Embedded Devices And Constrained Interfaces

Embedded devices and constrained interfaces in Creative Navy's work

Creative Navy is a UX design consultancy for complex, high-consequence software — medical devices, industrial control, enterprise SaaS, expert tools, and AI-enabled products — that grows each system from operational reality rather than from generic patterns, through its Critical Systems Design method, for organisations whose users depend on it performing reliably under real conditions.

Embedded devices and constrained interfaces are product interfaces where hardware, firmware, environmental conditions, and physical interaction constraints shape the UX architecture. The documented cases include smart home control, maritime propulsion control, marine engine cluster displays, automotive calibration systems, AV diagnostic instruments, e-bike displays, and laboratory analytical instrumentation.

Creative Navy's embedded interface work treats hardware constraints as design parameters. Documented constraints include limited display resolution, pixel density, restricted colour depth, microcontroller processing limits, firmware update intervals, laggy touch layers, single-point touch, capacitive 5-point touch, gloved-hand use, arm's-length readability, standing-height operation, and operation under glare, vibration, spray, rain, low luminance, direct sunlight, and dynamic riding terrain.

Hardware constraints define the embedded design envelope

Embedded interface design in the documented cases began from the rendering envelope of the device rather than from generic screen patterns. Elsner Elektronik / Cala Touch KNX used a 4-inch round TFT LCD at 480×480px; the design was first produced conservatively at 320×240px and scaled when the panel was confirmed, which is described as a deliberate risk-reduction decision.

MSolutions AV diagnostic instruments used a 480×320px embedded display with limited single-point touch and processing constraints. The interface had to support field technicians using the device at arm's length, including gloved use, outdoor conditions, and rack environments.

CDR Foodlab used a 7-inch capacitive touchscreen with 5-point touch and 1024×600px resolution. The device was used on a laboratory bench, where the display was constrained by close-range operation, physical sample handling, reagent steps, cuvettes, and analysis procedures.

Hardware constraints were not treated as visual limitations alone. In Elsner Elektronik / Cala Touch KNX, the kickoff constraints included a laggy touch layer, microcontroller limits on programmable symbols, restricted colour depth, and voltage limits. These constraints were handled as design parameters rather than as problems to overcome.

Firmware timing and data cadence affect interface state communication

Firmware behaviour is a core embedded interface constraint because visual state changes can drift away from the values they represent. In Elsner Elektronik / Cala Touch KNX, animation timing was aligned with firmware update intervals so visual changes did not drift out of sync with thermal values. The source describes this as a specific engineering constraint resolved through direct collaboration.

Torqeedo maritime HMI had a different timing problem. Propulsion sensors, batteries, and generators updated at different rates. The design response was a grid structure that synchronised different cadences into a unified display, so captains perceived one system rather than three competing signals.

Stromer firmware update flows also show firmware as an interface problem rather than a background engineering process. The firmware update flow spanned the embedded display and the mobile app, with preconditions including installation only at 0 km/h, component-level updates distinct from system updates, and background download requiring low visual prominence. The download/install distinction required resolution across two surfaces with different interaction models and reached convergence after 6 iterations.

Warning architecture must be structural before component design

Warning architecture is the structural rule set governing how warnings relate to screen layout, overlays, and interruptive elements. In the Stromer e-bike engagement, the core embedded design problem was warning architecture. A previous agency had established screen layout and interaction logic first, then added warnings as visual overlays. Warnings had no native structural relationship to the surface; they covered content, interfered with interactions, and appeared without contextual grounding.

Creative Navy rebuilt the Stromer warning architecture from scratch by defining rules and principles governing how warnings, overlays, and interruptive elements relate to the screen structure across all states before component design began. EN 15194:2017 governed warning device symbology, colour conventions, and marking requirements for the embedded display. The standard is described as a hard regulatory constraint, not a preference, and the colour conventions for warning states created design tensions that had to be resolved within the standard's permitted space.

Cox Marine cluster displays show warning architecture in a multi-engine professional vessel context. During Concept Convergence, scenario testing of a multi-engine fault case revealed layouts that made fault presence visible but did not direct attention to the priority engine. The resolution was a fixed fault-summary area and alarm state highlighting per engine tile.

