HCL Retrofit · Thermal Engineering · ANSI/IES TM-21
Thermal Management
for HCL Retrofits
Heat kills LEDs. Not immediately — slowly, predictably, and in a way that is entirely calculable before you install a single fixture. Junction temperature is the single variable that determines whether your retrofit lasts 12,000 hours or 60,000. This guide gives you the complete engineering methodology: how to calculate temperature at every node in the thermal chain, how to choose the right board and interface material, how to verify your work in the field, and how to project rated L70 life using the ANSI/IES TM-21 standard rather than manufacturer spec sheets.
Why Junction Temperature Controls Everything — The Physics of LED Degradation
LED degradation is not random wear. It follows predictable chemical kinetics — the same Arrhenius rate law that governs every thermally-activated chemical reaction. Once you understand this, the entire thermal engineering problem becomes tractable: measure the temperature, look up the degradation rate, calculate the life.
Inside every LED package, the active region — the semiconductor junction — is where electrons recombine with holes to produce photons. This process is not perfectly efficient. Some of the electrical energy is released as heat rather than light, and that heat accumulates at the junction. The temperature at this point — called the junction temperature Tj — is the primary variable governing how quickly the LED degrades.
The degradation mechanism is primarily chemical: at elevated temperatures, crystalline defects form in the III-V semiconductor lattice, point defects migrate to the active region and act as non-radiative recombination centres (they capture electron-hole pairs and release heat instead of photons), and the phosphor binder (if present) undergoes thermal yellowing. All of these processes follow the Arrhenius equation — their rate doubles for approximately every 10°C increase in temperature. This is not an approximation or a rule of thumb. It is the fundamental physical chemistry of thermally-activated processes.
The practical consequence: an LED running at 75°C junction temperature degrades roughly 4× faster than the same LED running at 55°C. A source rated for 50,000 hours at 55°C will reach its L70 endpoint at approximately 12,500 hours at 75°C. Same LED, same driver, same photons — different temperature, completely different lifespan.
This is why the choice of housing matters enormously for the HCL retrofit. A die-cast aluminium Portfolio housing with its large thermal mass and surface area keeps junction temperatures low. A cheap plastic IoT fixture with minimal heatsinking runs hot. The housing does not just provide the structural platform — it is an active thermal component that directly determines system life. We cover the specific housing data in the housing-as-heatsink section below.
Most manufacturer LED datasheets publish an L70 rating — typically "50,000 hours at Tc = 65°C" or similar. What this means: the LED was tested on an LM-80 test rig at a controlled solder-point temperature of 65°C, and the TM-21 extrapolation projects L70 at 50,000 hours under those conditions. If your installation runs the LED at a different solder-point temperature — and most field installations do — the published spec does not apply directly. You must recalculate, which is exactly what this guide teaches.
A common misconception: "If I use a quality constant-current driver, I can make any fixture last 50,000 hours." False. The driver controls current, which controls power input and therefore heat generation. But the housing controls where that heat goes. A quality driver in a plastic housing that can't shed heat still runs the junction hot. You need both: a good driver to control current, and a good housing to dissipate the resulting heat. Die-cast aluminium housings solve the heat side. The driver solves the current side. Neither alone is sufficient. See Why Legacy Metal Beats New Plastic for the thermal resistance comparison.
The Thermal Chain — Four Nodes in Series, Each One Measurable
Heat flows from the LED junction to the surrounding air through a series of thermal resistances, exactly like electrical current flows through resistors in series. Each node has a resistance (in °C/W) and a temperature. Understanding the chain lets you calculate the temperature at every point — and identify where to intervene.
The thermal path in a retrofitted LED fixture has four distinct nodes, each with its own resistance:
Node 1: Junction to Solder Point (RθJS)
This is internal to the LED package — the resistance from the semiconductor junction to the solder pad on the bottom of the package. It is determined by the LED package design and reported in the LED datasheet. For MR16-format emitter packages, typical values range from 3–8°C/W. You cannot change this — it is a material property of the package you choose.
