What this page solves

Online SOH diagnostics addresses a simple but critical question for ESS, UPS and microgrid operators: how much power and energy each rack, string or module can still deliver safely, without shutting down the system for long offline tests.

Traditional SOH estimation based only on cycle count and OCV–SOC trends often detects problems late, when capacity has already fallen or internal resistance has increased enough to force derating, nuisance trips or premature replacement. Online SOH diagnostics adds an extra layer of insight by using targeted narrowband or wideband excitation and impedance measurement to estimate key model parameters while the system remains in service.

In this context, online SOH diagnostics means in-system impedance or response measurements that operate within the normal operating envelope of an ESS: small, controlled excitation on the DC bus or at module nodes, carefully timed windows that avoid disturbing PCS operation, and processing chains that convert raw waveforms into usable SOH metrics for BMS and EMS decision-making.

This page focuses on the hardware and IC roles that make online SOH diagnostics practical: excitation stages, impedance AFEs, ADCs, edge MCUs and the interfaces into BMS, EMS and cloud analytics. Detailed SOC algorithms and conventional fuel-gauge implementations are covered on the fuel gauge and SOH/SOC estimation page, and production cycler power stages are covered on the cell formation and cycler AFE page.

The goal is to help system designers understand how an online SOH module fits into a real ESS cabinet, which signals it observes, how it communicates SOH metrics upstream, and where impedance-based insight adds value beyond standard BMS measurements.

Scope, placement & interfaces in an ESS

An online SOH diagnostics function can live at different levels inside an energy storage system. It may be integrated on the pack BMS board, implemented as a dedicated diagnostics module connected to pack terminals, or placed on the DC bus or station side to observe the behaviour of multiple racks. Each placement option changes the observable signals, the measurement granularity and the safety constraints that the hardware must respect.

When the online SOH module sits on the pack BMS, it can share cell or module taps, temperature sensors and protection state information, providing tight coupling and precise timing but also facing stricter EMC and layout constraints. An add-on diagnostics module connected through harnesses and a communication link is attractive for retrofits, but introduces longer analog paths, more noise pickup and a need for carefully defined ownership of protection decisions. A station-level implementation on the DC bus has the best view of overall station health, yet offers less resolution at string or module level and must be coordinated with PCS and microgrid stability requirements.

Upstream, online SOH modules interface to cell or module measurement nodes, DC bus nodes, temperature channels, contactor status signals and EMS scheduling constraints. Downstream, the module exposes SOH metrics, derating suggestions and maintenance indicators to pack BMS, ESS EMS and site gateways, and optionally streams compressed impedance features or trends to cloud analytics. These interfaces define the required measurement range, bandwidth, isolation strategy and the minimum time window available for safe excitation in each deployment.

This section stays at system level: it maps typical placement options and the associated signal interfaces. Detailed BMS communication stacks and EMS architectures are described on the pack BMS, ESS EMS edge controller and site gateway pages, so the focus here remains on how online SOH hardware attaches to the ESS and which signals it must exchange.

Measurement methods: EIS, pulse, PRBS and variants

Online SOH diagnostics can use several excitation and measurement methods, each with different implications for the power stage, sensing AFE and data converters. The most common families are frequency-domain impedance spectroscopy with single or multiple sine tones, time-domain response analysis based on controlled load steps, and wideband techniques such as PRBS or multi-sine signals that compress more frequency content into a short measurement window.

Sine-based EIS injects small AC perturbations around a chosen operating point and measures voltage and current phasors at one or more frequencies. Laboratory implementations can sweep many points across a wide band, but online systems usually select only a few key frequencies and keep the excitation amplitude tightly constrained to avoid disturbing PCS operation or protection logic. This approach places strong demands on AFE bandwidth, ADC sampling stability and phase accuracy between voltage and current channels.

