0. If you are seeing a specific symptom, start here
This page is written to be read end to end. If you are here because something is already wrong, start below instead. The table maps reading the datasheet symptoms to the three causes that account for most of them in the field, and to the one measurement that separates them. If the measurement contradicts the table, the cause is outside the range this page covers.
| Symptom you are seeing | The three most likely causes | The one measurement that separates them |
|---|---|---|
| Two parts with the same specification behave differently on the board | The figures are stated at a different test current or junction temperature | Read the test conditions line before the value line |
| Efficacy quoted by us does not match your own measurement | Different spectral integration or a different measurement geometry | Compare measurement geometry first, then the integration function |
| Flux looks high but the part looks dim under a camera | Radiant flux and luminous flux are being compared; the units differ by wavelength | State which quantity is being used and at what wavelength |
Two things to do before you continue production
- Lower the drive current or the duty cycle before anything else. Almost every symptom below is current- or heat-driven, and reducing both is reversible within minutes.
- Measure before you swap parts. A forward-voltage measurement across the assembly, taken with the board powered, separates a driver problem from a LED problem in under five minutes and costs nothing.
When to contact us
Contact us when the measurement contradicts the table, when the same batch shows the same symptom in more than one assembly, or when you need a failure analysis to close a customer case. Bring the bin label, the forward voltage at operating current, and the ambient temperature at the moment the symptom appeared. Send the details to our engineering team.
The chapters below explain the underlying physics and specification in full.
1. What these characteristics are and why the reading order matters
A ceramic LED datasheet carries four families of characteristic: the optical output (luminous flux, radiant flux, wavelength or colour point), the electrical input (forward voltage), the beam description (viewing angle), and a derived figure that combines them (efficacy, in lumens per watt). The derived figure is the one buyers quote and the one most often misread, because it is not an independent measurement — it is a quotient of the other three, and it moves whenever any of them moves.
That is why this page is ordered the way a datasheet is read rather than the way a catalogue is written. Forward voltage first, because it is set by how the dies are wired and it determines the driver. Then flux, because it is published as a range at a stated test current. Then wavelength and colour, because they determine the binning scheme. Then viewing angle, because it interacts with any secondary optics. The efficacy column is checked last, against the other four, rather than believed on its own.
Every figure on this page is taken from the published package pages and from the ceramic handbook sections 13 to 16, which cover the 5050 single-die monochromatic, ball-top multi-die, flat multi-die and white-laser series. For the ratings themselves — the absolute maximum table — see ceramic absolute maximum ratings. For what the ceramic package changes relative to a plastic one, see the ceramic technical overview.
2. Forward voltage is a wiring outcome, not a device property
The single most useful thing to understand about the forward voltage column is that it describes the configuration, not the part number. In the 5050 package the same outline appears with three distinct forward voltage bands, and each band corresponds to a specific die count and wiring arrangement stated in the series remarks.
Forward voltage bands in the 5050 package, by die configuration
| Die configuration | VF band (V) | Cool white flux at 700 mA | Warm white flux |
|---|---|---|---|
| 45 mil x 4 dies, 3 V | 2.8 – 3.4 | 400 – 500 lm | 300 – 400 lm |
| 55 mil x 4 dies, 3 V | 2.8 – 3.4 | 450 – 550 lm | 350 – 450 lm |
| 45 mil x 4 dies, 6 V | 5.8 – 6.4 | 600 – 700 lm | 500 – 600 lm |
| 55 mil x 4 dies, 6 V | 5.8 – 6.4 | 700 – 800 lm | 550 – 650 lm |
| 35 mil x 9 dies, 9 V | 8.6 – 9.2 | 900 – 1,000 lm | 750 – 850 lm |
Those bands are white-die bands. The forward voltage on these packages also depends on which colour the die is, and the package pages group the colours into two forward voltage families.
