Zero-Drift / Chopper Current Sense Amplifiers
← Back to: Current Sensing & Power / Energy Measurement
Zero-drift, chopper-stabilized current sense amplifiers trade a little bandwidth for ultra-low offset and drift, so current readings stay trustworthy over years instead of weeks. This page shows where they are worth the extra cost, how to design and lay them out correctly, and how to write their requirements clearly into your BOM.
System Role & When to Choose Zero-Drift Current Sensing
Zero-drift and chopper-stabilized current sense amplifiers are designed to remove most of the offset and 1/f noise that limits conventional current sense stages. Instead of living with hundreds of microvolts of offset and slow drift over temperature and time, a properly designed zero-drift front-end can push offset down to a few microvolts and keep long-term drift predictable over years.
A traditional CSA or op amp based current shunt amplifier often shows 200–500 µV of input offset, noticeable 1/f noise at low frequency and a temperature coefficient in the 10–50 µV/°C range. For rough monitoring or simple over-current decisions this can be acceptable, but for energy billing, fuel gauging and long-term telemetry these errors accumulate directly into measurement and billing inaccuracy. Zero-drift or chopper-stabilized CSA architectures attack exactly these weak points.
Typical use-cases include precision energy metering, long-life remote telemetry, EV and industrial BMS fuel-gauges, current integrators that build state-of-charge or usage histograms and sensor bias or leakage monitoring channels. In these applications offset, 1/f noise and long-term drift are not just data sheet numbers—they map into real charge, energy or safety margins. Where the main requirement is fast protection with microsecond response time rather than accurate long-term measurement, a fast non-zero-drift current sense path is often a better fit and is covered in the sibling Fast Current Sense for Protection page instead of this one.
Chopper / Auto-Zero Architecture & Error Sources
A zero-drift or chopper current sense amplifier does not magically make all errors disappear. Instead, it modulates the input signal and the amplifier’s own offset so that low-frequency offset and 1/f noise are shifted to a higher frequency band, then filtered out by a low-pass stage. At DC and low frequency the user sees a much smaller residual offset and a noise spectrum that stays flat instead of rising as 1/f.
Internally, most architectures share a similar block structure: sampling switches or an auto-zero front-end, a chopper modulator, the main amplifier, a demodulator and one or more low-pass filters. The useful shunt signal is modulated, amplified and demodulated back to baseband, while the quasi-static offset and 1/f noise components are translated into narrowband ripple around the chopper clock and then suppressed by the filter. This is why the data sheet quotes a few microvolts of input offset and very low tempco, even over years.
The price paid is a dependence on chopper frequency, bandwidth and ripple management. Residual errors now include finite-gain artifacts, switching charge injection and clock feedthrough, as well as coupling between the chopper clock and supply noise, switching regulators or ADC sampling clocks. These residuals do not show up as a simple “DC offset” but as small ripples or spurs around the chopping frequency. The next sections on design, layout and self-test describe how to align shunt value, gain, filtering and sampling so that chopper artifacts stay under control without sacrificing measurement bandwidth.
Shunt, Gain, Bandwidth & Filtering for Chopper Current Sense
The design of a zero-drift current sense chain starts with the basic measurement requirements: minimum and maximum current, allowable full-scale drop across the shunt, acceptable power loss and the resolution and reference range of the ADC. From these values you can derive a reasonable shunt resistance, estimate the resulting shunt self-heating and compare the effective shunt signal level against the few-microvolt offset and low drift promised by the zero-drift amplifier.
Typical full-scale drops of 50 mV or 100 mV are often a good compromise between efficiency and resolution. For example, at 10 A full-scale a 5 mΩ shunt produces 50 mV; a 10 µV input-referred offset in the amplifier then corresponds to only 0.02 % of full-scale. Making the shunt ten times smaller to save power would shrink the signal to the point where the same offset becomes a visible part of the error budget, while making the shunt larger increases power loss and shunt self-heating, which adds its own temperature coefficient and mechanical stress errors.
Once the shunt is defined, the amplifier gain and bandwidth are planned together with the ADC. The gain is normally chosen so that the full-scale shunt drop uses a comfortable fraction of the ADC input range, leaving some headroom for overload and drift. Bandwidth should cover the required current content by a factor of three to five while remaining well below the chopper clock frequency so that modulation artifacts stay outside the useful band. The low-pass pole or poles that follow the demodulator must sit low enough to suppress chopper ripple, yet not so low that control loops or telemetry updates become sluggish.
