Multi-Axis Servo Drive with ΣΔ Sensing & PWM Control
← Back to: Industrial Robotics
This page focuses on the power stage of a multi-axis servo drive: how PWM outputs, gate drivers, isolated current and voltage sensing, isolated DC-DC rails and fault signals fit together. The goal is to make drive performance, protection behaviour and safety hooks reviewable in a single, practical block diagram.
What problems does a multi-axis servo drive architecture really solve?
Industrial robots and positioning systems push the servo drive hard: multiple axes share a DC bus, torque steps are fast and EMI margins are tight. The power stage must keep phase current sensing low-latency and symmetric across axes, while surviving high dv/dt and repeated short-circuit events.
Typical field issues include:
- Phase current measurements lag the PWM update, so over-current protection reacts too late.
- PWM dv/dt couples into sense lines, making calibration and tuning inconsistent across axes.
- Isolated DC-DC rails brown out or restart under dynamic load, causing random drive trips.
- Fault pins and status flags are inconsistent across boards, so safety and diagnostics are hard to audit.
During sourcing and design review, several questions must be answered clearly:
- Can the ΣΔ current and voltage sampling be synchronised to the PWM scheme and switching frequency?
- Do the isolated DC-DC converters and isolation components withstand the worst-case PWM dv/dt and surge events?
- Are fault and diagnostic outputs clean, digital and routeable all the way to the main or safety MCU?
The rest of this page translates these review questions into a concrete multi-axis servo drive block diagram that can be used as a checklist when selecting ICs and modules.
Multi-axis servo drive system diagram: PWM, gate drive, isolated sensing and DC-DC
The block diagram below groups the multi-axis servo drive into four main elements: PWM outputs from the controller, gate drivers and power stages, isolated current and voltage sensing, and isolated DC-DC supplies. Fault and diagnostic signals are routed on a dedicated path so they can be used by the main controller or a separate safety MCU.
The intent is to separate what must be decided at the drive level from what belongs in the robot controller, feedback encoders or safety PLC pages. Networking, kinematics and high-level safety logic are deliberately kept out of this diagram.
PWM control schemes in a multi-axis servo drive
Before selecting gate drivers and current sensing devices, the PWM scheme for each axis needs to be fixed. The choice between unipolar, bipolar and space-vector modulation has direct impact on loop bandwidth, EMI behaviour, switching losses and the minimum microcontroller or motion-controller performance required.
The table below compares the main options used in industrial multi-axis servo drives. The focus is on practical trade-offs: how wide the usable bandwidth is, how hard it is to pass EMC testing, and which motion platforms each scheme typically ships in.
| PWM mode | Key strengths | Main limitations | Typical applications |
|---|---|---|---|
| Unipolar PWM | High effective bandwidth, simple timer setup, easy to scale across many axes with shared carriers. | Generates strong common-mode dv/dt and EMI, higher harmonic currents in the DC bus. | Compact multi-axis drives where cost and response speed are prioritised over EMC headroom. |
| Bipolar PWM | Better EMI behaviour, lower common-mode swing and more predictable filter design. | Control implementation is more complex, effective bandwidth can be lower than unipolar for the same switching frequency. | High-precision machines and test stands where torque ripple and acoustic noise matter. |
| SVM (Space Vector) | High DC bus utilisation and efficiency, smooth current waveforms, good compromise between losses and performance. | Requires higher MCU or motion-controller performance, implementation and debugging are more demanding. | Medium to high power industrial robots, gantries and machine tools with tighter efficiency and thermal budgets. |
Regardless of which PWM mode is selected, the rest of the power stage must be sized around it: dv/dt and common-mode voltage drive the isolation and DC-DC design, while the effective update rate defines how fast the current sensing chain must respond.
In a multi-axis servo drive, it is common to mix schemes across product families but keep one scheme per platform. The diagrams and tables on this page focus on how PWM choices interact with current and voltage sensing, rather than on motion-control algorithms.
Current and voltage sensing options for multi-axis servo drives
Once the PWM scheme and switching frequency are known, the next decision is how to sense phase currents and DC bus voltage. The sensing method sets the achievable protection latency, measurement noise floor, isolation strategy and total bill-of-materials cost.
