10.2 Frequently Asked Questions

The currently shipping XGRIDS scanners are the L3, L2 Pro, K2, and PortalCam. They differ in scan range, accuracy, imaging, storage, output type, and intended project scale, and no one of them leads on all of these.

L3 leads on published accuracy and imaging. XGRIDS states 5 mm post-processed relative accuracy at 1 sigma, measured in a dedicated 5,382 sq ft (500 m²) indoor test environment, and 2 cm RMSE absolute accuracy under open sky, with survey-grade RTK built in and a 1/1.3-inch camera sensor with 4K HDR. Both of its configurations reach 120 m (394 ft). It has 1 TB UFS storage and 100 minutes operating time. It mounts to a backpack, a vehicle, or a standard RTK survey pole, and a drone mount for the L3 is in development for the DJI Matrice 400. The L2 Pro's accuracy figures come from a different test; the L3 buying guide sets the two side by side.

L2 Pro leads on range and drone mounting. Its 32-channel / 300 m configuration reaches 300 m (984 ft), which no L3 configuration does, and it is the only device that mounts to a drone. It covers campus-scale environments (tens of thousands of square meters per battery), has a 1 TB internal SSD, and delivers 3 cm RMSE absolute accuracy with RTK. Runs on LixelStudio (point cloud) and LCC Studio (3DGS).

K2 is the current-generation handheld. 40 m working range at 10 percent reflectivity with a 100 m maximum, 200K points per second, 360° x 59° LiDAR field of view, 512 GB internal eMMC, built-in RTK with no external module needed, and real-time true-color point cloud rendering. Handheld only. Runs on LixelStudio for point clouds and on LCC Studio for 3DGS.

PortalCam is the 3DGS-focused device with the highest visual fidelity. 96 LiDAR channels, a four-camera array, 512 GB internal eMMC. Runs on LCC Studio only. It does not produce point clouds and does not output coordinate-registered data. Projects larger than one battery are stitched with control points placed during the scan, so a multi-segment PortalCam job needs planning before you start rather than more batteries in the bag. If your project requires measured deliverables, the PortalCam cannot fulfill that need.

For every device side by side, see 10.7 Device Specifications. For the L3 against the L2 Pro, see the L3 buying guide.

No, and the reason differs by device.

L3, L2 Pro, and K2: all three are rated IP54. That protects against light water splashes from any direction. It does not cover operation in active rain. Use them only after rain has stopped.

PortalCam: XGRIDS has not published an IP rating for this device. Treat it as unrated. Keep it out of rain, high humidity, and dusty environments entirely, rather than assuming it matches the other three. See 10.7 Device Specifications.

Beyond the hardware risk, scanning in rain degrades data quality on every device. Raindrops cause laser refraction and reflection that introduce complex noise into the scan and can corrupt SLAM tracking. Data from a rainy session is rarely recoverable to a usable standard. Any malfunction caused by water ingress is not covered under warranty.

L3, L2 Pro, and K2: -4°F to 122°F (-20°C to 50°C). PortalCam: -4°F to 113°F (-20°C to 45°C), in indirect sunlight only. Sustained temperatures above the rated maximum may affect reliability.

The PortalCam sunlight qualifier matters more than the air temperature on a clear day. Do not scan in direct sun for extended periods, and do not leave the device on a dashboard, on an exposed rooftop, or in a closed vehicle between segments. Surface heat soak exceeds the rated range long before ambient air does.

Cold weather reduces battery runtime by approximately 20% below 41°F (5°C). Add one extra battery to any cold-weather kit.

Runtime: The L3 has 100 minutes of total operating time, on a 14.4 V, 3,350 mAh quick-swap battery charged in the supplied charging hub. L2 Pro and K2 approximately 90 minutes per charge. PortalCam approximately 60 minutes. Cold weather reduces runtime by approximately 20%.

