123 Main Street, New York, NY 10001

Isolated Feedback for Offline Power Supplies

← Back to: Power Supplies & Adapters

Isolated feedback lets you regulate tightly, meet safety standards and carry protection signals across the isolation barrier, far beyond what simple primary-side regulation can do. On this page you can choose the right optocoupler, isolation amplifier or ΣΔ path, design stable multi-output loops and verify layout, fault and IC selections with a practical checklist.

What this page solves

This page explains when simple primary-side regulation (PSR) is no longer sufficient and why a true isolated feedback loop is required for modern offline adapters and PSUs. The focus is on regulation accuracy, long-term stability and safety across demanding AC-DC and DC-DC use cases.

When primary-side regulation starts to break down

PSR works by inferring the secondary output from transformer auxiliary windings and primary current waveforms. This approach can be cost-effective in simple, single-output adapters, but its internal assumptions become fragile when the system is pushed beyond those conditions.

  • Multi-output supplies introduce cross-regulation: only one rail aligns with the PSR model, while other rails drift with load distribution.
  • Remote loads and long cables add losses and voltage drops that PSR does not directly “see”, leading to inaccurate regulation at the actual load point.
  • Component tolerances, transformer variation and aging slowly move the real system away from the originally tuned PSR model.

Pain points in multi-output, strict CV/CC and safety-critical designs

Many adapters and PSUs must now hold tight CV/CC limits across a wide input and load range, while meeting stringent safety and compliance constraints. In these systems, regulation based on approximate primary-side signals is often not enough.

  • Fast transient loads and protocol-driven profiles (such as fast charging or digital loads) expect predictable voltage and current behavior at the secondary output pins.
  • Strict CV/CC requirements demand that the control loop tracks the real output node, not a proxy waveform on the primary side.
  • Long service life in industrial and medical equipment requires regulation schemes that tolerate component drift and transformer aging without falling out of spec.

Why true isolated feedback is needed

A true isolated feedback path samples the critical secondary node directly, processes the error on the secondary side and transfers only the necessary feedback information across the isolation barrier. This enables:

  • Accurate CV/CC regulation at the real load point, even with multiple outputs and remote wiring.
  • Clear separation between safety isolation and control: isolation devices can be chosen to meet specific creepage, clearance and surge ratings.
  • Predictable loop behavior over temperature and lifetime when optocouplers, isolation amplifiers or ΣΔ-based feedback devices are selected and compensated correctly.

Common pitfalls with traditional optocoupler loops

Optocoupler-based feedback is often treated as a simple add-on, but careless design can create subtle reliability and stability issues:

  • CTR spread and aging: wide current transfer ratio variation across devices and over lifetime changes loop gain and regulation accuracy.
  • Bandwidth and phase lag: limited optocoupler speed adds delay that can erode phase margin and cause ringing after load steps.
  • Noise injection: poor PCB layout around the feedback path couples high dv/dt switching noise into the COMP node, causing jittery output and EMI sensitivity.

Later sections of this page describe how isolated amplifiers, ΣΔ modulators and carefully compensated optocoupler loops can address these issues in a systematic way.

What this page does not cover

To keep the scope focused, this page does not attempt to cover every type of isolation used in power supplies. Instead, it concentrates on the isolated feedback path for CV/CC regulation.

  • Gate-drive isolation for MOSFETs, IGBTs, SiC or GaN devices is handled in the separate “Isolated Gate Driver” topic.
  • Precision isolated measurements of current and voltage for telemetry and monitoring are discussed in “Current/Voltage Sensing”.
  • Reference voltages and bias rails for controllers, amplifiers and comparators are covered in “References & Bias”.
From primary-side regulation to true isolated feedback Comparison diagram showing a primary-side-only regulation scheme alongside a true isolated feedback loop, highlighting the secondary sense point and isolation device in the regulation path. Primary-side regulation PWM controller Aux winding sense model Model-based only Single output, simple loads Limited lifetime accuracy True isolated feedback Sense point at output Error amp TL431 / op amp ISO opto / amp / ΣΔ PWM COMP / FB Tight CV / CC

Where isolated feedback sits in the PSU

Isolated feedback is one of several signal paths that cross the isolation barrier in an offline adapter or PSU. It is dedicated to closing the regulation loop between the secondary output and the primary-side controller. Other isolated paths carry gate-drive information or measurement data and are covered in separate topics.

System-level view of the regulation loop

A typical AC-DC or isolated DC-DC supply can be viewed as three major domains: the primary power stage with its controller, the isolation barrier, and the secondary power stage with output filtering and sensing. The isolated feedback path connects a carefully chosen sense point on the secondary side back to the primary-side feedback and compensation pins.

  • Primary domain: rectifier or PFC stage, transformer primary, and a PWM or digital controller with FB/COMP and current-sense inputs.
  • Isolation barrier: optocoupler, isolation amplifier, ΣΔ modulator or digital isolator that transports error information across the safety boundary.
  • Secondary domain: rectification, output LC filter and an error amplifier that compares the real output to a reference.

