What this page solves

This page is for situations where the motor already runs closed-loop and meets basic performance targets, yet the drive still feels noisy, rough or cheap. The cabinet may whistle at certain speeds, the axis may crawl with a cogging sensation, and customers complain about acoustic tone or vibration even after safety and efficiency requirements are met.

Instead of re-opening motor mechanics or re-spinning the entire power stage, the focus here is on what can be improved in the control and drive chain: torque-ripple suppression, randomized PWM strategies and phase-current balancing. These levers turn a “works on the bench” drive into one that sounds and feels like a premium product in the field.

The page distinguishes three kinds of user-facing issues:

• Acoustic noise from PWM switching and magnetic forces exciting the frame.
• Structural vibration where resonances in mounts, housings or fixtures are amplified.
• Current and torque ripple that shows up as low-speed roughness or ripple in motion.

The scope is strictly within the Control & Algorithms branch of the motion system. Mechanical design, motor geometry and detailed power-stage layout are acknowledged but treated as mostly fixed. Whenever a topic belongs to those domains, this page points to their dedicated pages instead of duplicating content.

Where do acoustic noise & vibration come from in motor drives?

Noise and vibration in a motor drive always come from a combination of the motor mechanics and the way the drive excites that hardware. Some contributions are baked into the motor: cogging torque from slot and pole choices, winding distribution, shaft and bearing design and structural resonances in the mounts and housing. Once these are set, the remaining leverage lives in how the drive generates currents, voltages and switching patterns over speed and load.

On the drive and control side, the first contributor is the PWM carrier and its harmonics. Fixed-frequency switching with high dv/dt creates periodic electromagnetic forces in the stator and frame. When the carrier or its sidebands land in the 2 kHz to 16 kHz range, human hearing is very sensitive to the resulting tones. If any of those frequencies coincide with mechanical modes in laminations, end bells, ducts or cabinet panels, the structure amplifies the sound even when electrical ripple looks acceptable.

A second source is torque ripple created by current and angle errors. Limited encoder resolution, misalignment during mounting, imperfect FOC tuning or an underdamped current loop all introduce periodic torque components tied to electrical angle. At low speed this shows up as a cogging sensation or speed modulation; at higher speed it shifts into vibration and acoustic excitation. The motor’s intrinsic geometry sets a base level of ripple, but the drive can either exacerbate it or suppress it depending on current quality.

The third contributor is timing and sampling error in the control chain. Poor coordination between PWM updates and ADC sampling, coarse timer resolution or per-phase S/H skew lead to small but repeatable current and voltage errors. In multi-axis systems with separate clocks, the same effect appears between axes as phase drift and beat patterns. Although these errors are small on a single cycle, they form stable spectral lines that the mechanical system can lock onto and amplify.

From an IC perspective, these mechanisms translate into concrete requirements on the motion MCU, the gate driver and the sensing front-end. High-resolution timers and ADCs with tight PWM-ADC synchronisation are needed to push audible content out of sensitive bands. Precision current and voltage AFEs reduce drive-induced torque ripple so that the remaining ripple is dominated by unavoidable motor geometry. Gate drivers with controllable dead-time and dv/dt let designers trade off acoustic behaviour, EMI and switching loss. The rest of this page builds on these roots and focuses on how to use torque-ripple suppression, spectrum shaping and phase-current balancing to pull noise and vibration down.

IC-side levers for torque-ripple suppression

Torque ripple in a modern motor drive is not only a mechanical artefact; it is strongly influenced by how the control ICs generate voltage vectors, measure currents and reconstruct electrical angle. With the motor geometry largely fixed, the remaining levers sit inside the FOC or motion MCU, the gate driver and the sensing front-end. These devices decide whether the drive excites cogging and structural ripple or smooths them out to the point where the system feels premium.

From the FOC and motion controller side, the first lever is PWM and ADC resolution. Limited timer resolution forces coarse duty-cycle steps, which in turn quantise the commanded voltage vector and show up as low-order current harmonics. In the same way, an ADC that cannot sample phase currents with sufficient resolution or speed restricts current-loop bandwidth and makes it harder to track an ideal d–q current trajectory. Higher-resolution timers and ADCs, tightly coupled through PWM-triggered sampling, allow finer control of the stator field and reduce electrically induced torque ripple.