Node 2: Solder Point to Board (RθSB)
This is the interface between the LED package solder pad and the metal-core PCB (MCPCB) it's mounted on. It includes the solder joint itself and the thermal path through the MCPCB's copper layer. Typical values: 0.5–2°C/W with good solder joint, thermally conductive adhesive, or proper reflow soldering. This is the node most affected by workmanship — a cold solder joint or improper adhesive here can add 5–10°C to junction temperature.
Node 3: Board to Housing (RθBH)
This is the interface between the MCPCB mounting surface and the die-cast aluminium housing. It depends on the thermal interface material (TIM) used — thermal compound, graphite pad, or phase-change material. Typical values: 0.3–2°C/W with quality TIM applied correctly, up to 5°C/W with poor contact or air gaps. This is the node most improved by attention to TIM selection and application.
Node 4: Housing to Ambient (RθHA)
This is the largest resistance in the chain for most small fixtures — the thermal resistance from the outside surface of the housing to the surrounding air. It depends on housing size, surface area, surface finish, orientation, and air movement. For die-cast aluminium Portfolio housings, typical values are 2–5°C/W. This is the node that differentiates metal from plastic — plastic housings often have RθHA of 8–15°C/W, three to five times higher.
Total thermal resistance: RθJA = RθJS + RθSB + RθBH + RθHA
Junction temperature is then: Tj = Tambient + (Pdissipated × RθJA)
Calculating Tj — Three Worked Examples You Can Apply Directly
Abstract formulas only become useful when you can apply them to real components. Here are three worked examples covering a good installation, a marginal one, and a failing one — each with the specific resistance values, calculation chain, resulting junction temperature, and projected L70 life.
More thermal compound is not better. Thermal interface materials work by filling microscopic air gaps between surfaces — not by adding bulk thermal material. The optimal bond-line thickness (BLT) for most silicone-based TIMs is 0.05–0.15mm. Squeezing a thick blob onto the MCPCB, then bolting the fixture shut, creates a thick bond line with high thermal resistance. A thin, even layer — just enough to fill surface roughness — is optimal. Excess TIM on Node 3 can add 3–5°C to junction temperature — equivalent to running 3–5W more power than specified. Apply it like you're icing a cake with a very thin frosting layer, not spreading peanut butter.
MCPCB Selection — The Board That Moves Heat Away from the Junction
The circuit board your LED mounts on is not just a structural substrate — it is an active thermal component. Standard FR4 fibreglass PCB material is a thermal insulator. In LED applications, using FR4 between the emitter and the heatsink dramatically increases RθSB and defeats the purpose of a good heatsink.
A Metal-Core PCB (MCPCB) is built around an aluminium or copper core layer rather than the fibreglass weave of FR4. The core provides a low-resistance thermal path from the LED solder pad through the board to the mounting surface. The typical structure is: copper circuit layer (35–140µm) → dielectric insulator layer (50–150µm) → aluminium or copper core (1–3mm). The dielectric is the thermal bottleneck — it must be thin enough for good heat conduction but thick enough for electrical isolation.
Why not use FR4? FR4 has a thermal conductivity of approximately 0.25–0.35 W/m·K. A standard MCPCB dielectric achieves 1–3 W/m·K, and high-performance MCPCB dielectrics reach 3–8 W/m·K. This means the MCPCB transfers heat through its dielectric at 3–25× the rate of FR4. At 3W drive power, this difference translates directly into junction temperature: an FR4 board may add 5–8°C to RθSB; a quality MCPCB adds 0.5–1.5°C.