Forward voltage by colour family, as published on the package pages
| Colour family | VF at 300 mA, 5050 | VF at 350 mA, 3535 and 7070 |
|---|---|---|
| Red / Orange / Yellow | 2.0 – 2.4 V | 1.8 – 2.4 V |
| Green / Blue / White / Purple | 2.8 – 3.6 V | 2.8 – 3.6 V |
So the 2.8 to 3.4 V band is not a lower-power band in a generic sense — it is specifically the band the green, blue, white and purple dies occupy. Red, orange and yellow sit at a lower voltage, which is why an RGBW board built from these packages does not have one forward voltage: the red channel and the green channel are on different voltage bands and need to be driven accordingly.
The pattern across the bands is straightforward once seen. Dies in series multiply the forward voltage: four dies in series at roughly 3 V each give the 9 to 10 V band, while the same four dies in parallel give the 3 V band. The flux rises with the configuration because more die area is being driven, and the package is unchanged throughout.
Two practical consequences follow. A driver designed for the 3 V band will not drive the 9 V band at all, so the forward voltage has to be confirmed before anything else on the electrical side. And a part described only as "5050 white" is not yet a complete specification — it may be any of five electrical configurations inside that one outline, with fluxes differing by a factor of more than two.
3. Luminous flux is a range at a stated test current
Luminous flux is published as a minimum and a maximum, not as a single figure, and always at a stated test current and junction temperature. Both matter. The test current is what makes two parts comparable, and it is frequently not the current the part will be driven at in production.
Test currents published on the package pages
| Package | Table heading as published | Flux measured at |
|---|---|---|
| 3535 | at 350 mA and 700 mA | 350 mA and 700 mA |
| 5050 | at 700 mA and 1000 mA | 700 mA and 1000 mA |
| 7070 | at 700 mA and 1400 mA | 700 mA and 1400 mA |
| 7070 UVA | at 500 mA, 700 mA and 1400 mA | 500 / 700 / 1,400 mA |
The 5050 pages illustrate the consequence well. The same part number family is published at 700 mA and again at 1,000 mA, with a higher flux figure at the higher current. Comparing a 5050 measured at 700 mA against a competitor quoted at 1,000 mA flatters the competitor by roughly the ratio of the two currents, and that error is easy to make because both tables look equally authoritative.
Junction temperature is the other half of the condition. All figures on these pages are stated at 25 °C, which is a bench condition, not a luminaire condition. Flux falls as the junction warms, so a design that targets lumen output in a sealed fixture has to work back from the 25 °C figure through its own thermal budget. The thermal management design guide sets out that method; the point here is only that the datasheet number is a starting point and not a delivered figure.
4. Wavelength, colour temperature and CRI: three different colour questions
Colour is specified in one of three ways depending on the die. A monochromatic die is specified by a dominant wavelength with a tolerance band. A phosphor-converted white die is specified by correlated colour temperature. A high-CRI white die additionally carries a Ra value. These answer different questions and are not interchangeable.
How colour is specified across the die types
| Die type | Colour parameter | Example value | What it tells you |
|---|---|---|---|
| Monochromatic, 5050 | Dominant wavelength | 625 nm red, 520 nm green, 460 nm blue | Which emission line the die produces |
| Monochromatic, 5050 | Dominant wavelength | 595 nm amber | The amber emission line |
| White, phosphor-converted | CCT | 3000 K / 6500 K | Where the white point sits |
| White, high CRI option | Ra | 70 standard, 80 high-CRI | How true the rendering is |
The monochromatic rows come from the 5050 single-die series in the handbook, where four colours are offered with the same electrical configuration: red at 625 nm, green at 520 nm, blue at 460 nm and amber at 595 nm. Note the forward voltage split — red and amber are 2.0 to 2.6 V while green and blue are 2.8 to 3.4 V, because the die material differs. A four-channel RGBW board built from these does not have one forward voltage, it has two.
For phosphor-converted white the same package is offered from 3000 K to 6500 K, with the luminous efficacy falling as the colour temperature rises — the 5050 page quotes 100 to 180 lm/W depending on CCT and bin, and the 7070 page quotes 120 to 180 lm/W. This is physical rather than a manufacturing compromise: the same flux is spread over a wider spectrum at higher CCT, so fewer lumens per watt. Specifying 6500 K and expecting the efficacy figure quoted for 3000 K is not possible.