The second lever is current-loop speed and the ability to apply feedforward compensation. A fast current loop with deterministic latency can react to load disturbances, cogging patterns and known torque harmonics instead of letting them propagate into motion. FOC engines with hardware accelerators, predictable interrupt timing and support for current and angle feedforward allow engineers to implement harmonic cancellation and notch strategies that target the dominant ripple components without consuming excessive CPU headroom.

Electrical angle accuracy forms the third leg of the chain. Resolver and encoder interfaces with sufficient resolution, low jitter and robust interpolation reduce the angle error that directly maps into torque ripple in a FOC system. When the control IC exposes alignment routines, offset calibration and filtering options, the effective electrical angle becomes smoother, and the torque contribution from angle quantisation and jitter is pushed down. Resolver-to-digital converters and encoder interface ICs on the sensing side therefore play a key role and are treated in more depth on the feedback and sensing pages.

On the power-stage side, gate drivers and bridge control add another set of levers. Dead-time that is too long, unbalanced between legs or drifting with temperature creates phase-dependent voltage errors and therefore torque harmonics. Gate drivers that support fine dead-time steps, temperature-aware adjustment and on-line calibration hooks make it possible to equalise the effective conduction period of each switch and pull these harmonics down. When the power stage and driver also expose multiple, well-synchronised phase-current feedback paths, the motion MCU can balance currents between phases instead of relying on reconstruction from a single shunt.

Typical applications make these IC-side decisions visible. In low-speed servo axes that need smooth crawling moves, small improvements in angle resolution, current-loop bandwidth and dead-time symmetry translate directly into a more uniform torque profile and better positioning quality. In high-end appliances such as drum motors and compressors, PWM and current quality determine whether a “quiet mode” actually sounds premium in a sleeping room. AGVs and collaborative robots operating near people depend on the same IC levers to remove roughness and tonal artefacts from motion so that the machine feels safe and refined rather than industrial and harsh.

Randomized PWM & spectrum shaping

Conventional fixed-frequency PWM concentrates switching energy into narrow spectral lines at the carrier frequency and its harmonics. When these lines fall inside the most sensitive region of human hearing or align with mechanical resonances in the motor, housing or cabinet, the result is a distinct tone or whistle. The electrical waveform may look clean, yet the acoustic impression is dominated by one or two loud notes rather than a neutral background. Spectrum shaping techniques aim to redistribute this energy without changing average power.

Randomized and spread-spectrum PWM methods start with the carrier itself. By allowing the carrier frequency to wander within a defined window around its nominal value, the energy that was previously concentrated at a single frequency is spread across a band. To the ear, this turns a sharp tone into a broader, less intrusive sound. Power-stage stress does not change significantly, but the drive must tolerate a small amount of carrier variation inside the current control loop without destabilising it.

Duty-cycle dithering introduces a second layer of spectrum control. In this case the average duty ratio is preserved, but small, controlled perturbations are applied cycle to cycle. The purpose is to push quantisation and non-linearity artefacts away from low-order harmonics toward higher frequencies where they are less audible or easier to filter. This approach relies on sufficient timer resolution so that the dither stays small compared with the commanded duty and does not undermine torque accuracy or efficiency.

A third dimension is speed-dependent carrier planning. Different operating regions benefit from different carrier strategies: low-speed operation may require frequencies well above the most sensitive hearing bands, mid-speed operation may prioritise efficiency and switching loss, and high-speed operation may be constrained by device limits. Planning carrier bands and spectrum-shaping strategies as a function of speed and load sometimes removes problematic tones without resorting to extreme switching frequencies.

These techniques translate into clear requirements on the control and power ICs. Motion MCUs need timer blocks that support programmable jitter or spread-spectrum modes, flexible PWM alignment and precise triggering of ADC sampling so that current control remains coherent. Gate drivers and digital power controllers must track the varying frequency without losing protection integrity or desynchronising multi-phase outputs. In tightly integrated systems, spread-spectrum behaviour may also need to coordinate with front-end PFC stages and EMI filters to avoid shifting conducted or radiated emissions into more problematic bands.