Time-domain methods rely on step or pulse changes in current or power and track the voltage response over short and medium time scales. A controlled current step, or even a naturally occurring load change, can reveal internal resistance and dominant time constants when sampled with sufficient rate and resolution. These methods relax the requirements on explicit waveform synthesis but require fast ADC sampling, accurate time stamping and an AFE with a predictable transient response so that curve-fitting and model extraction remain reliable in a narrow measurement window.

Wideband excitation approaches such as PRBS, multi-sine and chirp signals aim to cover a broader frequency span in a short time. A pseudo-random sequence on current or voltage can be processed with correlation and spectral analysis to estimate an effective frequency response. These techniques demand higher sampling rates, wider AFE bandwidth, stable low-jitter clocks and more DSP capability in the edge MCU or SoC, but they can be attractive where only short measurement windows are available between dispatch events or protection constraints.

This section focuses on how each measurement method translates into concrete hardware requirements for excitation, AFE bandwidth, ADC sampling rate and timing. Detailed electrochemical modelling and curve-fitting strategies are treated later in the metrics and models section, while long, full-cycle laboratory and formation tests remain in the cell formation and cycler AFE topic.

Excitation & impedance AFE building blocks

An online SOH front end combines a controlled excitation path with a measurement chain that can resolve small AC components on top of large DC currents and voltages. The core building blocks are a waveform source and driver, a current and voltage sensing AFE, precision ADCs, an optional analog switch matrix for multiplexing between strings or modules, and protection structures that keep the circuitry safe under fault conditions. The architecture must be optimised for bandwidth, phase accuracy and synchronisation, not only for static DC accuracy.

On the excitation side, a DAC or dedicated waveform generator produces sine, pulse or PRBS signals within the limits defined by the ESS and PCS. In low-voltage nodes, a small-signal driver may be sufficient, while excitation on a high-voltage DC bus typically requires isolated drivers or isolated amplifiers with carefully controlled output impedance. The excitation path must avoid destabilising control loops or tripping protection, which often means limiting amplitude and choosing frequency bands that do not interfere with the dominant control dynamics.

The measurement chain usually starts with a precision shunt or current sensor, followed by a differential or instrumentation amplifier and a programmable gain stage. Its task is to extract small AC variations from high DC levels with low noise and high common-mode rejection, while preserving a well defined bandwidth and phase response. An anti-alias filter then constrains the spectrum seen by the ADC, using simple, predictable topologies so that any impact on impedance estimates can be modelled and compensated.

ADC selection depends on the chosen measurement methods. Sigma-delta converters offer excellent resolution and noise performance for low-frequency EIS points and quasi-DC internal resistance measurements, at the cost of latency. SAR converters provide higher sampling rates and short conversion delays for pulse and wideband excitation, but must still deliver sufficient ENOB and channel matching. In both cases, synchronised sampling of voltage and current channels is essential to maintain phase accuracy, and the digital interface and clock quality must support the required data rate and timing precision.

Analog switch matrices allow a single excitation and AFE chain to be shared across multiple strings or modules. The on-resistance, leakage, bandwidth and common-mode range of these switches directly influence measurement accuracy, especially for small-signal EIS. Protection elements such as series resistors, surge clamps and ESD structures protect the AFE and ADC inputs during faults. Compared to a typical BMS AFE, an online SOH AFE places more emphasis on controlled frequency response, phase accuracy and repeatability under narrow measurement windows, while basic cell sampling and balancing functions remain with the BMU/CMU chain.

Edge MCUs, timing and synchronization for online SOH

Edge MCUs and SoCs act as the control and processing core of an online SOH function. They generate or supervise excitation waveforms, coordinate ADC sampling and DMA, perform front-end digital signal processing and publish SOH metrics to BMS, EMS and cloud systems. The same devices also enforce timing constraints so that excitation, sampling and station-level control remain coherent across thousands of measurement windows over the lifetime of an ESS.

On the waveform side, the edge controller may use lookup tables, numerically controlled oscillators or dedicated CORDIC engines to drive DACs with single-tone, multi-tone or PRBS excitation. Frequency, amplitude and phase settings are programmed according to safety limits, PCS operating modes and the target impedance model. Timer blocks align waveform generation with ADC triggers so that each measurement window captures a consistent number of cycles or a well-defined step response profile.