Ra deserves separate emphasis because it is the parameter buyers most often treat as optional. The 7070 series is published with CRI options up to 90, and the standard parts in the tables below are 70. A design that specifies only a CCT and a flux has not specified rendering quality at all. The handbook rows carry Ra 70 for cool white and Ra 80 for warm white as standard, with higher CRI available on request.
5. Viewing angle: 120 degrees, and what it does not tell you
All these packages publish a viewing angle of 120 degrees, described as Lambertian emission. That figure is the full angle at which the intensity has fallen to half its on-axis value, and it is a property of the silicone lens profile rather than of the die or the phosphor.
The consequence is that a datasheet viewing angle is an input to the optical design, not a finished photometric result. The lumens a package produces and the lumens a fixture delivers are different quantities, and the difference is set by the secondary optics, the reflector or lens, and the spacing of the emitters. The 5050 page notes that secondary optics for 90 or 150 degrees are available, which is the more useful number for a lighting designer: the emitter supports a range of beam angles, and the choice is made downstream.
One related caution. Because the lens is the optical element, mechanical handling that damages or contaminates the silicone changes the beam profile without changing any electrical characteristic. A part that still measures correctly on the bench can still be optically out of specification in the assembled luminaire.
6. Radiant flux: the UVA case, where the units change
Ultraviolet and other non-visible emitters do not produce lumens at all, because lumens are a photometric unit weighted by the eye response. Those parts are published in radiant flux, in milliwatts, and the comparison to a white part stops there.
Radiant flux in the 7070 UVA series, by wavelength
| Wavelength | At 500 mA | At 1,000 mA | At 1,400 mA |
|---|---|---|---|
| 365 nm | 300 – 780 mW | 780 – 1,060 mW | 1,560 – 2,460 mW |
| 375 nm | 500 – 780 mW | 500 – 780 mW | – |
| 385 nm | 780 – 1,060 mW | 780 – 1,060 mW | – |
| 395 nm | 780 – 1,560 mW | 1,060 – 2,460 mW | – |
| 405 nm | 780 – 1,560 mW | 1,060 – 2,460 mW | – |
Two things are visible in that table and both matter for UV curing or inspection applications. First, radiant flux scales with current, and the scaling is not linear from 500 to 1,400 mA — the long-wavelength parts roughly double between 500 and 1,000 mA while the 365 nm part grows by a similar factor over the same interval, so a design at one current does not transfer to the other by simple scaling. Second, the forward voltage band is common across the series at 2.80 to 4.50 V, which is what makes a single driver usable for all five wavelengths.
The 7070 white page gives the same information in photometric terms, and the two should not be mixed: red at 1600 mA is quoted as 300 to 450 mW of radiant flux, while white at the same current is quoted in lumens. A UV-curing calculation and a lighting calculation run on different quantities entirely, even when both parts share the package.
7. Efficacy is derived, and this is how to check it
The lm/W column on a ceramic datasheet is a derived figure. It is luminous flux divided by electrical power, and because both flux and forward voltage are published as ranges, the derived figure is a range too. This has a direct practical consequence: recomputing lm/W from a single nominal flux and a single nominal voltage will not reproduce the handbook value, and the discrepancy is arithmetic rather than error.
The method is to take the extremes. Using the 5050 ball-top series, the 45 mil four-die 3 V row is published as 400 to 500 lm at 700 mA across 2.8 to 3.4 V. The most conservative combination gives 400 divided by 3.4 times 0.7, which is 168 lm/W. The most optimistic gives 500 divided by 2.8 times 0.7, which is 255 lm/W. The handbook nominal for that row is 207 lm/W, which sits inside the derived band rather than at either end.
