A practical guide for SME office owners: how to upgrade an existing 100–500 sqm office to human-centric lighting using wireless Matter drivers — no rewiring, no structural changes, measurable results from week one.
A human-centric retrofit for a small office usually means swapping the driver inside each luminaire, not replacing the luminaires — and then loading one schedule that runs every working day. For a 200 sqm open-plan office that is roughly 24 driver nodes, one wall panel and a handful of presence sensors, commissioned in about one electrician-day with no new control cabling. This guide gives the schedule, the evidence behind it, the arithmetic for energy and payback, and the assumptions each number rests on.
| Key fact | Source |
|---|---|
| Scheduled office lighting was measured at ~18% better task performance and ~12% better work accuracy versus static illumination. | CBRE × University of Twente / VU Amsterdam healthy-offices study (2017) |
| The retrofit swaps drivers, not luminaires — most commercial LED panels accept a replacement driver. | Design consequence; driver replacement is the intervention |
| A 200 sqm office schedule runs 5700K at 08:00 down to 2700K at 18:30, six nodes. | SUMLINK Office Circadian Light Line programme |
| Presence-linked dimming typically reduces lighting energy 15–25% in low-traffic zones. | Internal estimate — method and assumptions shown below |
| The retrofit sits in the Tier S — Standard Programme; the design fee is offset against hardware once the BOM reaches the agreed threshold. | SUMLINK programme structure |
Vocabulary used below
| Term | What it means | Why it matters here |
|---|---|---|
| CCT | Correlated colour temperature, in kelvin — how warm or cool the white looks | The variable a tunable white schedule moves |
| Illuminance | Light on a surface, in lux | The variable the schedule dims |
| M/P ratio | Melanopic ÷ photopic illuminance | Governs the circadian signal at a given lux |
| m-EDI | Melanopic equivalent daylight (D65) illuminance, per CIE S 026:2018 | The unit WELL and CIE thresholds are written in |
| Driver | The power supply and dimming interface inside each luminaire | The component this retrofit replaces |
| Node (schedule) | One time-stamped step in the daily programme | Six nodes make the office day |
A conventional office runs fixed 4000K panels from 08:00 to 20:00. It is defensible at 10:00. By 14:30 it is doing nothing to help the post-lunch dip, and by 18:00 the people still at their desks are under the same blue-weighted spectrum that will suppress melatonin for the rest of the evening.
The mechanism is photobiological, not a matter of taste. A third class of retinal photoreceptor — intrinsically photosensitive retinal ganglion cells (ipRGCs) — peaks in sensitivity in the short-wavelength (blue) region around 480 nm and projects directly to the suprachiasmatic nucleus, the body's master clock. That channel is what sets melatonin suppression, cortisol timing and alertness. A static 4000K spectrum delivers the same dose of it at 08:00 as at 20:00, which is exactly the wrong pattern.
Source: the ipRGC discovery (Berson, Dunn & Takao, 2002, Science) and the CIE S 026:2018 α-opic action spectra; "around 480 nm" is the sensitivity peak of the melanopsin photopigment, not a specification for a lamp.
Human-centric lighting changes the spectrum and the level across the day so the signal varies the way daylight does. Nothing about this requires replacing your luminaires.
The most-quoted figure in office lighting is the CBRE study, and it is usually quoted without its details. Here are the details.
| Attribute | Value |
|---|---|
| Study | Healthy-offices field study run at a CBRE office workplace |
| Partners | CBRE, University of Twente, VU Amsterdam |
| Year | 2017 (published/announced October 2017) |
| Intervention | LED lighting with a tuned spectral curve and brightness variation, versus the conventional installation |
| Sample | ~124 participants over roughly 7 months |
| Reported outcomes | Perceived task performance ~+18%; objective work accuracy ~+12%; self-reported wellbeing measures (energy, mood, health) also improved |
| Metric definitions | "Performance" is self-reported task performance; "accuracy" is the objective test measure. They are not interchangeable. |
| Source type | Field study reported by the lighting manufacturer and trade press; read as media/industry reportage, not as a peer-reviewed paper |
Treat those numbers as directional evidence, not a guarantee. They are specific to one building, one organisation, one task mix and one baseline. What they establish is that a scheduled tunable-white programme is a defensible specification choice rather than a lifestyle flourish — not that any given fit-out will reproduce an 18% gain.
Two further references are worth knowing when a client asks "what does the standard say". CIE S 026:2018 is the metric system (α-opic quantities, including melanopic EDI) that quantifies the non-visual content of light; it is a measurement standard, not a threshold source. The thresholds themselves come from elsewhere — WELL v2 Feature L03 for workplace circadian lighting, EN 12464-1:2021 for illuminance, uniformity and glare limits. Quoting CIE S 026 as if it set the limits is a common and checkable error.