Three main information flows across isolation

From a control perspective, the isolation barrier carries three distinct categories of information in a modern PSU:

  • Regulation loop (this page): error or duty-cycle information that keeps the output voltage or current within specification under changing conditions.
  • Gate-drive path: drive signals for primary-side MOSFETs, IGBTs, SiC or GaN devices, handled in the “Isolated Gate Driver” topic.
  • Measurement and telemetry: isolated voltage, current or temperature data used by digital controllers and PMBus/SMBus interfaces, covered in “Current/Voltage Sensing” and “Digital PSU Controller (PMBus)”.

The rest of this page focuses only on the regulation loop path and how to design its isolated segment without compromising accuracy, stability or safety.

Where isolated feedback sits in a power supply Block diagram of a power supply showing the primary power stage, isolation barrier and secondary stage, with separate paths for regulation feedback, gate drive and telemetry. Primary domain Controller FB / COMP / gate drive Primary power stage PFC / half-bridge / LLC Isolation barrier Regulation feedback opto / iso amp / ΣΔ Gate-drive isolation isolated drivers Measurement & telemetry ADC / ΣΔ / digital isolators Secondary domain Error amp & sense TL431 / op amp at output Secondary power stage rectifier, LC filter, loads Regulated outputs

Device families for isolated feedback

Several isolation device families can be used to close the feedback loop between the secondary output and the primary controller. Each family offers a different trade-off between accuracy, bandwidth, cost and lifetime behavior, and not every isolated device is equally suitable as a loop element.

Optocoupler with TL431 or error amplifier

The classic combination of a TL431-style shunt regulator and an optocoupler LED and phototransistor remains the most common feedback solution in offline adapters and auxiliary supplies. The TL431 acts as a reference and error amplifier, while the optocoupler transports the error current across the isolation barrier into the primary COMP or FB node.

  • Well suited for low-to-mid power adapters and modules where cost and footprint are tightly constrained.
  • Loop behavior is strongly influenced by current transfer ratio (CTR), which varies across devices, temperature and aging.
  • Limited bandwidth and additional phase lag must be considered when setting compensation and target loop bandwidth.

Isolation amplifiers with analog outputs

Isolation amplifiers provide an isolated analog representation of the sensed secondary voltage or current. They behave like precision amplifiers with a defined gain and isolation barrier, and are typically followed by an ADC or directly connected to an analog controller input.

  • Offer higher linearity and tighter gain accuracy than optocouplers, which improves CV/CC accuracy over temperature and lifetime.
  • Provide wider and more predictable bandwidth, enabling higher loop crossover frequencies where the power stage supports it.
  • Require careful consideration of offset, gain drift and supply headroom to avoid subtle regulation errors at extreme operating points.

Isolated ΣΔ modulators and ADC-based feedback

Isolated ΣΔ modulators and ADC-based solutions convert the sensed quantity into a bitstream or digital code that crosses the isolation barrier and is processed by a digital power controller. They are widely used in high-performance PFC, LLC and multi-rail digital PSUs.

  • Enable flexible, digitally tuned control loops and non-linear control strategies that adapt to operating conditions.
  • Introduce additional latency from modulation, digital filtering and control computation, which must be included in phase margin analysis.
  • Quantization noise and resolution determine how finely output voltage or current can be regulated without introducing audible or visible artifacts.

Digital isolators for PWM, duty-cycle and fault flags

Digital isolators can also participate in the feedback loop when error information is encoded as duty cycle, pulse width or logic level on the secondary side. PWM-style signals, DAC-coded levels or discrete fault flags can be transported across isolation and decoded by the primary controller.

  • Suitable for architectures where a small secondary-side controller interprets sensing and sends a coded feedback signal to the primary.
  • Loop performance depends on timing resolution, propagation delay and pulse-to-pulse jitter through the digital isolator.
  • Well suited for dedicated fault and status signals that need robust isolation but do not require high analog resolution.

Across all families, suitability for feedback is defined by linearity, delay, temperature stability, noise behavior and how the device fails under stress. Later sections translate these device characteristics into practical loop design decisions.

Device families used in isolated feedback loops Block diagram showing four device families used for isolated feedback: optocoupler plus TL431, isolation amplifiers, isolated sigma-delta modulators and digital isolators. Isolated feedback device families Feedback loop element Opto + TL431 classic analog loop cost-focused designs Isolation amplifiers precise analog feedback higher bandwidth ΣΔ / ADC-based digital controllers programmable loops Digital isolators PWM / duty / flags coded feedback

Classic optocoupler and TL431 feedback loops

Optocoupler plus TL431-style reference remains the benchmark solution for isolated feedback in many offline PSUs. The topology uses a TL431 or similar shunt regulator as the error amplifier on the secondary side, and an optocoupler LED and phototransistor to convey the error signal back to the primary controller COMP or FB pin.

From secondary sense point to primary COMP pin

The loop begins at the chosen secondary sense point, usually the main regulated output after the LC filter. A resistor divider maps the output voltage to the TL431 reference input. The TL431 compares this scaled voltage to its internal reference and adjusts its cathode current accordingly.

  • The sense divider defines the nominal output voltage and contributes to overall tolerance and drift.
  • The TL431 cathode current is routed through a series resistor and into the optocoupler LED, turning output error into LED current.
  • The optocoupler phototransistor on the primary side converts LED current into a corresponding current at the controller COMP or FB node.

Together, the TL431 transconductance, the optocoupler CTR and the impedance of the COMP network define the effective loop gain and dynamic behavior.

