Control it
any state in → one chosen state out
Something downstream needs a defined input: coherent mixing, a modulator, a polarimetric measurement.
Cost: Power, feedback, and a device that can drift or need resetting.
← All guides
Photonics · 9 min
Fiber polarization control
Single-mode fiber rotates the polarization state slowly with temperature and abruptly with every touch. Pin it, erase it, or carry it in PM fiber — this guide is about telling which one your measurement needs.
Poincaré-sphere labResponse timeExtinction ratio budgetWhat to specify
The governing choice
Three strategies, and only three.
Scramble it
sweep the sphere faster than the detector integrates
The problem is polarization dependence, not a specific state: PDL test, sensitive gratings and detectors.
Cost: You give up any defined state — the average is all you get.
Maintain it
launch on a PM axis, keep every joint aligned
The state must survive a whole link passively, with no electronics in the path.
Cost: Extinction ratio becomes an installation problem: one bad splice caps the system.
Interactive
Watch the state settle — or run away.
Pick the disturbance and the controller, and the trajectory tightens onto the target or smears across the sphere. Then check the other two questions: whether scrambling gets you there instead, and what extinction ratio survives your splices.
Polarization lab
Can you track it, average it, or must you maintain it?
Set what is disturbing the fiber and watch the state on the Poincaré sphere tighten or run away — then check whether scrambling gets you there instead, and what extinction ratio survives your splices.
state trajectory on the Poincaré sphere — target marked, spread is the residual error
Motorized paddles keep up, but with only 1.0× margin the residual error is significant.
Disturbance
2 Hz
60° free wander
Bandwidth margin
1.0×
f_control / f_disturbance
Residual error
60°
state wander left after control
Insertion loss
0.5 dB
finite range, needs reset
Scrambler cycles
1k
in 10 ms integration
Residual DOP
3.16%
1/√N averaging
System PER
22.3 dB
from 25 dB parts
Lost to joints
46%
4 joints at 1.5°
Or scramble instead
10 ms
Extinction ratio budget
Working lab — people, airflow, cables
Handling, doors and airflow. Still slow, but fast enough that manual paddles stop being a fix and become a chore.
Four stepper-driven fiber loops. Broadband, high power, lowest loss — and the cheapest thing that solves a thermal-drift problem.
Tracking uses a first-order suppression model (residual ≈ wander × f_disturbance / f_control), scrambling uses 1/√N averaging over the cycles inside one integration time, and the PER budget adds leakages in linear power with tan²θ cross-coupling per PM joint. These are the right order-of-magnitude relations for choosing a construction, not a substitute for measuring your own link — real disturbances are broadband rather than single-frequency, and real splices vary joint to joint.
The speed decision
Buy the time constant you are fighting.
| Construction | Speed | Loss | Best for | Watch out |
|---|---|---|---|---|
| Motorized paddles | ~2 Hz | 0.5 dB | Thermal drift, visible wavelengths, high power | Too slow to track; finite range needs reset |
| Piezo squeezers | 100 kHz | 0.5 dB | Active tracking, closed-loop stabilization | Hysteresis; needs feedback for absolute states |
| Electro-optic rotators | 10 ns | Higher | Field transients, fast switching, endless control | Narrow band, drive electronics, temperature |
| Rotating waveplates | Seconds | Low | Reference states, polarimetry, high power | Slow, bulky, alignment-sensitive |
| Passive depolarizer | Passive | Low | SLED and FP sources, PDL suppression | Needs >1 nm source; erases, never sets |
| Multimode scrambler | 10–150 kHz | ≈ none | Speckle removal, illumination uniformity | Multimode only; needs exposure to average |
Diagnosis
What you are seeing → what fixes it.
