A phone that can scan in 3D changes what a site visit produces. Instead of a notebook of dimensions and a handful of photos, an engineer can walk out of a Zone 1 or 2 area carrying a measurable point cloud of the equipment they just inspected, which is only worth having if the depth data behind it is accurate.
However, in hazardous areas there is always something between the sensor and the equipment - because the phone sits inside a protective enclosure there is a window over the camera array. So we ran the obvious test: the same surface scanned twice with the same phone, once bare and once inside the enclosure, then compared point by point.
Scanning through the window of an Xshielder explosion-proof cover does not measurably reduce iPhone LiDAR accuracy. Average deviation from the reference surface was 0.72 mm bare and 0.75 mm through the window, a difference of under 0.04 mm. Standard deviation was 0.57 either way, and both error distributions had the same single peak with no second cluster of bad returns. The gap is smaller than the variation you get from repeat scans with a bare phone.
Now, onto the full blog...
The LiDAR scanner on an iPhone Pro model emits a grid of invisible near-infrared laser pulses and times how long each takes to return. Those returns are fused with the wide camera image and accumulated as you move, building a point cloud that can hold millions of points.
On a live process site, that capability gets used where a tape measure is slow, awkward or unsafe: as-built documentation of pipe racks and skids, clearance checks before a retrofit, capturing the geometry of equipment due for replacement, and taking dimensions off items above head height or behind a barrier. Scanning also gives the phone a depth reference that helps the camera focus quickly in the dim, high-contrast conditions typical of tank interiors and enclosed plant, which is one of several reasons optical performance matters on an intrinsically safe camera.
Accuracy also carries weight for reasons that have nothing to do with the software. Every entry into a Zone 1 or Zone 2 area has a cost attached: a permit to work, gas testing, an escort, sometimes a slot in a shutdown window that closes in three days. A scan nobody trusts has to be captured again, and in hazardous areas that is an expensive thing to repeat.
Depth error rarely announces itself. A point cloud with a systematic bias looks like any other: it renders cleanly and the measurements come back to three decimal places. It then carries into whatever it feeds, whether that is a fabrication drawing or a digital twin of the site.
The consequences appear later. A spool piece is fabricated to a dimension that was slightly wrong and does not fit on the day it is installed. A clash check clears a route that is too tight for the insulation. Documentation captured for an insurer gets challenged, and a number with a questionable source is hard to defend. Each of those ends in rework, which usually means another permit and another entry.
It is a fair assumption that a window which is optically clear to the eye lets the scanner see straight through it. However, near-infrared light does not behave identically to visible light, and transmission at the wavelengths a LiDAR module uses is a property of the specific material and any coatings on it.
So, in principle, several things can go wrong. The window can scatter part of the outgoing pulse, which broadens the returning signal and makes each measurement less certain. It can reflect light back toward the receiver, adding a false population of very short returns. Refraction through the window can offset ranges slightly. Each of these produces a noisier point cloud rather than an obviously broken one, which is why it is worth measuring. For these reasons, we decided to test the LiDAR accuracy on Xshielder enclosures.
It’s first worth mentioning that no LiDAR scanner is absolutely perfect. The returned points never land exactly on the surface. Each sits slightly in front of or behind it, by a fraction of a millimeter, more or less at random, and that scatter is the sensor's normal noise floor. So, our question was narrow: does scanning through the window make the scatter worse?
To find out, we scanned the same surface twice with the same phone under comparable conditions, once with the phone bare and once with the phone inside the enclosure. Nothing else changed between the two captures.
The software measured how far every individual point strayed from the reference surface, producing millions of small distances that are then summarized with two numbers.
This answers how far a typical point strays from the surface. Without the enclosure, the average deviation was 0.72 mm. With the enclosure, it was 0.75 mm. The gap between them is under 0.04 mm. For scale, a human hair is roughly 0.07 mm thick, so the difference between scanning bare and scanning through the window is negligible - around half the thickness of a hair.