Copper vs. aluminium core: Copper has higher thermal conductivity (~390 W/m·K vs. ~205 W/m·K for aluminium) but is heavier and more expensive. For most retrofit applications up to 5W, aluminium-core MCPCBs with high-conductivity dielectric are the correct specification — they achieve adequate thermal performance at lower cost and weight. Copper-core MCPCBs are warranted for higher-power applications or situations with particularly constrained heatsink area.
The "thermal vias" shortcut: Some cheaper LED modules use FR4 boards with thermal vias — copper-filled holes through the board intended to conduct heat. At low power, these can work adequately. At 3–5W sustained drive power, they are insufficient — the effective thermal resistance through a via array is still several times higher than a proper MCPCB dielectric. Do not use FR4 with thermal vias as a MCPCB substitute for HCL retrofit applications.
| Board Type | Dielectric k (W/m·K) | R_θSB at 3W | HCL Retrofit Use |
|---|---|---|---|
| Standard FR4 | 0.25–0.35 | 5–8°C/W typical | Never |
| FR4 with thermal vias | Effective ~0.8–1.2 | 2–4°C/W typical | Avoid |
| Al-core MCPCB, standard dielectric (1 W/m·K) | 1.0 | 1.5–2.5°C/W | Marginal at >3W |
| Al-core MCPCB, high-k dielectric (2–3 W/m·K) | 2.0–3.0 | 0.8–1.5°C/W | ✓ Recommended |
| Al-core MCPCB, premium dielectric (4–6 W/m·K) | 4.0–6.0 | 0.4–0.8°C/W | ✓ Best for >5W |
| Cu-core MCPCB, high-k dielectric | 3.0–5.0 | 0.3–0.6°C/W | ✓ High-power spec |
When ordering MCPCBs for a retrofit application, ask for: (1) Dielectric thermal conductivity in W/m·K — minimum 2.0 for LED retrofit applications. (2) Dielectric thickness — thinner is better thermally, but minimum ~75µm for safe electrical isolation at 12V. (3) Copper weight of the circuit layer — 35µm (1oz) is standard; 70µm (2oz) reduces trace resistance for higher currents. (4) Board dimensions and hole pattern matching your emitter footprint. Never order an MCPCB without the dielectric k value specified — "LED board" without a thermal specification is not a specification at all.
Thermal Interface Materials — Choosing and Applying the Layer That Joins Board to Housing
Node 3 of the thermal chain — between the MCPCB and the die-cast housing — is where most retrofit installers make their biggest mistake. The thermal interface material (TIM) fills microscopic surface roughness between two metal surfaces. Choose the wrong material, apply it incorrectly, or skip it entirely, and you can add 5–15°C to junction temperature.
Step 1: Clean both surfaces (MCPCB back and housing mating surface) with isopropyl alcohol — remove all oil, flux, and previous TIM residue. Allow to dry completely.
Step 2: Apply a small amount of compound — approximately the size of a grain of rice for an MR16-format MCPCB. Place it at the centre of the board.
Step 3: Press the MCPCB against the housing surface with even pressure and a slight twisting motion to spread the compound. The compound should spread to cover the full contact area with a visible but very thin layer — you should almost be able to see the metal surface texture through it.
Step 4: Fasten with the specified fasteners to the specified torque. Over-torquing squeezes the compound out entirely. Under-torquing leaves gaps. The optimum is just enough pressure to maintain full contact — consult the TIM manufacturer's recommended closure pressure, typically 40–150 psi for silicone compounds.
What you should NOT see: Compound oozing out significantly around all edges means you applied too much. A clear outline of the MCPCB with no compound visible means you applied too little or are missing contact in the centre.