Randomized PWM and spectrum shaping are therefore not simple on–off features but system-level design decisions anchored in IC capabilities. The choice of motion MCU, gate driver and power controller determines how much freedom is available to move switching energy away from human hearing and structural modes without sacrificing control stability, EMI margin or thermal headroom. Detailed EMC planning and compliance aspects are covered in the EMC subsystem page, while this section concentrates on the control and acoustic consequences.

Phase-current balancing & calibration

Even when a motor and power stage are well designed, small mismatches in phase currents can turn into visible torque ripple and noise. The drive may deliver the commanded power, yet one electrical sector feels rougher than the others or certain speeds exhibit a persistent hum. The root cause often lies in how accurately each phase current is measured and reconstructed, and how consistently those measurements are used inside the control loop.

Hardware gain and offset errors are the first source of imbalance. Shunt resistors carry tolerance and temperature drift; three nominally equal values rarely stay identical over current and temperature. Any mismatch is amplified by the current-sense amplifier or isolated amplifier, which adds its own offset, gain error and bandwidth limitations. Layout asymmetry and reference routing can further bias one phase relative to the others, especially when measuring small currents near zero crossings where signal levels are low.

Timing errors form the second major contributor. Multi-channel ADCs often sample different phases at slightly different instants; if PWM waveforms are changing quickly, each channel can observe a different point on the current ripple. Single- and dual-shunt schemes depend on carefully placed sampling windows, and any deviation in sample-and-hold timing or PWM sector detection causes systematic reconstruction errors. Around zero crossings, dead-time and device delays skew effective conduction times, creating asymmetry between positive and negative half-cycles and injecting low-order harmonics into the torque.

Digital calibration closes much of this gap. During production test, known currents can be driven through each phase while the ADC codes are recorded, allowing per-phase offset and gain coefficients to be derived. These coefficients are then stored in OTP, internal flash or an external EEPROM and applied on every power-up so that subsequent control computation operates on corrected values. Additional temperature sensors or periodic self-check conditions make it possible to track slow drift over lifetime and refine the coefficients without interrupting normal operation.

Reconstruction algorithms also play a role. Designs that measure only one or two shunts rely on mathematical reconstruction of the third phase current based on modulation patterns and line currents. This approach can be highly cost effective, but only if the control IC provides deterministic PWM–ADC synchronisation, sufficient computation headroom and stable numerical behaviour across all sectors. Where budget and space allow, three-shunt architectures with properly calibrated front-ends simplify phase balancing and reduce the burden on reconstruction logic.

On the IC side, several data-sheet parameters map directly into how well phase-current balancing can be achieved. Current-sense amplifiers and isolated AFEs are judged by input offset, gain error, linearity and temperature drift, as well as noise spectral density that might convert into jitter. ΣΔ modulators with matched digital filters provide excellent accuracy and tracking across phases, at the cost of additional digital processing. Motor control MCUs that expose per-channel calibration registers, built-in trim engines and motor-control-optimised ADC triggering make it much easier to implement and maintain these calibration schemes while keeping the control loop deterministic.

Design trade-offs vs efficiency, EMI & cost

Acoustic and vibration performance rarely improves in isolation. Decisions that push a drive toward quieter, smoother operation often increase switching loss, change EMI behaviour or raise bill-of-materials cost. The key is not to chase absolute minimum noise, but to select a combination of PWM strategy, sensing quality and control IC capability that meets acoustic targets without compromising energy use, thermal margins or project budget.

Raising PWM frequency is a straightforward way to move switching artefacts out of the most sensitive hearing band and reduce low-order harmonics in torque. However, every additional kilohertz increases switching losses in the power devices, the gate driver and sometimes the front-end supply. At high power levels this translates directly into higher junction temperatures, larger heatsinks and tighter thermal design. Devices such as SiC modules may tolerate higher speed, but the efficiency and lifetime targets for the end application still impose limits on how far this lever can be pulled for acoustic benefit.

Slew-rate control provides another trade-off axis. Slower edges reduce dv/dt, lowering common mode currents, softening mechanical excitation and helping EMI filters. The downside is increased switching loss and potential interaction with control linearity, especially in topologies that rely on clean, fast transitions. Gate drivers with programmable dv/dt and per-leg settings allow careful tuning so that only the most critical transitions are softened, balancing acoustic and EMI improvements against efficiency.