CTR spread, aging and loop gain margins

Optocouplers are specified with a current transfer ratio (CTR) that spans a wide range. The actual CTR depends on device lot, LED current, temperature and operating time. In a feedback loop, this variation directly affects the available loop gain.

  • CTR_min at end-of-life is the critical parameter when checking that the loop still has enough gain to meet regulation targets.
  • CTR_max can increase high-frequency gain, potentially amplifying noise and reducing phase margin if compensation is too aggressive.
  • CTR drift over temperature and lifetime should be treated as a design input, not just a production test parameter.

Establishing loop gain and phase margins with realistic CTR limits helps prevent borderline behavior where units pass at room temperature but drift out of specification after years of operation.

Compensation structure around TL431 and COMP

The TL431 and COMP networks together implement the compensation that shapes closed-loop response. The most common approach uses capacitors and resistors around the TL431 to create a type II or type III compensator, while the primary COMP pin includes an additional RC network that interacts with the optocoupler current.

  • Secondary-side components around the TL431 set low-frequency gain and introduce zeros to boost phase margin near the crossover frequency.
  • Primary-side COMP components help filter optocoupler current, limiting high-frequency noise and defining the dominant pole of the loop.
  • Compensation values should be chosen with the power stage poles and zeros in mind, avoiding excessive bandwidth or marginal phase margin.

Detailed compensation design is usually handled in controller-specific documentation, but understanding how TL431 and optocoupler characteristics feed into that design is essential for robust loops.

Biasing and light-load behavior

The TL431 and optocoupler must remain in a controlled operating region across all loads, including standby and light-load modes. Operating too close to minimum TL431 cathode current or very low LED current can push both devices into non-linear regions where loop behavior becomes unpredictable.

  • TL431 cathode current should stay above the minimum specified for accurate reference and gain behavior, even in light-load conditions.
  • Optocoupler LED current should be chosen so that CTR remains in a relatively linear region while still meeting efficiency and standby targets.
  • Controller burst or skip modes must be verified with the feedback network to avoid oscillation, jitter or large output droop.

Noise and EMI considerations for optocoupler loops

The optocoupler loop spans sensitive low-level circuitry on both the secondary and primary sides. If layout and filtering are not treated carefully, high dv/dt and switching edges can couple into the feedback path and disturb the loop.

  • Keep TL431 and optocoupler routing away from primary switch nodes and fast rectifier transitions to avoid capacitive coupling into the loop.
  • Use small RC filters where appropriate to attenuate high-frequency noise without compromising required loop bandwidth.
  • Minimize COMP node trace length and provide a clean reference return to avoid picking up common-mode noise.

A well-laid-out optocoupler and TL431 loop can provide robust, predictable regulation; a poorly treated loop can become the most fragile part of an otherwise solid PSU design.

Classic TL431 and optocoupler isolated feedback loop Block diagram showing a secondary sense point and TL431 driving an optocoupler LED, with the optocoupler phototransistor connected to the primary controller COMP pin. Secondary side Output sense point Divider to TL431 sets Vout TL431 / error amp ref + gain + comp LED optocoupler Primary side Photo transistor PWM controller COMP / FB node isolated error current Sense → TL431 → optocoupler → COMP

Isolation amplifiers and ΣΔ modulators for feedback

High-power PFC and LLC stages increasingly rely on isolation amplifiers and isolated ΣΔ modulators to transport accurate voltage and current information across the isolation barrier. These devices support higher loop bandwidths, tighter regulation and close integration with digital controllers compared to traditional optocoupler feedback.

Isolation amplifiers with analog outputs

Isolation amplifiers behave like precision amplifiers with a built-in isolation barrier. The secondary-side voltage or current is sensed, scaled and transferred as an analog signal to the primary side. That signal can feed an ADC in a digital controller or an analog FB/COMP pin in a high-performance analog controller.

  • Well suited for high-power LLC or PFC stages where tight CV/CC accuracy is required over temperature and lifetime.
  • Provide defined gain and bandwidth, simplifying prediction of loop crossover frequency and phase margin.
  • Remove dependence on optocoupler CTR and LED aging, improving long-term consistency of regulation behavior.

From the feedback-loop perspective, the most important isolation amplifier parameters are gain accuracy, linearity, offset and gain drift, bandwidth and group delay, output noise and fail-safe behavior during fault or power-down conditions. These characteristics directly determine steady-state regulation accuracy and dynamic response.

Isolated ΣΔ modulators and ADC-based feedback

Isolated ΣΔ modulators and ADC-based feedback solutions convert the sensed secondary quantity into a bitstream or digital code that crosses the isolation barrier. A digital power controller then reconstructs the signal with a decimation filter and applies a control algorithm to generate PWM or gate-drive signals.

  • Support fully digital control of PFC and LLC stages, enabling non-linear control laws and adaptive behavior.
  • Offer high effective resolution and excellent linearity when combined with appropriate oversampling and filtering.
  • Provide isolation ratings and creepage/clearance performance suitable for telecom, server and industrial PSUs.

For ΣΔ-based feedback, total latency must include modulator delay, digital filter group delay and controller computation time. This latency constrains the achievable loop bandwidth and must be accounted for when targeting specific crossover frequencies and phase margins. Quantization noise and filter design also shape output ripple and small-signal resolution.

