A laser Doppler vibrometer measures how fast a surface moves towards or away from it, without touching that surface. It is the instrument of choice when the object is too small, too hot, too light or too delicate to carry an accelerometer, and when adding a few grams of sensor would change the very resonance you are trying to measure.

What follows is the engineering view: what the optical head needs from the light source, which numbers come out of the physics rather than out of a datasheet, and where the usual specifications stop being useful.

From Doppler shift to velocity

Light reflected from a moving surface comes back with its frequency shifted. The shift is proportional to the velocity component along the beam, and because the light makes the trip twice, out and back, the factor of two appears in the formula.

A megahertz-scale shift cannot be read directly against an optical frequency of several hundred terahertz. The instrument therefore mixes the returning light with a reference beam on a detector and reads the beat, which is the same trick an interferometer uses, and laser Doppler velocimetry runs on the same Doppler relation applied to particles in a flow rather than to a solid surface.

Wavelength	Doppler shift per 1 m/s	Displacement per interference cycle (λ/2)
532 nm	3.76 MHz	266 nm
633 nm	3.16 MHz	317 nm
785 nm	2.55 MHz	393 nm
1064 nm	1.88 MHz	532 nm
1550 nm	1.29 MHz	775 nm

Read the table in both directions. A shorter wavelength gives you more hertz per metre per second, which is useful when the vibration is slow and you want signal. It also means fewer nanometres per interference cycle, which is where displacement resolution comes from if you integrate the velocity or count fringes. Practical systems interpolate within a cycle and reach well below that figure, so treat the right-hand column as the scale of the problem rather than as the resolution limit.

The carrier that gives you direction

A plain interferometer cannot tell approach from retreat. Both produce the same beat frequency, and the sign of the velocity is lost. Commercial vibrometers solve this by shifting the reference beam with an acousto-optic modulator, the Bragg cell, typically by 30 to 40 MHz.

A stationary surface then produces a beat at exactly the carrier frequency. Motion towards the sensor pushes the beat above it, motion away pulls it below. Demodulating that frequency-modulated signal gives velocity with its sign intact.

The upper measurement bandwidth is bounded by the same carrier, since the demodulator needs room on both sides of it. In specialised systems, slow-shear-mode Bragg cells have pushed carriers into the hundreds of megahertz and measurement bandwidths towards the gigahertz range, which is what surface acoustic wave devices and fast MEMS resonators require. Standard instruments sit well below that.

Coherence length follows your standoff

This is the specification most often taken from the wrong end. The requirement does not come from the laser, it comes from the geometry of your measurement.

The measurement arm runs to the object and back. The reference arm stays inside the instrument. The path difference the light has to bridge is therefore roughly twice the standoff distance, and the coherence length has to exceed it by a comfortable margin, or the interference contrast collapses and the signal disappears into noise.

Linewidth	Coherence length	Standoff this supports	Typical source (order of magnitude)
several hundred GHz, roughly 1 nm	0.1 to 1 mm	bench setups only	free-running Fabry-Perot diode
10 GHz	9.5 mm	a few mm	grating-stabilised diode
1 GHz	95 mm	centimetres	DFB diode, depending on design and drive noise
100 MHz	0.95 m	tens of cm	stabilised single frequency
10 MHz	9.5 m	metres	narrow line DPSS, frequency stabilised
1 MHz	95 m	tens of metres	reference-grade single frequency

The source column is an orientation, not a specification. Linewidth varies considerably between designs within each of these families, and with drive current noise and temperature, so a datasheet beats a category every time.

A bench measurement at 100 mm standoff asks very little, and a narrow linewidth module covers most laboratory geometries with room to spare. A scanning head looking at a car body across a hall, or a vibrometer pointed at a building, pushes the requirement into the single frequency regime. Deciding this early is cheaper than discovering it during commissioning.

Choosing a wavelength

Vibrometry grew up around the helium-neon laser at 633 nm, and a large share of installed instruments still works there. The wavelength is visible, which makes alignment on a real structure straightforward, silicon detectors are sensitive and inexpensive at that wavelength, and the historical availability of long coherence lengths from gas lasers made it the obvious choice.

