A manufacturer-sliced Benchy shows that one selected machine, file, material and environment produced an attractive object. It does not prove that the printer is accurate, fast, quiet, safe or reliable for a buyer's work. LAYERTIMES therefore uses an open, repeatable and application-specific protocol rather than collapsing every observation into one score.

This method is for editorial reviews. It draws on the measurement logic of the NIST additive-manufacturing test artifact, ISO/ASTM 52900 terminology and the geometric test-artifact approach of ISO/ASTM 52902, but it does not claim standards certification. Conformity requires the full standard and a competent laboratory working within its scope.

Define the question first

There is no universally best printer. A school, dental laboratory, prop shop and mechanical workshop need different systems. Before testing, define the target profile: typical part size, material, accuracy, volume, supervision, environment and permissible post-processing. Only then can a measurement be classified as an advantage or a limitation.

Every claim needs a pre-defined question. Speed can mean short elapsed time, high volumetric flow or many accepted parts per shift. Accuracy can mean mean dimensional error, variation or whether a critical hole fits. Reliability is not one long successful print; it is repeated work with documented intervention and scrap.

Sample provenance and conflicts

The article states whether the printer was bought, loaned or retained, and discloses payment, travel, affiliate links and pre-publication access. A supplier may correct factual errors but may not select the tests, suppress a negative result or rewrite the conclusion before publication.

Production status matters. Beta firmware, a pre-production hotend, a hand-selected unit or unreleased profile changes validity. Record every deviation with version and date. If a supplier replaces a component or setting during review, both the original failure and the result after correction remain part of the evidence.

Freeze the environment

Record room temperature and humidity, printer location, support surface, ventilation, power and network. For filament, document manufacturer, lot, colour, diameter, drying and storage. For resin, add batch, temperature, mixing, vat film, washing and post-cure. The goal is not to impose one environment but to let readers judge the effect of differences.

List instruments with identifier, resolution and calibration or check date. A low-cost calliper can support workshop comparisons, but cannot make a micrometre claim credible. Measurement uncertainty must not disappear behind display digits. When surface texture dominates, contact point and measuring force matter as much as nominal resolution.

Versions and baseline

Save serial number, hardware revision, firmware, slicer, network plug-in and profile identifier. Cloud service and mobile-app versions also matter. The evidence package includes the original model, project, exported toolpath or resin job, screenshots and settings so a later update cannot silently rewrite history.

Test factory condition first. Install necessary safety updates but avoid internet tuning modifications. Manufacturer fast mode and an editorial normalised profile are separate results. One shows the buyer's immediate experience; the other explores hardware under comparable quality constraints.

Setup and usability

Measure active labour and waiting from sealed box to the first completed user-sliced model. Separate unpacking, assembly, calibration, account registration, update, material loading and transfer. A supplied demo file does not finish setup; the reviewer must slice, start and remove an independent model.

Usability is task performance, not icon aesthetics. Can the operator change material, cancel a failed job, export logs, work locally without internet and reach the nozzle, extruder, bed or vat? Documentation errors are recorded with the failed step and screen state.

Safety and data

An editorial review is not certification, but checks visible protections: enclosure, interlocks, hot and moving access, wiring, emergency stop, thermal runaway messaging and resin or powder handling guidance. Dangerous faults are not induced without controls. For connected machines, document account requirements, cloud dependency, camera, remote start, telemetry choices and real local-mode behaviour.

AI cameras create a separate privacy question. Presence of a feature does not prove local processing. Review consent, settings, deletion and the public privacy description. A preview or telemetry stream containing confidential geometry can be a procurement risk independent of print quality.

Geometric artifacts

Use separate artifacts rather than one all-purpose torture model. A dimensional artifact covers outer and inner dimensions, holes, pins, walls, steps and planes. A detail artifact includes positive and negative text, gaps and projections. A process artifact covers overhangs, bridges, islands, tall slender features and changing cross-section. Separation makes causes easier to identify.

Tests must not be limited to a machine-friendly small object. Add a broad flat part and a tall part using meaningful build volume. On large beds, repeat a coupon at multiple positions to expose flatness, heat distribution and optical variation. Centre-to-corner comparison is especially important for resin systems.

Accuracy, distortion and repeatability

Check exports and slicer compensation before treating nominal CAD as truth. Measure pre-defined points on cooled, conditioned parts, repeat readings and print multiple specimens. Report mean and range or standard deviation rather than the best number. Treat axes separately because XY and Z errors can have different causes.

