How to Evaluate LED Chip Quality: Beyond the Datasheet
Same part number, same datasheet, same bin code — two batches measured 38mW/cm² and 31mW/cm², a 23% gap. The supplier was right that both were “within specification,” because the datasheet only promises a range. Wavelength and irradiance are the two variables that decide whether a device does anything at all,[3] and neither can be taken from a datasheet: you measure incoming batches on an integrating sphere traced to a calibrated reference,[1] and you specify bin codes tight enough that “660nm” means the same thing in every unit you ship.
Table of Contents
- 1. The Key LED Specifications for Therapy Devices
- 2. What the Datasheet Doesn’t Tell You
- 3. Our LED Evaluation Process
- 4. The Integrating Sphere: Your Most Important Tool
- 5. Binning: The Secret to Consistency
- 6. LED Supplier Tiers
- 7. Counterfeit LED Detection
- 8. What We’d Tell a New Brand
- 9. Partner With Rainbow Technology for Custom LED Devices
- 10. References & External Links
We received two batches of LEDs from the same supplier. Same part number, same datasheet, same bin code. But when we measured them, Batch A produced 38mW/cm² at 20mA and Batch B produced 31mW/cm² — a 23% difference.
The supplier said both batches were “within specification.” They were right — the datasheet specified a range of 28-42mW/cm². But a 23% output difference between batches means inconsistent treatment efficacy.
LED chip quality is the foundation of LED therapy device performance. And datasheets don’t tell the full story. Here’s how we evaluate LED chips beyond what’s on paper.
1. The Key LED Specifications for Therapy Devices
Wavelength (nm):
- Center wavelength ± tolerance (e.g., 660nm ±5nm)
- Full width at half maximum (FWHM) — how narrow the spectral peak is
- Narrower FWHM = more targeted treatment
- Typical: Red 630-660nm (FWHM 15-25nm), NIR 830-850nm (FWHM 25-40nm), Blue 415-470nm (FWHM 15-20nm)[3]
Radiant power (mW):
- Total light output at specified drive current
- The most variable specification between batches
- Must be measured at the same drive current you use in your product[4]
Viewing angle:
- Half-angle of the light emission cone
- Determines coverage pattern and uniformity
- 30-60° for focused applications, 90-120° for broad coverage
Forward voltage (Vf):
- Operating voltage at specified current
- Affects driver circuit design and power consumption
- Tight Vf binning = more consistent current in series strings
Forward current (If):
- Maximum rated continuous current
- Overdriving LEDs reduces lifespan dramatically
2. What the Datasheet Doesn’t Tell You
| What the Datasheet Omits | What It Means in Practice |
|---|---|
| 1. Batch-to-batch consistency | Datasheets give you the range. They don’t tell you where within that range a specific batch falls. A 660nm LED with ±10nm tolerance could be 650nm in one batch and 670nm in the next. Both are “in spec” but they produce different treatment effects. |
| 2. Aging characteristics | How much does output drop over 1,000 hours? Datasheets rarely specify this. Our testing shows 5-15% degradation at 1,000 hours for mid-range LEDs and 2-5% for premium LEDs.[2] |
| 3. Thermal behavior | LED output drops as temperature increases. The datasheet specifies output at 25°C junction temperature. In a mask or panel running for 20 minutes, the junction temperature might be 60-80°C. Output at operating temperature can be 10-20% lower than datasheet values. |
| 4. Spectral shift with temperature | LED wavelength shifts with temperature — typically 0.3nm/°C for red LEDs. At 50°C above room temperature, your 660nm LED is actually emitting at 675nm. This matters if you’re targeting a specific absorption peak. |
| 5. Mechanical reliability | Solder joint fatigue, wire bond integrity, and encapsulant yellowing are real failure modes that don’t appear on datasheets. |
3. Our LED Evaluation Process
| Step | Duration | What We Do |
|---|---|---|
| Step 1: Datasheet screening | (1-2 days) |
|
| Step 2: Sample ordering | (1-2 weeks) |
|
| Step 3: Incoming measurement | (2-3 days) |
|
| Step 4: Thermal testing | (1-2 weeks) |
|
| Step 5: Reliability testing | (4-8 weeks, can overlap with production) |
|
Total evaluation cost: $3,000-8,000 per LED supplier (equipment amortized over many evaluations)
Total evaluation time: 6-10 weeks for complete evaluation
4. The Integrating Sphere: Your Most Important Tool
An integrating sphere measures total radiant power and spectral distribution. Without it, you’re guessing at LED performance.[1]
What we measure:
| Measurement | What It Confirms |
|---|---|
| Peak wavelength: | Confirm it matches the datasheet specification |
| Spectral bandwidth (FWHM): | Narrower = more targeted therapy |
| Total radiant power: | Measured in mW at the specified drive current |
| Spectral power distribution: | The complete emission spectrum, not just the peak |
Equipment cost: $5,000-15,000 for a basic integrating sphere + spectrometer setup. For small brands, third-party testing labs can do this for $50-150 per LED model.[6]
Why it matters: We once found that a supplier’s “660nm” LEDs actually peaked at 648nm — outside their own ±10nm specification. Without measurement, we would have built and shipped an entire production run with the wrong wavelength.