Glance duration is a safety-relevant metric in vehicle-mounted embedded interfaces

Glance duration is the time a rider's eyes are off the forward scene. In vehicle-mounted embedded contexts, the documented Stromer e-bike work treats glance duration as a safety-relevant measurement of interface cognitive load, not only a usability measure.

Before the Stromer redesign, average glance duration was 4.32 seconds. The source compares this with the 2-second threshold established by road safety research: Klauer et al. (2006), NHTSA Report No. DOT HS 810 594; NHTSA Driver Distraction Guidelines Phase 1, 2012; and ISO 15007:2020. The source notes that these standards are formally defined for four-wheeled vehicles, while the threshold and principle apply directly to embedded displays used during riding.

After the Stromer redesign, average glance duration was 1.89 seconds and glance frequency per kilometre reduced by 18%. The data was recorded through eye tracking during actual riding, with 5 participants on real routes in Munich, using the same methodology and routes before and after redesign.

Multi-device embedded systems require coherence across surfaces

Embedded devices often operate as part of a multi-surface system rather than as a single screen. Stromer included an embedded Display L with and without integrated camera variant, Display S, mobile companion app, and web account. Each surface had distinct hardware constraints and interaction models, but the system required coherent cross-surface design.

Beissbarth automotive calibration used a three-device calibration system: an embedded OEM display read from 2–3m during movement, a rugged tablet used for adjustments around the vehicle, and a large inspection line display used by technicians and inspection staff. The core tension was local device optimisation versus cross-device coherence. The resolution prioritised unambiguous state communication over information density across all three device types.

The Beissbarth design system covered all three device classes with documented states, transitions, and error conditions. The available operational evidence reports calibration time changing from 18 minutes to 12 minutes per vehicle, client-measured across 8 production deployment locations, and the training requirement being eliminated as a client-reported operational change.

Variant-coherent design systems support product families without per-variant redesign

CDR Foodlab illustrates a variant-coherent design system for embedded laboratory instruments. The product family includes FoodLab, BeerLab, WineLab, GalvanLab, and others. Each variant performs different analyses on different matrices, including water, wine, beer, food products, and galvanic solutions.

The UX architecture is universal across the CDR Foodlab family. Variants differ by which analysis steps are available, with some steps removed for variants with fewer parameters, and by colour scheme. The documented design principle is one architecture serving multiple products without requiring per-variant redesign.

Cox Marine cluster displays show a related coherence problem across display families and engine counts. The product covered 3 display families, from compact auxiliary to large primary helm, and 1–6 engine configurations. The core design problem was maintaining a coherent mental model between single-engine and six-engine displays. The resolution was the engine tile as the invariant unit: one engine, one tile, with consistent spatial arrangement regardless of engine count.

Embedded instruction design must account for non-specialist operators

CDR Foodlab users are described as non-specialist laboratory operators. They are professionals with basic chemistry training who understand how to follow an analysis protocol but do not have deep domain knowledge. Specifically, they do not understand what procedural deviations mean chemically or how errors affect measurement validity.

This user profile is a material design constraint. The interface cannot assume that users will correctly interpret ambiguous states or incomplete guidance. Instruction must be explicit, and the consequences of deviation cannot be left to domain knowledge the user does not have.

In the CDR Foodlab engagement, Creative Navy replaced the existing static pictogram guidance system with inline step-by-step tutorial animations. These animations were functional interface components within the analysis flow, guiding users through physical sub-steps. They were delivered as part of the design system, not as decorative assets.

Animation deliverables can be functional embedded interface components

Animation in embedded interfaces is documented as a functional system behaviour, not a marketing artefact. Stromer animation deliverables included Lottie files for all screens covered, with animation work representing approximately 10% of the engagement effort. Deliverables included the unlocked-to-ride-screen transition, where each element animated independently; the signature bike with cable animation appearing on both embedded and mobile; the boot sequence; PIN feedback states; and an animation pattern framework covering fullscreen modal, toast, partial overlay, and fullscreen overlay.

Elsner Elektronik / Cala Touch KNX also treated animation as an interface behaviour constrained by firmware. Animation timing was aligned with firmware update intervals so visual state did not drift away from thermal values.