Runtime is not the same as the capture ceiling, even though the numbers match. The L2 Pro and K2 stop recording at 90 minutes and the PortalCam at 60, as firmware limits that apply whatever the charge state. Those are separate facts that coincide, and neither is a scan target. See 10.7 Device Specifications.

Lifespan: Rated 300 to 500 charge cycles before capacity drops to 70 to 80% of original. Batteries remain functional below that threshold but runtime is reduced.

Storage rules:

• Fully charge after each day of use.
• If fully charged but unused for an extended period, recharge within 3 months.
• If critically low or fully depleted, recharge within 3 days. Failure to do so causes irreversible damage not covered under warranty.
• Store in a cool, dry environment.

Do not start a session on a nearly flat battery. The PortalCam will not power on at all with a battery below 10%. If the charge falls below 10% during a recording it force-stops and saves, and below 5% it shuts itself down. The K2 warns at 10% and expects you to stop and save at that point; continuing past the warning is outside normal operation and the data is not guaranteed to save or process correctly.

No to both on the L2 Pro, K2, and PortalCam. None of the three can pause a scan or resume one from a breakpoint: they offer start and stop only. Stopping ends the recording and writes the project file, and the next scan is a new segment.

Pause does exist elsewhere in the ecosystem, which is where the confusion comes from:

• LCC Studio can pause a reconstruction. Click the pause button during processing; the progress ring greys out and the job shows as Paused. Resuming continues from the breakpoint rather than restarting.
• LixelStudio can pause the live point cloud transmission in the Remote Transmission view. This pauses the stream to your workstation, not the scan. Data skipped while transmission is paused is not saved in LixelStudio, so use Save before closing if you need the real-time cloud.

Field guidance also uses the word in its plain sense: pausing at a doorway, pausing at a control point target, pausing in a large space to run a slow figure-8. Those mean stop walking, not press a button. If you need to break a capture into pieces, that is Map Fusion, and Map Fusion needs overlap you must have already captured. See the next answer.

Yes, through Map Fusion. Scan in segments, swap batteries between segments, and stitch the segments together in post-processing. Plan the segment break points before scanning, because a segment boundary discovered mid-scan almost never has the overlap fusion needs.

Two different fusion systems exist and their limits are not the same. Use the one that matches your device and deliverable.

• LixelStudio Map Fusion (L3, L2 Pro, and K2, point cloud output): up to 10 maps, and each capture session must be under 20 minutes. The PortalCam does not run LixelStudio.
• LCC Studio Map Fusion (L3, L2 Pro, K2, and PortalCam, 3DGS output): up to 10 segments, with a cap of 200 minutes of scan data combined across the whole fusion. There is no per-segment time limit in LCC Studio, provided the job stays inside those two numbers.

Each segment is still bounded by the scanner itself. The L2 Pro and K2 stop recording at 90 minutes, and the PortalCam at 60, as firmware limits. On the PortalCam, XGRIDS additionally advises against captures over 30 minutes in any mode, for single-scene reconstruction as well as fusion segments, because longer captures tend to produce drift and layering artifacts in the result.

Segments must overlap by 50 to 100 ft (15 to 30 m) at their boundaries, in a feature-rich area rather than an open plaza, a long corridor, or a smooth tunnel. No segment may sit entirely inside another. All segments must come from the same device type, and in LixelStudio the L2 Pro 16-line and 32-line count as different types and cannot be mixed. On the K2, LixelStudio Map Fusion requires current K2 firmware.

None of these are scanning targets. Reconstruction memory demand is the real constraint on segment length. See the next answer.

Plan on 12 to 15 minutes per segment. That is the working target, and it leaves room for unexpected coverage without approaching any of the limits below.

The limit that actually binds is workstation memory, not the battery and not any software ceiling. In LCC Studio, 64 GB reconstructs 30 minutes of capture data reliably and risks failing past 45 minutes, and 128 GB reaches 60 minutes of capture data and risks failing past 90 minutes. LixelStudio publishes no memory-to-duration figure: it specifies 64 GB at both the Minimum and the Recommended configuration.