Filtering and sampling are tightly tied to the converter architecture. A simple RC filter is often enough for a SAR ADC, provided that the time constant is long enough to average out most of the chopper ripple but short enough to settle within the sampling window. More aggressive multi-pole or digital filters can reduce ripple further at the cost of latency. When the same rail also feeds a fast over-current protection path, that fast path should bypass the heavy filtering and may even use a different amplifier, as covered in the Fast Current Sense for Protection page.
Finally, switching noise and EMI must be considered alongside chopper frequency. Switching regulators, motor drives and digital clocks can mix with the chopper clock to create low-frequency beat tones or spurs that land in the measurement band. Choosing a chopper frequency that is well separated from major system tones, avoiding simple integer ratios with ADC sampling and adding modest RC filtering on the input and output stages help keep these artifacts under control before layout and grounding do the rest of the work.
Kelvin Routing, Shielding & Chopper-Specific Layout Tips
Layout for a zero-drift current sense amplifier is all about preserving a few tens of microvolts of signal integrity in a board full of ampere-level currents and fast switching edges. The first rule is to treat the shunt as a precision sensor, not just a power component. Kelvin sense traces should leave directly from the dedicated sense pads or terminals of the shunt, carry almost no load current and arrive as a tight, balanced pair at the amplifier inputs without wandering through noisy copper.
The ground and return paths around the amplifier must form a quiet island. Avoid slotting the reference plane under the input pair or placing the zero-drift amplifier directly in the middle of high di/dt loops such as switching MOSFETs, rectifiers or motor phases. The local reference for the amplifier and ADC front-end should tie into the main return at a well defined point close to the shunt, not at the far end of a high-current copper pour that carries large load variations.
Chopper-specific effects add another layer of constraints. Any clock or synchronization signals associated with the chopper engine should be routed away from the sensitive input pins, and digital control lines such as I²C, SPI or GPIOs should not cross directly over the input pair or shunt sense region without a solid reference plane in between. Guard rings or shield traces tied to a quiet reference around the input pins can help intercept leakage and capacitive coupling, especially where input impedance is high or where the board may see contamination and humidity over its lifetime.
Package orientation and local thermal gradients also matter. Placing the zero-drift amplifier immediately next to a hot inductor or FET can create asymmetric heating across the package, which looks like a slow, directional drift. It is often better to keep a modest distance from hot power parts, route the Kelvin sense pair cleanly into the amplifier, and allow the shunt and amplifier to share a reasonably uniform temperature field. Full-board layout strategies for power stages and energy measurement are covered at the platform level; this section focuses on keeping the immediate shunt and amplifier neighborhood clean, quiet and thermally well behaved.
Self-Test & Calibration for Zero-Drift Current Sensing
A zero-drift current sense amplifier removes most of the static offset and 1/f noise that would otherwise dominate low-frequency measurements, but high-integrity systems still rely on explicit calibration and health checks. Factory or production-line calibration removes device-to-device variation in the complete chain of shunt, amplifier and ADC, while in-field auto-zero and health monitoring detect slow drift and emerging faults over the lifetime of the product.
During production, the system is placed in a controlled state where known currents or voltages can be applied to the shunt. Measuring the corresponding digital output at 0 A and one or two non-zero points allows the offset and gain of the entire measurement path to be identified. These correction coefficients are then stored in non-volatile memory such as MCU flash, external EEPROM or the NVM of a digital current monitor so that each unit boots with its own calibrated transfer function instead of relying only on typical data sheet values.
In the field, many systems naturally pass through known states such as sleep, open-load or well-defined maintenance conditions. These windows can be used for periodic auto-zero: the controller temporarily treats the expected current as a reference, averages multiple samples to estimate the residual offset and then updates a calibration term or logs the deviation from factory values. Auto-zero cycles must be scheduled so that they do not interfere with critical measurement windows or control decisions; where necessary a fast protection path can run in parallel, as covered in the Fast Current Sense for Protection page.
Health monitoring closes the loop by interpreting offset drift and noise growth as diagnostic indicators. Comparing the current auto-zero offset against the original calibration, or tracking noise levels under quiet conditions, helps reveal ageing devices, contamination, solder cracks or unexpected interference. When thresholds are exceeded, the system should raise a warning, log an event and, in safety or billing applications, potentially degrade to a safer operating mode. The full system-level diagnostics framework, including fault classes and telemetry logging, is defined in the dedicated diagnostics and current-monitor pages; this section focuses on the hooks at the zero-drift front-end.