Comparison to optocoupler-based feedback

Optocouplers, isolation amplifiers and ΣΔ modulators all provide galvanic isolation, but their feedback behavior differs significantly. Choosing the right device family requires evaluation of accuracy, delay, lifetime stability, design complexity and cost at the system level.

  • Linearity and accuracy: optocouplers exhibit non-linear CTR and significant spread, while isolation amplifiers and ΣΔ solutions deliver much tighter gain and offset specifications.
  • Delay and bandwidth: optocoupler bandwidth and delay are strongly device- and current-dependent; isolation amplifiers offer predictable analog bandwidth; ΣΔ chains add more delay but can be designed to meet a specific dynamic target.
  • Lifetime behavior: optocoupler CTR and LED output degrade with time, whereas isolation amplifiers and ΣΔ devices primarily exhibit semiconductor drift, typically smaller and more predictable.
  • Design complexity: optocoupler loops require careful compensation and aging analysis; isolation amplifiers simplify loop modeling but add analog interface design; ΣΔ-based solutions demand digital filter and firmware expertise.
  • Cost and integration: light-load adapters favor optocouplers; high-power, feature-rich PSUs can amortize the cost of isolation amplifiers or ΣΔ devices through improved performance and programmability.

Detailed current-sensing schemes and multi-channel measurement architectures belong to the “Current/Voltage Sensing” topic. The focus here is on how isolation amplifiers and ΣΔ modulators behave when they sit directly inside the main voltage or current regulation loop.

Isolation amplifier and ΣΔ paths in the feedback loop Block diagram showing secondary sensing feeding either an isolation amplifier with analog output or an isolated sigma-delta modulator, both driving a primary-side controller. Iso-amp and ΣΔ feedback options Secondary side Output sense point Sense and scaling divider / shunt front-end Isolation amplifier ΣΔ modulator bitstream output Primary side controller domain Analog feedback iso-amp output to ADC ΣΔ bitstream and filter digital reconstruction Digital power controller loops, limits, protections PWM / gate drivers PFC / LLC power stage

Sample/hold and multi-output feedback paths

Many offline adapters and PSUs provide multiple output rails while only one rail is directly controlled by the main feedback loop. The remaining rails rely on magnetic coupling, post-regulation or software supervision. Correctly structuring feedback paths is essential to avoid unstable interactions between outputs.

Master and secondary outputs in multi-output supplies

In a typical multi-output design, one rail is designated as the master output and is tightly regulated by the feedback loop. Other rails are derived from the transformer or from the master rail through post-regulators. The master rail sets the duty cycle or switching frequency, while secondary rails are tuned to meet their targets without trying to drive the main loop directly.

  • The master rail usually carries the highest power or the most critical load, such as a 12 V bus in a server or a 5 V main rail in an adapter.
  • Secondary rails often use magnetic coupling, DC-DC post-regulators or LDOs to reach their required voltage levels.
  • Only one feedback path should command the switching stage at any given time to prevent conflicting control actions.

Analog sample/hold and multiplexed feedback schemes

When several outputs need periodic attention, designers sometimes use analog multiplexing and sample/hold circuits to inspect more than one rail through a single error amplifier or isolation device. A multiplexer connects different outputs to the error amplifier in turn, and a hold capacitor maintains the previously sampled levels between updates.

  • Sampling periods must be short compared with the expected drift or disturbance on each rail to avoid large excursions between updates.
  • Switch resistance, leakage and hold-capacitor droop introduce additional error that must be budgeted in the regulation accuracy.
  • Loop compensation must tolerate the step changes that occur when the multiplexer switches from one output to another.

Analog sample/hold architectures can be powerful but demand careful verification of transient behavior; they are usually reserved for designs with clearly defined operating modes and priorities among outputs.

Digital controllers with multi-channel ADC and time-multiplexed sampling

Digital power controllers commonly employ multi-channel ADCs to monitor several output rails and internal nodes. One channel usually feeds the main regulation loop, while other channels provide supervisory information, cross-regulation compensation or protection thresholds.

  • The primary regulation loop is tied to a specific ADC channel that samples the master rail at a defined rate and phase.
  • Additional rails are scanned periodically to detect overload, undervoltage or thermal stress and to adjust offset or trim values where needed.
  • Time-multiplexed sampling introduces latency and potential skew between channels; both must be considered when multiple readings influence shared control decisions.

For digitally managed multi-output PSUs, the combination of ADC scheduling, filtering and control logic determines how effectively secondary rails can be supervised without destabilizing the master loop.

Avoiding unstable interactions between multiple feedback sources

A common mistake in multi-output designs is to let several outputs drive the same error amplifier or TL431 in an attempt to regulate all rails directly. When multiple dividers or compensation networks compete on one feedback node, the loop can exhibit unpredictable behavior and may oscillate or show strong cross-coupling during load transients.

  • Only one rail should set the reference for the main feedback loop; other rails should rely on post-regulation or controlled cross-regulation.
  • Shared TL431 networks should be avoided unless explicitly designed as part of a multi-rail compensation scheme with clear dominance of one master rail.
  • Protection thresholds can share certain components, but regulation loops must remain clearly partitioned to preserve stability.

