Current-Sense Amplifier Selection: Common Mode, Shunt Loss, and Dynamic Range

Table of Contents

• Place the Shunt Where the System Can Tolerate It
• Select the Shunt from Loss and Signal Level
• Match Common-Mode Range Before Gain
• Build the Error Budget at Minimum Current
• Check PWM Response and Output Headroom
• Validate the Complete Measurement Chain

A current-sense amplifier measures the small differential voltage across a shunt resistor while rejecting the much larger voltage shared by both inputs. That makes the whole chain inseparable: shunt value determines loss and signal, amplifier gain determines output range, common-mode behavior determines whether the signal remains valid, and the ADC and layout add their own error.

Do not start by selecting the highest gain or the lowest offset. First decide where current can be measured without breaking the system’s grounding and fault-detection objectives, then size the shunt and common-mode range. Precision specifications only matter after the amplifier remains in its guaranteed operating region.

Place the Shunt Where the System Can Tolerate It

A low-side shunt sits between the load and ground. It simplifies amplifier common-mode requirements but lifts the load ground by the shunt voltage and may hide a short to ground downstream of the shunt. A high-side shunt preserves the load’s ground connection and can detect current delivered into several fault paths, but the amplifier must tolerate the rail voltage.

In-line motor-phase sensing places a rapidly switching common-mode waveform at the inputs. That requires behavior beyond a static maximum voltage: negative excursions, high dV/dt, common-mode rejection over frequency, and recovery after switching edges all affect the control loop.

Document the measurement purpose. Energy reporting may accept slower filtered data, while current-mode control and overcurrent protection have different delay and bandwidth limits. One shunt channel should not be assigned every purpose without confirming that its range and response cover them.

Select the Shunt from Loss and Signal Level

At a given current, the shunt signal is VSHUNT = I × RSHUNT and its dissipation is P = I² × RSHUNT. Raising resistance improves the low-current signal but consumes efficiency and adds heat. Lowering it reduces loss but makes offset, noise, thermoelectric voltage, and ADC resolution more significant.

As a simple design example, a 2 mΩ shunt at 20 A develops 40 mV and dissipates 0.8 W. These calculated values do not include tolerance, pulse overload, PCB heating, or temperature coefficient. Use the actual current waveform and the resistor manufacturer’s power-derating and pulse data.

Use a four-terminal shunt or Kelvin connections where the required accuracy justifies it. Measurement traces should sense at the intended resistor terminals and avoid carrying load current. A few milliohms of shared copper can be comparable to the shunt itself.

The shunt’s initial tolerance, temperature coefficient, long-term stability, and thermal gradient belong in the error budget. If calibration removes initial gain error, record the calibration current, temperature, equipment uncertainty, coefficient storage, and traceability.

Match Common-Mode Range Before Gain

Common-mode voltage is approximately the average voltage at the two amplifier inputs with respect to its ground. It can be far above the amplifier supply in a purpose-built current-sense architecture. It can also move below ground during switching, inductive events, reverse battery, or startup.

TI’s current-sense design brief identifies common-mode range, offset, gain, temperature stability, and output type as key parameters. The cited examples illustrate specific parts; use the selected device’s guaranteed minimum and maximum range across temperature.

Build a common-mode table for normal operation, minimum and maximum input, startup, shutdown, reverse connection, PWM edges, load dump or surge where applicable, and an output short. If any state exceeds the specified range, redesign the input protection or measurement architecture. Absolute-maximum survival does not make the measurement valid.

Build the Error Budget at Minimum Current

Refer errors to one point before adding them. Input offset can be converted to equivalent current by dividing by shunt resistance. Gain error acts on the measured signal. Common-mode rejection contributes an input-referred error as common-mode voltage changes. The shunt, ADC reference, ADC offset/gain, and external filter add separate terms.

Error term	Usually most visible at	Can calibration remove it?
Amplifier offset	Low current	Initial value often; drift remains
Gain error	High current	Initial value often; drift remains
Common-mode rejection	Changing or high rail voltage	Only under characterized conditions
Shunt tolerance and TCR	Full range and temperature	Initial tolerance; temperature requires model/data
Output/ADC headroom	Range endpoints	No; architecture must leave margin
Noise	Low current and high bandwidth	Filtering trades response time

Separate typical and maximum data, and do not mix conditions from different grades or packages. The TI engineer’s current-sensing guide discusses how common-mode voltage, rejection, bandwidth, and offset affect selection.

The minimum useful current often exposes an unrealistic range requirement. If one channel must measure microamp standby current and a large fault current, consider switched ranges, separate channels, an integrated digital power monitor, or a different sensing architecture rather than expecting one gain setting to resolve both.

Check PWM Response and Output Headroom

Bandwidth should cover the information needed by the controller, not simply exceed the switching frequency by the largest possible ratio. More bandwidth increases noise and does not guarantee clean recovery after a fast common-mode edge. Review common-mode transient response, settling, input filtering recommendations, propagation delay, and output slew.

For bidirectional current, a reference input may set the zero-current output above ground. Calculate output at maximum positive and negative current using gain tolerance and offset, then leave headroom for the amplifier output swing and the ADC input range. Confirm the reference source can drive the pin and remains valid during power sequencing.

Coordinate any RC input filter with the amplifier’s input architecture. Mismatched series resistance can degrade common-mode rejection. An output filter avoids that imbalance but adds settling time and interacts with ADC acquisition. Follow the manufacturer’s stability and filtering guidance for the chosen device.

Where loss or galvanic separation dominates, compare the shunt chain with a Hall-effect current sensor. The Hall solution removes the precision shunt but adds magnetic geometry, offset, and mechanical-error terms.

Validate the Complete Measurement Chain

Test known currents at the low, mid, and high ends of the range, in both directions where used. Repeat at rail-voltage extremes and operating temperatures. Add the relevant common-mode steps, PWM states, startup, overcurrent, and recovery events while observing amplifier output and digitized results.

Compare measured error with the same budget and conditions used for approval. Include probe and source uncertainty so a test does not appear more precise than its equipment. Inspect the shunt temperature and Kelvin-node waveforms; a correct amplifier output can still represent the wrong voltage if layout includes load-current copper.

Use the TI or Analog Devices catalog to organize candidates, then control the exact gain option, package, temperature grade, qualification, and pinout. If the measurement feeds a high-resolution converter, apply the same exact-code discipline used for data-converter sourcing. An alternate is approved only when the full current-to-code chain meets its specified range, error, and timing limits.