Aging, Long-Term Drift & Lifetime Planning
Long-term drift is often the last source of uncertainty left once offset, 1/f noise and gain errors have been tamed by a zero-drift architecture and careful calibration. Data sheets typically describe it with parameters such as long-term drift in microvolts per thousand hours, temperature coefficient in parts per million per degree Celsius and sometimes additional notes on hysteresis or reflow-induced shifts. Reading these lines in context is essential when designing a current measurement channel that must stay within a given error budget for ten years or more.
A practical way to think about lifetime stability is to decompose the total drift into contributions from the amplifier, the shunt, the PCB and connectors. The amplifier contributes temperature-related drift and ageing effects captured by TC and long-term drift specifications. The shunt adds its own TC and self-heating behaviour, which can easily dominate if power dissipation is high. Copper traces, vias and contact resistances may change over time due to corrosion, contamination or mechanical stress, and the PCB stack-up itself can introduce hysteresis when cycled over a wide temperature range.
Using typical numbers from the data sheet, designers can build an order-of-magnitude estimate for total drift over the target lifetime. For example, a 10 ppm/°C temperature coefficient across a 100 °C span on a 50 mV full-scale drop corresponds to about 50 µV of temperature-related shift, while a long-term drift of 5 µV per 1000 hours over 50,000 hours yields roughly 250 µV of ageing component. Comparing the combined figure with the original full-scale signal and the required billing or safety accuracy helps determine whether periodic recalibration, tighter-grade components or derating are needed.
Finally, reliability and qualification grade frame what can realistically be expected from a device in the field. AEC-Q100 automotive components are tested over wider temperature ranges and more aggressive life tests than general industrial parts, while metrology or precision instrumentation devices may document even tighter drift and hysteresis statistics. System architects can decide whether to rely on component grade alone, add extra burn-in or temperature cycling at module level, or combine moderate-grade devices with robust calibration and monitoring strategies as described in the self-test and diagnostics chapters.
Application Scenarios for Zero-Drift / Chopper Current Sense
This section focuses on where zero-drift / chopper current sense is truly justified—not on every possible current sensing application. The goal is to map real systems where offset, 1/f noise and long-term drift make a conventional current sense amplifier (CSA) risky, and where zero-drift architectures pay back their extra cost.
Energy Metering & Billing
- What current? Long-term average and bidirectional charge/discharge currents in revenue-grade meters, DC chargers and sub-metering nodes.
- Why plain CSA is not enough? 200–500 µV offset and 1/f noise hide low currents and accumulate into billing error over years, even if one-time calibration is performed.
- Zero-drift payoff. Microvolt-level offset and low drift keep low-current readings usable and allow longer verification intervals while still meeting revenue-grade accuracy targets.
EV & Industrial BMS / Fuel Gauge
- What current? Battery pack charge/discharge current, coulomb-counting averages and leakage currents in EV and industrial BMS.
- Why plain CSA is not enough? Small offset integrated over months shifts SOC by several percent; wide temperature range and aging make a single factory trim insufficient.
- Zero-drift payoff. Slow SOC drift and predictable long-term error budgets, especially when the zero-drift CSA is combined with in-field auto-zero windows defined in the self-test & calibration section.
Data Center & Server Power Monitoring
- What current? VRM and bus rail currents on low-resistance shunts with only tens of millivolts full-scale, including light-load telemetry for capacity planning.
- Why plain CSA is not enough? Offset becomes a large fraction of a 10–20 mV shunt drop, making light-load and idle power poorly resolved and poorly comparable between racks.
- Zero-drift payoff. Stable, comparable telemetry across thousands of boards and long lifetimes, enabling realistic power, thermal and TCO analysis without frequent recalibration downtime.
Precision Sensor Excitation & Bias Monitoring
- What current? 4–20 mA loops, bridge excitation current and precision sensor bias currents that directly define measurement gain and offset.
- Why plain CSA is not enough? Drift in excitation or bias current directly appears as span or zero error in the sensor output, so amplifier drift translates into apparent sensor drift.
- Zero-drift payoff. A zero-drift CSA close to the sensor enables slow health checks and long-term stability, reducing how often ratiometric sensor systems must be re-trimmed.
Medical & Safety-Critical Channels
- What current? Patient-side leakage, actuator and pump currents, and safety-rail currents in life-support and safety-critical controllers.