Longer scans also cost disproportionate time. Reconstruction runs at roughly 20 minutes of processing per minute of capture, so a 45-minute scene is most of a day of compute before you find out whether it succeeded. A failed reconstruction cannot be partially recovered.

LCC Studio itself imposes no time limit at all. There is no timer and no duration-based rejection, so the software will accept and attempt a scan far larger than the machine can finish. What it does do is estimate memory demand from the uploaded data and show you the result before the run. A red Estimated Memory reading means the scan is too large for that workstation, and that is the moment to split the capture, drop Reconstruction Quality to Standard, or add memory.

The 90-minute and 200-minute figures you may see elsewhere are not scan targets, and they come from different places. Ninety minutes is the device capture ceiling on both the L2 Pro and the K2, a firmware limit on one recording session that happens to match their battery runtime. Two hundred minutes is the combined cap on an LCC Studio Map Fusion job. Neither describes what a reconstruction can handle.

XGRIDS recommends against capturing a full-length dataset in one session on any hardware. A single fault inside a long capture, whether SLAM drift, environmental interference, or a battery problem, makes the whole dataset difficult to recover and may force a complete rescan. For anything larger than a comfortable single scene, split the capture and stitch it with Map Fusion. See 6.2 Scan Splitting for the full segmentation strategy and 10.7 Device Specifications for the full workstation requirements.

The current scanners carry large internal storage. The L3 has 1 TB UFS, the L2 Pro a 1 TB internal SSD, and the K2 and PortalCam 512 GB internal eMMC each. On the L2 Pro, K2, and PortalCam that storage is not field-swappable. Plan capacity before the day rather than managing it in the field.

Budget 60 to 80 GB per hour of scanning on the L2 Pro, K2, and PortalCam. For a full day, confirm the device has enough free space for the planned sessions before you leave, and transfer and clear completed work between visits if the schedule is tight.

If a long project approaches the storage limit on site, stop the scan cleanly, wait for solid green, and transfer the project to your workstation. Then start a fresh segment that overlaps the end of the previous one by at least 50 ft (15 m), and stitch the segments in processing with Map Fusion.

LixelGO (L3, L2 Pro, and K2): Snapdragon series processor, Snapdragon 8 or above recommended. 8 GB RAM minimum. A discrete display chip is required, which rules out a lot of otherwise capable mid-range phones. The phone must support Bluetooth, a wireless hotspot, and USB tethering. The L3 also runs a scan from its 4.3-inch built-in display with no phone connected.

LCC Scan (PortalCam): Android 12 or above with a processor equivalent to a Snapdragon 765G or a MediaTek Dimensity 8000-MAX or higher. On iOS, version 13 or above.

Both apps use a personal hotspot for Direct Connect. LixelGO also uses the hotspot to pull RTK and NTRIP corrections. Use simple hotspot names with no special characters or spaces to avoid connection failures. On iPhone, the device name under Settings, General, About becomes the hotspot name, so set it to plain English characters before the first connection. If the phone cannot create a hotspot, use Bridge Mode, where the phone and scanner both connect to the same existing WiFi network with internet access.

It depends entirely on your deliverable.

LixelStudio produces point clouds: georeferenced, measurable, compatible with Revit, AutoCAD, Navisworks, and GIS platforms. If your project requires dimensional verification, BIM modeling, or survey deliverables, you need LixelStudio.

LCC Studio produces 3D Gaussian Splat models: photorealistic, immersive, suited for client presentations, virtual tours, web sharing, and digital twins. You cannot take accurate measurements from a Gaussian Splat model. If your client needs a visual experience rather than measured data, you need LCC Studio. Current pricing is on the XGRIDS Store.