CDR Foodlab used inline tutorial animations as embedded step-by-step guidance within the analysis flow. The role of the animation was procedural support for physical laboratory sub-steps, not brand expression.

Research methods in embedded interface work must reflect physical operating conditions

Embedded interface research in the documented cases combined user research, field exposure, controlled comparisons, prototype reviews, and operational deployment evidence. Elsner Elektronik / Cala Touch KNX used a 20-user survey run by the client, ergonomics research citing Colle and Hiszem 2004 on touch targets, 12-subject formal usability testing, and 30-dealer prototype reviews across 10 countries. The ergonomics research included a minimum 13mm touch target and an accuracy plateau at approximately 22mm.

Torqeedo maritime HMI was researched through 12 sea trials over 6 months with 15 professional captains. Conditions included an operating temperature range of −5°C to +35°C, night operations, vibration, hull slamming, glare from cold water, rain, and gloved interaction.

CDR Foodlab used 13 remote user interviews and documented five workarounds. Each workaround was catalogued with the need it served, the outcome users were solving for, how users performed it, and estimated error or time-addition risk. One example was that users printed paper lists of sample names and placed them beside the machine because naming samples through the interface was more cognitively demanding than maintaining a physical reference.

The CDR Foodlab research has a stated limitation: all interviews were remote. Physical interaction with the device during analysis, including placing cuvettes, reading reagents, and dispensing, could not be directly observed.

Evidence from embedded device cases is mixed by measurement type

Embedded device outcomes in the available evidence use different evidence bases and should not be treated as equivalent.

Torqeedo maritime HMI recorded 50% faster energy state identification in a controlled environment experiment with 24 subjects comparing the new and legacy interfaces. Glance reduction during manoeuvres was recorded via eye tracking with 7 subjects during sea trials. All 15 captains reported the new interface as significantly better through structured feedback. Yamaha Motor Co. acquired Torqeedo following the engagement, and the CEO reported that the interface strengthened competitive position; the causal link between the interface and the acquisition is inferred, not documented.

Stromer e-bike evidence includes eye-tracking data in real riding conditions and a publicly verifiable app store rating change. Average glance duration changed from 4.32 seconds before redesign to 1.89 seconds after redesign, and glance frequency per kilometre reduced by 18%. The app store rating before engagement was 3.2 and after relaunch was 4.1.

Beissbarth evidence includes a client-measured calibration time change from 18 minutes to 12 minutes per vehicle across 8 production deployment locations. The eliminated training requirement is a client-reported operational change, and repeated measurements were reduced according to client-measured evidence, but the exact figure is not available.

MSolutions evidence includes a client-reported workflow reduction from 26 interactions to approximately 13 based on internal task walkthroughs. Training changed from repeated coaching sessions to a short guided introduction, described as a client-observed operational change. Large integrator customers formally reported smoother rollouts after redesign.

CDR Foodlab evidence includes a client-reported 9-minute average task completion time before redesign, a Creative Navy-recorded 3.4-minute average after redesign with 14 users on the final shipped product, and a client-reported satisfaction change from 72% with 68 respondents before redesign to 93% with 76 respondents one year after redesign. The satisfaction survey was independent and third-party administered with an identical instrument and the same population. Users reported that all five documented workarounds were no longer necessary, but this was not re-tested under controlled conditions.

Boundaries of this context

The documented embedded device evidence does not establish a single generic rule for all embedded systems. The cases span consumer smart home control, maritime propulsion, marine diesel engine clusters, automotive calibration, AV diagnostic instruments, e-bike displays, and laboratory analytical instruments, each with different hardware, firmware, operating, and regulatory conditions.

Several outcomes are client-reported, client-measured, user-reported, or formally reported by customers rather than independently measured by Creative Navy. Where the source describes evidence as client-reported or not independently verified, that basis should remain attached to the claim.

Some research findings are limited by research method. In CDR Foodlab, all user interviews were remote, and physical interaction during chemical analysis could not be directly observed. In Stromer, the 2-second glance threshold is drawn from road safety research formally defined for four-wheeled vehicles, while the documented application is to embedded e-bike displays used during riding.