For data acquisition, the edge MCU configures ADC sampling patterns, hardware triggers and DMA transfers. Voltage and current channels must be sampled synchronously or with well-characterised timing offsets so that amplitude and phase can be reconstructed accurately. Once a buffer is filled, the MCU or SoC applies window functions, DFT or FFT operations, simple filtering and temperature compensation, extracting the features needed by the SOH model while discarding excess raw data to keep memory and bandwidth under control.

Timing and synchronization are as important as numerical precision. A stable sampling clock limits jitter-induced phase error, while a real-time clock or time-synchronised counter assigns absolute time stamps to each SOH update. When coordination with EMS, PMU or microgrid controllers is required, edge devices may participate in PTP or TSN time distribution over Ethernet so that SOH events, power dispatch changes and fault records share a common time base. Even without full PTP support, aligning local RTCs with station controllers allows credible long-term trend analysis and fleet-wide comparisons.

Communication blocks then expose SOH metrics and quality indicators over CAN, Ethernet, Modbus or higher-level protocols such as gRPC. Safety, secure boot and OTA update mechanisms are handled at system level, while this section focuses on the compute, timing and interface requirements that keep online SOH measurements repeatable, phase-accurate and traceable over years of operation.

Turning impedance into usable SOH metrics

Online SOH diagnostics ultimately depends on converting impedance or response measurements into meaningful indicators that planners, operators and asset managers can act on. The processing flow starts from raw waveforms or impedance points, extracts a small set of model parameters that can be tracked repeatably, and maps those parameters to capacity, power capability and resistance-driven stress indicators. The goal is not to recreate full laboratory-grade EIS analysis in the field, but to provide a stable, interpretable health layer on top of conventional gauge and SOC estimation.

Measured data from sine-based, pulse or PRBS methods is first reduced to a compact parameter set. Typical examples include ohmic resistance, charge-transfer related resistance and one or more diffusion-related terms inferred from low-frequency behaviour or slower time constants. For each cell chemistry and pack design, engineering teams select a simplified model structure and identify which parameters are robust enough to estimate online within the limited bandwidth and excitation allowed in a live ESS, avoiding over-fitted representations that are difficult to maintain in production.

These parameters are then linked to SOH metrics that influence system-level behaviour. Rising resistance and changes in characteristic time constants correlate with reduced usable capacity and reduced safe power capability at given temperature and SOC conditions. By combining parameter trends with thermal limits and allowable voltage sag, the system can derive recommended maximum discharge and charge power, highlight strings or modules that are drifting away from the fleet, and flag assets that should be removed from peak-duty service or scheduled for replacement.

Edge devices typically use compact mapping functions or lookup tables to convert the latest impedance parameters into decisions such as “normal operation,” “derate this rack by a defined margin” or “request maintenance.” Cloud analytics platforms can combine many such data points across time and across sites to estimate remaining useful life windows, identify chemistries or operating regimes that age faster than expected, and update the model coefficients used in each fleet. SOH metrics are therefore split between fast, local decisions and slower, fleet-level optimisation loops.

Traditional fuel-gauge and SOC algorithms remain responsible for energy accounting and state estimation based on current integration, OCV curves and temperature corrections. Online impedance adds an extra health dimension that can highlight cells, modules or racks with abnormal resistance evolution even when SOC behaviour still appears adequate. Together, these views help operators decide how to allocate duty cycles, when to derate, and when to retire or refurbish assets in a way that balances safety, availability and lifecycle cost.

Safety, standards and operational constraints

Online SOH diagnostics injects small perturbations or leverages natural load changes on an operating ESS. Every excitation and measurement must remain inside a clearly defined safety envelope so that protection functions, PCS stability and grid-code compliance are never compromised. The objective of online SOH is to add an extra health layer on top of existing battery and PCS protections, not to replace or bypass any safety mechanism already required by standards or by the system-level design.

