An active vibration isolation system uses sensors, actuators, and real-time control electronics to detect incoming vibrations and generate counteracting forces that cancel them out. Unlike a passive mount made of rubber or springs, which can only absorb energy above its own natural frequency, an active system can suppress vibrations across a much wider band, including the difficult low-frequency range below a few hertz where passive methods actually amplify motion. These systems now achieve reductions on the order of 60 to 78 decibels in transmitted vibration, and they underpin some of the most demanding precision instruments on Earth and in orbit.
Why Passive Isolation Is Not Enough
A passive isolator works on a simple principle: mount your sensitive equipment on something compliant, like a spring or an air bearing, and vibrations at frequencies well above the mount’s natural resonance get filtered out. The catch is that at and just below that resonance frequency, vibrations are amplified rather than reduced. To push the resonance frequency lower and isolate against slower disturbances, you need a softer mount, but a softer mount also means the payload sways more under any steady force, like its own shifting weight. This is the fundamental trade-off that active systems exist to break.
An active system can behave as if it has an impossibly soft spring with impossibly high damping, because the “spring” is really an electronic feedback loop generating whatever force profile the controller demands. The payload can be stiff enough to stay put under static loads yet appear virtually decoupled from ground vibrations at low frequencies. That dual capability is why active isolation has become standard in gravitational-wave observatories, semiconductor lithography, satellite optical payloads, and precision metrology labs where nanometer-level stillness matters.
How the Hardware Works
Every active isolation system has three essential elements: sensors that measure either the incoming vibration or the payload’s residual motion, actuators that push or pull in response, and a controller that decides how much force to apply and when. The details of each component vary widely depending on the application, but the basic loop is always the same: sense, compute, act.
Sensors
The sensor’s job is to tell the controller what the payload is doing in absolute terms, not just relative to the vibrating floor. Inertial sensors like geophones and accelerometers are the workhorses here because they measure motion against their own internal inertial reference rather than against the shaking ground. A geophone designed for active isolation can cover a bandwidth from about 0.1 Hz up to 28 Hz with high sensitivity, providing the absolute-motion information the controller needs to generate what engineers call “skyhook” damping, a force that acts as though the payload were hanging from an imaginary hook fixed in space.1Mechanical Systems and Signal Processing. An active geophone based on a magnetic spring design for low-frequency vibration measurement For the most extreme applications, optical inertial sensors with resolutions on the order of a fraction of a picometer per root-hertz at 1 Hz have been developed, enabling isolation platforms that hold displacement to less than 100 nanometers from 0.3 Hz onward.2Journal of Low Frequency Noise, Vibration and Active Control. A six-degrees-of-freedom platform for deci-hertz active isolation using optical inertial sensors
Actuators
The actuator converts the controller’s electrical signal into a physical push or pull on the payload. Two broad families dominate the field. Lorentz-force (voice-coil) actuators work like a loudspeaker: a coil carrying current sits in a magnetic field, and the resulting force is directly proportional to the current. They offer long stroke, smooth controllability, and no hysteresis, which means the relationship between input signal and output force stays linear and predictable.3Mechanical Systems and Signal Processing. Two-axis Lorentz actuator for active vibration isolation system in optical payloads Piezoelectric actuators, by contrast, use crystals that deform when voltage is applied. They produce very high forces in a compact package but have shorter strokes and some inherent nonlinearity. A given system’s choice between the two often comes down to whether wide-band low-frequency motion or high-frequency precision force is more important. Some hybrid platforms use both types on the same structure, pairing a Lorentz actuator for the low-frequency heavy lifting with a piezoelectric element for stiff, fast corrections at higher frequencies.
Beyond electromagnetic and piezoelectric options, pneumatic actuators using pressurized air have been combined with electromagnetic force in hybrid architectures. One early demonstration paired the two in a single isolation mount and used a neural-network controller to coordinate them, exploiting the pneumatic element’s ability to carry the static payload weight while the electromagnetic actuator handled dynamic corrections.4Journal of Sound and Vibration. A HYBRID-TYPE ACTIVE VIBRATION ISOLATION SYSTEM USING NEURAL NETWORKS
Control Strategies That Make It All Work
Hardware alone does not isolate vibrations; the controller’s algorithm determines how well the system actually performs. Several control philosophies are used, and the best modern systems often blend more than one.