Servo drives typically combine several approaches: shunts and amplifiers near the power stage, sigma-delta modulators for isolated feedback, Hall devices in low-bandwidth paths and fast transimpedance amplifiers in specialised sensing channels. The table below highlights where each option fits best in an industrial robot drive.
| Method | Latency | Noise level | Cost | Isolation | Typical range / use |
|---|---|---|---|---|---|
| Sigma-delta (ΣΔ) modulators | Low latency when clocked and filtered correctly; directly tied to oversampling ratio. | Medium noise, but shaped out of band and well suited for digital filtering. | Medium (✓✓) | Requires digital isolation or isolated clock and data framing. | Up to several MHz of signal bandwidth; isolated phase current and DC bus sensing in high-performance drives. |
| Shunt + instrumentation amplifier (INA) | Medium latency, dominated by amplifier bandwidth and ADC sampling strategy. | Low noise when layout and Kelvin sensing are done carefully. | Low (✓) | No intrinsic isolation; needs separate isolation if used on the high side. | Up to tens of kHz of bandwidth; local phase current and bus-sense in grounded sections of the drive. |
| Hall-effect sensors | Higher latency and slower response; dominated by magnetic and signal-conditioning time constants. | Higher intrinsic noise and offset drift than precision shunt solutions. | High (✓✓✓) | Provides galvanic isolation by nature of the sensing principle. | Up to a few kHz of bandwidth; suitable for slower axes, current monitoring and protection in lower-speed channels. |
| Transimpedance amplifier (TIA) based sensing | Medium latency, tuned by feedback components and front-end design. | High-frequency noise must be managed carefully; very sensitive to layout and parasitics. | Medium–high (✓✓) | Requires optical or dedicated isolation if used in high-voltage paths. | Up to multi-MHz; specialised current or light-sensing channels in laser profilers and vision-related modules. |
In the core servo drive, sigma-delta modulators and shunt-plus-amplifier chains cover most needs. Hall sensors and TIA-based front-ends are usually reserved for lower-bandwidth axes, auxiliary monitoring channels and specialised sensing modules, rather than for the main high-speed current loops.
This page keeps the discussion at the level of sensing hardware and isolation topology. Motor control algorithms, observers and sensor fusion for encoders and resolvers belong to the robot controller and feedback pages, not to the power-stage view shown here.
Isolated DC-DC design essentials for a multi-axis servo drive
In a multi-axis servo drive the isolated DC-DC converters are often quiet single points of failure. They feed gate drivers and sensing AFEs that must survive high dv/dt, short-circuit events and rapid load transients across several axes. Poor DC-DC choices easily show up as random trips, noisy measurements or unexplained resets in the field.
The design task is not limited to picking an isolation rating and power level. Isolation strength must match real PWM dv/dt, magnetic components must support transient currents, synchronous rectification must be justified against complexity, and the layout must keep high-edge currents away from sensitive sensing circuits. Protection devices such as eFuses bridge the DC bus and the DC-DC islands and therefore deserve a deliberate placement strategy.
Isolation rating vs. real PWM dv/dt stress
Isolation data sheets usually quote a kilovolt rating for a one-minute test, but multi-axis servo drives stress the insulation every switching cycle. High dv/dt edges on the bridge generate common-mode currents through the stray capacitances of transformers, capacitors and isolation devices. The selected DC-DC and digital isolators should not only meet clearance and creepage requirements, they should tolerate the highest expected dv/dt of the chosen PWM scheme with margin.
When unipolar PWM or aggressive SVM is used, the common-mode swing can be large and fast. In such cases it is often safer to use isolation components with specified CMTI ratings and characterised common-mode behaviour instead of ad-hoc wound parts. For drives that share a DC link between many axes, these dv/dt events add up; the DC-DC isolation must be selected as a system element, not only per axis.
Magnetic core sizing vs. transient current
Average DC-DC power is usually modest compared with the motor phases, but peak currents can be surprisingly high. Gate drivers charge and discharge gate capacitances in bursts, clamp networks absorb fault energy, and multiple axes may start or re-enable simultaneously. Core size and winding design therefore need to be checked against worst-case transient conditions, not only against nominal power at steady-state.
A practical review includes short tests where all axes are enabled at once, the drive is recovered from a stall or short-circuit condition, and the transient Vout and Iout of each DC-DC rail are captured. Any saturation, deep droop or prolonged overshoot is a sign that the magnetics or control loop need adjustment before the design is frozen.
Synchronous rectification vs. diode rectification
Synchronous rectification is attractive in higher power rails or in low-voltage outputs where diode losses dominate. In a multi-axis servo drive this typically applies when a single higher-power DC-DC feeds many gate drivers or AFEs. The efficiency gain, however, comes with tighter control requirements and more sensitivity to layout and timing errors, which can translate into ringing or even oscillation during transients.