- Why plain CSA is not enough? Regulatory limits are tight and must hold for many years; drift can quietly move trip thresholds away from safe margins.
- Zero-drift payoff. Enables credible lifetime safety analyses with documented drift and allows rare, planned recalibration instead of frequent field service visits.
Low-Drift Calibration Loops
- What current? Reference, DAC or stress-test currents used inside factory or in-system calibration loops and long-term drift loggers.
- Why plain CSA is not enough? If the monitor path drifts, the calibration loop will incorrectly “blame” the DUT, baking amplifier drift into every new correction.
- Zero-drift payoff. Keeps the calibration monitor chain stable so that recorded drift and correction factors mostly reflect DUT and reference behavior, not the sense amplifier.
Topology choices such as low-side, high-side or bidirectional sensing, and fully digital monitors are covered in their own pages (for example, Low-Side Shunt Current Sense Amp, High-Side Shunt Current Sense Amp and Digital Current Monitor). This section only answers where zero-drift current sense is worth the cost.
Seven-Brand Zero-Drift / Chopper Current Sense Options
This table highlights typical families of zero-drift or ultra-precise current sense solutions from seven major vendors. It is not a complete catalog; instead it gives a short list of representative parts to anchor BOM discussions. Digital current monitors with integrated ADCs are covered in the Digital Current Monitor page.
| Brand | Family / PN (datasheet) | Type / Key Specs | Notes (why this choice) |
|---|---|---|---|
| Texas Instruments |
INA21x (INA210–INA214) INA28x (INA282–INA286) INA240-Q1 INA254 |
High-precision and ultra-precise current sense amplifiers with wide common-mode ranges (up to 80–110 V), PWM rejection options, and integrated shunt variants for compact layouts. | Go-to choice for general-purpose and automotive high-side sensing. INA21x and INA28x families cover 26 V and −14~80 V buses, while INA240 adds enhanced PWM rejection for motor and inverter rails. INA254 integrates a low-value shunt for tight board space. |
| Analog Devices (incl. Maxim) |
AD8418 MAX49925 |
High-voltage, high-resolution current shunt amplifiers for −2~70 V and −40~+76 V common-mode ranges, with low gain error and strong PWM rejection suitable for motor, BMS and 48 V HEV rails. | AD8418 is a classic automotive high-side current sense amplifier used in motor and battery management. MAX49925 extends CM range deeply negative and positive, easing transient and TVS design on 48 V systems and simplifying HEV front ends. |
| STMicroelectronics |
TSC21x (TSC210–TSC215) TSC201x / high-voltage CSA |
Precision current sense amplifiers with low offset and up to 26 V or higher common-mode range, supporting both low-side and high-side configurations for automotive and industrial rails. | TSC21x family gives a compact, precision front end for low-to-medium voltage shunts, while high-voltage CSA series extend to automotive and industrial bus voltages. Good match when the rest of the design already uses ST power and MCU devices. |
| NXP |
NAFE93352 N-AFE family |
Integrated analog front-ends with low-offset-drift buffers, high-precision amplifiers and 16/24-bit ADCs, with input protection up to 65 V for voltage and current measurement channels. Designed for multi-channel energy and condition monitoring. | Suitable when you want a tightly integrated measurement chain: protection, low-drift amplification and ADC in a single device. Used as the zero-drift current / voltage measurement backbone for metering, power and industrial diagnostics inside NXP-centric platforms. |
| Renesas (incl. Intersil) |
ISL28005 ISL28634 |
ISL28005: micropower high-side / low-side CSA with rail-to-rail sensing for 1.35–28 V rails. ISL28634: 5 V zero-drift instrumentation amplifier with programmable gain and differential output, suitable as a precision current sense front end. |
Good fit for telecom / industrial power platforms with existing Intersil heritage. ISL28005 covers classical high-side sensing, while ISL28634 addresses very high precision, low-drift requirements when configured for shunt current measurement. |
| onsemi |
NCS21671 / NCV21671 NCS333 (zero-drift op amp) |
NCS21671: current sense amplifiers with −0.1~40 V common-mode, precision gain options and low bias
current, configurable for low-side or high-side sensing. NCS333: zero-drift op amp family (10 µV offset, near-zero drift) often configured as a custom current sense front end. |
Strong option when you already use onsemi MOSFETs, drivers and protection devices. NCS21671 offers a clean CSA for 12–48 V systems; NCS333 covers ultra-precise low-drift conditioning where a custom shunt front end is acceptable. |
| Microchip |
MCP6C02 MCP6C04 TC7650 (chopper op amp) |
MCP6C0x: zero-drift high-side current sense amplifiers with 3–65 V common-mode range, preset gains and
strong EMI performance. TC7650: classic chopper-stabilized op amp used where ultra-low offset is required and the system builds its own shunt and gain network. |
Natural match for designs already based on Microchip MCUs and mixed-signal parts. MCP6C0x devices give a ready-made high-side zero-drift CSA, while TC7650 and newer zero-drift op amps support custom precision current measurement chains. |
This table is intentionally limited to a few representative series per brand. For digital current monitors (I²C/SPI with integrated ADC and alert logic), see the dedicated Digital Current Monitor page; for protection-centric devices such as eFuses and hot-swap controllers, refer to the separate protection hub.