Once the hardware is in, the schedule is programmed once and runs every working day. The published SUMLINK office programme has six nodes.
| Time | CCT | Illuminance | M/P ratio | m-EDI at the work plane | Intent |
|---|---|---|---|---|---|
| 08:00 | 5700K | 500 lx | 0.92 | 460 | Morning activation — counter grogginess, sharpen early decisions |
| 09:30 | 5000K | 480 lx | 0.81 | 389 | Peak focus window — sustain alertness through the morning |
| 12:00 | 4500K | 450 lx | 0.74 | 331 | Midday transition — neutral through the lunch break |
| 14:00 | 4000K | 420 lx | 0.66 | 277 | Afternoon steady — counter the post-lunch dip |
| 17:00 | 3500K | 350 lx | 0.58 | 203 | Wind-down — begin reducing the circadian signal |
| 18:30 | 2700K | 250 lx | 0.45 | 112 | After-hours protection — low melanopic load for late work |
How the last two columns were produced, in full. The CCT and illuminance columns are the published programme. The M/P column is the SUMLINK luminaire design-target curve at each CCT step (0.45 at 2700K rising to 0.92 at 5700K); the 4500K value is a linear interpolation between the 4000K and 5000K design targets. The m-EDI column is then `M/P ratio × illuminance` — a multiplication, not a laboratory measurement. The whole table is therefore design-target arithmetic, and it is labelled that way so that nobody mistakes it for measured photometry. On a real project the numbers come from an SPD measurement of the actual driver-and-engine combination.
Two things the table makes visible. First, the melanopic dose falls by roughly a factor of four across the day (460 to 112) while the illuminance falls by only half (500 to 250 lx) — because both variables move together. If you dim a static 4000K luminaire from 500 lx to 250 lx, you get 165 m-EDI in the evening, not 112. That difference is what the spectrum change buys you, and it is the entire reason the schedule is not simply a dimmer.
Second, the evening node lands at 112 m-EDI. That is far above the ≤10 m-EDI evening guidance in the peer-reviewed literature, and above what a 2700K source can reach at any usable illuminance. The honest framing: a 2700K end-of-day node reduces the evening signal substantially but does not deliver a pre-sleep environment. It is the right setting for an office at 18:30 under occupancy; it is the wrong setting to describe as melatonin-neutral.
The good news for an SME office is that the intervention is the driver, not the luminaire. Most commercial LED panels and linear luminaires in a 100–500 sqm Hong Kong office already accept a replacement driver.
| Component | Product | Role | Protocol |
|---|---|---|---|
| Tunable white driver | LTECH `SE-40-300-1050-W2MA` (40 W, 300–1050 mA, dual-channel) | Replaces the existing driver; controls CCT and level | Matter over Thread, Matter 1.4 |
| Tunable white driver (smaller fixtures) | LTECH `SE-12-100-500-W2MA` (12 W, 100–500 mA, dual-channel) | Same role for lower-wattage luminaires | Matter over Thread, Matter 1.4 |
| Wall panel | LTECH Super Panel 4S or 12S | Stores the schedule locally; app bridge | Bluetooth 5.0 SIG Mesh + Wi-Fi + Ethernet |
| Presence sensor | LTECH `MR03-B3P` microwave sensor | Auto-dim unoccupied zones; daylight compensation | Bluetooth SIG Mesh |
| App | LTECH Smart Life | Schedule setup, zoning, overrides | Free, no licence |
| Thread Border Router | Not supplied by SUMLINK | Required for Matter-over-Thread reachability | Thread |
Three specification notes that the published article omits. First, the driver model matters: the `SE-12-100-500-W2MA` does not carry IEEE 1789 or CIE SVM certification in our product data, while the `SE-40-300-1050-W2MA` does. If a flicker certification is a tender requirement, it must be written against the specific driver, not "the LTECH driver". Second, the catalogue CCT range for these engine-and-driver combinations spans 1800–6500K — the widest `cct_range` in `bom.csv` (`H032-LE016-D007-PENDING-25W`), with other rows at 2200–6500K and 2700–6500K — which is wider than the 2700–6500K band the schedule uses. Third, the `MR03-B3P` presence sensor currently exists in our catalog without a datasheet record, so no detection range or hold-time figure should be written into a tender specification for it yet.
No new control cabling, no structural changes — one electrician, approximately one day for a 200 sqm open-plan office. That labour figure is a planning estimate, not a quotation.
