Can LED point lights be repaired? What you can fix in the air, and when repair costs more than replacement
A practical guide from a facade lighting factory: the four checks that find the fault before you open a fixture, how to replace a single pixel without destroying its IP rating, and the arithmetic that usually makes replacement cheaper than repair.
Yes — a single LED point light can almost always be repaired or replaced. But across 22 years of field service records since 2004, the majority of “dead pixel” call-outs are not dead pixels at all: they are power, signal or connector faults, and the fixture is fine. That distinction matters, because a 12 mm pixel costs $0.50–2.00 while one visit to a 20-storey facade costs $1,000–1,800 in access before anyone touches a light. The real question is never “repair or replace one light”. It is “what do I do in a single visit”.
Diagnose in order: power, signal, connector, fixture. Opening the fixture first is the most common reason a repair gets done twice.
The connector is the weak point, not the LED. Water enters at the joint, oxidises the pins, and the symptom looks exactly like a failed pixel.
Repair loses on labour, not on parts. Replacing a bead takes 25–40 minutes; swapping a whole pixel takes 5–8. On hired access, that ratio is the entire cost.
A repaired pixel will not match its neighbours. The old pixels have already lost 10–20% of their output. You can reduce the mismatch, not erase it.
Private tooling locks you to the original factory. A new housing mould costs $8,000–35,000, so no supplier will open one for a repair order — which is a decision you make at purchase, not at failure.
Start with the order of checks, not the tool kit
Almost every point light we are asked to look at arrives with the same history: someone saw a dark pixel, climbed to it, opened the housing, found nothing obviously burnt, sealed it again, and the pixel stayed dark. The fixture was never the problem.
A point light sits at the end of a chain: switch-mode power supply, power trunk, signal source, connector, fixture, next fixture. Faults can start anywhere on that chain, and the symptom at the top of the building is identical in every case — one pixel, or a run of pixels, goes dark. So the checks have to run outward-in, from the shared infrastructure to the individual device. Each step below eliminates a whole class of causes, and each one is cheaper than the step after it.
Step 1 — Confirm the branch has power, measured under load
A 24 V constant-voltage run measured with nothing connected will read 24 V even when the supply is failing. Measure it with the load on. If a 24 V rail sags below about 21.6 V at the far end — a 10% drop — you will see flicker, pink or amber colour shift, and eventually pixels that refuse to address. That is a voltage-drop problem, not a pixel failure.
Voltage drop is a function of current and cable, and point light runs are current-heavy because the pixels are dense. A 24 V run of 60 pixels at 0.3 W each draws 0.75 A at the head, but the tail sees the accumulated drop of every joint before it. Two things cause most of it: cable that is thinner than the run length justifies, and power injected at one end only. Injecting at both ends of a long run is the cheapest fix in facade lighting, and it is routinely missed.
Keep a 24 V point light run under 50–60 pixels per power feed, and feed both ends above 40. On 5 V systems the number is far lower — 5 V drops four times faster for the same current, which is why 5 V pixels belong on short runs only. If a client reports “the last twenty pixels are dimmer”, this is the cause nine times out of ten.
Step 2 — Find the break in the signal chain
Point lights are addressed over one of two architectures, and the difference decides how a single failure propagates:
| Protocol | Wiring | Effect of one failed pixel | How to locate the fault |
|---|---|---|---|
| DMX512 | Bus. All fixtures share one differential pair; each listens for its own address. | Usually isolated — other pixels keep working, unless the failed unit shorts the bus. | Read addresses with a DMX tester; the missing address is the failed unit. |
| SPI (clockless serial) | Daisy chain. Data shifts through each pixel to the next. | Cascading. Everything after the failed pixel goes dark or shows shifted colours. | Bypass or replace the first dark pixel; if the run returns, that unit was the break. |
Table 1. This is why “the whole top half is out” is a normal SPI symptom and an abnormal DMX512 one.
The trap on SPI is that the first dark pixel is not necessarily the broken one. Data enters a pixel, is regenerated, and leaves on the data-out line. If the pixel is dead but passes data through, the next one works and only this one is dark. If data stops here, everything downstream dies. So the unit to replace is either the first dark pixel (it stopped regenerating data) or the one immediately before it (its data-out is dead). Carry two spares up, not one — otherwise you climb twice.