The Housing as Heatsink — What Portfolio Die-Cast Aluminium Actually Delivers
The fourth and largest thermal resistance node is the housing-to-ambient interface. For die-cast aluminium fixtures, this is determined by surface area, surface finish, wall thickness, and orientation. Here is measured RθHA data for common Portfolio fixture formats — the numbers that go into the calculation.
| Fixture Format | Housing Mass | R_θHA (vertical) | R_θHA (horizontal) | Max T_j at 3W |
|---|---|---|---|---|
| PAR16 / MR16 spot, medium body | 280–380g | 3.2–3.8°C/W | 3.8–4.5°C/W | ~51–57°C ✓ |
| PAR16 / MR16 spot, large body | 380–550g | 2.5–3.0°C/W | 3.0–3.5°C/W | ~48–53°C ✓ |
| Path / bullet light, compact | 150–240g | 4.5–6.0°C/W | 5.0–7.0°C/W | ~58–68°C ⚠ |
| Well light / in-ground, medium | 400–600g | 2.8–3.5°C/W | N/A (fixed) | ~50–56°C ✓ |
| Flood / wide-beam, large housing | 500–800g | 2.0–2.8°C/W | 2.4–3.2°C/W | ~44–51°C ✓ |
| Decorative mini spot, small body | <120g | 6.5–9.0°C/W | 8.0–11.0°C/W | >70°C ✕ Limit to <2W |
R_θHA values measured by thermocouple on housing exterior surface relative to still-air ambient at 25°C. T_j estimates assume quality MCPCB (R_θSB = 1.0°C/W), quality TIM (R_θBH = 0.8°C/W), datasheet R_θJS = 6°C/W, 3W drive at 72% efficiency (0.84W dissipated). Add ~5–8°C for outdoor summer installations. Natural convection — add airflow correction factor of 0.6–0.8× for fixtures with active air movement.
Several practical implications arise from this data:
Orientation matters significantly. A fixture mounted with the heatsink body horizontal (lamp pointing up or down) loses 15–25% of natural convection effectiveness compared to vertical orientation. If mounting orientation is constrained, use the horizontal RθHA value in your calculation — not the vertical.
Small compact fixtures have physical limits. A decorative mini-spot with less than 120g of aluminium cannot dissipate more than about 2W at safe junction temperatures in still outdoor air. Do not attempt to drive these fixtures at 3W or above for HCL retrofit applications — the housing mass is simply insufficient. Use a larger housing or reduce drive current.
Soil contact helps well lights. In-ground well lights benefit from soil thermal conduction — the ground acts as a supplemental heatsink with seasonal temperature variation but significant thermal mass. In summer with dry soil, assume still-air performance. In spring with moist soil, RθHA can be 20–35% lower.
Surface finish matters less than mass. Anodising or painting a housing in flat black does improve emissivity and slightly reduces RθHA — but the effect is secondary to housing mass and surface area. Don't repaint fixtures expecting a significant thermal improvement; the effect is 5–15% at most in typical residential applications.
A comparable-size cheap plastic IoT smart spotlight has a typical R_θHA of 8–15°C/W. A Portfolio medium PAR16 housing runs at 3.2–3.8°C/W. That 4–5× difference in heatsink resistance is worth approximately 15–25°C of junction temperature at 3W drive power. That temperature difference translates — via the Arrhenius rule — to a factor of 4–8× in system life. The metal housing is not a nostalgic preference. It is the primary thermal asset of the retrofit.
TM-21 L70 Life Projection — The Correct Method vs. the Spec Sheet Number
Every LED manufacturer publishes an L70 rated life. Almost none of those numbers apply directly to your field installation. TM-21 gives you the methodology to project L70 at your actual operating conditions — which can be better or worse than the datasheet number, depending on how well your thermal design performs.