Basis: `driver_lt.csv`, `panel.csv` and `bom.csv` for the hardware; the one-day labour figure is a SUMLINK internal planning estimate.
A six-node schedule is only as good as its alignment with the actual daylight window. And in Hong Kong that window is not a constant.
| Daylight anchor | Winter (21 Dec) | Summer (21 Jun) | Difference |
|---|---|---|---|
| Day length | 10 h 46 m | 13 h 30 m | 2 h 44 m |
| Sunrise | 06:59 | 05:40 | 1 h 19 m earlier |
| Sunset | 17:45 | 19:10 | 1 h 25 m later |
| Solar-noon altitude | 44° | 89° | Winter sun rakes deep into the floor plate; the summer sun is effectively overhead |
Source: SUMLINK calculation (standard solar geometry, refraction-corrected); the Hong Kong values cross-check against Hong Kong Observatory published sunrise and sunset times.
The consequence for programming is concrete. Sunset moves by 1 h 25 m against a fixed 09:00–18:00 office day, so an evening warm-up that starts at 17:00 is roughly right in December and about an hour and a half early in June. The six-node programme below is therefore specified as a two-season variant for Hong Kong — a winter node set and a summer node set sharing the same shape — so that the low-melanopic evening phase tracks local sunset rather than the clock. A single year-round curve is the correct answer at equatorial latitudes, where day length varies by only about nine minutes across the year; it is the wrong answer here.
If a single fixed schedule is all the budget allows, put the flexibility in the evening node and leave the morning alone. The morning activation window (08:00–10:30 at high CCT) lands inside the first working hour in both seasons, so it needs no adjustment; it is the 17:00-onwards phase where a fixed clock time visibly mismatches the sky.
Most SME offices in Hong Kong are leased, which turns a lighting upgrade into a question about the tenancy agreement as much as about photometry. The retrofit is favourable here because the intervention is a component swap.
| Question | Answer |
|---|---|
| Does it need landlord consent as a building-services modification? | It is ordinary electrical work inside the demised premises: no new control cabling, no ceiling grid changes, no landlord switchgear. Confirm the lease's electrical-works clause anyway. |
| Is the change reversible at lease end? | Yes, if you keep the original drivers. The replacement driver sits in the same fixture; refitting the originals restores the previous installation. |
| Are the luminaires permanently modified? | No. The fixture housing, the ceiling grid and the circuit wiring are untouched. |
| What about the wall panel and sensors? | Surface-mounted and removable. Sensor mounting points leave minor ceiling marks, in line with normal fit-out practice. |
| Who owns the hardware? | The tenant, unless the fit-out agreement says otherwise. Because the drivers are removable components, they can move with you to the next premises. |
None of this is legal advice, and reinstatement clauses vary. The practical point is that a retrofit built on driver substitution leaves far less to argue about at handover than one that replaces luminaires or adds a control bus.
Basis: general Hong Kong commercial lease practice; not a review of any specific tenancy agreement and not legal advice.
| Check | Why it decides the project |
|---|---|
| Driver compatibility per fixture type | Not every luminaire accepts a replacement driver; wattage, current range and physical form factor must all match |
| Existing circuit layout | Circuits should let you zone by area — corridors, meeting rooms and pantry separate from the open plan |
| Panel location | The panel is mains or USB-C powered and stores the schedule; it needs a practical mounting point and a power source |
| Network availability | Matter-over-Thread drivers need a Thread Border Router; Wi-Fi and a WAN connection are required for remote app access and updates |
| Zoning ambition | The energy case below scales directly with how much of the floor can be auto-dimmed when unoccupied |
Basis: SUMLINK retrofit planning checklist (internal); the driver and gateway requirements are taken from `driver_lt.csv` and `panel.csv`.
| Mistake | Why it costs money | Do this instead |
|---|---|---|
| Specifying "DALI-2" out of habit | The SE-series drivers in this programme are Matter over Thread, not DALI. Writing DALI forces a different driver and new control cabling | Write the protocol you are actually installing, and check it against the driver model |
| Requiring a CCT range wider than the schedule uses | The catalogue spans 1800–6500K at its widest (most rows are 2200–6500K or 2700–6500K); the programme uses 2700–6500K. Paying for range you never programme buys nothing | Specify the band the schedule actually traverses |
| Applying a flicker certification requirement to "the LTECH driver" | Certification is model-specific: in our product data the `SE-40-300-1050-W2MA` carries IEEE 1789 and CIE SVM, while the `SE-12-100-500-W2MA` does not | Name the driver model in any certification clause |
| Quoting a detection range for the `MR03-B3P` sensor | No datasheet record exists in our catalogue, so any figure in a tender would be unverifiable | Leave the parameter out until the datasheet exists |
| Writing HCL performance as a measured outcome | Our luminaire data records M/P and m-EDI as open `HCL_PENDING` gaps — the values are design targets | Write the design target and the verification method, and commission with a lux meter |
| Ignoring zoning | The energy case scales with how much floor can be auto-dimmed when unoccupied | Zone corridors, meeting rooms and pantry separately from the open plan |
Basis: each row's factual basis is the source cited in the corresponding section.