Accuracy and repeatability differ. A machine consistently producing a hole 0.2 mm large may be compensable; another may average correctly while varying widely. A fit ladder adds functional evidence by showing which clearance slides, binds or snaps. Results are tied to material and orientation.

Surface and detail

Photograph surfaces under fixed light, distance and magnification. Assess top, side and downward faces, seams, corners, ringing, layers and support marks separately. Include a scale and uncropped overview because a macro image amplifies defects invisible in normal use.

Use pre-defined acceptance for detail: readable letters, open gaps and separated lines. Excess exposure or flow can look smooth and strong while closing holes. Appearance, geometric fidelity and mechanical integrity remain separate judgments rather than one subjective star score.

Speed and flow

Report slicer estimate, printer-reported time and wall-clock cycle. The cycle runs from heat-up or preparation until safe part removal. Maximum toolhead speed is not throughput; acceleration, volumetric flow, cooling, minimum layer time and geometry determine usable motion.

For FDM, a stepped flow test identifies the range where mass, surface and bonding stay acceptable, not a momentary peak before failure. For resin, measure exposure, separation, return and settling across different cross-sections because advertised millimetres per hour depend heavily on the model.

Normalised comparison

Compare machines with the same task and acceptance threshold: model, orientation, layer, walls, infill, material, temperature and quality. A larger nozzle or thicker layer can be a valid application race, but must not be labelled equal-quality speed.

Factory profiles remain valuable and are reported separately. A normalised profile answers another question. Both tables include compromises and failures. A machine that wins once but requires repeated restarts can lose shift-level productivity.

Noise and vibration

Measure sound with the same instrument, distance, height and room, reporting background and settings. Capture idle motion, typical printing, fast mode, fan peak and material change. One dBA value does not describe character, so note tonal motors, resonance and repeating clicks.

An uncalibrated phone app is indicative, not laboratory evidence. Bench and wall distance alter resonance and should be photographed. Vibration effects are examined on tall parts and rapid reversals; input shaping cannot substitute for sound mechanics.

Energy and resources

Measure at the outlet across standby, heating, printing, cooling, drying and mandatory post-processing. Peak power matters for electrical capacity; kilowatt-hours matter for cost. Resin wash and cure or powder heat cycles cannot be omitted without understating the process.

Material includes supports, purge, prime tower, loading waste, failed parts and unrecoverable resin or powder. For colour systems, report waste in grams and relative to accepted-part mass. Energy or material alone is not a life-cycle assessment; manufacture, life and disposal are also required.

Failure handling

In safe reversible scenarios, test runout, sensor response, power loss, network interruption and door opening. Do not create fire or damage risk. Record detection, alert, stop and recovery separately. A phone notification is not the same as automatic prevention.

AI failure detection requires repeatable harmless fault scenes. Report true detections, false alarms, misses and latency. One dramatic spaghetti detection does not establish general accuracy. The feature assists supervision; it does not justify unattended operation in an unsafe environment.

Endurance and maintenance

An endurance run is a representative workload, not one hundred hours of easy ornaments. Mix sizes, materials and cycles. Record starts, accepted parts, interventions, cleaning, bed preparation, calibration, consumables and downtime. Reprint the reference artifact at the end to reveal drift.

Maintenance labour is part of the product. Time nozzle, extruder, film or filter service and note tools and lost calibration. Record spare price and availability on the test date without assuming permanence. A sealed module may be quick to exchange but expensive and unrepairable.

Repetition and uncertainty

Repeat key measurements at least three times when practical. Do not create false certainty from small samples: disclose n, mean, minimum, maximum and relevant spread. A one-off run is a case study. Missing data are not zero and cannot be filled with supplier claims.

Uncertainty can come from the instrument, operator, texture, temperature, material batch and location. A workshop cannot quantify every contribution, but can list them. The defensible claim describes what three parts did in this system, not universal capability.

Raw data and reproducibility

Publish models or permitted links, profiles, times, measurements, failures and changes in a table or download. Link each photo to its specimen and show measuring points. Respect model licences; do not republish third-party files without permission. Reproducibility means enough detail to explain differences, not necessarily identical numbers.

A review is versioned. When firmware fixes an important fault, repeat the affected test and add a dated update without erasing the original result. Readers then see launch capability, improvement and early-adopter risk instead of a silently rewritten history.