For smaller, axis-local DC-DC rails the extra complexity is often unnecessary. Diode rectification delivers lower efficiency, but the simpler control loop can be more tolerant of layout compromises and production variation. The choice between synchronous and diode rectification should be driven by thermal budget and derating analysis, not by habit or by copying reference designs from unrelated power levels.
Layout: gate-driver distance and common-mode return paths
The physical placement of the DC-DC modules relative to the gate drivers determines how high-frequency currents flow. Gate driver supply loops should be compact and close to the switching devices, using short VDD and return paths with tight coupling. The DC-DC itself does not always need to sit directly beside the driver; local decoupling capacitors and carefully routed traces can bridge moderate distances without turning the supply loop into an antenna.
At the same time, the common-mode return paths between primary and secondary must be controlled. High dv/dt edges will drive current through parasitic capacitances regardless of intent. The aim is to keep that current confined to regions that are tolerant of noise and to avoid crossing sensitive sensing traces or reference nodes. This topic links naturally into layout and EMC guidelines, but from the drive perspective the minimum requirement is to keep gate-driver and DC-DC loops defined and compact.
Where to place eFuses around the DC-DC rails
Electronic fuses and smart high-side switches sit between the shared DC link and the isolated subsystems. Placing an eFuse on the DC bus side of a DC-DC module protects the bus against a shorted converter and allows each DC-DC island to be disconnected without collapsing the entire drive supply. Placing eFuses on the secondary side improves fault granularity by limiting the impact of a failure in one gate-driver cluster or AFE channel, but puts more stress on the protected devices and requires ratings that match the isolation level.
Many robust multi-axis drives therefore combine both approaches: a primary-side protection stage for each DC-DC and secondary-side current limiting or eFuses for the most critical loads. The detailed selection and coordination of eFuses and smart high-side switches belongs to the dedicated protection and eFuse topic, but this page defines where those devices should sit relative to the DC-DC blocks in the servo drive power-stage view.
All of the above decisions shape how faults will later be detected and reported. Under-voltage lockout, overload and thermal faults from the DC-DC and gate-driver rails must feed into a consistent fault-detection framework so that host and safety controllers can act on them predictably.
Fault detection paths in a multi-axis servo drive
Once the power stage, sensing and isolated DC-DC rails are defined, the remaining task is to expose their failures in a consistent way. Over-current, short-circuit, under-voltage, over-temperature and sigma-delta status bits all originate in different ICs, yet host and safety controllers should see a predictable set of fault lines and status maps.
The table below groups typical drive-level faults by source and by interface format. Each entry indicates whether it is realistic to feed the signal directly to a safety MCU or STO chain, or whether it is better handled by the host controller as part of diagnostics and derating schemes.
| Fault type | Source block | How it is read | Suitable for safety MCU? | Typical reaction |
|---|---|---|---|---|
| OC / SC (over-current / short-circuit) | Gate driver, desaturation detection or fast current limit. | Dedicated digital fault pin, active-low or active-high, often latched until reset. | Yes. Ideal for direct connection to safety MCU and hardware gate-disable or STO path. | Immediate hardware gate-off, log event, optional controlled restart after cool-down. |
| UVLO (under-voltage lockout) | Isolated DC-DC converters, gate driver VDD rails, internal bias supplies. | Digital UVLO flag, analogue VDD monitor or status bits in a control or power IC. | Yes, when de-glitched and combined into a clean “supply-good” or “supply-fault” line. | Inhibit drive outputs to avoid half-on gate states, then decide between retry and safe stop. |
| OT (over-temperature) | Gate drivers, shunt resistors, DC-DC modules or PCB thermal sensors. | Digital OT flag, analogue temperature output, NTC into an ADC or comparator threshold. | Yes, typically after being classified into warning and shutdown levels. | Derate currents or frequency first, then trigger a protective stop if temperatures keep rising. |
| FS / status from ΣΔ AFE or isolated ADC | Sigma-delta modulators, isolated ADCs and digital isolation devices. | Serial status bits, CRC error flags, frame sync loss indicators, sometimes a summary pin. | Yes, but usually via host-controlled logic; often treated as “data quality” rather than hard fault. | Mark measurements invalid, shrink torque limits or stop the affected axis if errors persist. |
The fastest faults, such as over-current and short-circuit detection, are best wired into a dedicated hardware chain that can disable gate drivers without waiting for software. Supply and temperature faults often pass through simple logic that filters chatter and groups related events before signalling to host and safety controllers. Data-quality faults from sigma-delta AFEs sit slightly apart: they indicate that control decisions should be based on validated data only.