On L3, L2 Pro, and K2 projects, the same raw scan data feeds both pipelines. Plan for both before you scan: the pipelines cannot be retroactively applied from raw data that was not designed for them. The PortalCam produces 3DGS output only and does not connect to LixelStudio.

No. Both applications require an NVIDIA GPU with CUDA support. AMD and Intel graphics cards are unsupported with no workaround. NVIDIA is the only supported GPU brand for XGRIDS work.

This applies to graphics cards only. AMD CPUs are fully supported and appear in the recommended processing configurations; pairing an AMD Ryzen processor with an NVIDIA GPU is a standard XGRIDS workstation build.

Three constraints matter when specifying a machine. For LixelStudio the card must be NVIDIA RTX 20 series or above with CUDA 11.6 or later, running NVIDIA driver 580 or later. For LCC Studio the published floor is NVIDIA RTX 3060 12 GB or NVIDIA Quadro RTX A4000 16 GB, running NVIDIA driver 581.90 or later. Either way, older generations are excluded regardless of how much VRAM they carry. Neither application uses more than one GPU: multi-GPU parallel processing is not supported and the software selects a single card automatically, so a second card adds nothing to processing. And project data must sit on a local internal disk, because external drives, USB flash drives and network paths are a documented cause of post-processing crashes.

For short scans, yes. For production work, specify a desktop.

XGRIDS publishes no laptop-specific configuration, and the published guidance is to use a desktop for large datasets because laptops thermal-throttle under sustained load and can crash mid-reconstruction. A desktop runs at full TDP for the hours a large reconstruction takes; a laptop does not. On large scan projects, multi-segment fusion, or aerial-ground fusion, the difference is not speed but whether the job completes.

A laptop that meets the published GPU and memory minimums will handle short scans. Both applications are Windows only, and neither runs on macOS, Linux, Windows Server, or a virtual machine. See 10.7 Device Specifications for the full requirement tables.

Both consume significant RAM. Run one and then the other rather than both at once: no XGRIDS source gives a figure for running them together. Monitor RAM usage in Windows Task Manager during the first few minutes to confirm available headroom before committing to an overnight run.

A timeout error means the SLAM optimization exceeded the maximum allowed processing time for a segment. This typically occurs when the scan data has high complexity: large open spaces with repetitive features, very long corridors, or difficult lighting transitions that cause the solver to run many iterations without converging.

Solutions: break the scan into shorter segments, switch the processing Mode to Robust, add control points to anchor the trajectory in problematic areas, or review scanning technique to identify areas where coverage was insufficient for reliable SLAM tracking.

Geometry drift is a SLAM tracking failure. The processed model reflects the accumulated trajectory error that the solver could not correct. Work through these in order before concluding a rescan is needed.

• Switch the Mode dropdown to Robust and reprocess. In the LixelStudio Project Processing dialog, change the Mode from Standard to Robust and rerun processing on the same raw data. This resolves the majority of drift artifacts without returning to site. See 10.4 Error Messages for LIO Drift specifically.
• Check for insufficient loop closures. SLAM corrects accumulated drift when the scanner passes through previously mapped areas. Long linear corridors scanned end-to-end without returns produce progressive drift that Robust mode cannot fully correct. The underlying field technique needs to change. See Module 3.2 Route Planning.
• Check how the scan was stopped. If the operator did not return to the start point and stand still before pressing stop, the final loop closure never ran and everything the session accumulated stayed in the model. See 10.8 Accuracy Reference.
• Check scan speed. Above 3 ft/s (1 m/s) in repetitive environments such as corridors and storage aisles, the SLAM solver loses confidence and drift increases. If the LixelGO preview showed reduced tracking confidence in specific areas, those areas are the source of the distortion.
• Add GCPs in poor-structure areas. White-on-white rooms, mirror-covered surfaces, and long featureless corridors give SLAM very little to track. Ground control points in these areas anchor the coordinate frame and prevent drift from propagating forward from the problematic zone.