Excitation amplitude and frequency content are therefore constrained. Current or voltage perturbations around the operating point must stay well below over-current, over-voltage and under-voltage thresholds, and must not trigger nuisance trips in arc detection, insulation monitoring or other protective functions. Frequency planning must avoid exciting PCS control loops or creating additional harmonic content in frequency bands that are controlled by grid codes. In practice, engineering teams validate excitation profiles through a combination of simulation and laboratory testing before enabling them in the field.

Measurement cycles and time windows must also be coordinated with power scheduling and protection actions. Online SOH is typically disabled during severe faults, intentional islanding, reconnection and other transients where primary control and protection have priority. In normal operation, diagnostics windows are scheduled to avoid large dispatch ramps or critical support modes, and measurement frequency is tuned according to asset age, seasonal conditions and operational needs, rather than running continuously at maximum rate.

Battery and ESS safety standards, such as cell and system standards in the IEC 62619 and IEC 62933 families along with relevant UL and national regulations, define allowable operating regions for voltage, current, temperature and fault handling. Online SOH excitation and measurement must remain within these operating regions and may need to be demonstrated not to degrade compliance. Grid codes and utility interconnection rules limit voltage flicker, harmonics and dynamic support behaviour, so any diagnostic activity must be shown to have negligible impact on these parameters or be confined to approved operating modes of the PCS and microgrid controller.

In fault or alarm conditions, online SOH typically transitions into a degraded mode. Severe faults and fire-related events require all active excitation to stop, with only event logging maintained for root-cause analysis. For mild warnings, diagnostic excitation may be reduced in amplitude or replaced with purely opportunistic measurements based on normal load steps. In every case, online SOH remains subordinate to the safety architecture and coordination rules defined for PCS, microgrid and UPS subsystems; it operates as a complementary diagnostic layer inside a clearly bounded safe envelope.

IC roles and BOM hooks for online SOH diagnostics

Online SOH diagnostics relies on a coordinated set of IC functions rather than a single device. A complete solution typically combines a waveform or DDS generator, precision AFEs and programmable gain amplifiers, high-performance ADCs, references and timing sources, edge MCUs or SoCs with DSP capability, and isolation devices that keep high-voltage measurement domains safely separated from control logic and communications. Understanding these roles helps translate system requirements into BOM entries that cannot be silently downgraded during cost optimisation.

Waveform and function generator ICs provide controlled excitation for EIS, pulse and PRBS methods, with adjustable frequency, amplitude and phase and low distortion. AFEs, instrumentation amplifiers and programmable gain stages extract small AC components from large DC currents and voltages with high CMRR, controlled bandwidth and predictable phase response. Sigma-delta and high-speed SAR ADCs digitise voltage and current channels with sufficient resolution, sampling rate and channel synchronisation to support the chosen measurement methods without excessive jitter or mismatch errors.

Precision references, low-drift clocks and RTCs anchor measurement accuracy and time alignment over years of operation. Edge MCUs or SoCs supply waveform control, trigger and DMA management, front-end DSP and communication stacks and often benefit from dedicated DSP extensions or FFT accelerators. Digital isolators, isolated amplifiers and isolated power devices provide galvanic isolation at high-voltage nodes and across communication links, allowing online SOH measurements on HV strings and DC buses without compromising safety or EMC behaviour.

BOM hook statements define which of these functions are mandatory and how they may not be reduced. Specifications can state that the design shall include an online SOH module capable of updating impedance-based health metrics, and that it shall not be replaced by fuel gauges that only support SOC estimation without resistance or impedance tracking. Hooks can further require that ADC channels, AFE bandwidth, timing quality and isolation resources reserved for online SOH must not be reassigned to unrelated auxiliary functions without a formal reassessment of SOH capability.