Decision profile instead of one score

A total score can hide a slow but precise quiet machine or a fast one demanding supervision. LAYERTIMES reports evidence by setup, quality, geometry, speed, noise, energy, failure handling, maintenance and ecosystem. The conclusion names suitable users and disqualifying constraints rather than a universal winner.

Verified facts are documented measurements and public specifications. Supplier claims are values not reproduced under matching conditions. Editorial inference is the reasoned application judgment built from both. Keeping these layers separate prevents marketing from masquerading as measurement.

Minimum evidence for a full review

A full LAYERTIMES review requires known hardware and firmware, documented environment and material, at least three geometric tasks, time and repeatability evidence, failure-handling and maintenance experience, and conflict disclosure. A shorter session can be labelled first impressions, but cannot receive a definitive test title or buying verdict.

The method is not intended to make testing endless. It requires evidence proportional to the strength of each claim and enough context for readers to judge transfer to their work. A transparent measurement with limits is more useful than a many-decimal ranking that nobody can reproduce.

FDM material coverage

A PLA baseline shows the easiest operating point, not full material capability. If PETG, TPU, ABS, ASA or fibre-filled compatibility is advertised, test a relevant profile and document enclosure, nozzle, drying, bed and filtration requirements. A polymer name does not establish functional suitability; the exact material data and part requirement do.

Flexible filament tests feed path and retraction stability. Engineering thermoplastics test chamber temperature and warping. Abrasive composites introduce nozzle wear and clogging. The review never forces temperatures outside specification. Brand-locked material support is disclosed under ecosystem and operating cost.

Resin-specific protocol

For MSLA or DLP, pixel size is not actual XY detail. Measure positive and negative features, holes, gaps and texture. Start from a supplier profile, calibrate exposure and record every change. Inspect film, resin temperature, mixing and optical condition because these influence release and image transfer.

The workflow includes drainage, removal, washing, drying, support removal and post-cure. Dimensions are compared at a defined post-cure state. Record gloves, solvent, wipes, resin loss, liquid-waste handling and active labour. A short layer cycle is misleading when post-processing is slow or scrap is high.

Multi-colour and multi-material systems

Do not judge colour change from one two-colour logo. Measure transition time, purge mass, tower or waste demand, jams, moisture protection and spool compatibility. Use one model with frequent short changes and another with long regions because efficiency can differ greatly.

For multi-material printing, test adhesion and differing temperature needs. Soluble support counts as an advantage only when the pair prints reliably, dissolution time is documented and residue is handled safely. Advertised colour count is not the same as the number of materials reliably combined in one job.

Software and format fidelity

Import STL, 3MF and a CAD-adjacent format where supported. Check units, object structure, colour, painted support and profile portability. The same 3MF extension can be a neutral geometry package or a vendor project; silent loss of settings is a serious workflow risk.

Inspect toolpath previews for unsupported starts, thin-wall loss, collision, unexpected colour changes and excessive flow. Preserve the original and repaired mesh after automatic correction. Successful slicing does not prove geometric fidelity, so check a critical dimension after export.

Network and offline operation

Test local mode without internet: transfer, start, camera, logs, firmware behaviour and an already activated machine. Document lost features, periodic login and service outage behaviour. A physical USB port or SD slot does not prove complete cloud independence.

Do not perform unauthorised penetration testing. Document user-visible ports, encryption, updates and permissions, and handle vulnerability claims through responsible disclosure. Remote start has different risk in a supervised workshop and beside combustible materials in a home.

Service and support

Use a real support question. Record channel, response time, language, diagnostic quality and unnecessary data requests. An instant automated email is not resolution time. Evaluate knowledge-base search, diagrams, error codes, parts lists and warranty transparency.

Regional availability changes product value. Price is dated and includes tax, freight and required accessories. A launch discount is not a durable value comparison. Local service, stock and consumer protection can outweigh a small hardware advantage.

Stop conditions

Stop for electrical danger, smoke, uncontrolled heat, resin or powder release, repeated collision or injury risk. The interruption is evidence, but safety and supplier notification come first. A dramatic failure video never justifies continued hazardous operation.

Disqualifiers can include a reproducible safety fault, persistent failure inside specification, unusable mandatory cloud service or unavailable critical spares. One defective sample does not define an entire production run, but quality control and support response are part of ownership. Report replacement-unit results separately, not over the original.