The power-stage topology and magnetic design for multi-output supplies are covered in other topics. This section concentrates on the feedback paths and sampling strategies that allow multiple rails to coexist without compromising loop stability.

Sample/hold and multi-output feedback organization Block diagram showing a master output rail with direct feedback and secondary outputs monitored through analog or digital sampling paths in a multi-output power supply. Multi-output feedback and sampling paths Power stage transformer and rectifiers Master output rail primary feedback target Secondary output 1 post-reg / cross-reg Secondary output 2 post-reg / cross-reg Feedback and sampling domain Master feedback TL431 / iso-amp / ΣΔ Analog MUX and sample/hold periodic multi-rail sampling Digital controller ADC multi-channel, time-multiplexed Control core loops, priorities, limits master rail sense

Using the feedback path for protection and reporting

The isolated feedback path can do more than regulate output voltage or current. It also provides a convenient channel to transfer fault information such as OVP, UVP, OCP and OTP across the isolation barrier so that the primary controller can shut down the power stage and log events for diagnostics.

Injecting fault information into a TL431 and optocoupler loop

In a classic TL431 and optocoupler loop, secondary-side fault comparators can influence the error amplifier and indirectly the primary COMP node. Overvoltage, undervoltage or overtemperature detectors can adjust the TL431 reference or cathode current so that the loop forces a safe reduction in power or a complete shutdown.

  • OVP comparators often pull the TL431 reference or cathode in a direction that drives the optocoupler hard, pulling COMP toward a shutdown state.
  • UVP and severe load faults can be used to clamp the error amplifier so that the controller does not attempt to boost output under abnormal conditions.
  • OTP circuits can reduce duty cycle or trigger latch-off by overriding the normal error signal when critical components approach their temperature limits.

Protection logic must avoid simply nudging the setpoint near a threshold where the control loop and comparator fight each other. Robust designs use clear pull-down, clamp or latch behavior that moves the loop into a defined safe state instead of allowing repeated over- and undershoot around the fault threshold.

Dedicated isolated channels for fault flags and status reporting

Many systems separate regulation and protection information by using a dedicated optocoupler or digital isolator channel for fault flags. The regulation loop then continues to operate normally until the primary controller receives a discrete fault signal and issues an orderly shutdown or power derating command.

  • Secondary-side protection ICs or comparators drive a separate optocoupler or digital isolator, producing fault, power-good or ready signals on the primary side.
  • The primary controller interprets these isolated flags, disables PWM outputs, manages restart timers and logs the event over PMBus or another system interface.
  • Dedicated fault channels avoid distorting the main feedback loop and simplify compensation and stability analysis.

For complex PSUs, separating regulation and fault signaling paths helps achieve predictable loop behavior while still providing rich status reporting and fault logging at the system level.

Detecting optocoupler open or short failures with fail-safe behavior

Because the optocoupler sits directly in the regulation path, its failure modes must be considered as part of the safety strategy. An open LED or open transistor can remove feedback entirely, and without safeguards the primary controller might increase duty cycle and drive the output into a dangerous region.

  • COMP or FB bias networks on the primary side should be arranged so that loss of optocoupler current forces the controller into a low-duty or disabled state rather than full-power operation.
  • Startup self-tests and supervisory logic can monitor COMP voltage and output behavior to detect optocoupler faults and latch the supply off until service.
  • Shorted optocoupler behavior is also important; the design should treat a stuck-low COMP or forced feedback as a reason to enter a safe shutdown rather than continue partial operation.

Detailed trip thresholds, timing and restart strategies are addressed in dedicated protection topics such as “OV/OC/SCP Protection” and “eFuse and Hot-Swap”. The emphasis here is on how the feedback channel carries protection and reporting information across the isolation barrier while remaining stable and fail-safe.

Feedback path used for regulation, protection and reporting Block diagram showing a TL431 and optocoupler feedback loop with additional fault comparators and a separate isolated fault flag channel feeding a primary-side controller and event logger. Regulation, protection and reporting via feedback Secondary side Output rail sense point TL431 loop regulation path OVP / UVP comparators OTP / OCP comparators Fault comparators steer TL431 to pull COMP Fault flag isolator dedicated fault channel Primary side controller and reporting COMP / FB node regulation input PWM / digital controller duty, limits, restart Fault logging and reporting PMBus / system host Power stage PFC / LLC / flyback regulation feedback isolated fault flag

Isolation safety, creepage and EMI for feedback components

Feedback components such as optocouplers, isolation amplifiers and isolated converters must satisfy both control-loop requirements and safety requirements. Isolation rating, creepage and clearance, and PCB layout around these devices determine whether the PSU passes safety and EMI tests while maintaining stable regulation.

Isolation levels and creepage requirements

Safety standards distinguish between basic, reinforced and double insulation. Feedback components that bridge mains-referenced primary and SELV secondary must be selected with suitable insulation ratings, working voltage and surge capability, and their creepage and clearance distances must match the system requirements.

  • Optocouplers, isolation amplifiers and digital isolators specify internal creepage, clearance and working voltage ratings in their datasheets.
  • Medical, industrial and EV applications often require reinforced insulation, higher working voltages and larger creepage distances.
  • PCB creepage paths between primary and secondary copper must respect the same isolation requirements as the components themselves.