If Robust mode does not resolve the drift and field technique was correct, the segment requires a rescan. Processing cannot recover geometry that the scanner did not collect correctly.

Fusion misalignment has three documented causes. Work through them in order.

• Insufficient overlap. Segments must share at least 50 ft (15 m) of the same scanned geometry at their boundary. Less than 33 ft (10 m) causes fusion failure. An overlap placed in an open area, a long corridor, or a smooth tunnel can also fail even at full length, because there is nothing distinctive to match. Review the field app preview from each segment to confirm coverage in the overlap zone.
• Trajectory drift in one segment. If a segment has internal SLAM drift before fusion begins, the fusion algorithm aligns two distorted trajectories and the result is distorted. Reprocess the affected segment in Robust mode. If it does not resolve the internal drift, that segment data is not usable and must be recollected.
• Control point naming mismatch. Connection point names must match exactly between adjacent segments, including capitalization. A point entered as "CP1" in one segment and "cp1" in another will not connect. Check the control point log from each segment before processing. Naming errors are the most common cause of fusion failure on otherwise well-executed scans. The L3 does not require control points for Map Fusion: its sessions are matched by region selection.

Also confirm the segments are eligible to fuse at all. All segments must come from the same device type, and in LixelStudio an L2 Pro 16-line and an L2 Pro 32-line count as different types.

See Module 5.6 Troubleshooting Fusion Problems for the complete diagnostic process.

Four causes account for almost all of it.

• The capture was too short. Colorization needs a minimum duration with continuous movement: 3 minutes on the K2, 2 minutes on the L2 Pro. A short capture, or one where the operator stood still for much of it, produces poor colour regardless of lighting.
• Camera synchronization issue. In LixelStudio, verify that the correct video coloring option is selected and that the panoramic camera video files transferred completely from the device. A missing or corrupted video file causes incorrect timestamp mapping.
• Inconsistent lighting during the scan. Scanning through areas where lights switched on or off, or through large window areas where outdoor light shifted significantly, creates exposure changes that misalign with the LiDAR geometry.
• Walking speed too high. At speeds above 3 ft/s (1 m/s), camera frames and LiDAR points are captured at slightly different moments. The temporal offset creates color bleed at fast-moving transitions between surfaces.

K2 note: real-time true-color rendering happens during scanning rather than during post-processing, and only in RGB Mode. K2 color problems are typically a scan-time issue rather than a video-file or processing issue.

Accuracy depends on the device, positioning method, and scan execution quality. See 10.8 Accuracy Reference for the full breakdown, including how to verify and document what you actually achieved on a given job.

No, not from a standard reconstruction. The LCC Studio measurement tool provides approximate distances but these are not survey-grade and should not be used for construction, permitting, or dimensional verification.

The Gaussian Splat model is a visual representation. The underlying ellipsoids represent appearance, not precise geometric surface positions. For measured deliverables, process the same scan data through LixelStudio to obtain the point cloud, which is the appropriate source for dimensional measurement.

Ask the client what software they use before you scan. The format choice affects how you configure processing.

• Autodesk Revit (BIM modeling from point cloud): RCP from LixelStudio, or E57 via Recap.
• Navisworks or AutoCAD: RCP or E57.
• ArcGIS, QGIS, or GIS delivery: LAZ with georeferenced coordinates.
• 3DGS visualization or web sharing: LCC2 (smallest, XGRIDS ecosystem) or PLY (broadest compatibility).
• Unreal Engine or Unity: LCC via XGRIDS SDK, or PLY via community plugins.
• Cesium or WebGIS digital twin: 3D Tiles (requires absolute coordinates and model under 4 million Gaussian points).
• NVIDIA Omniverse or Isaac Sim: USD.

Several of these carry a minimum host application version, and 3D Tiles and USD in particular will not load in an older release. Confirm the client's version before committing to a format. See 10.6 Platform Compatibility for the full decision reference.