How buyers should read the review

Start with personal exclusions: material, size, noise, offline mode, safety and service. Then examine normalised data and factory experience. Task similarity matters more than the highest score. A resin printer excellent for miniatures is not automatically suitable for heat-resistant engineering components.

Date and version belong to every conclusion. Profiles can improve and cloud functions can regress within months. Before purchase, check current release notes, price, regional warranty and recent faults. A review is decision evidence, not a timeless guarantee.

Production sampling

A build-volume test includes both one large object and multiple identical parts. This reveals first-layer, thermal and dimensional variation by position and tests batch handling. Yield is accepted parts divided by all starts, with every failed position shown rather than averaged away.

Unexpected states are logged even without scrap: recalibration, filament unload, resin refill, camera reconnection or profile warnings. Small interventions accumulate into labour at workshop scale. Alongside machine time, report operator minutes and interruption count.

Thermal behaviour

Record warm-up from a stable starting point and inspect whether reported bed or chamber temperature matches an independent check at relevant locations. A surface sensor and embedded thermistor do not measure exactly the same thing, so the comparison is interpreted with method limits. The goal is spatial and temporal stability, not accusing one display from a single reading.

For enclosed FDM, note chamber temperature across a long build, door or lid rules and heat exposure of electronics and filament. A nominally enclosed shell is not an actively controlled chamber. For resin, low liquid temperature can alter viscosity and release; the review records starting and operating conditions rather than silently warming the bottle.

Quality at throughput

After finding an attractive quality point and a high-speed point, run a representative batch at each. This shows whether a small benchmark advantage survives repeated starts, heat accumulation and material changes. Report accepted parts per elapsed hour, not merely shortest successful specimen time.

A quality threshold is defined before the batch: dimensional limits, visible defects, support damage and required post-processing. A faster mode does not win by lowering the threshold after inspection. If a customer could reasonably rework the part, the labour is included rather than hiding the defect.

Evidence retention

Preserve raw logs, photographs, measurement sheets and file hashes after publication. Do not discard a failed part before documentation; retain a reference success under controlled conditions when useful for later comparison. Remove personal or commercial data while keeping technical detail needed for reproduction.

A second editor compares article numbers with source tables and checks units, captions, profile versions and the scope of conclusions. If independent internal review was unavailable, disclose that limitation. Publication speed cannot override traceability.

Updating the protocol

Version the method when a new machine category, measurement error or reader reproduction reveals a gap. A new requirement does not retroactively rewrite an old review; it identifies which results used the earlier procedure. The change log states reason and expected effect, allowing improvement without pretending that tests years apart are perfectly comparable.

Reader reproductions are evidence only when their conditions are documented. A contrary result can expose batch variation, environmental sensitivity or an error in the original article. The editorial response is to compare files and conditions, correct proven mistakes openly and add uncertainty—not to select whichever result supports the preferred verdict.

The active public protocol version therefore appears in every review header and its downloadable measurement appendix, keeping later audits tied to the procedure actually used.

Measuring micro-precision resin systems

Optical resolution in micrometres, layer thickness, stage positioning accuracy and finished-part tolerance must be treated as separate quantities. A 25 µm projected pixel does not demonstrate a 25 µm hole or general dimensional accuracy. Photopolymer spread, release, washing, support, post-cure and the measuring system all shape the final geometry.

The test artefact should include positive and negative features: pins and holes, walls and gaps, raised and recessed text, vertical and horizontal channels, and repeated elements at the centre and edges of the build area. Instead of reporting the smallest visible detail, state which size was reproduced consistently, remained open or fell inside a predeclared tolerance.

A calliper is useful only for quick checks of larger external dimensions. Micro-features require suitably resolved optical measurement, a measuring microscope or, where justified, CT with known calibration and uncertainty. Rounded edges, translucent material and compliant thin walls can create false precision, so the measurement location and evaluation rule are fixed before printing.

At least three repetitions and several tray positions distinguish a one-off success from a stable process. Washing, drying and post-cure time, temperature and orientation are recorded, and parts are measured in the same defined state. For microfluidic or mating applications, a functional test — flow, leakage or fit — may matter more than one nominal dimension.

Acceptance of a micro-precision printer therefore cannot rest on the smallest catalogue number. Accepted-part yield, spatial variation, open negative features, surface condition, post-processing and operator time together describe usable capability. Supplier data provide a starting point; the editorial conclusion must rest on the documented measurement chain.