| Symptom | What fixes it | Specify |
|---|---|---|
| Power through a polarizer drifts over minutes | Paddle or piezo controller, ideally with polarimeter feedback | Drift and repeatability, insertion loss, USB/RS232 control |
| State must track a deployed or vibrating fiber live | Piezo squeezer (100 kHz) or electro-optic controller (10 ns) | Response time, endless operation, drive electronics |
| A detector behaves differently per input state | Scrambler — average the dependence away instead of chasing it | Scramble rate vs integration time, residual DOP |
| Fringes or PDL modulate a sensor reading | Passive depolarizer — no power, no electronics | Source linewidth >1 nm, input fiber, output DOP |
| Grainy, unstable multimode illumination | Mode scrambler / de-speckler | Core size and NA, scramble rate vs camera exposure |
| System PER worse than the components promise | Inline polarizer to reset the floor; PM splicing to fix the cause | PER grade, splice angular alignment, connector keying |
Before you ask for a quote
Six things to pin down.
Wavelength & band
All-fiber parts are broadband, visible to 2000 nm. Electro-optic and micro-optic parts are built around one design wavelength.
Speed vs disturbance
Buy the time constant you are fighting, with a decade of margin — not the fastest number on the page.
Extinction ratio
Polarizers come as >18 and >25 dB; splitters ~22 dB fused, 26–29 dB micro-optic. Specify above what you must measure.
Fiber, jacket, connector
Fix both fiber types, jacket, pigtail length and connector — FC/APC where back-reflection matters — or quotes are not comparable.
Power handling
All-fiber takes high power; micro-optic and EO parts cap at hundreds of mW to a few W. High-power polarizers dump the rejected axis out a third port.
Control interface
USB/RS232 with a Python API, analog SMA drive, or TTL enable — this decides whether polarization is a setup step or part of the loop.
Where it pays for itself.
Coherent & heterodyne detection
Mixing efficiency follows the angle between signal and local oscillator — a wandering state is a wandering signal.
PDL & component test
Present every state faster than the meter integrates and a state-by-state sweep becomes one reading.
Fiber sensing & interferometry
Sagnac and interferometric sensors read polarization drift as signal; depolarizing removes it at the source.
Spectroscopy illumination
Gratings and detectors are polarization-sensitive — scrambling upstream makes day-to-day comparison meaningful.
Sequencing & biophotonics
A >150 kHz mode scrambler flattens speckle into uniform illumination within one camera exposure.
Quantum optics & QKD
Polarization encoding needs endless, reset-free control: no discontinuity while qubits are flowing.
From decision to part number
One component per role.
Slow drift, lowest loss
Fiber Polarization Controller – 0.5dB Loss, 2Hz, USB
- ~2 Hz
- 0.5 dB
- USB / RS232
- 488–1550 nm
Four stepper-driven fiber paddles on one continuous fiber — the default answer to thermal drift on a bench.
Active tracking
Fiber Polarization Controller Piezoelectric – 0.5dB, 100kHz
- to 100 kHz
- 0.5 dB
- 2–7 plates
- 850–2000 nm
Piezo squeezers at 45°, fast enough to follow acoustic and handling transients.
Closed loop, hands off
Automatic Polarization Controller With Buit-In Polarimeter
- Built-in polarimeter
- SM in / PM out
- Plug and play
Squeezers plus detectors and firmware in one module: any state in, one fixed state out, held without an operator.
Endless, nanosecond
High Speed Fiber Polarization Controller, 10ns
- 10 ns
- 4 EO plates
- Analog SMA
- 1310 / 1550 nm
No moving parts, no resets — for deployed fiber and links that cannot glitch.
Erase the state
Ultra-Fast Fiber Polarization Scrambler/Depolarizer
- to 5 MHz
- 3–4 rotators
- to 5 W
- 750–2000 nm
EO rotators each driven at a different frequency, sweeping the sphere faster than any detector integrates.
No electronics at all
Passive Optical Fiber Depolarizer: >1nm Linewidth
- Passive
- >1 nm source
- High power
- 780–1550 nm
Cascaded couplers decorrelate a broadband source across its own spectrum. Nothing to drive, nothing to drift.
Kill speckle
Piezoelectric Fiber Scrambler/De-Speckler – 150kHz
- >150 kHz
- 50/125 µm
- TTL input
- Benchtop
Multimode, not polarization: randomizes modal interference so a camera sees uniform illumination.