From the perspective of a multi-axis servo drive board, the key outcome is a small, well documented set of fault lines and registers that can be reviewed during design and sourcing. Higher-level safety functions, such as how a safety PLC combines those lines into STO and safe-stop categories, are defined in the safety-controller pages, not in the power-stage view here.
IC mapping for a multi-axis servo drive power stage
This section gives a light-touch mapping between the blocks on this page and representative IC families from common vendors. The goal is to show which kinds of devices typically sit in the PWM, gate driver, sensing and isolation roles of a multi-axis servo drive, not to copy data-sheet content or lock the design to any specific part number.
Each family below stands for a class of solutions. During sourcing and design reviews you can look for functionally similar options from the same vendor or from others, as long as the interface and robustness match the drive architecture defined in this topic.
| Vendor | IC family | Role in this page |
|---|---|---|
| Texas Instruments (TI) | DRV8412 / AMC1306 | Bridge driver and isolated sigma-delta AFE pairing for phase current and DC bus sensing. |
| Renesas | HIP4086 | High-side and half-bridge gate driver family for discrete multi-phase bridges. |
| Analog Devices (ADI) | ADuM7701 / MAX14866 | High-CMTI digital isolation for sigma-delta data, clocks and fault/status lines. |
| Infineon | IRS2007S | Robust half-bridge gate driver family for compact, repeatable power-stage layouts. |
TI – DRV8412 and AMC1306: bridge driver plus sigma-delta sensing
Devices such as DRV8412 and AMC1306 illustrate how a bridge driver can be paired with an isolated sigma-delta AFE to implement the power and sensing chain shown in this topic. The driver handles PWM bridge switching, while the sigma-delta AFE delivers isolated current or bus-voltage feedback into the digital domain. This combination aligns directly with the PWM-mode and current-sensing choices discussed earlier on this page.
In practice you can look for equivalent pairings that offer stable gate-drive behaviour, sigma-delta outputs and clear fault pins, even if they come from different vendors or newer families. The key is that the interface and isolation scheme match the servo drive’s timing, dv/dt and fault-reporting requirements, not that any single part number is used.
Renesas – HIP4086: high-side gate driver for multi-phase bridges
HIP4086 represents a class of high-side and half-bridge gate drivers used when the servo power stage is built from discrete MOSFET bridges. These devices are well suited to the multi-phase bridge blocks in a multi-axis servo drive, where several legs must share a common DC link but still be driven and protected independently.
When selecting this type of gate driver, the focus is on how well the device supports the required PWM frequency, dv/dt levels and desaturation or over-current signalling, so that its fault outputs can plug cleanly into the fault-detection framework described on this page. Other vendors offer comparable high-side driver families; the same evaluation criteria apply.
ADI – ADuM7701 and MAX14866: high-CMTI digital isolation for signals
Families such as ADuM7701 and MAX14866 are examples of high-CMTI digital isolation devices used to carry sigma-delta data, clocks and fault lines across the isolation barrier. In a multi-axis servo drive they sit on the paths between the sensing AFEs, gate drivers and the host or safety MCUs, and must tolerate the same dv/dt environment as the isolated DC-DC rails.
When mapping this role to your own design, it is helpful to treat the isolator as a “communication fuse” between the drive board and the controller board. Robust CMTI, predictable common-mode behaviour and a channel count that matches the sigma-delta and fault interfaces on this page are more important than any specific publication number or package.
Infineon – IRS2007S: proven half-bridge gate driver families
IRS2007S stands for a family of half-bridge gate drivers that are often chosen for compact, repeatable power-stage layouts. In a multi-axis servo drive, parts from this class are used where designers prefer simple, well understood gate drivers that have already seen extensive use in similar voltage and dv/dt ranges.
The main selection criteria are long-term availability, clear protection behaviour and clean fault-pin interfaces. As long as those requirements are met, you can substitute other half-bridge driver families without changing the overall architecture shown in this topic. The mapping here is meant to illustrate the type of device that fits each block, not to prescribe a fixed bill of materials.
Request a Quote
FAQs for multi-axis servo drive power stage and sensing
I use these twelve questions to sanity-check my own multi-axis servo drive designs. Each answer keeps the focus on power stage, isolated sensing and fault handling, so I can talk clearly with suppliers and reviewers without drifting into motor-control algorithms or safety PLC details that belong on other pages.