Randomized PWM and spread-spectrum techniques, introduced earlier in this page, reshape the switching spectrum to avoid sharp tones. From an EMI standpoint they may reduce narrowband peaks but can raise the broadband noise floor. Conducted and radiated emission limits, filter topologies and interactions with front-end PFC or communication bands all need to be considered. A spectrum that sounds quieter to the ear is not automatically easier to certify, so EMC planning must be aligned with any spectrum-shaping decisions and is covered in more depth in the EMC subsystem section.

Sensing and control silicon add a cost and complexity dimension. Moving from single-shunt to multi-shunt or ΣΔ-based current measurement improves phase-current balancing and torque smoothness, but adds components, isolation channels and layout effort. Selecting a higher tier motion MCU with motor-control-optimised timers, ADCs and spread-spectrum PWM support makes advanced mitigation techniques practical but may impact the cost structure of lower-volume variants. In cost-sensitive platforms, it is often more realistic to prioritise calibration and control algorithms on existing hardware before committing to a full IC refresh.

When budgets are tight, a staged approach helps. Firmware-only changes such as revisiting current-loop tuning, enabling available calibration features and modestly adjusting PWM frequency or alignment usually come first. Small BOM changes like upgrading shunts or AFEs and adding temperature sensing can be reserved for designs that still fall short of their acoustic goals. New platform designs can then justify higher-spec MCUs, gate drivers and sensing ICs by tying their features directly to measurable improvements in noise, vibration and energy use over the entire product line.

Application playbooks

The same IC and control techniques from earlier sections land very differently in real applications. Each segment has its own expectations for how quiet a drive must feel, how much efficiency can be traded away and how aggressively BOM cost needs to be held. The playbooks in this section group torque-ripple suppression, spectrum shaping and sensing strategies into practical combinations for common motion markets.

The three examples below do not attempt to fix exact numbers for PWM frequency or filter corner points. Instead they define bands and priorities: which ranges are typically used, which levers tend to give the largest acoustic benefit in that range and what kind of MCU, driver and sensing chain is normally selected. Design teams can then refine these guides using their own mechanical, thermal and regulatory constraints.

Low-noise servo axes (collaborative robots and semiconductor tools)

Collaborative robot joints and semiconductor handling axes operate close to people in relatively quiet workspaces. Users expect smooth low-speed crawling with minimal cogging sensation, very little tonal noise and consistent behaviour across multiple axes. In this segment acoustic and vibration quality are often part of the perceived safety and precision of the machine, so motion quality and noise usually outrank small differences in controller cost.

Servo axes in this class typically use PWM carriers in the high-kilohertz region so that switching artefacts sit above the most sensitive part of human hearing. High-resolution FOC or motion MCUs with fast current loops, precise encoder or resolver interfaces and strong phase-current calibration are favoured. Torque-ripple suppression combines accurate angle sensing, per-axis phase-current balancing and finely tuned dead-time in the gate driver. Light spread-spectrum PWM may be used to soften residual tones, but always within limits that preserve multi-axis synchronisation and tight positioning. Suitable IC combinations pair motor-control MCUs with hardware FOC and ADC–PWM synchronisation, isolated gate drivers with fine dead-time and dv/dt control, and three-phase shunt or ΣΔ current-sensing AFEs matched across channels.

Premium appliances (washers, air conditioners and compressors)

High-end household appliances run in bedrooms, living rooms and open-plan apartments, and many include dedicated night or quiet modes. End users notice repetitive tones and cabinet vibration much more than small shifts in efficiency, yet long-term energy consumption and efficiency ratings still matter. Cost pressure is also intense in this volume-heavy segment, so most designs share a common control platform across several product tiers.