Set a hard PER floor
Fiber Inline Polarizer
- >18 or >25 dB
- SM→PM, PM→PM
- 200–500 mW
Throw one axis away to guarantee the other — at the cost of up to half the power.
Going deeper
Either side of the polarization stage.
- Polarization & phase control components
The full family, with datasheets and configurable options.
- Laser specifications decoded
Linewidth and coherence decide whether a passive depolarizer works at all.
- Isolators & circulators
Keep reflections out while polarization control keeps the state in.
- Component finder
Filter by wavelength, fiber type and connector.
- Fiber alignment stages
Where extinction ratio is usually won or lost.
- Pump–probe spectroscopy
A system guide where polarization is one budget item among many.
A system built from 25 dB parts is not a 25 dB system. Connector keying tolerance, splice angular error and tight bends all subtract, and the result is set by the worst joint — measure it after installation rather than inheriting it from a datasheet.
Common questions
A few important nuances.
What is the difference between a polarization controller, a scrambler and a depolarizer?
A controller is deterministic: any input state becomes one chosen state, held there — what you want when something downstream needs a defined input. A scrambler is the opposite, sweeping the state over the whole Poincaré sphere so anything slower sees only an average, which is how polarization-dependent loss is measured out rather than fought. A depolarizer produces light with a low degree of polarization: actively in time, or passively by decorrelating a broadband source across its own spectrum. Choose by what your detector integrates over.
How fast does my polarization controller need to be?
Match the disturbance, with about a decade of margin. Thermal drift in a lab fiber is millihertz, so 2 Hz paddles track it easily and cost far less. Vibration and handling reach kilohertz — squeezer territory. Nanosecond electro-optic control is for transients on deployed fiber, fast switching and endless control in coherent or quantum links; on a bench it buys bandwidth the disturbance never reaches.
Do I still need PM fiber if I have a polarization controller?
They solve the same problem at different points. PM fiber passively preserves a state already aligned to one of its axes — but only if every splice and connector is rotationally aligned, and one bad joint quietly caps the whole path. A controller fixes the state actively at one point, and needs power, feedback and occasional resets. Most real systems use both: a controller or inline polarizer to launch a clean state, PM fiber to carry it.
What extinction ratio should I specify?
Work back from the error you can tolerate: 20 dB PER means 1% of the power is in the wrong axis — invisible in a power measurement, fatal in a high-contrast polarimetric one. Inline polarizers come as >18 dB and >25 dB; splitters run ~22 dB fused and 26–29 dB micro-optic, the latter holding better over temperature. PER only degrades downstream, so specify a few decibels above what you need to measure at.
Why would I put a de-speckler on a multimode fiber?
Hundreds of modes interfere at the output, producing a grainy pattern that shifts with temperature, bending and handling — a drifting systematic for anything measuring intensity. A mode scrambler modulates the fiber fast enough (10 kHz voice-coil, or above 150 kHz piezoelectric) that the pattern averages into a smooth, stationary distribution over a typical camera exposure. The optical path is unbroken, so insertion loss barely changes.
Do these components work at visible wavelengths, or only in telecom bands?
All-fiber constructions do: paddles, squeezers and mode scramblers work on stress birefringence, which is inherently broadband, and are supplied with visible-band fiber such as SM450, 460HP and 780HP — roughly 360 nm to 2000 nm. Micro-optic and electro-optic hardware does not transfer as easily; polarizers, splitters and EO rotators are built around a design wavelength. Always state wavelength together with fiber type.
What does “endless” or “reset-free” polarization control mean?
Paddles and squeezers have finite mechanical range. Track a continuously rotating input long enough and an element hits its end stop and must be unwound — a reset, during which the output state jumps. Harmless on a bench, unacceptable on a live coherent or quantum link. Endless control uses elements that keep advancing phase indefinitely, with firmware handing over between them. If your link cannot tolerate a glitch, make it an explicit requirement.
Not sure whether to control, scramble or maintain?
Send the wavelength, the fiber on both ends, and what the polarization is doing that it should not. We will shortlist two or three parts and say where the cheaper one is good enough.