Isolation integrity depends on components and PCB layout together. Copper fills, vias and routing near isolation packages can unintentionally reduce creepage or create shortcuts if keep-out regions and slots are not planned carefully.

Layout guidelines for a quiet feedback loop

The feedback loop processes small signals and is sensitive to noise. High dv/dt nodes and large current loops in the power stage can couple into feedback traces and destabilize the control loop if layout does not provide enough separation and controlled return paths.

  • Keep sense dividers, TL431 or isolation amplifier inputs and reference grounds away from switching nodes and rectifier junctions.
  • Use compact loops and short traces for feedback signals, and route them over a quiet reference plane to minimize magnetic coupling.
  • Define a local analog ground region for feedback components and connect it to the main ground at a controlled point rather than through long shared paths.

Y capacitors, common-mode noise and feedback reference

Y capacitors create an AC reference between primary and secondary for EMI control. Their placement determines which part of the secondary ground is most strongly tied to primary common-mode voltage. If the feedback reference sits too close to a noisy Y-cap node, common-mode currents can disturb the sensed voltage.

  • Connect Y capacitors to a relatively quiet region of secondary ground rather than a point that already carries high ripple currents.
  • Place feedback reference components near the chosen “quiet” ground region so that the sensed voltage remains stable against common-mode swings.
  • Verify both EMI performance and control-loop stability when experimenting with alternative Y-cap placements or values.

Filtering common-mode interference in the feedback path

Parasitic capacitances between primary and secondary allow fast edges to couple into feedback networks. Without filtering, these disturbances can modulate the error amplifier input or COMP node and cause jitter, audible noise or outright oscillation in the control loop.

  • Use small RC filters between the sense divider and the TL431 or isolation amplifier input to attenuate high-frequency noise while preserving the desired loop bandwidth.
  • Add series resistors or small RC networks at isolation amplifier or optocoupler outputs to limit the impact of sharp spikes on the COMP node.
  • Partition feedback-related traces and components into a clearly defined low-noise zone, separate from power-switching loops and gate-drive routing.

System-level EMI filters, surge protection and medical isolation design are discussed in other topics. This section focuses on keeping isolated feedback components safe, layout-compliant and quiet enough to maintain stable regulation under real-world noise and stress conditions.

Isolation, creepage and EMI around feedback components Block diagram showing primary and secondary regions separated by isolation components, with creepage arrows, Y capacitors and a quiet feedback zone away from high dv/dt nodes. Isolation safety and EMI for feedback path Primary side (mains referenced) Secondary side (SELV) Opto / iso-amp Creepage distance along PCB Clearance in air High dv/dt zone switches and drains Controller and gate drive primary ground plane Y-cap Quiet feedback zone sense, TL431 / iso-amp Secondary ground and Y-cap reference low-noise area for feedback Avoid routing feedback through high dv/dt Clean route into quiet zone

Design checklist and IC role mapping for isolated feedback

This section provides a compact checklist for isolated feedback design together with an IC role map. The goal is to validate isolation strength, feedback architecture, dynamic behavior, failure modes and PCB layout, then match each requirement to a suitable class of isolation components.

Isolation, creepage and surge capability

  • Verify that the feedback path uses basic or reinforced insulation that matches the AC front-end and application standard.
  • Check that the working voltage, surge rating and isolation voltage in the optocoupler, isolation amplifier or ΣΔ modulator datasheet exceed the required mains or DC bus conditions.
  • Confirm that PCB creepage and clearance between primary and secondary copper around the feedback components meet or exceed system-level requirements.

Feedback target and multi-output strategy

  • Identify a single master feedback output rail and confirm that the divider, TL431 or isolation device senses that rail directly.
  • Document how secondary outputs are handled: magnetic cross-regulation, DC-DC post-regulators or LDOs, rather than multiple rails sharing one TL431 reference node.
  • If using analog MUX or sample/hold, verify that the sampling interval and hold time are compatible with worst-case drift on each rail.
  • If using a multi-channel ADC in a digital controller, ensure that the master rail’s sampling rate and phase support the intended loop bandwidth, while other channels are primarily used for supervision and trim.

CTR, gain and bandwidth across the operating range

  • For optocoupler loops, account for minimum CTR, CTR spread and CTR aging at high temperature and low LED current; confirm that the loop still has enough control range at light load and worst-case conditions.
  • For isolation amplifiers, confirm that gain error, offset and temperature drift are compatible with CV/CC accuracy targets and that bandwidth and group delay support the intended crossover frequency.
  • For ΣΔ or ADC-based feedback, include modulator delay, digital filter group delay and computation latency in the loop phase margin calculation, and verify that OSR and filter order deliver both the required bandwidth and noise performance.

Failure modes and fail-safe behavior

  • Evaluate optocoupler LED open or transistor open faults and verify that bias networks drive the COMP or FB node toward a safe low-duty or shutdown state, not toward uncontrolled full power.
  • Check optocoupler short behavior and confirm that a stuck-low COMP voltage is treated as a fault condition by the controller, not as a valid regulation state.
  • For isolation amplifiers and isolated error amplifiers, understand output behavior during loss of supply or internal fault and ensure that the power supply transitions into a defined safe mode.
  • For ΣΔ modulators and digital isolators, implement watchdogs, timeout counters or CRC checks so that frozen bitstreams or missing edges force PWM shutdown and fault logging.

