Modul LED COB (Chip-on-Board) pentru Iluminat

CRI, R9 și Calitatea Spectrală: Ce Nu Vă Spune Fișa Tehnică

Color Rendering Index Ra is a weighted average of color fidelity across eight test color samples (R1–R8). The problem is that R9 — saturated red — is excluded from the Ra calculation entirely. A COB LED module can report Ra≥90 on the datasheet and still render saturated reds very poorly if R9 is below 20.

This matters in practice for retail display lighting, hospitality, museum cases, and horticulture. Under a fixture with Ra 90 but R9 of 15, a red product appears brownish, cooked meat looks grey, and red-spectrum photosynthesis efficiency in a grow light is compromised. For these applications, Ra≥90 with R9≥50 is the minimum workable threshold. For high-end museum or gallery applications, specify Ra≥95 with R9≥70.

What to request from a Chinese COB factory:

Ask for the IES LM-79 test report for the specific bin and CCT you are buying, not a generic datasheet figure. LM-79 is a photometric test of the complete module, measured by an accredited laboratory with a calibrated integrating sphere. The report will include a spectral power distribution (SPD) graph — the shape of this curve tells you more about color quality than any single Ra number. A narrow-spike SPD (typical of phosphor-converted blue-die LEDs optimized purely for efficacy) will show a red-channel depression that translates directly to low R9.

TM-30-20 is a more complete metric than Ra. TM-30 reports two numbers: Rf (fidelity index, equivalent in intent to Ra but calculated across 99 test color samples instead of 8) and Rg (gamut index, where 100 = neutral, >100 = saturated, <100 = desaturated). A COB with Rf 88 and Rg 98 renders color accurately and neutrally. A COB with Rf 88 and Rg 105 renders color accurately but with slight over-saturation — sometimes preferred for retail. Some Chinese manufacturers now include TM-30 data in their premium CRI datasheets; if it is absent, ask. If the factory cannot produce TM-30 data, that is a signal about their measurement capability.

A common pattern in Chinese COB datasheets: Ra is listed prominently; R9 is absent; LM-79 reports are available only on request and sometimes only from the domestic factory’s own integrating sphere rather than an accredited third-party lab. For our sourcing service engagements, we require accredited LM-79 data (CNAS or NVLAP lab) before recommending a module for high-CRI applications.

Managementul Termic și Designul Radiatorului

COB modules are high-flux-density light sources. A 100W COB with 150 lm/W efficacy converts 33W of input power into heat — all conducted through a ceramic or aluminum substrate roughly 38–43mm in diameter. Getting that heat out of the junction is the primary engineering challenge.

Thermal resistance θjc (junction to case) for COB modules typically ranges from 0.5°C/W (large-footprint, high-die-count modules on alumina substrates) to 2.0°C/W (small footprint, single-die concentrations). The datasheet value for θjc assumes a perfectly flat, thermally conductive mating surface. Real-world heatsink contact is never perfect.

Junction temperature calculation:

Tj = Tc + (Pth × θjc)

Where Tc is the measured case (substrate back) temperature, Pth is the thermal dissipation in watts, and θjc is the junction-to-case thermal resistance. For a 50W COB with 150 lm/W efficacy (Pth ≈ 17W) and θjc of 1.2°C/W running at Tc = 60°C:

Tj = 60 + (17 × 1.2) = 60 + 20.4 = 80.4°C

This is well within the 125°C Tj maximum. But if the heatsink contact is poor and Tc reaches 90°C at the same power, Tj hits 110.4°C — still within spec but shortening L70 lifetime significantly. Every 10°C increase in Tj above 60°C approximately halves lumen maintenance life.

Heatsink contact surface requirements:

• Direct metal-to-metal contact (no TIM): heatsink mating surface flatness ≤25µm Ra (roughness average). Most machined aluminum heatsinks achieve 0.8–3.2µm Ra — adequate. Die-cast heatsinks without secondary machining typically run 6–12µm Ra — borderline.
• Thermal interface pad or grease: contact surface flatness ≤50µm Ra. Phase-change pads (e.g., Bergquist GP3000) conform to moderate surface irregularity and are preferred for field-replaceable modules — no mess, consistent application thickness. Thermal grease offers slightly lower bulk resistance but requires controlled application volume and becomes a problem in systems requiring periodic lamp replacement, because re-greasing requires cleaning and re-application.

Efficacy derates with temperature. Typical COB lumen output drops 0.2–0.4%/°C as junction temperature rises above the rated test point (usually 25°C case). A module rated 15,000 lm at 25°C Tc will deliver approximately 13,500–14,250 lm at Tc = 70°C — a 5–10% reduction. Factor this into fixture photometric calculations. Efficacy derating data should be available in the module datasheet as a normalized lm/W vs Tc curve. If it is not, request it; the factory’s test lab will have measured it.

Our inspection service includes thermal imaging of COB heatsink assemblies during pre-shipment QC to verify contact quality and flag cold spots indicating poor TIM application.

