Tech-led.com 
Summary: A custom LED module for a research instrument is an engineering specification, not a catalog pick. Before a supplier can build one, you have to define the optical, mechanical, electrical, and thermal targets that make the instrument's measurement repeatable. This guide walks through the spec sequence: peak wavelength and spectral bandwidth (FWHM) chosen against the chromophore or absorption feature you are measuring; radiometric units (radiant flux in mW, radiant intensity in mW/sr, or irradiance in mW/cm2 at a stated working distance); package selection across through-hole, SMT, SMBB, EDC/EDCC, and COB; die count and array geometry for single-band or multispectral modules; binning tolerance for unit-to-unit consistency; and forward-current, thermal, and wavelength-drift limits. It covers the full catalog from UV-A (365, 405 nm) through visible, NIR (780 to 940 nm), and SWIR (1050 to 1750 nm). An engineer will finish able to write a module spec a supplier can quote and build against.
What "custom module" actually means for an instrument
Off-the-shelf LEDs are specified for general use. A research instrument is specified for a measurement, and the two rarely line up without work. A custom LED module is the bridge: a defined assembly of one or more emitters, in a chosen package, on a substrate you can mount and cool, driven to a known operating point, and characterized tightly enough that two units behave the same.
The reason to go custom is almost always one of four constraints the instrument imposes. The wavelength you need is narrowband and off the mainstream (a specific SWIR peak, a UV-A excitation line). The geometry is fixed by the optical path, so the emitter has to sit on a particular footprint at a particular pitch. The radiometric budget is set by the detector and integration time, so output has to be specified, not assumed. Or the measurement demands stability, so drive and thermal behavior have to be bounded. Spec those four well and the module is buildable. Leave them vague and you get a part that lights up but does not measure.
Step 1: peak wavelength and FWHM
Wavelength is the first decision because everything optical downstream depends on it. Pick the peak where the property you are measuring shows strong, repeatable contrast: a fluorophore's excitation band, a molecular absorption feature, a hemoglobin difference, a reflectance edge. Do not pick by what is easy to source.
Spectral bandwidth matters as much as the peak. FWHM (full width at half maximum) is the width of the emission band, and for spectroscopy, fluorescence, and multispectral work a narrow band is what separates your signal from adjacent features. LEDs are naturally narrower than incandescent or white sources but wider than lasers, typically 20 to 50 nm FWHM in the visible and NIR and somewhat wider in SWIR. If your instrument needs to resolve two features 30 nm apart, a 40 nm FWHM emitter will blur them and you either move to a narrower die, add an optical filter, or reconsider the band. State both the target peak and the maximum acceptable FWHM in the spec.
| Band | Typical research use | Selection note |
|---|---|---|
| UV-A 365, 405 nm | Fluorescence excitation, optical detection | Excite at the peak; image the longer-wavelength emission |
| Visible 450 to 660 nm | Colorimetry, absorbance, cytometry | Narrowband beats broadband white for stable readings |
| NIR 780 to 940 nm | fNIRS, PPG, tissue optics, sensing | Sits in the tissue optical window; pair wavelengths for chromophore separation |
| SWIR 1050 to 1750 nm | Moisture, hydrocarbons, hyperspectral | Requires an InGaAs detector, not silicon |
Step 2: radiometric specification and units
"Bright enough" is not a spec. The detector, its integration time, the optical throughput of the instrument, and the working distance together set how much optical power you need at the sample, and you have to state it in radiometric units, not lumens. Lumens are photometric, weighted to human vision, and meaningless in the UV, NIR, and SWIR where most research instruments operate.
Use the unit that matches how the module illuminates:
- Radiant flux (mW or W): total optical power the emitter puts out. Use it when the optics collect most of the output.
- Radiant intensity (mW/sr): power per solid angle. Use it for a directional module feeding a defined acceptance angle.
- Irradiance (mW/cm2) at a stated working distance: power per unit area on the sample. This is usually the most useful for an instrument because it maps directly to what the sample and detector see. Always pin it to a distance.
Specify the operating point too. Radiant output scales with forward current, so a flux number without the drive current behind it is incomplete. State the target output, the drive current it is measured at, and the minimum you can tolerate as the emitter ages or warms.
Step 3: package selection
The package sets the footprint, the thermal path, and how the module integrates with the instrument's optics and board. The Marubeni catalog spans several package families, and the right one follows from the mechanical and thermal constraints, not preference.
| Package family | Character | Fits when |
|---|---|---|
| Through-hole | Leaded, larger, easy to prototype | Bench rigs, low density, hand assembly |
| SMT | Standard surface-mount footprint | Board-level integration, moderate density |
| SMBB | Surface-mount with defined optical face | Directional output on a compact footprint |
| EDC / EDCC | High-power, ultra-compact | Tight optode or probe geometry, close pitch |
| COB (chip-on-board) | Multiple die on one substrate | Arrays, flood illumination, multi-die modules |
For a probe or optode where several emitters sit close to a detector, the ultra-compact EDCC package earns its place because it shrinks source-to-detector spacing. For a wide, uniform field, a COB array spreads die across a substrate you can heatsink as one unit. Match the package to the optical geometry first, then confirm the thermal path can carry the dissipation at your drive current.