Clear BOM wording also protects the interface and model aspects of online SOH. Requirements can state that galvanic isolation and timing sources needed for high-voltage measurements remain part of the safety concept, and that interfaces between the online SOH module and BMS or EMS shall expose health metrics and derating recommendations, not just a hard-coded percentage value. Detailed IC selection, vendor recommendations and parametric comparisons can then be handled on dedicated component pages, while this section anchors the categories and non-negotiable hooks that keep online SOH diagnostics intact across design iterations and cost-down cycles.

Design checklist & IC mapping for online SOH diagnostics

This checklist summarises the key design decisions that determine whether an online SOH function is feasible for a given ESS, UPS or buffer application and what level of depth it can reach. Each item maps back to earlier sections on measurement methods, AFE and ADC design, edge compute, safety envelope and system integration. The goal is to confirm scope, methods and constraints before component selection and PCB partitioning are frozen.

• Measurement nodes and voltage domain. Decide whether online SOH will observe cell, module, rack or DC-bus level behaviour. Cell and module nodes enable fine-grain diagnosis but require many channels and careful AFE design, whereas rack and DC-bus nodes provide aggregate health trends with fewer channels but less ability to isolate a single weak string.
• Frequency range and excitation methods. Select the usable frequency band and excitation type based on safety and grid-code limits. Narrow-band sine or multi-sine methods focus on a few EIS points; pulse and step methods leverage existing load changes; PRBS and other wideband schemes require higher sampling rates and more careful filtering. Choices here drive AFE bandwidth and ADC specifications.
• Measurement time windows and derating constraints. Define how long each diagnostic window may last, whether tests may run during full dispatch and whether small reductions in power or temporary restrictions on operating modes are acceptable. Very short windows push the design towards higher bandwidth and more complex excitation, while purely opportunistic schemes accept slower convergence but minimise interference with scheduling.
• Target SOH resolution and model complexity. Clarify whether the requirement is limited to coarse health bands, approximate remaining useful life, or detailed power capability curves. Higher ambition demands more frequency points or more accurate time-domain fits, larger data sets and closer alignment between edge and cloud models.
• Edge compute and memory budget. Estimate the number of channels, sampling rates and window lengths, then confirm that the chosen MCU or SoC has sufficient processing throughput, RAM and non-volatile memory for waveform control, FFT or DFT, parameter extraction and logging. If resources are marginal, either relax method choices or allocate a more capable controller before layout is locked.
• Interfaces to BMS, EMS and cloud. Decide which systems will consume SOH metrics and at what update rate. CAN or RS-485 may handle summary indicators and derating flags, while high-rate fleet analytics may require Ethernet, TSN or cellular links. Bandwidth and latency constraints determine whether raw or lightly processed data can leave the edge, or whether only compact metrics are shared.
• Safety and grid-code boundaries. Confirm the safe excitation envelope defined by battery and ESS standards, PCS protection settings and interconnection rules. Online SOH designs must document that diagnostic activity remains inside these limits and define how measurements are inhibited or degraded when protection mechanisms, islanding or reconnection sequences are active.

Once these questions are answered, IC categories can be mapped to the application class rather than selected in isolation. The following summary shows typical combinations of waveform, AFE, ADC, timing, compute and isolation resources for several online SOH deployment patterns.

Application mini-stories: ESS, UPS and fast-charging buffer

The following application mini-stories illustrate how online SOH diagnostics changes day-to-day decisions in container ESS, UPS battery systems and fast-charging buffer installations. Each story highlights typical IC roles and example device families without turning the narrative into a full component selection guide.

Container BESS: detecting a weak string before summer peak

A multi-MWh container BESS supports energy arbitrage and grid services through several racks of lithium-ion modules. Standard pack-level BMS and fuel gauges track SOC, temperature and basic SOH, but early divergence in internal resistance across racks is difficult to observe from aggregate current and voltage readings alone. During summer peaks, a single weak string can limit the power capability of an entire container and trigger unexpected derating.