Request a Quote
BOM & Procurement Notes for Zero-Drift Current Sense
Use this checklist to describe your zero-drift or chopper current sense requirements clearly in the BOM. By filling in range, drift, bandwidth, common-mode and quality level up front, you make it much easier for suppliers to propose the right parts instead of downgrading the design to a generic “0.5% current sense amplifier.”
1. Measurement Range & Shunt Policy
- I_min / I_max and direction: minimum useful current, maximum continuous current and whether the path is unidirectional or bidirectional (charge and discharge).
- Shunt value and full-scale drop: target shunt resistance and maximum allowed full-scale drop (for example 50 mV or 100 mV) to control efficiency and headroom.
- Shunt power and self-heating sensitivity: indicate whether shunt self-heating and temperature rise must be minimised and if derating rules apply.
2. Accuracy & Drift Targets
- Total error budget: target accuracy over temperature and lifetime, such as “±x% of reading ± y mA across −40~+85 °C and ten years”.
- Offset level: required input-referred offset range in microvolts, to distinguish zero-drift devices from conventional 200–500 µV CSAs.
- Temperature coefficient and long-term drift: maximum acceptable input-referred TC in ppm per degree Celsius and long-term drift in microvolts per thousand hours, or an explicit lifetime drift target.
3. Bandwidth, Chopper Limits & Sampling
- Measurement bandwidth: required signal bandwidth in hertz and whether you care about slow averages, moderate dynamics or fast transient peaks.
- Chopper frequency constraints: any preference for minimum or maximum internal chopper frequency and whether ripple inside the ADC band is acceptable.
- Filter and delay tolerance: maximum acceptable group delay introduced by analog or digital filtering and whether a separate fast protection path is used for microsecond events.
4. Common-Mode Range & Topology
- Topology flag: state whether the sense amplifier is used in low-side, high-side or bidirectional mode and if direction or polarity information is required.
- Operating common-mode window: nominal and worst-case common-mode voltage range, including expected transients and negative swings after protection elements.
- Isolation or intrinsic safety domain (if relevant): identify which side of isolation barriers the zero-drift front end resides on; full system architecture is handled in the corresponding parent pages.
5. Package, Temperature Grade & Second Source Strategy
- Package and height: preferred package types, maximum mounted height and any cleaning or moisture sensitivity constraints.
- Temperature and qualification: required operating temperature range and grade, such as AEC-Q100 grade, industrial or precision instrumentation level.
- Second source expectations: whether you require at least two vendors with compatible electrical performance and pinouts and if cross-brand validation data is needed.
6. Risk Notes for Vendors & Distributors
- Chopper behaviour differences: highlight sensitivity to chopper frequency, residual ripple and settling, and ask for typical waveforms or application notes when EMI or ADC aliasing is critical.
- Realistic long-term drift: request clarification on test conditions, sample size and statistics behind any long-term drift claims, not just typical bench numbers.
- MOQ and lead time: for small-batch and early production, call out minimum order quantity, packaging options and lead-time constraints, especially on automotive grades.
When you prepare an RFQ or submit a bill of materials, use the checklist above instead of a single “precision current sense amplifier” line. Clear numbers for range, drift, bandwidth and qualification help suppliers avoid downgrading your design. To share your requirements with our team, use the /submit-bom form and paste these fields into your message.
Zero-Drift Current Sense FAQs
The questions below collect the most common doubts engineers have when they first use zero-drift or chopper current sense amplifiers. Each answer is short enough to reuse in design reviews, internal notes or customer documentation, but detailed enough to guide real choices on architecture, layout, calibration, lifetime drift and BOM planning.
