Fast-axis collimation and fibre coupling: getting the beam from the die to the target

The problem is a stripe, not a spot

Take a bare 905 nm edge emitter: 180 W peak, emitting aperture 110 µm by 30 µm, divergence 9° parallel to the junction and 25° perpendicular to it. That perpendicular axis is called the fast axis, and it is the one that causes trouble.

At 1 km, a 25° full-angle fan is a band about

tall. The slow axis, at 9°, is about 160 m wide over the same distance. So the illuminated footprint is 440 m by 160 m - a stripe nearly three times taller than it is wide, most of which is sky, ground, or a wall that is not where your receiver is looking.

The energy is not lost, it is simply spread where you cannot use it. A receiver with a 1 mrad field of view collects returns from a patch about 1 m across at 1 km. If the transmitter's footprint is 440 m tall, then roughly one part in 440 of the vertical fan lands inside that patch.

What a fast-axis collimator changes

A fast-axis collimator (FAC) is a cylindrical lens placed a few tens of micrometres from the facet. It compresses the fast axis and leaves the slow axis untouched. On our own parts, the effect is documented in the datasheets:

At 1 km the footprint goes from 440 m x 160 m to 105 m x 160 m - roughly square, and 4.2 times smaller in area. Since the return signal for an extended target is proportional to the fraction of transmitted power that lands inside the receiver's field of view, concentrating the beam by 4.2x in area recovers up to 4.2x of return power, which is equivalent to about 2x the range at the same diode.

That is a large gain for one lens, and it is why every serious long-range 905 nm design is a collimated design.

The gain has a ceiling. Once the illuminated footprint is smaller than the receiver's field-of-view footprint, concentrating further does not help: the receiver is already seeing all of the light that comes back. The useful question is not "how tight can I make the beam" but "is my transmitter footprint smaller than my receiver footprint".

When a wider beam is the right answer

• Aiming error. A handheld rangefinder at 6° divergence covers 105 m at 1 km, which is forgiving. Going to 1° would put the spot at 17 m - and require the user to aim to within a fraction of a degree.
• Non-cooperative targets. A narrow beam can pass between branches and miss the target entirely; a wider beam increases the chance that something in the field of view returns a signal.
• Speckle and scintillation. Coherent light on a rough surface produces speckle; a wider beam averages over more independent patches of the target, which stabilises the return. The mirror image of this problem, where speckle has to be suppressed rather than averaged over, is dealt with in visible and RGB laser illumination.
• Eye safety, in the wrong direction. Collimation makes the source look smaller from a distance. Apparent source size is one of the parameters that sets the permitted exposure, and a small, collimated source has the lower limit of the two. This is covered in laser safety classes for ranging: collimating a transmitter makes the finished product harder to certify, not easier.

Fibre coupling: the same problem, in reverse

Coupling a laser into a fibre is an exercise in matching two numbers: how much beam the fibre can hold (its core diameter times its numerical aperture) and how much the diode produces.

The 905 nm edge emitter's slow axis, taken as a 110 µm stripe at 9° full angle, already has a beam parameter product of roughly 4 mm·mrad. That is twenty times what a single-mode fibre can accept, so coupling a 180 W emitter into single-mode fibre is not a matter of better optics - it is impossible without discarding about 95 % of the power. That is exactly why our single-mode 905 nm modules are rated at 80 mW, not 180 W, and why high-power single-mode sources are DFBs or narrow-stripe devices rather than broad-area emitters.

Multi-mode fibre is different. A 105 µm, 0.22 NA fibre has a beam parameter product of about 11 mm·mrad, comfortably above the emitter's 4 mm·mrad - with one condition. Numerical aperture converts to a full angle as

The bare fast axis diverges at 25° FWHM, which is beyond the fibre's acceptance angle in either convention. A FAC is not an optimisation for fibre coupling - it is the precondition. Compressing the fast axis to 6-8° brings it inside the fibre's acceptance, at which point a 905 nm bar can be coupled into a 105 µm or 400 µm multi-mode fibre and delivered to a remote head.

Choosing between the three architectures

Free-space, collimated. FAC plus a receive lens. Best for long range, handheld products and anything with a strict size and cost budget. The whole 905 nm family is available in a collimated version.

Fibre-delivered, multi-mode. FAC first, then a 105-400 µm fibre. Use it when the optoelectronics cannot sit next to the optics: a rotating head, a sealed enclosure, a multi-channel instrument fed from one box. You pay a coupling loss and you gain mechanical freedom.

Fibre-coupled, single-mode. Only for low power - tens of milliwatts to about 150 mW in our catalogue - but you get a diffraction-limited beam, which is what interferometry, fibre sensing and single-mode beam combining need.

A practical note on measuring divergence

Divergence specifications are quoted at FWHM (full width at half maximum) in our 905 nm datasheets, and at 1/e² in the VCSEL datasheets. Numerical aperture and beam parameter product are usually 1/e² quantities. The numbers are not interchangeable, and a fast axis specified at 25° FWHM is considerably wider than 25° at 1/e². When you are checking whether a beam fits inside a fibre's acceptance angle, measure it, or use the same convention on both sides of the comparison.