On DMX512 the failure mode is different and often misdiagnosed. A single failed unit does not usually break the bus, so a dark pixel in the middle of a working run is a device fault, while a whole universe going dark points at the controller, the terminator, or the signal ground. Missing or lifted signal ground is the classic cause of “works when the controller is on the ground, flickers when it is in the rack”.
Step 3 — Open the connector before you open the fixture
This is the step that saves the most money and gets skipped the most. On outdoor facade systems the connector, not the pixel, is the highest-failure component, because it is where mechanical load, electrical contact, waterproof sealing and installer technique all meet in a 20 mm moulding.
The failure sequence is predictable. A seal compresses and hardens under UV, or the locking ring is left a quarter-turn short, or the cable is pulled tight so the joint sits under tension. Water enters, sits against the pins, and starts electrolytic corrosion. Contact resistance climbs, the pixel sees a voltage sag, and it flickers or drops out. Then it dries, works again for a week, and fails after the next rain. By the time someone climbs to it, the fixture looks fine and the joint looks “a bit green”.
It is the most common field mistake we see, and it makes things worse. Wrapping a joint traps the moisture already inside, keeps it against the pins, and hides the corrosion from the next inspection. If a joint has taken water, the correct action is to dry it and replace the seal — or replace the connector. If the cable has taken water into the conductor strands, no amount of external sealing will help; that cable end has to be cut back and re-terminated.
What to check, in this order: pins blackened or carrying white-green corrosion products; O-ring hardened, cracked, flattened or missing; locking ring not fully seated; cable jacket nicked or pulled out of the gland; joint mounted face-up so it holds water; joint buried in a curtain-wall groove with no drainage. Any of the first four means replace the connector, not clean it.
Cleaning is only worth doing when the corrosion is light and recent: power off, separate the connector, photograph the pin order before anything moves, flush the pins with anhydrous ethanol, blow dry with clean compressed air, fit a new O-ring, and seat the ring fully. Do not use heat to dry a connector — a heat gun hardens the silicone seal and you have re-created the fault you just fixed.
Step 4 — Only now open the fixture
If power is good, signal is good, and the joint is dry and clean, the fault is inside the unit. What you find decides whether it is repairable at all:
| What you see inside | Verdict | Why |
|---|---|---|
| Light condensation, no corrosion | Repairable | Dry it, replace the seal, re-test. The board is intact. |
| One LED bead with a black burn spot | Repairable in a workshop | Requires a matching bin and a controlled re-seal. Not a wind-and-ladder job. |
| Single colour missing (red dead, green and blue fine) | Replace the unit | One channel of the package has failed; the bead is not field-replaceable in practice. |
| Driver IC bulged or cracked | Replace the unit | On a point light the driver is integrated and potted; board-level repair costs more than the pixel. |
| PCB blackened, copper lifting, green corrosion across traces | Replace the unit | The corrosion is under the solder mask and will keep spreading. |
| Housing fully vacuum-potted | Not repairable | Opening it destroys the encapsulation. See the note below. |
Table 2. Field verdicts. Anything in the bottom three rows is a swap, not a repair.
The better a point light is sealed, the less repairable it is. A vacuum-potted unit has no air path and no serviceable joint — that is precisely why it survives ten years on a facade, and precisely why you cannot open one without destroying its IP rating. A pixel that comes apart easily in your hand is a pixel that will let water in on its own. When a supplier promises both “fully potted” and “field serviceable”, one of those claims is marketing.
Replacing a bead without wrecking the IP rating
It can be done, and it is worth describing honestly, because most guides stop at “solder a new LED in”. The soldering is the easy part.
- Match the bead, not just the part numberSame model, different production batch is a different colour. Ask for the original bin: wavelength within 2.5 nm, luminous flux in the same brightness bin, and forward voltage in the same VF bin. Without that, you are installing a visible patch.
- Control the iron350 °C ±10, contact under 3 seconds per joint. Longer than that and the phosphor and the silicone over the die take heat damage that shows up as a colour shift three months later.