5. Binning: The Secret to Consistency
LEDs are manufactured in large batches and then sorted (binned) by key parameters. Tighter binning = more consistency = higher cost.
| Bin Parameter | Widest Bin | Middle Bin | Narrowest Bin |
|---|---|---|---|
| Wavelength binning: | Standard: ±10nm (e.g., 650-670nm for 660nm nominal) | Tight: ±5nm (e.g., 655-665nm) | Ultra-tight: ±2nm (e.g., 658-662nm) |
| Brightness binning: | Standard: Range of ±30% from nominal | Tight: ±15% | Ultra-tight: ±7% |
| Vf binning: | Standard: ±0.2V | Tight: ±0.1V | Ultra-tight: ±0.05V |
Our specification for therapy devices:
- Wavelength: ±5nm (tight bin)
- Brightness: ±15% (tight bin)
- Vf: ±0.1V (tight bin)
This adds approximately 15-25% to the LED cost compared to standard binning. On 200 LEDs per mask at $0.05 each, that’s $1.50-2.50 more per unit. Worth it for consistent treatment output.
What happens with standard binning: If you use ±10nm wavelength bins, some units in a production run will have 650nm LEDs and others 670nm. Both are “in spec” but they deliver measurably different treatment. A customer who reads our spec sheet and sees “660nm” has a reasonable expectation that every unit delivers 660nm.
6. LED Supplier Tiers
| Tier | Typical Suppliers | What We See |
|---|---|---|
| Tier 1 (Premium): | Cree, OSRAM, Lumileds, Nichia |
|
| Tier 2 (Mid-range): | Everlight, Lite-On, Seoul Semiconductor, Chinese branded (Epistar, San’an) |
|
| Tier 3 (Budget): | Unbranded or generic Chinese LEDs |
|
Our approach: We use Tier 2 LEDs (Epistar) for consumer products and Tier 1 (OSRAM) for clinical products. The performance difference is real — Tier 1 LEDs show 30% less output degradation at 1,000 hours.[2]
7. Counterfeit LED Detection
Counterfeit LEDs are a real problem in Shenzhen sourcing:
Common counterfeit patterns:
- Rebranded Tier 3 LEDs with Tier 1 markings
- Re-binned LEDs (rejects from tight bins relabeled as tight-bin product)
- Reworked LEDs (used/pulled LEDs resold as new)
How we detect counterfeits:
| Check | How We Apply It |
|---|---|
| 1. Buy only from authorized distributors or factory-direct. | No trading companies, no Alibaba marketplace sellers without factory verification. |
| 2. Measure every batch. | If the measured performance doesn’t match the datasheet, we reject the batch. |
| 3. Check packaging and markings. | Tier 1 LEDs have consistent marking quality. Sloppy markings are a red flag. |
| 4. Verify lot traceability. | Legitimate LEDs have lot codes that trace back to the wafer. If the supplier can’t provide lot traceability, walk away. |
8. What We’d Tell a New Brand
- 1. Don’t choose LEDs based on price alone. The cheapest LED that meets your spec on paper may not meet it in reality. Test before committing.
- 2. Specify tight binning. The 15-25% cost premium for tight bins pays for itself in reduced customer complaints and warranty claims.
- 3. Measure incoming LEDs. Even trusted suppliers have bad batches. A quick measurement of 10-20 LEDs from each incoming batch catches problems before they reach your production line.
- 4. Budget for an integrating sphere. It’s the single most valuable piece of test equipment for an LED therapy brand. If you can’t afford one, use a third-party lab — the $150 per LED model is cheap insurance.
The LED is your product. Everything else — housing, battery, packaging — is supporting infrastructure. Invest in verifying the LEDs, and you’ve protected the core of your product.
Choosing the wavelength is the decision that comes before all of this. For the near-infrared end of that choice, see our
Near-Infrared 810nm vs 830nm vs 850nm: Choosing the Right Wavelength for Your Product.
9. Partner With Rainbow Technology for Custom LED Devices
“660nm” on a spec sheet is a claim about a distribution, not a single value. This is what we do to make the claim true in every unit:
- LEDs specified by bin code on the BOM — wavelength, brightness and Vf — so a batch swap cannot happen silently.
- Incoming measurement of every batch on an integrating sphere and spectrometer: peak wavelength, FWHM, radiant power and Vf, at the drive current your product actually uses.[1]
- 100-hour burn-in plus temperature cycling and 85/85 damp heat to published JEDEC test methods, run before a part is released to production.[5]
- Authorized-distributor or factory-direct sourcing only, with lot codes traced back to the wafer — no trading companies, no marketplace sellers.
- Tier 1 parts for clinical claims, Tier 2 for consumer lines, and an honest answer about which one your price point actually needs.
10. References & External Links
[1] National Institute of Standards and Technology (NIST).
Spectral Comparator Facilities — spectral power responsivity calibrations for photodiodes and radiometers, and the measurement of irradiance and spatial uniformity on which integrating-sphere LED measurement depends.
[2] U.S. Department of Energy, Federal Register.
L70 lifetime definition and the IES LM-80-2008 / TM-21-11 test and projection methods for measuring how much light output a solid-state source loses over time — the basis for the aging and degradation figures quoted here.
[3] PubMed, U.S. National Library of Medicine.
Wavelength-specific effects of red (630nm) versus blue (463nm) light on epidermal proliferation and mitochondrial activity — why centre wavelength and spectral bandwidth are therapeutic variables, not cosmetic ones.
[4] PubMed Central (PMC).
Photobiomodulation dosing parameters — the irradiance window (recommended < 100 mW/cm²), the biphasic Arndt-Schultz dose response, and why a batch-to-batch swing in delivered power changes the dose rather than just the brightness.
[5] JEDEC Solid State Technology Association.
JESD22-A101 (Steady-State Temperature-Humidity Bias Life Test) — the standard accelerated method behind the 85°C/85% RH damp-heat condition used to qualify non-hermetic packaged devices.
[6] International Laboratory Accreditation Cooperation (ILAC).
The ILAC Mutual Recognition Arrangement (MRA) and Signatories — how to identify a testing laboratory accredited to ISO/IEC 17025 whose results will be recognised, the check to run before sending LED samples to a third-party lab.