An online SOH function is added at the rack DC-bus level. Each rack uses shunt-based current sensing and an isolated amplifier chain, for example devices similar to AMC1302 or AD8410-class current sense amplifiers, feeding a multi-channel sigma-delta ADC such as ADS131M04 or AD7175-2. A mid-range MCU with floating-point and FFT support, such as STM32F4 or NXP Kinetis K6x families, controls low-amplitude multi-sine excitation, synchronises sampling and extracts resistance-related parameters. Digital isolators and isolated CAN or Ethernet transceivers, such as ISO1042 or ADuM style devices, connect the rack to the container BMS and EMS.

Over several months, impedance trends from all racks show one string with faster R_ohmic growth and abnormal low-frequency behaviour. Before summer peak tariffs begin, maintenance staff schedule a controlled outage to replace that rack or reconfigure dispatch so that it no longer carries peak-duty service. As a result, the container avoids surprise derating events during high-revenue periods, and the cause can be traced directly to resistance trends instead of generic SOC deviations.

UPS battery system: managing high C-rate performance with resistance monitoring

A data centre UPS uses battery strings that must support high C-rate discharge for seconds to minutes during power interruptions and ride-through events. Capacity and SOC estimates suggest that the strings are adequate, yet during test discharges, some strings show excessive voltage sag and thermal stress, indicating internal resistance growth that conventional monitoring has not captured in time.

Online SOH diagnostics is integrated into the UPS battery cabinets. High-bandwidth current measurement uses Hall or TMR sensors and shunt amplifiers, for example combinations of devices like Allegro ACS770 sensors with fast amplifiers, or INA240-class current shunt monitors. Synchronous multi-channel SAR ADCs, such as ADS8588S or AD7606 families, digitise voltage and current during scheduled self-tests and carefully controlled high C-rate steps. A control-oriented MCU or DSP, such as devices from the TI C2000 F2837x family or comparable high-performance microcontrollers, runs time-domain analysis to extract resistance and dynamic response parameters per string.

Impedance data from repeated self-tests reveals that certain strings experience faster resistance growth under specific temperature and loading patterns. Maintenance planning uses this information to shorten service intervals for affected strings, adjust test schedules and prioritise battery replacement budgets. During real outages, the UPS sees more predictable voltage behaviour because high-risk strings have either been replaced or removed from critical load support.

Fast-charging buffer ESS: optimising dispatch and retirement order

A city fast-charging station uses several buffer ESS cabinets on the DC-bus side to reduce grid stress and provide dynamic support for multiple high-power chargers. Each cabinet experiences a different duty cycle because parking patterns, charging habits and pricing schemes vary over time. Relying only on installation date to decide which cabinet to retire first would waste health on lightly stressed assets and keep heavily aged cabinets in service longer than necessary.

Online SOH modules are installed at the DC-bus interface of each cabinet. Isolated voltage and current sensing chains use isolated amplifiers and sigma-delta modulators, such as AMC1301 or AD740x devices paired with multi-channel ADCs like ADS131A04 or AD713x. Edge controllers with Ethernet or TSN capability, for example STM32H7 or NXP i.MX RT series devices, manage step-response and PRBS-based measurements, apply parameter extraction and push compact SOH metrics and health flags to a site controller or cloud analytics platform over standard Ethernet PHYs.

The analytics layer aggregates online SOH metrics and produces a ranked list of cabinets by remaining useful life and power capability. Dispatch schemes are adjusted to allocate high-frequency dynamic power to healthier cabinets, while ageing cabinets are shifted toward gentler profiles. When capital budgets allow, the retirement sequence follows the health ranking rather than simple installation order, extracting more value from each cabinet and avoiding sudden loss of fast-charge performance.

Online SOH diagnostics – FAQs

This FAQ summarises common questions engineers and asset owners ask when adding impedance-based online SOH diagnostics to ESS, UPS or fast-charging buffer systems. Each answer links conceptually back to the sections on scope, methods, AFEs, edge compute, safety envelope, design checklist and application stories on this page.