Step 4: die count and array geometry
A single-band module is one die (or a small cluster of the same die) at one wavelength. Many research instruments need more. A multispectral illuminator carries several discrete wavelengths in one module so the instrument can read a target across bands (for example UV-A excitation plus NIR and SWIR absorption bands in one head). fNIRS and pulse-oximetry front ends pair two or more NIR wavelengths in a fixed geometry near the detector.
When you specify a multi-wavelength module, define the wavelength set, the physical arrangement (which die sits where, and the pitch), and whether the bands are driven independently or together. Independent drive lets the instrument time-multiplex bands and normalize each one, which is usually what a measurement wants. It also means the module needs the right number of anodes and cathodes broken out, so specify the electrical interface alongside the optical layout. For array uniformity across a field, state the acceptable irradiance variation across the illuminated area, because that tolerance drives die placement and count.
Step 5: binning and unit-to-unit consistency
Research instruments are often built in small runs or need to be replaced years later without recalibrating the whole system. That makes binning a first-class spec item. Binning is the sorting of production output into tighter ranges of wavelength, forward voltage, and radiant output than the datasheet's full spread.
Decide which parameters you need binned and how tight:
- Wavelength bin: critical for spectroscopy and multispectral work, where a few nanometers of peak shift moves your measurement.
- Radiant flux bin: keeps output consistent unit to unit so the instrument's calibration holds.
- Forward voltage bin: matters when several emitters share a driver or when thermal budget is tight.
Tighter bins cost more and can lengthen lead time because the supplier is sorting for a smaller fraction of production. Ask for the binning that the measurement actually needs, not the tightest available. State it as a numeric range, not "consistent."
Step 6: drive, thermal, and wavelength drift
An LED's peak wavelength and output both move with junction temperature. Peak wavelength typically drifts on the order of 0.1 to 0.3 nm per degree C depending on material system, and radiant output falls as the junction heats. For a measurement that depends on a stable band and stable intensity, that drift is error, so the module spec has to bound it.
Three things control it. The drive current sets the dissipation and the nominal operating point, so specify the forward current and confirm it is within the emitter's rating with margin. The thermal path (die to substrate to heatsink) sets how hot the junction runs for that dissipation, so specify the mounting and the maximum junction or case temperature you will hold. And the drive mode matters: constant-current drive keeps output stable where constant-voltage does not, and pulsed or time-multiplexed drive both limits average heating and lets a multi-band module read one wavelength at a time. If the instrument needs absolute radiometric stability, plan for a monitor photodiode or a temperature-referenced calibration, and say so in the spec so the module carries the sensor.
Putting the spec together
A module spec a supplier can quote and build reads as a short table of numbers with tolerances: peak wavelength and max FWHM per band; radiometric target with its unit, working distance, and drive current; package family and footprint; die count, arrangement, and electrical interface; binning ranges for the parameters that matter; and the forward-current, thermal, and stability limits. Write those down and the conversation with the supplier is about feasibility and lead time, not guesswork. For the underlying wavelength selection behind any of these bands, the LED wavelength guide covers the UV-through-SWIR range in one place, and multi-die and flood modules are typically built on the chip-on-board (COB) package.
FAQ
What are custom LED modules for research instruments?
They are LED light sources built to an instrument's measurement spec rather than picked from a catalog: one or more emitters at chosen wavelengths, in a defined package and geometry, driven to a known operating point, and binned tightly enough that units behave consistently. The point is a repeatable measurement, so the module is specified by peak wavelength, FWHM, radiometric output, package, and thermal and drive limits.
What do I need to spec before a supplier can quote a custom LED module?
At minimum: the peak wavelength and maximum FWHM for each band, the radiometric target with its unit and working distance, the drive current, the package family and footprint, the die count and arrangement for multi-wavelength modules, and the binning ranges plus thermal limits. Those numbers, with tolerances, are what turn a request into a buildable quote.
Which package should a custom research module use?
Match the package to the optical geometry and thermal load. Through-hole and SMT suit bench and board-level prototyping, SMBB gives directional output on a compact footprint, EDC and EDCC suit tight probe or optode geometry, and COB (chip-on-board) suits multi-die arrays and flood illumination. Confirm the thermal path can carry the dissipation at your drive current before finalizing.
How do I specify a multi-wavelength module for a research instrument?
Define the wavelength set, the physical layout and pitch of the die, whether bands are driven independently or together, and the electrical interface (anodes and cathodes broken out) that the drive mode requires. For a field illuminator, add the acceptable irradiance variation across the illuminated area, since that tolerance drives die count and placement.
Why does binning matter for research-grade LED modules?
Binning sorts production into tighter wavelength, forward-voltage, and radiant-output ranges than the full datasheet spread. Research instruments need that consistency so calibration holds across a build and across replacement parts years later. Specify numeric bin ranges for the parameters your measurement depends on, and expect tighter bins to add cost and lead time.
