Visible and RGB laser illumination: brightness, white balance and speckle

What actually changes below 700 nm

Four things, and only the first one is about the light itself.

The receiver changes. Silicon peaks near 900 nm. At 450 nm it still works - roughly 0.2 A/W against 0.6 A/W at 905 nm - but you are no longer sitting on the peak of the responsivity curve. For illumination work this rarely matters.

The hazard changes. Between 400 nm and 700 nm the cornea and the lens are transparent, so the light arrives at the retina and is focused by it. A 1 mm beam entering the eye at 450 nm is concentrated onto a spot of about 19 µm, which multiplies the irradiance by roughly 2700. Past 1400 nm almost nothing reaches the retina at all. That asymmetry is why 1535 nm is the eye-safe band - and why visible light is not the safe end of the spectrum. It is the opposite.

The eye reacts. Visible light triggers an aversion response: a blink and a turn of the head, completed in about 0.25 s. The standards give you credit for it, and the exposure basis for visible wavelengths is 0.25 s rather than the 10 s used at 905 nm. That is a real relaxation of the accessible emission limit. It is not a licence, because it assumes the beam arrives unexpectedly - not that you are holding a collimated 5 W beam steady on a bench.

And the numbers change. Which is the part that actually determines the design.

Equal watts are not equal brightness

The eye's photopic response V(λ) peaks at 555 nm and collapses at both ends of the visible band. At our three wavelengths it is approximately:

Read the third column as the power budget of an RGB engine. To produce the same luminous flux, a 450 nm source has to emit about eighteen times the optical power of a 520 nm source, and a 638 nm source about three and a half times.

That one fact explains the shape of the visible catalogue. Our blue parts are rated at 5 W to 5.5 W, the green parts at 1 W to 1.5 W, and the red parts at 500 mW to 1 W. Blue is not rated higher because it is easier to make. Blue is rated higher because blue is where the watts have to go.

Etendue: the ceiling that power cannot lift

Brightness - radiance, in W/m²/sr - is what a projector or a searchlight actually needs, and it is the one quantity no passive optic can increase. A lens can trade area for angle, and it can trade angle for area, but it cannot reduce the product of the two. That product is the etendue, and it is set by the emitter before any of your optics exist.

Two consequences follow, and both of them are visible in the source sheets.

Fast axes are expensive. A 450 nm emitter leaves the die at 4 - 16° in the parallel direction and 35 - 57° in the perpendicular direction. The wide axis is not merely harder to collimate - it is where the etendue lives. Compressing it down to the 0.8 - 1° line spot that the sheets also specify cuts the solid angle on that axis by nearly three orders of magnitude. That is the real product being sold when you buy a collimated part: not a tidier beam, but the same power packed into a much smaller angle.

A bank does not automatically win. Adding emitters adds power and adds emitting area in the same proportion, so radiance does not rise on its own. The Nichia green bank reaches 13.5 W from eight or more emitters at 1.65 W each, and it is specified collimated to 0.2 - 0.6° parallel and ±1.0° perpendicular, with pointing deviation held to 0.7°. The collimation is what converts the extra emitters into usable brightness. Without it, a bank is a bigger and hotter version of one emitter.

Three spot shapes from the same die

Every colour in the summary sheets is specified with three beam shapes, and the difference between them is only the shaping optics.

The square spot figure is not a different measurement. It is the perpendicular divergence compressed to match the parallel one, so the far field becomes a square rather than an ellipse. The line spot goes further and squeezes one axis to under a degree, which is what a sheet-of-light illuminator or a scanning system wants. Choose the shape first, then the source: the three options are not interchangeable, and asking a line-spot optic to produce a square is not a matter of re-focusing.

Speckle, and why it is cheaper to fix here than at 905 nm

Because a laser is coherent, an illuminated surface shows a granular pattern: speckle. A fully developed pattern has a contrast - standard deviation over mean - of 1, which is a 100 % modulation. On a projection screen it is the first thing an audience notices.

The remedies, roughly in ascending order of cost:

• More independent emitters. N uncorrelated speckle patterns summed together reduce contrast by about 1/√N. An eight-emitter bank takes it from 1 to roughly 0.35 for free.
• Wavelength diversity. Two emitters a few nanometres apart do not produce the same speckle pattern, so a bank assembled from binned emitters speckles less than one assembled from a single wafer batch.
• Polarisation diversity. Two orthogonal polarisations do not interfere. A PBS combiner gives you the second polarisation and a halved speckle contrast at the same time.
• A moving diffuser. The classical fix, and the only one that works for a single-mode source - at the cost of a mechanism and its reliability.

Note which way this argument points. Multi transverse mode, a wide spectrum and many emitters are all defects in a ranging laser. In an illumination laser they are the cure.

Reading the sheets: two rows worth arguing about

One driver will not cover all three colours. Red operates at 1.9 - 3.0 V, green at 4.8 - 6.5 V, blue at 3.5 - 5.3 V. On a single lithium cell or a 5 V rail, red can be driven from a linear current source and the other two cannot - they need a step-up. In a three-colour engine that means three driver topologies, or one boost converter per channel, and the green channel is the one that sets the requirement.

The efficiency units differ between sheets. Red is quoted in mW/mA, green and blue in W/A. They are the same quantity - 1 W/A is 1 mW/mA - but the numbers are not comparable at a glance. Blue at 1.45 - 2.0 W/A converts current to light roughly twice as well as green at 0.45 - 0.8 W/A and about one and a half times as well as red. That is a large part of why the blue ratings are four times the green ones.

Aligning three colours

The raw emitters are specified with a pointing deviation of ±5°. The collimated banks are specified at 0.7° or better. That difference matters more than it looks.

±5° is 87 mrad. Combine three colours through dichroics and launch them at a target 10 m away, and each colour walks by up to 87 cm - the three beams separate and the white spot becomes three coloured spots. Collimate first and the same figure becomes 12 cm.

That is the second reason to buy a collimated part, and it is the one that gets forgotten. The divergence number protects your spot size; the pointing number protects your colour mixing.

A selection rule

A single emitter when the engine has a homogeniser or a diffuser downstream, the etendue of one emitter is enough, and cost and thermal simplicity matter most. That covers most projection and most machine-vision illumination.

A collimated bank when the power has to arrive inside a narrow angle, when three colours have to stay on top of each other down the throw, or when speckle has to be reduced without adding a moving part. The green and blue banks exist for exactly that combination.

Check the colour ratio before the total power. In an RGB engine the blue channel owns the power budget - by a factor of eighteen against green if you size all three for equal brightness. Sizing all three the same is the most common way to arrive at a colour point that can only be corrected by throwing light away.

Then check the class. Every collimated part in this family is a Class 3B or Class 4 product. The arithmetic for visible wavelengths is more generous than for 905 nm; see laser safety classes for ranging for where the credit comes from and where it stops.

And check the etendue, not the wattage. The spot size calculator will tell you what your divergence does at range. A watt you cannot focus into the target is a watt you paid for and did not use.

Where to next

• FAC, collimated and fibre-coupled beam shaping - the optics that turn these numbers into a usable beam
• Laser safety classes for ranging - the class arithmetic, including the visible-light case
• VCSEL against edge emitter - the same trade in the infrared, where an array is often the round-beam answer
• Visible and RGB sources - the 450 nm, 520 nm and 638 nm heads with their per-source parameter matrices