LiDAR cover: Use the included lint-free cloth and wipe gently in one direction. Small amounts of dust generally do not affect LiDAR performance significantly. A neutral cleaner is acceptable on the window: soapy water, distilled water, or high-purity ethanol.

If material has dried onto the LiDAR window, do not dry-wipe it. Rinse or spray the area with clean water first to lift the debris, then wipe gently with a damp lint-free cloth, rinse off any cleaner residue, and dry. Dry-wiping mud or grit across the window scratches it, and a scratched window degrades every scan afterwards.

Camera lenses: Camera cleaning is essential. Clean before and after each use. A dirty camera lens directly degrades both point cloud color quality and 3DGS visual quality. This applies to all three cameras on the L3, all onboard cameras on the L2 Pro and K2, and all four cameras on the PortalCam. Check for condensation as well as dirt: devices fog when moved between hot and cold environments, and a fogged lens looks clean.

Device body: Do not use alcohol on the aluminum body. It causes paint damage and discoloration. Wipe with a dry or lightly damp cloth only.

There are three routes, and the first two are firmware updates driven from your phone. The app raises a prompt on its own when an update is available: LixelGO for the L3, L2 Pro, and K2, LCC Scan for the PortalCam.

• Minor updates install inside the app. XGRIDS also calls these incremental package upgrades. Accept the prompt, wait for the LED to return to solid green, and confirm the new version in device settings. Nothing to download by hand.
• Major updates do not. Also called full package upgrades. Download the package and leave it compressed, put the scanner into USB mode, drag the package to the root directory of the device drive, exit USB mode, then power off and back on. The scanner installs it automatically on startup. Major packages are usually larger than 1 GB, but go by what the app says rather than by file size. On the K2, connecting the cable switches it to USB drive mode by itself; on the L2 Pro you enable and disable USB Mode in the LixelGO settings.
• LixelStudio can install firmware directly. Connect the scanner to your workstation by USB, click the Device button at the lower right, then Update, then select the firmware file. This route covers the full handheld scanner range and is the one to use when no phone is available or an in-app update will not complete. It requires a battery above 60 percent and takes about 30 minutes.

All firmware and software comes from xgrids.com. From a desktop browser, open Support, choose Download from the dropdown, then select your scanner from the list on the left.

Charge above 70 percent on the L2 Pro and K2, above 50 percent on the PortalCam before you start, and never interrupt an update in progress. The LED alternates red and green throughout. Do not power the device off and do not remove the battery during that state. An update that loses power partway through can render the device unresponsive and require manufacturer recovery.

Do not unzip the firmware package. It is copied to the scanner in its compressed state, exactly as downloaded. Extracting it first is the most common reason a correct download will not install.

Transfer your scan data off the device first. Projects captured under older firmware may stop appearing in the app after a major upgrade. This is the most common reason an operator believes an update destroyed data.

Firmware versions cannot be skipped. A scanner several releases behind must be brought forward one published release at a time, in the order XGRIDS publishes. 10.5 carries the L2 Pro chain. Check for updates before every major project, and keep your processing software matched to the device firmware, because an outdated LixelStudio release cannot process data from newer firmware and the error it raises looks like corrupt scan data rather than a version problem. See 10.5 Firmware Updates for the full step-by-step procedures and the version compatibility rules.

Format through LixelGO (L3, L2 Pro, and K2) or LCC Scan (PortalCam). Do not format using Windows or macOS disk utilities. Formatting via the app ensures the correct file system structure is maintained for the scanner firmware.

Contact Alpine Reality Capture first. As your authorized XGRIDS dealer we handle hardware issues, firmware problems, and processing errors directly, and we escalate to XGRIDS when a case needs the manufacturer. Reach us through alpinerealitycapture.com.

Have the device serial number ready when reporting a hardware issue, and record a short video of any intermittent fault. That single step resolves more cases on the first exchange than any other.