Engineering · scale demonstrator

Two units exist and they work. They are one fifth the size, and we say so.

The rig has real modules, real cells with a BMS, a real USB-C output at each end and the 2 mm programmable COB lighting running the production dusk programme, all on the same firmware and the same telemetry frame as the full bench. That makes it a genuine control and software demonstrator, and the fastest honest route to real photography. What it is not is a strength, uplift or ingress test, and nothing on this platform will pretend otherwise: our whole argument against the render-led competition is that we do not do that.

Results that carry across unchanged

Verified

Results that scale with shown arithmetic

Estimated

Results a scale rig cannot settle

Verified

The rig

Not a model of a bench. A small bench that charges a phone.

Scale

Built specification: 1:5 linear

Array area ratio

Built specification: 1:25 by array area

Units built

Built specification: Two units, built to allow paired scene photography and A/B control tests

Chassis

Built specification: Printed PETG chassis and canopy shell to production proportions

Array

Built specification: Working small-format monocrystalline modules, 4 x 2 W nominal, 8 W array

Storage

Built specification: 2 x concealed thin lithium polymer cells, 3.7 V 2,000 mAh, 14.8 Wh total, with BMS

Outputs

Built specification: 1 x USB-C output at each end, 5 V regulated, real device charging

Lighting

Built specification: 2 mm programmable COB LEDs on the canopy underside, production dimming and dusk programme

Control

Built specification: Same load priority firmware and same telemetry frame as the production controller

Item	Built specification
Scale	1:5 linear
Array area ratio	1:25 by array area
Units built	Two units, built to allow paired scene photography and A/B control tests
Chassis	Printed PETG chassis and canopy shell to production proportions
Array	Working small-format monocrystalline modules, 4 x 2 W nominal, 8 W array
Storage	2 x concealed thin lithium polymer cells, 3.7 V 2,000 mAh, 14.8 Wh total, with BMS
Outputs	1 x USB-C output at each end, 5 V regulated, real device charging
Lighting	2 mm programmable COB LEDs on the canopy underside, production dimming and dusk programme
Control	Same load priority firmware and same telemetry frame as the production controller

Proof matrix

Fourteen questions, sorted by whether the rig can answer them.

Ratios and control logic are size independent, so they carry. Energy figures scale with published arithmetic. Structure, uplift, sealing, thermal behaviour and human use do not carry at all, and are held for the full-size programme.

MPPT tracking behaviour and conversion efficiency

Verdict: Carries across unchanged
Why: Efficiency is a ratio. A controller that tracks correctly at 8 W tracks correctly at 1,400 W with the same algorithm.

Load shedding order and thresholds

Verdict: Carries across unchanged
Why: The priority rules are expressed in percentage state of charge, so they are size independent by construction.

USB-C power delivery negotiation and session logging

Verdict: Carries across unchanged
Why: Same protocol, same negotiation, same session record. Only the number of concurrent ports differs.

Lighting programme, dimming curve and dusk trigger

Verdict: Carries across unchanged
Why: The COB LEDs run the production programme. Lumen output differs, the control curve does not.

Telemetry frame, buffering, retry and offline back-fill

Verdict: Carries across unchanged
Why: Identical 44-byte frame into Canopei Fleet, so the software is validated against real hardware rather than a simulator.

Fault detection, recovery and firmware update path

Verdict: Carries across unchanged
Why: Same firmware image and the same fault codes. Recovery after a forced brownout is the same test at either size.

Daily energy harvested

Verdict: Scales by area or capacity
Why: Measured Wh per day multiplied by 25 for the array area ratio, then adjusted for the production module efficiency. The arithmetic is shown, not hidden.

Usable storage and autonomy in days

Verdict: Scales by area or capacity
Why: 14.8 Wh against a 2.4 kWh production pack is a 162x capacity ratio. Autonomy in days is comparable only after both are normalised to their own daily load.

Winter generation shape across a month

Verdict: Scales by area or capacity
Why: The daily profile and the ratio of December to June generation carry across; the absolute figures scale.

Seat strength, bearer deflection and fatigue

Verdict: Does not carry across
Why: A printed chassis proves nothing about a 3.0 kN/m seat load. This waits for the production frame and the BS EN 581 programme.

Wind uplift and ballast

Verdict: Does not carry across
Why: Uplift scales with area and pressure in ways a 1:5 model in still air cannot represent. Calculation and full-size rig only.

IP65 ingress at production enclosure size

Verdict: Does not carry across
Why: Sealing behaviour depends on gasket length, face geometry and thermal cycling at production scale.