- Work grounded and supportedESD strap, and the board supported so the pad does not lift. On a 12 mm pixel the pads are small and the substrate is thin; a lifted pad ends the repair.
- Replace the seal, never reuse itA compressed O-ring does not re-seal. New gasket, correct torque on the ring, and re-pot the entry if the original was potted.
- Re-test the ingress protectionAt minimum a 30-minute water spray from above and a visual check for moisture at the gland. Better, a batch check in an IP test chamber. A repair without a re-test is a guess.
Now the honest part: this sequence needs a bench, still air, a grounded mat, a temperature-controlled iron, a supply of matched beads and a way to re-test ingress. On a gondola at the 18th floor, in wind, with a pouch of tools, you have none of it. That is why the practical division of labour is: replace whole units in the air, repair beads in the workshop. A unit swap is a five-minute operation with a connector. A bead swap is a thirty-minute operation with a bench. Trying to do the second one in the first one’s environment is how a small fault becomes a leak.
The cost that never appears in the repair quotation
Ask a contractor what it costs to repair a point light and you will get the price of a pixel and an hour of labour. That number is close to meaningless, because the pixel is not what you are buying. You are buying a window in which a person can reach a facade.
Access is the dominant line item in facade maintenance, and it does not care how many pixels you fix. Published rate cards for the equipment make this concrete: a boom lift capable of reaching a mid-rise facade starts around $1,200 a day, a two-technician rope access team runs roughly $600–1,100 a day depending on market, and a building maintenance unit or gondola mobilisation is in the same order again. Add permits and the day costs about $2,000 before anyone touches a light. On tall buildings, access is routinely 40–70% of the whole maintenance budget; the work done at height is the small part.
So the only cost figure that matters is cost per pixel replaced during a single access window. And that is where repair loses, not on parts.
| Per pixel, on hired access | Repair (swap the bead) | Replace (swap the unit) | Why |
|---|---|---|---|
| Part cost | $0.10–0.30 | $0.50–2.00 | Beads are cents. A whole pixel is a dollar or two. |
| Time on the facade | 25–40 min | 5–8 min | Open, desolder, clean, resolder, re-seal, re-test vs unplug and plug. |
| Units per 5-hour window | ~9 | ~50 | 300 productive minutes, less repositioning. |
| Access cost per pixel | ~$220 | ~$40 | $2,000 day ÷ units completed. |
| Re-test needed | Yes — the seal was broken | No — factory tested | A repair without a re-test is an assumption. |
| Risk of a second visit | High | Low | Seal failures show up weeks later, off the access window. |
| Realistic total per pixel | $220–260 | $41–45 | Roughly a 5× difference, and it is all labour. |
Table 3. Illustrative model: $2,000 for a day of access, five productive hours, one technician per task. Substitute your own access rate; the ratio barely moves.
This is the arithmetic behind a rule we apply to our own projects: never send a team up for one pixel. If a facade has three dead pixels and the crew is already mobilised for annual cleaning, fixing them costs a marginal hour and it is obviously worth doing. If the only reason for the access is the three pixels, the visit costs $2,000 to avoid a $6 part, and the sensible response is to schedule it with the next campaign and replace a whole section while the crew is there.
When access is already paid for. A marginal repair inside an existing window is cheap. When the unit is genuinely expensive. On a $120 high-output floodlight or a linear wall-grazer, board-level repair can beat replacement — but that is a different product class from a 12 mm pixel. When the failure is a connector. A joint that needs cleaning and a new O-ring is a ten-minute job, not a repair, and it is the one case where “fix” clearly wins. When the run is under warranty. Then the manufacturer carries the replacement, and your only cost is access.
And the repaired pixel will not match the ones beside it
This is the part that surprises people after the second visit. The pixel lights up, the colour is roughly right, and the facade still looks wrong — because the new pixel is brighter and slightly cooler than the pixels around it, and the eye is very good at finding a single bright dot in a dark grid.
Three physical effects stack up, and only one of them can be compensated in the controller.