Testarea LM-80 și Proiecțiile de Menținere a Lumenilor TM-21

LM-80 is the IESNA standard test method for measuring lumen maintenance of LED light sources. It requires measuring lumen output at three case temperatures (55°C, 75°C, and 85°C Ts) at intervals over a minimum of 6,000 hours. The result is a dataset of lumen maintenance percentages at each temperature over time.

TM-21 is the projection method that extrapolates from LM-80 data to estimate L70 (time to 70% of initial lumens), L80, or L90 lifetime. The critical constraint: TM-21 allows extrapolation to a maximum of 6× the LM-80 test duration without extended testing. This means that if a factory’s LM-80 test ran for 6,000 hours, TM-21 can project only to 36,000 hours — not the “50,000 hours L70” that appears in many Chinese COB datasheets.

What “50,000 hours L70” actually means in practice:

A datasheet claim of “L70 > 50,000 hours” on a domestic Chinese COB module typically means one of the following:

1. LM-80 testing ran for ≥8,334 hours at 85°C Ts, with a curve slope allowing TM-21 projection to ≥50,000 hours. This is legitimate. Major suppliers like Bridgelux and Citizen maintain 10,000+ hour LM-80 datasets.
2. LM-80 data was generated on a related (but not identical) die package and extended to this COB model without retest. This is common for minor package variants.
3. The number is a marketing figure with no LM-80 data behind it. This is common for sub-$1 COB modules from smaller domestic suppliers.

How to verify before ordering:

Request the LM-80 test report for the specific product code — not a generic “COB LED Module” report. Check: (a) test lab accreditation (NVLAP or A2LA for US market; CNAS or DAkkS for EU market — domestic Chinese CMA accreditation is acceptable for some applications but not for DLC or ENERGY STAR qualified fixtures); (b) the Ts temperature at which testing was conducted; (c) the test duration in hours; (d) whether TM-21 projection was performed and at which temperature it applies to your application.

If your fixture operates at 75°C Ts, the LM-80 data at 85°C is conservative — the actual L70 at your operating temperature will be longer. If your fixture operates at 90°C Ts (e.g., a tightly sealed IP65 downlight in a hot climate), the 85°C LM-80 data underestimates degradation. Design for the actual operating temperature.

Peisajul Furnizorilor Chinezi de COB și CQ la Recepție

International benchmarks manufactured in or near China:

• Citizen (Japan): The most widely specified high-CRI COB reference. Citizen CLU series (CLU028, CLU048) are available through distributors in China. Ra 90 and Ra 95 variants with R9 data; full LM-80 datasets at all three temperatures.
• Bridgelux (US, manufactured in China): Bridgelux BXRC and Vero series. Competitive efficacy, full LM-80 coverage, detailed TM-30 data on premium CRI lines. Price point between Citizen and domestic.
• Cree (US): Less commonly available as a COB for OEM fixture builders; Cree tends to sell through its own fixture channel. Useful as a performance benchmark.

Domestic Chinese COB manufacturers:

• San’an Optoelectronics (三安光电): China’s largest LED epitaxy manufacturer. Supplies wafers to many other domestic COB makers. San’an-branded COBs are competitive on efficacy (140–160 lm/W at Ra 80) and price.
• Nationstar (国星光电): Guangdong-based. Strong in commercial lighting COBs; Ra 90 line is well-regarded for the price point.
• Refond (瑞丰光电): Competes directly with Nationstar; similar quality tier.
• Shenzhen MK Semiconductor / Shenzhen COB-focused SMEs: Dozens of smaller assemblers source dies from San’an or imported wafers and package COBs. Quality is highly variable — incoming QC is mandatory.

Bin tolerance and matching:

COB modules are binned after test. Standard flux bin tolerance is ±7.5% of rated lumen output. CCT binning follows ANSI C78.377 7-step MacAdam ellipse — within a 7-step ellipse, color shift is visible to a trained eye under controlled conditions; 3-step ellipses (tighter binning) are available at premium pricing.

For multi-COB fixtures (e.g., a 4-COB downlight or a 12-COB grow light bar), request matched bins from a single production lot. Mixed bins from different lots can produce visible color variation across a fixture array — a common complaint in field returns. Specify on your purchase order: same flux bin (e.g., “bin H5”), same CCT bin (e.g., “ANSI 5-step 3000K”), same production batch.

Incoming QC protocol for COB modules:

1. Forward voltage at rated current. Measure Vf with a constant-current source at the rated drive current (e.g., 1,050mA for a 36V COB). Compare against the datasheet Vf ±0.5V tolerance. Vf significantly below spec indicates over-driving or die quality issues; Vf above spec indicates resistance problems.
2. Lumen output spot-check. Sample 5–10% of each shipment lot using an integrating sphere. Compare to the datasheet bin value. Reject the lot if mean output is below the lower bin boundary.
3. CCT verification. Measure CCT with a calibrated spectrometer (Sekonic C-800 or laboratory equivalent). Flag any modules outside the specified ANSI ellipse.
4. Visual inspection. Check for die attach defects (bubbles, partial detachment visible through the phosphor), contamination on the optical surface, and package cracking around the solder pads.

For commercial fixture qualifications, our inspection service can coordinate third-party lab photometric testing (LM-79) on production samples before full-volume acceptance.