Based on comparable installations, and stated as expectations rather than guarantees:
None of these is a measured outcome for your specific space. The first month is the period in which you find out which of them applies.
Basis: SUMLINK programme observation notes (internal); none of these is a measured result for any specific installation.
A savings percentage with no formula is a slogan. The relationship is:
`annual kWh saved = saving% × connected lighting load (kW) × daily operating hours × operating days`
Basis: SUMLINK internal estimate. The relationship is arithmetic; the 15–25% band is our planning assumption, not a measured result from a completed installation.
| Scenario | Saving | Connected load | Daily hours | Days/yr | kWh/yr saved |
|---|---|---|---|---|---|
| Small office, conservative | 15% | 5 kW | 10 h | 250 | 1,875 |
| Small office, mid | 18% | 5 kW | 10 h | 250 | 2,250 |
| Small office, optimistic | 25% | 5 kW | 10 h | 250 | 3,125 |
| Larger floor, conservative | 15% | 10 kW | 12 h | 250 | 4,500 |
Multiply the kWh figure by your own tariff to get currency. The tariff is deliberately left as a variable here: Hong Kong commercial tariffs vary by contract and by time-of-use band, and quoting a single rate would make the table wrong for most readers.
The saving percentage itself depends on how much of the floor is low-traffic — a fully occupied open plan with no zoning will realise far less than a floor with meeting rooms, corridors and a pantry. Treat 15% as the conservative floor and 25% as the ceiling that requires real zoning.
The retrofit sits in SUMLINK's Tier S — Standard Programme. The design fee is offset partially or fully against hardware purchase once the BOM reaches the agreed threshold, so the fee is not paid twice.
Rather than a headline payback figure, here is the formula:
`payback (months) = (design fee + hardware BOM) ÷ (monthly kWh saved × tariff)`
Worked through with the conservative row above — 1,875 kWh/yr, or about 156 kWh/month — the numerator is what your quotation says and the denominator is 156 × your tariff. At an illustrative HKD 1.4/kWh the monthly saving is roughly HKD 218, which is why the honest payback statement is "energy savings offset a portion of hardware cost", not "the system pays for itself in 18–24 months". The 18–24 month claim in the published article depends on a larger connected load or a higher tariff than the conservative case; it is presented here as arithmetic instead of as a promise.
The productivity and wellbeing effects sit on top of the energy case and are not included in any payback figure, because they are directional evidence rather than a contractual quantity.
Basis: SUMLINK internal planning estimate. The design-fee and BOM inputs are project-specific and quoted per project; the payoff arithmetic is the formula below applied to a project's own tariff and hours.
Usually not. The retrofit replaces the driver - the power supply and dimming interface inside or adjacent to each luminaire. Most commercial LED panels and linear luminaires accept a replacement driver, so the luminaire, ceiling grid and wiring stay.
Hardware is quoted as a bill of materials per project, and the design fee sits in SUMLINK's Tier S Standard Programme with the fee offset against hardware once the BOM reaches the agreed threshold. The useful figure is the payback formula rather than a headline number.
Independent field research points that way: the CBRE and University of Twente / VU Amsterdam study reported roughly 18% better perceived task performance and roughly 12% better objective work accuracy against a conventional lighting baseline. Treat them as directional evidence, not a guarantee.
For a 200 sqm open-plan office, approximately one electrician-day for the driver swap, panel mounting and sensor placement, plus commissioning time to load the schedule and verify illuminance at the work plane. No new control cabling is required.
Yes. The six-node schedule is stored locally on the wall panel, so the daily programme continues without a WAN connection. Remote app access, voice control and firmware updates require the internet.
Commission with a lux meter. Verify illuminance at the work plane and at approximate eye height, then check the CCT steps against the panel readout. The published M/P and m-EDI figures are design targets; the installed result has to be measured.
The Office Circadian Light Line programme is specified and ready to deploy — the schedule, the configuration data and the BOM. → View the Office HCL Program