1. Lumen depreciation has already happened to the neighbours
A new pixel starts at its rated flux. The pixels around it have been running for two to three years, which on a nightly facade cycle is 15,000–25,000 hours, and have lost 10–20% of their output. Under a white or single-colour test pattern, a 10% step is visible to an untrained eye at a few metres; against a dark facade it is visible from the street.
2. Binning is tighter than suppliers admit
LEDs are sorted after production, and the sorting is coarse relative to human perception. Wavelength is typically binned in steps of about 2.5 nm, flux in brightness bands, and forward voltage in VF bands. Two beads with the same part number but different bins can differ by a few nanometres and a couple of hundred kelvin. Individually that is invisible. In a grid showing one flat colour, the boundary between bins is exactly where the eye looks.
3. VF mismatch changes the current, not just the colour
On a constant-current driver, a bead with a lower forward voltage draws more of the shared current, so it runs brighter — and hotter, which accelerates its own depreciation. Original production holds VF within roughly ±0.05 V; mixed aftermarket stock can be ±0.3 V. That is not a cosmetic difference. It is a brightness difference that grows.
| What you can do | Effect | Practical note |
|---|---|---|
| Use spares from the original bin | Large | The only real fix. It has to be planned at purchase, because the bin is gone two years later. |
| Replace a whole section, not a dot | Large | Put the boundary on a structural joint or a shadow line so the step reads as a design edge. |
| Trim brightness of the new pixel in the controller | Partial | Works on DMX512 (per-address dimming). On SPI it means rewriting the pixel map. Brightness only — it does not correct colour temperature. |
| Re-calibrate the run | Partial | Needs a camera-based calibration setup and a night session. Justified on large media surfaces, not on three pixels. |
| Do nothing and accept the patch | — | Honest option on a facade that is viewed from 50 m. Not honest on a lobby ceiling at 3 m. |
Table 4. There is no field trick that restores colour match. The mismatch is a property of time, and the only real answer is stock from the original batch.
This is why we ask clients to hold 3–5% spares from the original production run, packed with the bin and batch record. It is a small line on the purchase order and it is the difference between a facade that ages evenly and a facade that looks like a patchwork after year three. A pixel you cannot match is, in practice, a pixel you cannot repair — you can only replace it with something that is visibly different.
Common mould or private mould — why the call ends up with the original factory
Here is the step that decides whether a repair is even possible, and it has nothing to do with electronics. Two point lights can look identical in a photograph, take the same 12 mm pixel, run the same protocol, and be completely different in one respect: whether the housing is a common mould that half the industry uses, or a private mould cut for one customer.
With a common mould, the shell, the thread pitch, the cable entry and the connector footprint are standard parts. Any number of factories can supply a compatible unit, price is set by competition, and a spare arrives in a week. With a private mould, the housing exists in one place. When you need a replacement three years later, there is exactly one party who can make one that fits the hole in the facade.
And no factory will open a new mould for a repair order, because the arithmetic does not allow it. Tooling for a small injection-moulded housing runs $8,000–35,000 for a fully custom mould, against $1,500–5,000 for an insert in a standard frame; aluminium extrusion dies start around $2,000 and climb with section complexity; a precision optical lens tool in hardened steel can reach six figures. Add T1–T3 trials, gating corrections and surface texturing, and the engineering overhead is another fifth on top.
| Common mould (open tool) | Private mould (custom tool) | |
|---|---|---|
| Who can supply a spare | Any factory running that tool — several, normally | One. The owner of the tool. |
| Tooling cost to you | Zero, amortised across the market years ago | $8,000–35,000+ for injection; more for die-cast |
| Spare lead time | Days to weeks, from stock | Whatever the factory’s schedule allows — and only if the tool still exists |
| Unit price at volume | Low, competitive | Higher, and set by one seller once you are locked in |
| Visual differentiation | None — your product looks like everyone’s | Full control of the silhouette, bezel and fixing detail |
| Risk at year 5 | Interchangeable parts, easy to re-source | Tool scrapped, factory changed hands, or the model discontinued |
| Right for | Runs where serviceability and TCO matter more than form | Landmark facades where the fixing detail is part of the architecture |
Table 5. The private mould is not a mistake — it is a ten-year commitment. It should be signed as one.