When I scale from a single-axis design to a multi-axis servo drive, what really changes in the power-stage and sensing design?
When I move from a single-axis design to a multi-axis servo drive, the big change is how shared resources behave under worst-case events. I have to plan the DC link, isolated supplies, current sensing and fault lines as a system, so one axis fault does not drag the entire cabinet down.
Why does sigma-delta current sensing have to be synchronised with the PWM, and how tight does that timing alignment need to be?
I synchronise sigma-delta sampling with the PWM so the controller always sees consistent current information relative to each switching cycle. If the timing drifts, I get phase error, extra noise and slower protection. In practice, I plan clocking and filtering so the effective delay and jitter are well inside my current-loop bandwidth.
When do I actually need sigma-delta current sensing instead of relying only on shunt plus instrumentation amplifiers?
I reach for sigma-delta current sensing when I need isolated, wide-bandwidth feedback and high common-mode robustness on a shared DC link. For small drives with modest dv/dt, a shunt plus amplifier often covers the job. Once isolation, tighter accuracy and cleaner digital interfaces matter, sigma-delta AFEs start to earn their cost.
How do I choose between unipolar, bipolar and space-vector PWM for a multi-axis servo drive platform?
I pick the PWM scheme based on my EMC targets, efficiency and controller margin. Unipolar gives me simple, high bandwidth at the cost of more EMI. Bipolar helps when I care about acoustic noise and torque ripple. Space vector makes sense once bus utilisation and thermal limits dominate, usually in medium to high power robots.
Is it safe to let several axes share the same isolated DC-DC, and what limits or pitfalls should I watch for?
I only let axes share an isolated DC-DC after I have checked transient current, worst-case start-up and short-circuit recovery. If one axis event can pull the rail down or inject noise into neighbours, I split the supply. Sharing helps cost and efficiency, but I never accept a design where one fault collapses all gate-driver rails.
Should I place eFuses on the DC bus side or on the secondary side of each DC-DC, and how does that change fault behaviour?
I use bus-side eFuses to keep a failed DC-DC from dragging down the shared DC link, and secondary-side protection to localise faults in one gate-driver or sensing island. The exact split depends on power level and safety goals, but I always think in terms of limiting how far energy and faults can propagate.
When selecting gate drivers for a servo power stage, should I prioritise CMTI ratings or dv/dt capability, and how do they relate?
I treat dv/dt capability and CMTI as two sides of the same environment. The power stage sets the dv/dt I need to live with, and the gate drivers and isolators must survive that level without false triggering. In practice, I pick parts with proven CMTI margins above my worst-case dv/dt instead of relying on optimistic layout alone.
How can I tell whether a fault signal from the drive is reliable enough to feed a safety MCU or STO input?
I only trust a fault signal into a safety MCU after I have defined its source, filtering and test coverage. That means I know exactly which blocks drive it, how de-glitching is done and how it behaves under injected faults. If I cannot reproduce and log its behaviour on the bench, I treat it as an advisory signal only.
When a sigma-delta AFE reports FS, CRC or frame errors, should I immediately stop the axis or just derate the drive?
When a sigma-delta AFE reports data errors, I first treat them as a data-quality alarm, not an automatic emergency stop. Short bursts can be retried or filtered; persistent errors mean I should derate torque or slow the axis. If the errors remain, I plan to stop that axis gracefully and flag the drive for investigation.
Which bench tests should I run on early prototypes to validate the isolated DC-DC rails and the fault-detection framework?
On early prototypes I run combined tests that stress both the DC-DC rails and the fault network. I enable several axes at once, inject short-circuits, force UVLO and over temperature, and deliberately break sigma-delta links. I watch how each rail responds and whether every fault is reported to the host and safety controllers exactly as planned.
How can I sanity-check IC choices for drivers, AFEs and isolators without reading every data sheet in detail?
I start by mapping each candidate IC onto the blocks on this page: bridge driver, current AFE, isolated ADC or digital isolator. Then I check that its interfaces, isolation class and fault pins match the architecture I want. Only after that quick fit check do I dive into data-sheet detail, instead of comparing random parameters in isolation.
When I ask a supplier to quote or customise a multi-axis servo drive, which power-stage and sensing details must I specify up front?
When I talk to a supplier, I always specify axis count, DC bus voltage, power range and the PWM scheme I expect to use. I add my isolation, current sensing and fault interface expectations, including how OC, UVLO, OT and data errors should surface. This saves time and avoids receiving proposals that cannot meet my protection or diagnostics needs.