PCB layout and dv/dt exposure of the feedback loop

  • Route feedback traces away from high dv/dt nodes such as primary switch drains, rectifier junctions and half-bridge midpoints, and keep small-signal loops compact.
  • Place TL431, isolation amplifier or ΣΔ front-end around a quiet secondary ground region and connect this region to the main ground in a controlled manner.
  • Verify that Y capacitor placement does not inject excessive common-mode noise into the feedback reference point and that RC filters adequately suppress high-frequency spikes without overly reducing loop bandwidth.

IC role mapping for isolated feedback implementations

After the checklist is complete, each design choice can be mapped to an IC category. The following families illustrate how different devices support isolated feedback paths; the examples are indicative and can be replaced by equivalent parts.

Shunt regulator and optocoupler pair

A TL431-class shunt regulator paired with a feedback optocoupler implements a low-cost isolated loop for flyback adapters, LED drivers and auxiliary rails.

  • Typical shunt regulators: TL431, TL431A, TLV431, LMV431 and similar adjustable references.
  • Typical optocouplers: PC817 or EL817 families for general-purpose feedback, CNY17, LTV-817 and other linear feedback optocouplers where improved CTR behavior is desired.

Isolation amplifier with integrated reference or error amplifier

Isolation amplifiers and isolated error amplifiers transfer an analog feedback signal with defined gain and bandwidth, offering higher accuracy and more predictable dynamics than a simple optocoupler loop.

  • Isolated error amplifiers, such as ADuM3190-class or ADuM4190-class devices, are designed specifically for isolated feedback in AC-DC and DC-DC supplies.
  • Isolation amplifier families such as AMC1200, AMC1201, AMC1300 and similar devices can be used for isolated voltage feedback when the front-end scaling and loop compensation are chosen carefully.

ΣΔ isolated modulator feeding a digital controller

Isolated ΣΔ modulators generate a bitstream that crosses the isolation barrier and is reconstructed by a digital power controller, enabling fully digital loops for high-performance PFC and LLC stages.

  • Example families include AD7401A and related AD7400/AD7403-class devices for isolated ΣΔ conversion.
  • Other ΣΔ modulators, such as AMC1301, AMC1302, AMC1304 or Si8910/Si8920-class devices, provide isolated bitstreams for voltage or current feedback into digital controllers.

Multi-channel digital isolator for PWM and fault lines

Multi-channel digital isolators carry PWM, enable, fault and status signals across the isolation barrier. They are often used alongside analog or ΣΔ feedback to coordinate switching and protection behavior between primary and secondary controllers.

  • Digital isolator families such as ADuM120x/ADuM140x-class devices provide two or four isolated logic channels for PWM and feedback control signals.
  • Si86xx-class and ISO77xx-class multi-channel isolators offer options with higher channel counts, various direction configurations and high CMTI ratings for noisy power stages.

Together, the checklist and IC role mapping allow a designer to validate isolation feedback robustness and quickly align each requirement with an appropriate family of isolation components before freezing schematics and PCB layout.

Checklist and IC families for isolated feedback design Block diagram showing an isolated feedback design checklist on the left and IC role mapping on the right, with arrows connecting design questions to device categories. Isolated feedback checklist and IC role mapping Design checklist Isolation & surge • Insulation level and working voltage checked? • Creepage/clearance and surge ratings verified? Feedback target & multi-output • Master output rail clearly defined? • Other rails via post-reg or cross-reg only? CTR, gain and bandwidth • Worst-case CTR or gain and delay modeled? • Loop crossover and margins checked? Failure modes and fail-safe • Opto, iso-amp or ΣΔ faults lead to safe shutdown? • Watchdogs and fault flags implemented? IC role mapping Shunt regulator + optocoupler • TL431 / TLV431-class + PC817 / EL817-class • Low-cost CV/CC loops for flyback and LED drivers Isolation amplifier / error amplifier • ADuM3190-class, ADuM4190-class families • AMC1200 / AMC1300-class used with scaling networks ΣΔ isolated modulator to digital controller • AD7401A-class, AMC1301 / AMC1304, Si8910 / Si8920 • Bitstream feeds digital PFC / LLC loops Multi-channel digital isolator • ADuM120x / ADuM140x, Si86xx, ISO77xx families • PWM, enable and fault signal isolation

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

PSU current and voltage sensing – FAQs

This FAQ collects common questions about current and voltage sensing in switched-mode PSUs. Each answer is written from a practical design viewpoint so you can quickly cross-check sensor placement, bandwidth, isolation, error budget and digital-control impacts against your own design.

1. When is a high-side current-sense amplifier mandatory instead of a simple low-side shunt measurement?

A high-side current-sense amplifier becomes mandatory when you must keep the load referenced to system ground, detect open-load and short-to-ground faults, or monitor current in a node that already sits above ground, such as a synchronous buck input or battery rail. High-side sensing also reduces ground disturbance and improves EMC in multi-rail systems.

Related section: high-side vs low-side sensing topologies

2. How much bandwidth is really needed in the PFC current-sensing path to avoid degrading power factor and current-loop stability?