Put a number on it. A maintenance order for 200 replacement pixels against a private housing needs a tool that costs, say, $15,000 — which is $75 per pixel before a single part is moulded, and that assumes the factory is willing to run a 200-piece batch at all. Set that against a common-mould pixel at $2. The mould, not the LED, is why the conversation goes back to whoever supplied the original facade.
A facade installed with private tooling in 2016 needs 40 replacement pixels in 2026. The tool was scrapped when the model was discontinued in 2021. The factory will quote a new tool at $18,000 with a 10-week lead time, or offer a “similar” housing that does not fit the existing bracket. The client now faces re-drilling 4,000 fixing points, or accepting a visible mismatch on one elevation. None of this is a lighting failure. It is a procurement decision that was never written down.
If you are specifying private tooling — and sometimes you should, because the fixing detail is visible architecture — then the contract has to carry four clauses that cost nothing to agree and everything to omit:
- Tool ownership and custodyWho owns the mould, where it is stored, and what happens to it if the relationship ends. Ownership is negotiable; the absence of a clause is not.
- Spare supply horizonA stated number of years of spare-part availability after the last production run — seven to ten is a reasonable ask for a facade with a 15-year design life.
- Bin and batch retentionThe right to buy spares from the same LED bin and the same driver batch, not merely the same model number.
- Spare MOQ and lead timeWritten down. A factory that will not commit to a 50-piece spare order in year six has told you something important in year one.
What this means before you buy, not after
Almost everything above is decided at purchase and discovered at failure. That is the frustrating part of facade maintenance: the person who signs the purchase order and the person who books the gondola are usually three years and two departments apart. Six questions close most of the gap, and they fit in an RFQ:
| Question to put in the RFQ | Why it matters at year 3 |
|---|---|
| Is the housing a common mould or private tooling, and who owns the tool? | Decides whether you can buy a spare from anyone else. |
| How many years of spare supply do you guarantee after the last shipment? | Converts an open-ended risk into a documented one. |
| Can you supply spares from the same LED bin and driver batch? | Decides whether a replacement is invisible or a patch. |
| What is the spare MOQ and lead time? | A 500-piece minimum makes small repairs uneconomic. |
| Is the control protocol DMX512 or Art-Net, or proprietary? | Standard protocols let a third party integrate; proprietary ones do not. |
| What is the IP rating of the connector, not just the fixture? | The connector is where most outdoor failures start. |
Table 6. Six questions, no cost to ask, and they determine the maintenance cost of the building for a decade.
There is a design decision hidden in the last two rows as well. Fixtures that can be reached from a building maintenance unit, or that sit within a few metres of a structural edge, cost 30–50% less to maintain over their life than fixtures that require independent rope access. That is decided on the architect’s section drawing, long before a pixel fails.
How we build for this, since we are the ones who get the call
We have been making outdoor point lights and facade systems since 2004 — 22 years, 5,000+ projects across 40+ countries — from our own 8,000 m² plant in Guzhen, and the maintenance calls we get have shaped the product more than any specification sheet did. Three things changed as a result, and they are worth knowing when you compare suppliers.
Sealing is done in-house and tested in batches. Potting and gasketing are only as good as the process control around them, so the vacuum potting and the IP test chamber sit in the same building as the assembly line. Every batch is checked for ingress at IP67, not just the sample that went to the lab.
Bins are recorded, not just used. When a project ships, the LED bin and driver batch go on the record with the order, so a spare order in year four can be matched instead of approximated. This is the single most useful thing a factory can do for a facade, and it costs nothing but discipline.
Spares are quoted with the project, not after it. On private-tooling projects we recommend 3–5% spares from the first production run, packed and stored with the batch record. Clients sometimes strike it out to save a line on the PO. It is the cheapest insurance in the whole package.
None of this makes a pixel repairable. It makes the repair decision cheaper, because the part you need exists, matches, and arrives before the access window closes.
Questions we get asked about point light repair
Can a single LED point light be repaired on site?
Usually not at board level, and it is rarely worth trying. A unit swap — unplug the connector, fit a matching spare, plug back in — takes five to eight minutes and is a normal field task. Replacing an individual LED bead needs a bench, a temperature-controlled iron, a matched bin and a way to re-test the seal, none of which exist on a gondola. The practical split is: units in the air, beads in the workshop.