In these drives, typical PWM carriers sit at moderate frequencies during normal operation and move higher or use spectrum shaping in quiet modes. The focus is on reducing tonal content in speed ranges associated with spin cycles, start-up ramps and compressor operation. Torque-ripple mitigation relies on solid FOC tuning, reasonable angle resolution and basic three-phase current balancing and calibration. Randomized or spread-spectrum PWM is often reserved for premium modes where a slightly higher switching loss is acceptable. IC combinations usually involve appliance-optimised motor-control MCUs with integrated FOC libraries, compact high-voltage gate drivers with adjustable dv/dt, and cost-effective shunt-based current-sense amplifiers that still offer adequate matching and thermal stability for audible-noise control.

Commercial HVAC fans and pumps

Commercial HVAC systems serve office buildings, hospitals, data centres and public spaces. Fans and pumps often run continuously and are physically separated from occupants, yet structure-borne vibration and low-frequency noise can travel through ducting and building frames. At the same time, lifetime energy cost dominates the business case, so efficiency requirements are stringent and often regulated.

Typical PWM frequencies in this class are chosen as a compromise between efficiency, acoustic behaviour and EMI. Slew-rate control and careful gate-drive tuning are used to reduce dv/dt-induced noise and vibrations without pushing switching loss unreasonably high. Torque-ripple reduction focuses on phase-current balancing, dead-time calibration and avoiding operating points that excite duct or frame resonances. Randomized PWM may be introduced with narrow jitter ranges to soften tones while preserving predictable EMI characteristics. Preferred IC combinations include motor-control or PFC-capable MCUs with robust ADC triggering and protection features, industrial gate drivers with programmable dv/dt and comprehensive fault handling, and well-matched shunt or ΣΔ current-sensing front-ends suitable for higher power levels.

IC mapping & vendor examples

The mitigation techniques described on this page ultimately rely on a handful of IC building blocks: motion and FOC MCUs, gate drivers, phase-current and angle-sensing front-ends and, in some cases, isolation and safety monitors. Different vendors group these capabilities into families aimed at servo drives, appliances, HVAC and general industrial motion. This section does not list device ordering codes; instead it maps function blocks to the major suppliers that typically offer them.

For design teams, this mapping is a shortcut when shortlisting controller and driver families that can support noise and vibration improvements. For procurement and FAE discussions, it highlights which product lines and application notes are likely to be relevant when refining a drive toward quieter and smoother operation without losing efficiency or EMC margin.

Vendor portfolios evolve quickly, so any new design should still be checked against the latest product families, reference designs and application notes. However, the mapping above provides a starting point for aligning acoustic and vibration requirements with the appropriate classes of motion MCUs, gate drivers and sensing ICs from the major suppliers.

Internal linking & boundaries

Acoustic and vibration tuning makes sense only after the drive already runs reliably. This page is meant for situations where the motor can start, close the loop and deliver torque, but the noise level, tonal character or mechanical feel still fall below the application expectations. It is not a replacement for basic FOC control, power-stage design or EMC engineering and should be read as a dedicated optimisation layer on top of those topics.

When the motor cannot start robustly, stalls easily or shows unstable behaviour, the priority is to fix the control architecture and loop design. Likewise, if IGBT or MOSFET temperatures are already close to limits, or system efficiency is far from target, any increase in PWM frequency or spectrum shaping for acoustic reasons will only hide deeper issues. If conducted or radiated emissions fail compliance testing, the EMC subsystem and its filtering and shielding strategy set the boundaries for any spread-spectrum or random PWM features mentioned on this page.

Where to start for common field symptoms

The table below gives a quick routing guide for typical complaints. In each case the first stop is the page that owns the underlying discipline, and this acoustic and vibration section becomes the follow-up optimisation step once the base design is under control.

In summary, this acoustic and vibration section belongs to the “experience and refinement” branch of the Motor & Motion Control cluster. Fundamental loop stability is covered in the FOC Controller / Motion MCU topic, device stress and efficiency are handled in the Servo Power Stage and BLDC Driver topics, and compliance with emission limits is treated in the EMC Subsystem topic. Once those foundations are in place, the techniques on this page can be applied to push a working drive toward a quiet and refined user experience.

FAQs × 12 — Acoustic & vibration mitigation

This FAQ collects the most common questions that come up when a motor drive already runs but still feels noisy or rough. Each short answer points to the key levers for PWM planning, torque ripple, current balancing and IC selection so noise and vibration decisions are easier to make and justify.