Feedback control measures the payload’s residual motion and drives the actuators to push it back toward zero. The simplest version is skyhook damping, where the controller applies a force proportional to the payload’s absolute velocity so that the payload behaves as if it were damped against inertial space rather than against the shaking floor.5Journal of Vibration and Control. An analysis and design framework for robust control of a multi-axis active vibration isolation system with unknown payload More advanced feedback controllers use techniques drawn from robust-control theory, shaping the loop gain across frequency to maximize vibration rejection while keeping the system stable despite uncertainties in the payload’s mass or stiffness. One recent platform demonstrated up to 65 dB of vibration reduction in the 0.1 to 10 Hz band using this kind of loop-shaping approach combined with virtual sensor fusion.6Mechanical Systems and Signal Processing. Enhanced H∞ loop-shaping control with virtual sensor fusion for active seismic vibration isolation
Feedforward control takes a different tack. Instead of waiting to see the payload move and then reacting, it measures the disturbance at its source, on the ground or on a vibrating machine, and preemptively cancels it before it reaches the payload. When the disturbance is predictable, feedforward can achieve dramatic results. A self-tuning feedforward system targeting ultra-low-frequency vibrations reached 78 dB of attenuation vertically and 70 dB horizontally, cutting cumulative vibration power at 100 Hz by more than 60 percent in the vertical axis and nearly 48 percent horizontally.7Journal of Sound and Vibration. Ultra-low frequency active vibration isolation system with quasi-zero stiffness characteristic using self-tuning filter-based feedforward control The practical limitation is that feedforward needs a good reference signal of the incoming disturbance. If the vibration is random and no upstream sensor can see it in time, feedforward has nothing to work with and the system falls back on feedback alone.
In most high-performance installations, both strategies run simultaneously. Feedback handles the unpredictable broadband noise; feedforward knocks down the dominant known disturbances, such as a spinning reaction wheel on a satellite or a floor vibration at a known machine frequency.
Gravitational-Wave Detectors and the Gold Standard of Isolation
The most extreme active vibration isolation systems on the planet sit inside the Advanced LIGO observatories, where mirrors must remain still to within a tiny fraction of a proton’s diameter. LIGO’s isolation architecture stacks multiple layers of passive and active inertial isolation in series. Each test-mass suspension begins with a hydraulic pre-isolator that cancels large ground motions at very low frequencies, followed by internal active platforms, one single-stage and one two-stage, that progressively clean up what remains. The combined system meets the observatory’s stringent noise requirements across all frequencies and has operated reliably through multiple observing runs.8Classical and Quantum Gravity. Seismic isolation of Advanced LIGO: Review of strategy, instrumentation and performance
LIGO’s approach illustrates an important design principle: no single isolation stage can do everything. A hydraulic actuator can move large masses over centimeters to correct for slow tidal and seismic drifts, but it is far too noisy to provide the final sub-nanometer quiet. A precision voice-coil stage can deliver that quiet, but it cannot absorb the large motions from an earthquake. By chaining stages with complementary strengths, the full system covers a frequency range from below 0.01 Hz up to hundreds of hertz, something no single technology could manage alone.
Spacecraft and Satellites
In orbit, the vibration problem flips. There is no seismic ground noise, but spacecraft carry their own disturbance sources: reaction wheels that keep the satellite pointed, cryocooler compressors, and deployment mechanisms. These sources send microvibrations through the spacecraft structure into sensitive optical payloads and instruments. Even tiny oscillations at the wrong frequency can blur an image or shift a laser beam off target.
Active isolation platforms for spacecraft reaction wheels have demonstrated 65 dB of force reduction at the wheel’s primary harmonic near 270 Hz, using a soft suspension combined with self-sensing actuators that both apply force and measure displacement in the same device.9Journal of Spacecraft and Rockets. Multi-Degree-of-Freedom Active Isolation Platform for Microvibrations Earlier proof-of-concept work using piezoelectric struts and centralized harmonic control achieved 30 to 40 dB of attenuation for reaction-wheel unbalance forces, confirming that active cancellation is practical within the mass and power budgets of real missions.10Smart Materials and Structures. Passive and active microvibration control for very high pointing accuracy space systems More recent satellite isolation research has focused on broadband controllers that suppress both random wideband vibration and the sharp tonal disturbances that reaction wheels produce, aiming for a single control law that handles both without separate tuning.11Physica Scripta. Satellite micro-vibration active isolation control with broadband disturbance suppression capability based on generalized output feedback
Weight is always the constraint in space. Every kilogram of isolation hardware is a kilogram not available for instruments or fuel. That pressure drives the preference for compact actuators and lightweight sensor designs and motivates the push toward algorithms that extract more performance from the same hardware rather than adding more stages.
Six Degrees of Freedom and Why They Matter
Real payloads do not vibrate in just one direction. A sensitive instrument sitting on a table can be pushed up and down, side to side, and front to back, and it can also rock and twist around three rotational axes. A system that isolates only vertically leaves the payload exposed to horizontal and tilt disturbances, which can couple into the measurement just as easily. True high-performance isolation therefore requires control in all six degrees of freedom simultaneously.
