SENICO Electronics

Component sourcing and inventory guides

INA228 Server Energy Monitoring: Shunt Calibration and Timing

Make INA228 server energy readings traceable by defining the shunt, calibration, measurement boundary, conversion timing and accumulator reset policy.

Quick Answer

INA228 energy monitoring starts with a clearly defined measurement boundary and a known shunt. TI specifies a digital monitor that reports current, voltage, power, energy and charge. Those readings become useful server data only when software scaling, physical sense routing and measurement timing agree. A high-resolution converter cannot correct an undocumented resistor value or a dashboard that integrates the wrong units.

The discussion is at device-family level; the purchasing team must confirm the complete approved ordering code. It is a proposed server monitoring application, not evidence that INA228 appears on a particular accelerator manufacturer's board.

Key Specifications and Decision Points

TI describes a twenty-bit converter, common-mode support up to 85 V, programmable conversion time and averaging, and accumulated energy and charge reporting. It also provides selectable shunt input ranges. Choose those settings using the largest expected current and the required low-current resolution, while checking the datasheet's electrical limits.

For an illustrative calculation, 20 A through a 1 milliohm resistor produces 20 mV and dissipates 0.4 W. Those numbers are arithmetic, not a recommended shunt selection. They demonstrate why voltage drop, heating and measurement range should be evaluated together.

  • Define whether the shunt measures a complete board input, one regulator input or one output rail. Label dashboards with that boundary so unlike power readings are not compared.
  • Record the shunt resistance, tolerance, temperature behavior and connection method. Preserve the actual component code and layout revision in the calibration record.
  • Set conversion timing and averaging to match the measurement objective. Long averaging can improve the steadiness of a displayed value while hiding short variations important to another diagnostic task.

Application and Acceptance Review

Validate the current path before trusting energy totals. Use independent measurements at several operating points, including a low current where offset contributions matter and a higher current where shunt heating matters. Allow the assembly to reach relevant thermal conditions and observe whether the relationship between reported and reference values changes.

Energy reports need a start and end definition. Decide whether a total begins at power application, service initialization or an explicit measurement command. Preserve resets and outages in the data so a restarted counter is not interpreted as a sudden reduction in lifetime consumption. If software combines device readings with its own integration, document which method produces the published value.

Keep fast protection requirements separate from ordinary monitoring. The device includes an alert function, but its configured conversion behavior and the complete system response need verification before that path is assigned a protective role. A periodically refreshed dashboard is not automatically a fast fault detector.

  • Inspect the sense routing and connection points. A measurement that includes unintended copper resistance can depart from the calibration based on the nominal resistor alone.
  • Compare current polarity and power sign with the intended direction of energy flow. Test any permitted reverse-flow state deliberately instead of assuming software handles negative values correctly.
  • Check timestamps, accumulator behavior and reset handling. Define how a management restart affects the reported energy interval and whether the record can contain gaps.

Pricing and Lead Time

An RFQ for the monitoring circuit should separate the INA228 ordering code from the shunt and any required calibration service. The shunt's exact specification is a functional part of the design, not a purchasing detail to be substituted solely by nominal resistance. Ask engineering to approve changes that affect tolerance, thermal behavior, footprint or terminal arrangement.

For an assembled server control board, include firmware scaling and acceptance measurements in the manufacturing package. Clarify whether calibration constants are common to the design or measured for each unit. If per-unit values are used, define their storage and association with the assembly identity before requesting a quotation.

Compare commercial offers on equivalent scope: semiconductor packaging, shunt specification, assembly inspection and measurement verification. Obtain current pricing and delivery terms from the actual response. A precise device specification does not establish that a supplier has calibrated the buyer's assembly or that an offered board will meet the desired system-level accuracy.

Alternatives and Change Control

A lower-resolution monitor may be sufficient for a coarse operational indicator, while another interface may better fit the management architecture. Evaluate alternatives against the actual current range, accuracy budget, timing and energy-accounting requirement. Avoid selecting solely by the number of ADC bits, because total uncertainty includes the shunt and the physical implementation.

An isolated or magnetic sensing architecture can serve different voltage and isolation requirements, but it changes offset, bandwidth, power and mechanical considerations. Compare the complete signal chain and the required measurement boundary. Do not transfer calibration or error-budget assumptions from a resistive shunt implementation without reviewing them.

If readings disagree across tools, reconcile definitions before replacing hardware. One instrument may measure a regulator input while software displays an output estimate, or their averaging intervals may differ. Align the boundary, sign, units and timing, then investigate the remaining error. This sequence produces a useful diagnosis and protects procurement from unnecessary component changes driven by incompatible measurements.

FAQ

Does twenty-bit resolution guarantee twenty-bit system accuracy?

No. Resolution describes conversion granularity. Shunt tolerance, temperature, routing, calibration and other errors affect the complete measurement. Build an uncertainty budget and verify representative operating points on the assembled circuit.

Why must an energy report record resets?

An accumulated value depends on its measurement interval. A reset or power interruption can restart the total. Preserving those events prevents software from interpreting a new interval as continuous historical consumption.

Can I substitute any resistor with the same milliohm value?

Review the full shunt specification and physical connection. Tolerance, thermal behavior, power handling and terminal arrangement influence the measurement and its calibration, even when the nominal resistance matches.

Discuss your requirements with SENICO and request a quotation

Sources