Specifications and Test Data

Almost no part of this document was guessed. The numbers below were measured on the bench rigs and platforms in our atelier, recorded in logbooks that travel with each piece, and aggregated across roughly four hundred completed commissions. Where a number is averaged we say so. Where a number is from a single test we say so. Where we do not yet have a number we say that too.

We publish this because the field is small enough that the absence of public technical references slows it down. It also makes it easier for buyers to compare what we do against what other studios are starting to do, on something other than marketing language. If a competitor publishes a 600-degree melt point textile spec and we publish 280, the buyer can take that into their decision. We would rather lose a commission to better data than win one to softer language.

A short note on the origin of these numbers. The sensor-permeability rig was assembled in early 2024 from a pair of off-the-shelf NIR sources, a calibrated photodiode, and a fixed twelve-millimeter sample window. The thermal data was logged using embedded type-K thermocouples taped to the housing inside the cloth during programmed walking trials of the platforms named below. The articulation envelopes are taken from the published platform spec sheets, cross-referenced with our own measurements taken with a digital protractor against the chassis at rest. The construction hours are timed by the couturiers themselves and stored against the commission record.

The threshold we work to is below four percent attenuation through a single layer in the wavelengths each platform’s onboard cameras and LIDAR rely on. Most of the platforms we work with concentrate their sensors at peaks around 850 nanometers and 940 nanometers. The cloth has to clear both. We measure both and reject anything that fails either.

All measurements below are taken on a single layer of base cloth, with no facing or canvas. Construction adds reinforcement at zones where it does not interfere with the platform’s sensor cone. The sensor zones themselves are always single-layer.

A second test sweep is run at 940 nanometers for compatibility with depth cameras using structured-IR. Numbers are slightly higher across the board but the rank order does not change. Polyester satin is again the worst performer because of dye chemistry, not weave structure. Cotton and worsted wool clear at acceptable margins. The atelier uses worsted wool and the linen-silk twill as the default base cloths for any piece that crosses a sensor zone.

A small reproducibility note. The same fabric run can vary by half a percent in attenuation between bolts, depending on finish residue and humidity at the time of measurement. We re-measure every new bolt before cutting. Pieces destined for the same client across multiple commissions are cut from the same bolt where possible to keep behavior consistent.

The figures below are the upper end of what we have observed under a forty-minute continuous walking trial at the platform’s rated cadence. The thermocouple is placed inside the cloth on the housing surface, taped at the geometric centre of the contact zone. Three trials per platform, the published number is the highest peak observed across the three.

Atlas runs hottest. This was not a surprise. The platform is built for explosive movement and the actuator duty cycle reflects it. Our rule for Atlas pieces is no synthetic in any contact zone, and the contact face of the lining is always brushed cotton or felted wool. NEO runs coolest, which corresponds to the platform’s domestic-use design and lighter actuators. Pieces for NEO can use cashmere blends comfortably where the same blend would mark on Atlas.

The melt-point threshold for any candidate textile in our archive is 240 degrees Celsius. That is a wide margin against the 71 degree peak, but the margin is necessary because heat does not behave predictably in a small enclosed cavity between cloth and housing. We have seen localized hot spots near the elbow gimbal of Optimus reach 78 degrees on a third trial that the bulk-measurement missed.

The numbers in the table below are the operating ranges we cut against. Each is the published platform spec cross-checked with a measurement at the chassis with a digital protractor in our atelier. Differences between published spec and observed range are typically under three degrees on stationary platforms, but Atlas in motion routinely exceeds its published numbers by more than that during dynamic recovery, so we add a five-degree safety margin on Atlas seam allowances at every joint that the published spec touches.

A standard jacket sleeve cap clearance for a human torso allows around 165 degrees of shoulder pitch with comfort. Optimus exceeds that by five degrees and Atlas by thirty-five. Standard sleeve patterns will bind on either chassis. Our cut for Atlas uses a deeper armhole, a forward-shifted shoulder seam, and bias-cut panels at the shoulder cap. The same garment cut for a human shoulder would not bind under any everyday range of motion. The chassis is the constraint.

Hours are tracked by the couturier on a paper card that travels with the piece. The figures below are means across all completed commissions in each category, with the standard deviation shown in parentheses. The biggest variance is at the bespoke level where each piece is genuinely a one-off and the hours range from one hundred and forty to over six hundred.

A common question from new clients is why a Hospitality Noir piece, which photographs as a clean uniform jacket, takes nearly a hundred and fifty hours. The honest answer is that almost half of those hours are inside the piece. Internal canvas building, hand-padded chest piece, hand-set lining at the cuff to keep the cloth off the wrist gimbal, and the chassis fitting cycles. The visible exterior takes maybe sixty hours of cutting, sewing, pressing, and finishing. The remaining eighty are structural work no one ever sees.

Every piece runs through five validation gates before it leaves the atelier. We log every fail. The data below is from the trailing twelve months. The overall pass rate is high because most pieces that fail an early gate are revised and re-tested rather than scrapped, but the per-gate first-attempt pass rate tells the more honest story about where we still struggle.

The gate we fail most is the visual pass. The atelier director walks the finished piece on the chassis at rest and under articulation and decides whether the silhouette reads the way the commission intended. There is no rubric for this. Pieces that pass every quantitative test fail it routinely. They go back to the table for adjustments to the line, the canvas weight, the lapel break, the cuff finish. The director’s opinion is the last gate because it is the hardest to formalize and the most often correct.

The change-pass failure rate is also instructive. We routinely build pieces that look perfect at rest and then take an operator twelve minutes to dress instead of the four-minute target for hospitality clients. That sends the closure system back to the bench. Magnetic plackets and quick-release shoulder seams are the two most-revised systems in our archive.

Compiled and verified from atelier logbooks by the senior couturier and the atelier director. Sensor-rig bench measurements supervised by the textile technician. Numbers reviewed quarterly. Last revision: April 2026.