How much does it cost to repair one LED point light?
The part is trivial: a 12 mm RGB pixel sells for roughly $0.50–2.00, and a bare bead is cents. The cost is access. With a boom lift or rope access team at around $2,000 for a working day, a bead repair that takes 30 minutes amortises to roughly $220 per pixel, against about $40 for a unit swap completed in six. If the crew is already on site for cleaning or another task, the marginal cost collapses to labour only — which is the one situation where repair clearly wins.
Why do LED point lights fail after rain?
Because the water is usually in the connector, not the fixture. Rain does not create a fault; it reveals one that was already there — an O-ring that hardened under UV, a locking ring left a quarter-turn short, a cable pulled tight so the joint sits under tension, or a joint mounted face-up where it collects water. Moisture reaches the pins, corrosion starts, contact resistance rises, and the pixel sees a voltage sag. It dries, works for a week, and fails at the next storm. Replacing the connector and re-seating the seal fixes it; replacing the pixel does not.
Can I replace one pixel with a different brand or model?
Not on the same run, and usually not even on the same facade. Three things have to match: the control protocol and data rate, the supply voltage and current, and the physical fit in the existing bracket. Even when all three match, the LED bin will not — a different brand means a different wavelength band, a different brightness band and a different forward voltage, and the pixel becomes a visible patch. If you must mix, isolate the new units into a complete section rather than scattering them through the grid.
How long do LED point lights actually last?
Longer than the electronics around them, in most cases. The LED package itself is rated 50,000 hours or more, which on a five-hour nightly cycle is over twenty years of calendar life, but the integrated driver is typically the shorter-lived element and the connector is shorter-lived again. In practice the first failures on an outdoor facade appear at year three to five, and they are dominated by joints and drivers rather than by LED depreciation. A facade is usually re-lamped for appearance reasons — colour consistency across the surface — long before the LEDs reach end of life.
Is it worth repairing a six-year-old facade installation?
Judge it by failure density, not by age. Under about 5% of pixels failed, replace those units and keep going — the rest of the surface still has years left. Between 5% and 15%, replace whole sections so the colour boundary lands on a structural joint. Above about 15%, the population is ageing together and you are chasing failures with access days; that is the point where re-lamping the elevation is the cheaper decision, even though the LEDs are technically still working.
What is the difference between a common mould and a private mould?
A common mould is a housing tool shared across the market, so several factories can supply a compatible unit and spares are cheap and quick. A private mould is cut for one customer or one project, at a tooling cost of roughly $8,000–35,000 for an injection-moulded housing. That cost has to be recovered over a production run, not over a repair order, which is why a private-mould facade can only be maintained by the factory that owns the tool. It is a durability decision as much as an aesthetic one.
Send us the failure pattern and we will tell you which layer it is
If you are maintaining a facade — ours or someone else’s — we can usually identify the layer from three details: how many pixels, whether they are adjacent, and whether the fault follows rain. Send a photo of the grid and a photo of one connector, and you will get an answer rather than a quotation.
- Matching spares from the original bin and batch, where the record exists
- Common-mould equivalents when the original tool is gone
- Honest advice on whether to replace a unit, a section, or the elevation
Basis for the figures in this article. Access and equipment rates are drawn from published contractor and equipment-hire rate cards for rope access, boom lift and BMU mobilisation in European and Gulf markets; tooling ranges from published injection-mould and die costs for lighting housings; pixel pricing from current export quotations for 12–50 mm DMX/SPI point lights. Performance figures — lumen maintenance, binning intervals, VF tolerance, driver and LED rated life — follow standard LED industry practice and our own production records. Costs are indicative and vary by market; the ratios are what matter.
Project enquiry
Send your drawings. Get a specification you can defend.
Elevation DWG, facade area, closest viewing distance and site country — we reply within 48 hours with a fixture matrix, power and DMX topology, and a quotation. No price-only quotes.
5,000+ projects delivered · 40+ countries · ISO 9001 · CE / UL / RoHS