When do I really need a zero-drift or chopper current sense amplifier instead of a conventional CSA?
You really need a zero-drift or chopper current sense amplifier when offset, 1 over f noise and long term drift would accumulate into significant error over months or years. Typical examples include energy billing, battery state of charge tracking, medical leakage monitoring and long life industrial loggers where recalibration is difficult or expensive.
How does a chopper-stabilized current sense architecture actually remove offset and 1/f noise?
A chopper stabilized current sense amplifier periodically measures its own offset and modulates the signal so that low frequency errors are shifted up to the chopper frequency. After demodulation and low pass filtering, the offset and most one over f noise are removed from the measurement band, leaving a stable, low drift output.
What bandwidth and settling trade-offs should I expect when I switch to a zero-drift current sense front-end?
Compared with a conventional fast current sense amplifier, a zero drift front end usually has a finite settling time set by the chopper frequency and the output filter. You get excellent low frequency accuracy but limited usable bandwidth and more delay. For microsecond class protection, it is common to keep a separate fast sense path.
How do I choose the shunt resistor value and gain so that a few microvolts of offset still give me useful resolution?
Start from the maximum current, the acceptable shunt power loss and the desired full scale drop, for example fifty millivolts. Choose a shunt that meets this limit and then select a gain so that the smallest useful current produces several counts above the amplifier offset. Zero drift offset in the microvolt range lets you use smaller shunt drops.
What is the impact of chopper ripple and modulation artifacts on my ADC, and how do I filter or average them safely?
The internal chopper action produces small ripple and spikes at the modulation frequency and its harmonics. If the ADC sampling rate is close to simple ratios of this frequency, the ripple can alias into the measurement band. Use analog low pass filtering, coherent digital averaging and sampling frequencies that avoid simple integer relationships to the chopper tone.
Which layout practices are critical to keep a chopper current sense amplifier from picking up switching and clock noise?
With chopper based current sensing, layout must protect microvolt level inputs from large switching and clock edges. Use true Kelvin connections to the shunt, route the differential pair together over a quiet ground region and keep high di over dt loops away. Add guard or shield traces where necessary and avoid routing the chopper clock near the inputs.
How should I plan factory and in-field calibration for a zero-drift current measurement path?
Plan factory calibration around well controlled operating points with known currents at zero and one or two gain points, then store offset and gain coefficients in non volatile memory. In the field, reuse natural zero current windows such as sleep or open load states to run periodic auto zero cycles, updating calibration terms or logging detected drift trends.
What long-term drift should I budget for in a zero-drift current sense chain over a 10-year lifetime?
Rather than a single magic number, think about long term drift as the combination of temperature coefficient, specified microvolts per thousand hours and application temperature range. Use the data sheet values to estimate a ten year shift for the amplifier and then add shunt, self heating and interconnect effects. Finally, keep some safety margin in the budget.
How do I combine amplifier offset, shunt tolerance and temperature drift into a single current measurement error budget?
Build a current measurement error budget by converting each contributor into amperes or a percentage of full scale. Include amplifier offset, gain error, shunt tolerance, temperature coefficient and long term drift. Combine the terms using root sum of squares or worst case addition and compare against the allowed accuracy. This reveals which term dominates and deserves tighter limits.
Are there cases where a fast, non-zero-drift current sense amp is still better than a chopper device for protection?
Yes. For very fast over current or short circuit protection, a wide bandwidth, non zero drift current sense amplifier often remains the better choice. Protection loops may need microsecond level response and minimal filtering, while a zero drift device prefers slower bandwidth and heavier filtering. Many designs split the roles into a fast protection path and a precise metering path.
What minimum data-sheet parameters must I write into the BOM to avoid being swapped to a lower-precision current sense amp?
At minimum you should specify that the part is a zero drift or chopper current sense amplifier, plus input referred offset, gain error, temperature coefficient, long term drift expectations, common mode range, bandwidth and operating temperature grade. If you only request percentage accuracy, distributors can substitute ordinary amplifiers that will not meet long term performance needs.
How can I safely second-source a zero-drift current sense amplifier across different vendors without breaking my error budget?
Treat second sourcing as an error budget problem instead of a pinout match. Shortlist candidates whose offset, temperature coefficient, noise and long term drift are all within your limits, then compare chopper frequency, output ripple and bandwidth. Validate both vendors in the real circuit across temperature so that either device keeps the total current error inside the same budget.