Building a six-degree-of-freedom platform means arranging actuators and sensors so that every possible motion direction is both measured and correctable. A hexapod layout, with six struts connecting the payload to the base, is a common approach because it naturally spans all six axes. But the control problem grows considerably: the controller must coordinate six or more actuators at once, each influencing multiple directions, without exciting unwanted cross-coupling between axes. A six-degree-of-freedom deci-hertz platform using optical inertial sensors demonstrated isolation of up to two orders of magnitude between 0.1 and 10 Hz while holding the center-of-mass displacement below 100 nanometers from 0.3 Hz onward, showing that sub-hertz multi-axis active isolation is now achievable in laboratory settings.12Journal of Low Frequency Noise, Vibration and Active Control. A six-degrees-of-freedom platform for deci-hertz active isolation using optical inertial sensors
Where Active Systems Hit Their Limits
Active isolation is not a magic bullet. Several practical constraints can erode the performance you would expect from the control law on paper. The most persistent issue is phase loss at higher frequencies. Every real component in the loop, from analog-to-digital converters to sensor dynamics to actuator response times, introduces a small time delay. Those delays add up into a phase lag between the measured vibration and the corrective force. Once the phase lag reaches a critical threshold, the system’s force starts arriving too late and actually adds energy instead of removing it. In practice, the flexibility of the mechanical structure between sensor and actuator becomes especially hard to manage above roughly 100 Hz, and using larger, more sensitive sensors to improve low-frequency performance makes the high-frequency flexibility problem worse.13Journal of Sound and Vibration. Sensor fusion methods for high performance active vibration isolation systems
This is a fundamental engineering tension, not a solvable-with-better-parts kind of problem. Improving low-frequency isolation typically demands heavier inertial sensors with better noise floors, but heavier sensors create more structural flexibility and worsen the high-frequency phase margin. Sensor fusion, combining a heavy inertial sensor for low frequencies with a lighter, faster sensor for higher frequencies, is one proven workaround. The controller blends the two signals so that each sensor contributes only in the frequency band where it performs best. Both the LIGO isolation system and several laboratory platforms use this technique to extend their usable bandwidth.
Power consumption and heat are also real constraints, especially in space. Actuators that carry a static load continuously draw current and generate heat, which can distort the very structure you are trying to hold still. That is one reason many designs offload the static weight to a passive spring or pneumatic support and reserve the active actuator exclusively for dynamic corrections. Keeping the actuator’s job small keeps its power draw and thermal footprint manageable.
Machine Learning in Active Isolation
Classical control algorithms require a mathematical model of the system: you need to know the payload’s mass, stiffness, damping, and how these change with conditions. When the payload changes frequently, as on a general-purpose optical table that holds different experiments each week, retuning the controller is tedious and sometimes impractical. This has motivated interest in model-free control approaches that learn the right forces directly from data.
Deep reinforcement learning, where an algorithm explores different control actions and learns which ones minimize vibration over time, has been experimentally demonstrated on a structure equipped with an active mass damper. A recent framework combines imitation learning, in which a neural-network controller first copies a conventional reference controller’s behavior, with reinforcement learning that then fine-tunes beyond the reference controller’s performance. Pre-training with imitation learning sidesteps the risky random exploration phase that pure reinforcement learning would require, which matters when the “exploration” could mean slamming a delicate payload into its end stops.14Smart Materials and Structures. Model-free active vibration control based on imitation learning and deep reinforcement learning: experimental implementation in an SDOF structure
These approaches are still in the early demonstration phase, mostly single-degree-of-freedom lab setups rather than full six-axis production platforms. But the appeal is clear: a controller that adapts on its own when you bolt a new instrument onto the table, without anyone having to re-derive a transfer function or retune a filter. Whether reinforcement-learning controllers will match or exceed the performance of a carefully designed classical controller on a well-characterized system remains an open question. The more likely near-term path is hybrid: a classical controller handles the baseline isolation, and a learning algorithm adjusts gain schedules or feedforward filters as conditions drift.
Picking a System for a Real Application
If you are specifying an active isolation system for a laboratory or facility, a few practical considerations tend to dominate the decision beyond raw vibration-attenuation numbers.
- Frequency range: Know where your problem actually lives. A semiconductor lithography tool fighting floor vibrations from nearby traffic needs isolation mostly below 20 Hz. An optical bench suffering from pump vibrations at 60 Hz and harmonics needs narrowband rejection at those frequencies. The control architecture and sensor choice follow from the frequency range, not the other way around.
- Payload variability: If the payload mass and center of gravity change regularly, the controller must tolerate those changes without manual retuning. Robust control methods that explicitly account for uncertain payload properties are worth the added design effort in this case.
- Environment: Vacuum and cryogenic environments rule out certain actuator types and sensor technologies. Pneumatic actuators need air, which is unavailable in vacuum. Some sensor electronics generate too much heat for cryogenic settings.
- Residual noise floor: The controller can only cancel what the sensor can see. If your target is sub-nanometer residual motion, the sensor’s self-noise floor becomes the practical performance limit, and you are likely looking at interferometric or high-end inertial sensors rather than standard accelerometers.
Commercial active isolation tables for optics labs typically deliver around 20 to 40 dB of isolation improvement above their corner frequency, which is adequate for most microscopy and interferometry work. Custom research platforms, like those in gravitational-wave observatories, push well beyond that but at proportionally greater cost, complexity, and maintenance burden. The right system is the one that meets your actual noise requirement without overshooting into hardware you cannot maintain or afford.