Reading the check across all four configurations, the handbook nominal falls inside the derived range in every case. That is the test worth applying to any datasheet that quotes a derived column: recompute it from the limits and see whether the published value lands inside. If it does, the column is consistent. If it does not, one of the three source columns has been transcribed incorrectly — and the derived column is where that transcription error becomes visible.
One caution about the limits themselves. Forward voltage is a temperature-dependent and process-dependent quantity, and a minimum-to-maximum band across a bin spread is wider than the voltage range inside any single part. Treating the full band as applicable to one device produces a wider efficacy range than reality, which is another reason the derived value is a band and not a point.
8. Reading a part number to a full specification
The route from a part number to a complete set of characteristics is short but has to be taken in order. Identify the package, because that fixes the outline and the die arrangements available. Identify the colour, which fixes whether the colour question is answered by wavelength, by CCT, or by both with an Ra. Identify the die configuration, which fixes forward voltage and therefore the driver. Only then read the flux range, at the test current stated on the table heading, and check the resulting efficacy against the derivation above.
- Package sets the outline and the possible configurations. 5050 covers five electrical configurations on its own.
- Colour sets which of the three colour parameters applies: dominant wavelength, CCT, or CCT with Ra.
- Die configuration sets the forward voltage band, which sets the driver.
- Flux range is read at the test current in the table heading, never at an assumed one.
- Efficacy is checked against the derivation, not accepted on its own.
If any of these five is left open, the specification is incomplete in a way that surfaces later as a driver mismatch, a binning dispute or an efficacy claim that cannot be substantiated. Each of the three answers is available on the package pages, and the handbook sections carry the same figures in the form given above.
9. Frequently asked questions
Why does the same package appear with several forward voltage bands?
Because the bands correspond to different die counts and wiring inside the same outline. In the 5050 package the four-die 3 V configuration, the four-die 6 V configuration and the nine-die 9 V configuration are three different products sharing a footprint. The series remarks state the configuration for every row, so it has to be read from the table rather than assumed from the package.
Is a higher efficacy always a better part?
No, and efficacy alone will mislead on two of the axes that matter. A 9 V configuration may show a lower lm/W than a 3 V one in the same package while delivering substantially more flux, because more die area is being driven. And a high-CRI die generally gives up efficacy relative to the standard Ra option. Efficacy is a useful figure for comparing like with like; it is not a quality ranking.
Is a 2.8 to 3.4 V die always a low-power die?
No, and this is worth stating separately from the die count. The package pages group the colours into two voltage families: red, orange and yellow at 2.0 to 2.4 V, and green, blue, white and purple at 2.8 to 3.6 V. So the lower voltage band is a colour band, not a power band. A white die in the 3 V configuration and a red die in the same footprint are on different forward voltage, and a multi-channel board has to drive both.
Can I compare a UVA part’s mW with a white part’s lumens?
No. Radiant flux is optical power and luminous flux is power weighted by the eye response, so the numbers are in different quantities and there is no fixed conversion. A UV-curing design is sized on mW per area; a lighting design is sized on lumens per area. Quoting one as though it were the other is the most common specification error on mixed lists.
Why does flux fall as the colour temperature rises?
Because the same luminous output is spread over a wider spectrum. The 5050 page publishes efficacy as a range that depends on CCT and bin, and the 7070 page does the same at a higher band. A high-CCT part and a warm-white part of the same package are not the same product with a different label; they have different efficacy and should be compared on their own figures.
Does the viewing angle figure let me predict the fixture beam?
No. The 120 degrees figure describes the emitter’s own half-intensity angle, set by the silicone lens. The delivered beam depends on the secondary optics and on the emitter spacing, which is why the package pages describe available 90 and 150 degree options separately. Photometric modelling happens at the fixture level, not from the datasheet alone.
What does the derived efficacy column tell me if I cannot reproduce it?
That is exactly the diagnostic value. Recompute it from the flux and voltage limits as described in section 7; if the published nominal does not land inside the derived band, one of the source columns is wrong. In the configurations checked here the nominal falls inside the band in every case, so the column is internally consistent. A figure that cannot be reproduced is worth querying before it is quoted.