An average alone can hide a problem. If most points were near perfect but a subset landed several millimeters off, the average could still look acceptable while the cloud contained a group of badly wrong measurements. The standard deviation catches that, because it describes how consistently the points behave: a small number means they all behave alike, a large one means some are going rogue.
Without the enclosure: 0.57. With the enclosure: 0.57. Identical. The window did not create a hidden population of bad measurements.
We also plotted the deviations as a histogram, which counts how many points strayed by 0.1 mm, how many by 0.5 mm, how many by 1 mm, and so on. Both charts show the same profile: a large pile at almost zero deviation, tapering away to nothing by about 3 mm.
What we were looking for was a second bump further to the right. A second peak would mean a second source of error layered on top of the sensor's own noise, which is exactly how window scatter or an internal reflection would present itself. Neither chart contains one.
Put together, the two scans are equivalent within the sensor's own noise. Repeat captures of the same surface with a bare phone vary by a similar margin, so a 0.03 to 0.04 mm shift in average deviation is not a meaningful difference in accuracy. Scan quality in the field is governed by the scanner and by technique, not by the window in front of it.
Ordinary discipline matters more here than the enclosure does. Keep the window clean, since dust, oil film, glove smears and condensation all sit directly in the optical path. Hold a steady pace and a sensible standoff distance, and expect highly reflective and very dark surfaces to trouble any time-of-flight sensor. Those constraints exist with or without a protective cover.
On scope, the test quantifies point-level noise on a controlled surface, which is the specific mechanism by which a window would degrade a scan. Validating your own workflow at the ranges and on the materials you work with is still worth doing.
Data sheets in this category concentrate on certification, ingress protection and drop performance, all of which matter. Optical performance is described far less often, and usually in qualitative terms. It is worth remembering what an Ex certificate covers: ignition protection. It says nothing about whether the phone's cameras and depth sensors still perform once the phone is inside the housing.
If you are comparing enclosures for a team that scans, ask a few direct questions. Was the same device tested with and without the enclosure? What were the values for average deviation and standard deviation? Is the full error distribution available, or only a headline figure? Was the LiDAR scanner tested, or only the main camera? A supplier who says the window is optically clear has described the material rather than measured the outcome. The same applies when evaluating intrinsically safe phones and rugged handhelds with depth sensors.
Scanning the same surface twice with the same phone produced average deviations of 0.72 mm bare and 0.75 mm through the enclosure, with an identical standard deviation of 0.57 in both cases and no secondary error peak in either distribution. The scan taken through the window is as precise as the scan taken without it. As far as the LiDAR scanner is concerned, the window is invisible, so aniPhone Pro model can go into a Zone 1 area without downgrading what the team brings back.
LiDAR is a Pro-only feature, so if 3D capture is part of your inspection workflow, the model matters as much as the enclosure. Our explosion-proof covers for the iPhone 17 Pro Max and the iPhone 16 Pro Max keep the scanner, cameras and screen usable in Zone 1 and Zone 2 areas.
Get in touch if you would like to see the full test data before you standardize on a device.
In our testing it does not. The same surface scanned bare and through the enclosure gave average deviations of 0.72 mm and 0.75 mm with an identical standard deviation, a difference smaller than the width of a human hair.
LiDAR has only ever shipped on Pro and Pro Max iPhones, from the iPhone 12 Pro through to the iPhone 17 Pro and 17 Pro Max. Standard, Plus, Air, mini and SE models do not have it.
Yes, but only if the phone carries suitable Ex protection for that zone, which for a standard iPhone means an explosion-proof enclosure rated for Zone 1, used in line with the site's permit and gas testing procedures. If you are unsure how your areas are classified, start with what the ATEX zones mean for phones.
Point-level noise sits well under a millimeter on cooperative surfaces, though accuracy across a whole scan also depends on range, surface reflectivity, lighting and operator technique, so validate the workflow against a known dimension.
Some purpose-built intrinsically safe handsets include depth sensing, but capability varies widely by model, so confirm the sensor and the scanning apps you rely on before specifying a device.