Thermal behaviour of the pack and canopy

Verdict: Does not carry across
Why: Thermal mass scales with volume while surface area scales with the square. The model runs cooler than the product will.

Occupancy, dwell and real public use

Verdict: Does not carry across
Why: Nobody sits on a 1:5 bench. Dwell and session behaviour needs the first sited unit.

Question	Verdict	Why
MPPT tracking behaviour and conversion efficiency	Carries across unchanged	Efficiency is a ratio. A controller that tracks correctly at 8 W tracks correctly at 1,400 W with the same algorithm.
Load shedding order and thresholds	Carries across unchanged	The priority rules are expressed in percentage state of charge, so they are size independent by construction.
USB-C power delivery negotiation and session logging	Carries across unchanged	Same protocol, same negotiation, same session record. Only the number of concurrent ports differs.
Lighting programme, dimming curve and dusk trigger	Carries across unchanged	The COB LEDs run the production programme. Lumen output differs, the control curve does not.
Telemetry frame, buffering, retry and offline back-fill	Carries across unchanged	Identical 44-byte frame into Canopei Fleet, so the software is validated against real hardware rather than a simulator.
Fault detection, recovery and firmware update path	Carries across unchanged	Same firmware image and the same fault codes. Recovery after a forced brownout is the same test at either size.
Daily energy harvested	Scales by area or capacity	Measured Wh per day multiplied by 25 for the array area ratio, then adjusted for the production module efficiency. The arithmetic is shown, not hidden.
Usable storage and autonomy in days	Scales by area or capacity	14.8 Wh against a 2.4 kWh production pack is a 162x capacity ratio. Autonomy in days is comparable only after both are normalised to their own daily load.
Winter generation shape across a month	Scales by area or capacity	The daily profile and the ratio of December to June generation carry across; the absolute figures scale.
Seat strength, bearer deflection and fatigue	Does not carry across	A printed chassis proves nothing about a 3.0 kN/m seat load. This waits for the production frame and the BS EN 581 programme.
Wind uplift and ballast	Does not carry across	Uplift scales with area and pressure in ways a 1:5 model in still air cannot represent. Calculation and full-size rig only.
IP65 ingress at production enclosure size	Does not carry across	Sealing behaviour depends on gasket length, face geometry and thermal cycling at production scale.
Thermal behaviour of the pack and canopy	Does not carry across	Thermal mass scales with volume while surface area scales with the square. The model runs cooler than the product will.
Occupancy, dwell and real public use	Does not carry across	Nobody sits on a 1:5 bench. Dwell and session behaviour needs the first sited unit.

Conversion

The arithmetic, so you can redo it yourself.

Worked example

A measured 34 Wh harvested on the rig in a day becomes 34 x 25 x 0.94, or roughly 799 Wh at production array area. The 0.94 accounts for the small modelling modules running slightly better per square metre than a graded second-life module.

When it is not valid

The conversion is only valid when the rig and the production array sit at the same tilt, orientation and shading. Every logged run records all three.

Estimated

Test log

Empty is the honest answer today.

No runs logged yet

The log is empty until the first unit is powered and logging. When runs land they appear here in full, including the runs that go badly, because a log that only holds good days is a brochure.

Photography programme

Six scenes that replace a render with a photograph.

Daylight, occupied, urban parade

Real hardware in a composed street scene, canopy and deck reading as built product rather than a render.

Dusk, canopy lighting live

The COB programme photographed running, which is the shot no render can honestly claim.

Wet weather, standing water at the deck

Water behaviour on the composite profile and at the socket face, our own drainage proof.

Close frame, module glass and composite edge

Material honesty: the reclaimed glass surface and the hemp-plastic section, photographed not described.

Close frame, USB-C face with a phone charging

A device drawing real current from a real output. Session logged into Canopei Fleet at the same moment.

Paired units, A/B control test

Two rigs side by side, one with shading applied, evidencing the yield-against-model logic in the fleet view.

Imagery rule

The house rule that makes the competitor argument stand up.

We criticise render-led catalogues for implying a product exists. That criticism only holds if we never do it, so this rule is published and applies to every image on the platform.

·Prototype photography is real hardware and is presented as photography, not as a render.
·No caption, specification line or figure attributes a production dimension, wattage or storage capacity to a prototype image.
·Every absolute performance figure on the platform is sourced to the production specification or to a converted rig measurement with the conversion shown.
·Retouching, grading, sharpening and scene composition are normal product photography and are used freely.
·Renders that remain on the platform carry a render label at the image, with the date.

Where this sits

The rig validates the firmware behind Canopei Fleet, supplies the real photography tracked on the differentiation roadmap, and leaves the full-size structural programme on the compliance matrix.