The PFC current-sensing path should pass the fundamental line current and key switching-frequency harmonics with minimal phase lag up to at least several times the intended current-loop crossover frequency. You typically target a current-sense bandwidth in the tens of kilohertz range so that filter and amplifier delays do not erode phase margin or distort the sinusoidal input current.

Related section: PFC current-loop bandwidth and sensing design

3. In a server or telecom PSU, when does an isolated ΣΔ modulator offer a clear advantage over an isolation amplifier for current and voltage feedback?

An isolated ΣΔ modulator is especially attractive when your digital controller already includes decimation filters and you need high dynamic range, precise gain and inherent digital isolation. It simplifies multi-rail telemetry, eases calibration and aging compensation, and supports complex control algorithms, at the cost of higher latency and more involved digital filter design than a simple isolation amplifier.

Related section: isolated amplifier versus ΣΔ modulator

4. How should current-transformer based sensing be handled at light load to avoid distortion and false zero-crossing behavior?

To keep a current transformer accurate at light load, you need enough burden resistance and magnetizing inductance so that excitation current remains small compared with the sensed current. You also avoid deep core reset problems, provide a defined demagnetization path and, if necessary, blend CT-based sensing with shunt or amplifier signals in modes where the CT would otherwise saturate or drop out.

Related section: current transformer usage and limitations

5. How can the voltage-sense divider be chosen to balance accuracy, power dissipation and noise pickup in a multi-rail PSU?

You select the divider so that input bias currents and ADC errors remain small versus the divider current while resistor noise, PCB pickup and dissipation stay acceptable. Medium-value resistors, tight tolerance and low-TC parts on an analog ground island work well. For higher voltages, you can split the upper leg, add RC filtering and use Kelvin routing to reduce both noise and thermal gradients.

Related section: voltage-divider design for sensing

6. How should hotspot temperature and shunt TCR be combined into the current-sensing error budget over the full operating range?

You start from the shunt nominal value, estimated hotspot temperature rise and temperature coefficient of resistance to derive worst-case resistance variation. Then you add amplifier offset and gain errors, ADC errors and layout-induced drops to form a full-chain error budget. The result tells you whether a given shunt technology and footprint meet your accuracy target under maximum load and ambient conditions.

Related section: shunt TCR and error budgeting

7. For a USB-C PD adapter, which voltages and currents must be sensed to satisfy fast-charge protocol negotiation and safety requirements?

A USB-C PD adapter normally monitors VBUS voltage, output current and, for higher power levels, connector or cable temperature. You also need protections for overvoltage, overcurrent and short-circuit events, plus telemetry for fault logging. Coordinating these measurements with the PD controller ensures that advertised PDOs are respected, cable limits are not exceeded and protective actions are fast enough to avoid damage.

Related section: USB-C PD sensing requirements

8. For a PoE PD or 24 V industrial front-end, when is full power and energy metering required instead of simple current limiting and fault flags?

Simple current limiting and fault flags are usually enough for basic protection and classification. Full power and energy metering becomes necessary when you must implement precise power budgeting, billing, detailed diagnostics or standards that require reporting consumption to a host. At that point you move from rough OCP measurements to calibrated metering AFEs or ΣΔ-based solutions with well-defined accuracy over time and temperature.

Related section: PoE and 24 V front-end sensing strategy

9. With a digital controller and isolated ΣΔ modulator, how can sampling and filter delay be estimated to confirm that loop stability margins are still adequate?

You model the ΣΔ modulator, digital decimation filter and control update rate as a combined delay expressed in equivalent phase shift at the intended crossover frequency. Manufacturer data often gives group delay in clock cycles, which you translate into microseconds and degrees. You then include this delay in a Bode plot or simulation to verify that phase margin and gain margin remain within your design targets.

Related section: digital loop latency and stability

10. What happens to voltage and current measurements when a reference IC drifts or a bias LDO drops out, and how can these failures be detected by the PSU controller?

When a reference drifts, all ADC-based measurements share a proportional error, which can silently corrupt regulation and telemetry. If a bias LDO collapses, some sensors stop updating or rail readings move to unrealistic values. You can detect these faults by monitoring internal reference channels, comparing redundant measurements, checking for stuck codes and enforcing plausibility limits on reported voltage, current and power values.

Related section: references, bias rails and monitoring

11. How should the sensing-chain error budget be aligned with overall PSU efficiency, output-accuracy and power-reporting KPIs?

You start from system-level KPIs such as output regulation tolerance, billing accuracy or power-limiting thresholds, then allocate a fraction of this allowance to the sensing chain. Shunt, amplifier, ADC and reference errors are combined to form a total budget. If the sum exceeds your target, you either tighten component specs, improve calibration or relax noncritical KPIs that tolerate wider measurement uncertainty.

Related section: error budget and KPI alignment

12. In medical and industrial PSUs, how can safety isolation requirements be met while still preserving enough bandwidth for protection and control loops?

You select isolation components that provide the required reinforced insulation, creepage and CMTI while still offering sufficient analog or digital bandwidth for your loops. Often you separate high-speed protection flags into dedicated digital isolator channels and reserve the main isolated amplifier or ΣΔ path for regulation. Careful layout, Y-cap placement and filtering help you meet both safety and stability constraints simultaneously.

Related section: medical and industrial isolation considerations