Semiconductor and DPSS sources now reach the same coherence lengths in a package that needs no gas tube, warms up in seconds and tolerates being moved. The trade is no longer gas against diode but which wavelength suits the job:

Wavelength	Argument for	Argument against
532 nm	highest Doppler shift per m/s, finest displacement scale, visible green is easy to aim	more visible to anyone in the room; for the same power and beam diameter, permissible exposure in the visible is lower than at 1550 nm
633 nm	the established vibrometry wavelength, visible, silicon detectors at good sensitivity	nothing specific; the default unless a reason argues otherwise
785 nm	barely visible, less disturbing in an occupied lab, silicon still works	alignment needs a camera or a pilot beam
1550 nm	far higher permissible exposure, fibre components from telecom, good for long standoff outdoors	InGaAs detectors, lowest Doppler shift per m/s, invisible

Eye safety deserves a flag rather than a rule of thumb. Permissible exposure at 1550 nm is orders of magnitude above the visible range because the cornea absorbs the light before it reaches the retina. That is why long-range and outdoor vibrometry has largely moved there. The actual classification of your instrument depends on power, beam diameter and the applicable standard, and belongs in a safety assessment rather than in an article.

What the surface does to your signal

Most surfaces you will point a vibrometer at are optically rough. The returning light is a speckle field, not a clean wavefront, and the signal level varies from speckle to speckle. When the object moves laterally or rotates, the speckle pattern moves across the detector and the signal amplitude fluctuates with it. Where a dark speckle happens to sit on the aperture, the measurement drops out for a moment.

Three things help, and none of them is a property of the laser:

• Retroreflective tape or spray on the measurement point, which raises the return by orders of magnitude. It adds mass, which is exactly what you were avoiding on a light structure, so it is a trade rather than a fix.
• A larger receiving aperture, which averages over more speckles and softens the dropouts.
• Speckle tracking or signal-quality gating in the instrument, which discards the moments when the return is too weak to trust.

More laser power raises the return, but it does not remove dropouts: a dark speckle stays dark whatever you feed into it. Power helps when the surface is simply dark or distant, not when the speckle statistics are against you.

What to specify

Parameter	What decides it	Where to be generous, where not
Coherence length	twice your maximum standoff, plus margin	specify from the geometry; paying for 100 m when you work at 200 mm is wasted
Linewidth stability over time	mode hops shift the carrier and look like velocity	mode-hop-free operation over the working temperature range matters more than the absolute figure
Output power	surface reflectivity, standoff, receiving aperture	a few mW is plenty on tape, dark rough surfaces at distance need more
Beam quality	spot size on the object and coupling back into the head	TEM00 for a small measurement spot; multimode light does not couple back usefully
Polarisation	interference contrast at the detector	a stable, known polarisation state; a drifting one modulates your signal
Relative intensity noise	noise floor at the vibration frequencies you care about	specify with the bandwidth it was measured in, otherwise the figure means little

One point is worth stating plainly: the noise floor of a vibrometer is rarely set by the laser alone. Detector noise, electronics, the carrier demodulator, air turbulence in the beam path and the mechanical stability of the head all contribute, and they add through their variances rather than linearly. A quieter laser improves the result only once it stops being the dominant term.

Frequently asked questions

What is the difference between a vibrometer and an interferometer?

A vibrometer is a specialised interferometer. The optical arrangement is the same, but the signal processing is built around a frequency-shifted carrier and reads out velocity continuously, where a displacement interferometer counts fringes and reads out position.

Which wavelength do laser Doppler vibrometers use?

633 nm is the classical choice and still widespread. 532 nm appears where the finest displacement scale matters, 1550 nm where eye safety and long standoff dominate, for instance outdoors or on large structures.

How much coherence length does a vibrometer need?

More than twice the standoff distance, because the measurement arm travels to the object and back. At 1 m standoff that means more than 2 m, which corresponds to a linewidth below roughly 48 MHz using the convention L = c/(π·Δν).

Can I use a standard laser diode?

For very short standoff, possibly. A free-running Fabry-Perot diode has a line around a nanometre wide, which puts its coherence length between a fraction of a millimetre and about a millimetre. That rules out most practical geometries. A DFB or a frequency-stabilised module changes that picture.

Why does my signal drop out when the object rotates?

Rotation drags the speckle pattern across the receiving aperture. When a dark speckle lands on it, the return falls below the noise. A larger aperture, retroreflective treatment of the measurement point, or signal-quality gating in the instrument are the usual answers.

Does more laser power give a better measurement?

Up to a point. Power helps against dark or distant surfaces. It does not help against speckle dropouts, and once the noise floor is set by the electronics rather than by shot noise, additional power buys nothing.