Zivid uses harmful to people?

As with any bright light source, do not stare into the beam. The light source should not pose a hazard to the eyes, due to the aversion response (head turn or eye blink for example) to very bright light sources. However, to be on the safe side, one should avoid looking into the camera/projector during operation.

For detailed information, check Regulatory Information.

What light class certificate do Zivid 3D cameras have?

Zivid 3D cameras use a light source which is classified as Risk Group 2.

See the label at the back of the camera. For detailed information, check Regulatory Information.

What Safety Standards does Zivid follow?

To read about Safety before using the Zivid camera, check Regulatory Information.

How can I access the log files to forward to Zivid Support?

You can find the logs at %LOCALAPPDATA%\Zivid\API\Log on Windows and $XDG_CACHE_HOME/Zivid/API/Log on Ubuntu.

Caution

If $XDG_CACHE_HOME is not set use $HOME/.cache.