| T_j (°C) | Representative α (hr⁻¹) | t_L70 (hours) | Calendar life (8hr/day) |
|---|---|---|---|
| 45 | ~4.5 × 10⁻⁶ | ~79,000 | ~27 years ✓ |
| 50 | ~5.5 × 10⁻⁶ | ~65,000 | ~22 years ✓ |
| 55 | ~7.0 × 10⁻⁶ | ~51,000 | ~17 years ✓ |
| 60 | ~8.5 × 10⁻⁶ | ~42,000 | ~14 years — LumeCircadian minimum target |
| 65 | ~1.1 × 10⁻⁵ | ~32,000 | ~11 years |
| 70 | ~1.4 × 10⁻⁵ | ~25,000 | ~8.6 years |
| 75 | ~1.8 × 10⁻⁵ | ~20,000 | ~6.8 years ⚠ |
| 85 | ~3.0 × 10⁻⁵ | ~12,000 | ~4.1 years ✕ |
| 95 | ~5.0 × 10⁻⁵ | ~7,000 | ~2.4 years ✕ Typical smart bulb |
TM-21 life projection requires LM-80 lumen maintenance test data at the solder-point temperature closest to your operating T_s. If the emitter manufacturer cannot provide LM-80 data at T_sp ≥ 55°C for at least 6,000 hours, you cannot perform a valid TM-21 projection. The manufacturer's published L70 figure in this case is unverified — it may be a calculation estimate, an extrapolation from a different emitter family, or marketing content. Do not accept an L70 rating without LM-80 data to support it. See the emitter selection checklist for the full qualification criteria.
Field Verification — Measuring the Temperature You Built
After installation, you need to verify that your calculated junction temperature matches reality. You cannot measure Tj directly in the field — it is inside the LED package. But you can measure Thousing and back-calculate Tj using the known thermal resistance chain. Here is the exact protocol.
Junction temperature takes time to reach thermal equilibrium — typically 20–45 minutes depending on housing mass. Shorter measurement windows will give falsely optimistic readings. Always run the fixture at full rated drive current for at least 60 minutes before taking any measurements. For outdoor installations, do the steady-state measurement in the worst-case ambient condition — mid-afternoon in summer, not morning in spring.
Minimum runtime before measurement: 60 minutes at rated drive current in worst-case ambientUse a Type K thermocouple thermometer (accuracy ±1–2°C) or a calibrated infrared thermometer (accuracy ±2–3°C, ensure emissivity is set correctly for the surface finish — aluminium with natural oxide finish, ε ≈ 0.05–0.15 requires high-emissivity paint or tape for accurate IR readings). Measure the housing surface closest to the MCPCB mounting location — this is typically the upper body of the fixture or the area directly behind the emitter board.
Tools: Type K thermocouple (preferred) or calibrated IR thermometer with emissivity correction for aluminiumUsing the measured housing temperature Thousing and the known thermal resistances for Nodes 1–3, calculate Tj by adding the temperature rise through each node: Tj = Thousing + (Pdissipated × (RθJS + RθSB + RθBH)). This approach bypasses the uncertainty in RθHA by measuring Thousing directly rather than estimating it from ambient temperature plus RθHA.
Formula: T_j = T_housing + (P_diss × (R_θJS + R_θSB + R_θBH)) · R_θHA not needed when T_housing is measured directlyIf calculated Tj exceeds 60°C, the system is outside the LumeCircadian target. The most controllable adjustment is drive current — reducing drive current by 20% typically reduces Pdissipated by 20% and reduces junction temperature by the proportional amount. Secondary adjustments: verify TIM is applied correctly (re-open fixture and check bond line), verify MCPCB is fully in contact with housing, verify ambient measurement conditions. Do not accept an installation that calculates above 65°C Tj as a 50,000-hour system.
Pass criteria: T_j ≤ 60°C at worst-case ambient and full rated drive current. Document measured values for installation record.For every installation, record: fixture model and serial, emitter type and lot number, drive current, measured housing temperature, calculated Tj, ambient temperature at time of measurement, and date. This commissioning record serves two purposes: it is the basis for projecting actual L70 life under your specific conditions, and it provides a baseline for future maintenance — if the fixture shows early lumen depreciation, comparison against commissioning data will quickly identify whether it is a thermal problem or an emitter degradation issue.
Document: fixture ID, emitter lot, drive current, T_housing measured, T_j calculated, T_amb, measurement dateBare or anodised aluminium surfaces have very low emissivity (ε = 0.05–0.30 depending on finish). An infrared thermometer set to ε = 1.0 (the default for most consumer IR guns) will read 20–30°C lower than actual temperature on polished aluminium. This causes installers to incorrectly conclude the fixture is running cool. To get accurate readings from aluminium: either use a thermocouple in physical contact with the surface (no emissivity problem), or apply a small piece of matte black tape to the measurement point (ε ≈ 0.95) and measure the tape temperature. The tape reaches thermal equilibrium with the housing within 2–3 minutes and gives an accurate reading at ε = 0.95.
Common Thermal Failures — How to Diagnose and Correct Them
LED HCL retrofit systems fail thermally in predictable ways. Recognising the symptoms, identifying the root cause, and applying the correct fix is faster and cheaper than replacing components at random.
| Symptom | Likely Root Cause | Diagnostic Test | Correction |
|---|---|---|---|
| Rapid lumen depreciation — noticeable dimming within first 3,000–5,000 hours | Junction temperature too high — likely R_θBH or R_θHA problem | Measure T_housing at rated load after 60 min. Back-calculate T_j. If >70°C, thermal issue confirmed. | Re-open fixture. Replace TIM with quality compound, applied correctly. If still hot, reduce drive current 20–30%. |
| Colour shift toward red/orange over time in phosphor-converted sources | Phosphor thermal degradation — sustained T_j >80°C causing Ce:YAG yellowing | Spectral measurement of current output vs. commissioning SPD. Red shift confirms phosphor degradation. | Reduce drive current. Improve TIM. For InGaAlP amber emitters, colour shift is not expected — investigate if seen. |
| Premature open-circuit failure — LED dies completely | Thermal cycling fatigue at solder joints — rapid on/off cycles at high T_j cause solder crack propagation | Inspect solder joints under magnification for microcracks at pad edges. Check for cracked MCPCB-to-housing contact. | Reflow solder joints with quality lead-free solder. Avoid frequent rapid cycling. Consider larger MCPCB contact area. |
| Housing too hot to touch comfortably (>55°C surface) | Normal operation for high-power install, OR excessive drive current, OR poor TIM | 55°C surface = acceptable. 65°C+ surface = concern. Measure and back-calculate T_j. If T_j >70°C, intervene. | Do not assume "hot = bad" without calculation. But if T_j >70°C, reduce current or improve TIM before accepting. |
| System runs cool but L70 spec still not met — early failure despite good T_j | Emitter lot quality issue, or actual T_j higher than calculated (TIM gap or bad solder) | Open fixture and inspect TIM bond line for gaps. Verify solder joint under magnification. Re-measure T_housing. | Internal inspection usually reveals either TIM spreading fault or emitter quality below spec. Replace emitter lot. |
| High-frequency thermal oscillation — light flickers with thermal cycling | Driver thermal protection activating — driver or LED running hot enough to trigger auto-dimming | Check driver case temperature — should be <70°C. If driver in enclosed cavity with LED, both overheat. | Separate driver from LED thermal path. Mount driver outside the main fixture body if cavity is too confined. |
In compact fixture retrofits, the constant-current driver is often installed in the same wiring cavity as the LED emitter board. At steady state, the driver dissipates its own heat (1 − efficiency × power) inside the cavity. This raises the cavity air temperature above ambient — which raises the housing surface temperature — which raises Tj. A 90% efficient driver at 3W dissipates 0.33W of heat. In a small enclosed cavity, this can raise air temperature 8–15°C above ambient, adding the same amount to the effective ambient temperature for the LED calculation. Always mount the driver outside the main thermal mass of the fixture, or account for driver self-heating in your ambient temperature estimate. The best practice for compact retrofit is to route the driver to the transformer junction box or a separate mounting location, connected to the LED by a short 12V run.