Current-Sense Front Ends: Separate RF Rectification from Resistor Drift

System boundary

A drawing-defined current-sense resistor and its connections, analog amplifier, input filtering, nearby interference source and acquisition chain. The page concerns rectification diagnosis, not a system EMC qualification or a permitted energized test procedure.

Integration interfaces

Integration risks

Determine whether the offset exists before conversion

Hold the intended shunt current at a stable, independently observed value. Compare the amplifier's analog DC output with the digital reading while preserving the interference condition. If the analog mean already changes, altering ADC sample timing cannot remove the physical offset that the converter receives. A second digitizer with a different clock can provide supporting evidence only when its own loading and interference response are understood.

Keep this investigation separate from linear pickup sampled at an unfortunate phase. In that case a waveform can have an unchanged analog mean while selected digital samples look biased. Both effects can coexist. Retain the analog waveform and acquisition timing so the diagnosis does not force every slow-looking error into the same category.

Understand how a zero-mean disturbance creates a mean output

An ideal linear amplifier scales a zero-mean sinusoid without creating a DC term. A nonlinear input can behave differently. As a deliberately simplified local model, write the error contribution as a1v+a2v squared. With v=A cos(omega t), the average linear term is zero, whereas the average quadratic term is a2A squared divided by two.

This model illustrates rectification rather than specifying an amplifier. Real input stages have frequency-dependent coupling, protection structures and memory effects; their coefficients are not obtained from a printed resistor tolerance. The sign of the resulting offset can depend on circuit asymmetry. A negative measured shift does not rule out rectification merely because the illustrative coefficient was chosen positive.

Translate the offset into an apparent current

Assume the local model has a2=0.02 per volt and receives an interference amplitude of 0.10 volt peak. Its mean error is 0.00010 volt, or 100 microvolts. If that is an equivalent input error ahead of a gain of fifty, the output error is 5 millivolts. A current calculation using an assumed 20 milliohm shunt would interpret 100 microvolts as an extra 5 milliamperes.

These values are selected to demonstrate the conversion, not to describe a tested component. Reducing amplitude at the nonlinear boundary to one fifth gives a quadratic term one twenty-fifth as large: 4 microvolts input-equivalent and 0.20 millivolt at the gain-fifty output. A real circuit may not follow that scaling across a wide amplitude or frequency range, so the trend needs controlled verification.

Place mitigation before the unwanted frequency conversion

Once RF has become a DC offset, a low-pass filter placed only at the amplifier output passes that offset along with the wanted DC signal. A filter before the susceptible input can reduce the RF reaching the nonlinear boundary. Shielding or a corrected return path may reduce coupling earlier still. The appropriate choice depends on the identified source and circuit topology.

For a differential input, preserve the intended impedance balance and account for added capacitance, leakage and settling. A resistor change can affect both filtering and the normal error budget. Do not install a generic capacitor across an unknown high-energy sense interface. The circuit and EMC owners should review a complete safe solution with the selected active device and actual operating states.

Use observations that distinguish competing mechanisms

Start with a fixed mechanical and electrical configuration. Change one authorized factor at a time and record the original state before restoring it. The following contrasts identify where additional evidence is needed; no single signature proves a particular semiconductor mechanism.

Compare amplitude and frequency without losing the thermal state

A controlled frequency sweep can reveal a susceptible region that is invisible in an ordinary DC accuracy test. Record actual applied conditions and the observed input boundary where the approved method permits. Generator settings alone do not establish the voltage reaching the device because cables, coupling components and source impedance transform it. The qualified test procedure should address those paths.

An amplitude trend resembling a square law supports the local model but is not conclusive. Electrical heating also often scales with the square of a stimulus and can produce a thermal offset. Record temperature and response time, and include recovery after exposure is removed. Keep a sequence that separates fast electrical changes from slower thermal changes without assuming either timescale is universal.

Do not trim a correct resistor to cancel a receiver problem

Measure the resistor at its defined sense boundary using the accepted resistance method, independently of the suspect amplifier where practical. Compare that result with the raw shunt voltage and the receiver's current estimate. A stable resistor accompanied by an exposure-dependent receiver offset calls for a signal-chain investigation, not automatic adjustment of the printed value.

A zero stored under one interference condition can become wrong when that condition disappears or its frequency changes. Preserve the original correction and the conditions under which it was established. Recalibration may be appropriate after a validated hardware change, but it should not be used as evidence that immunity has improved. The uncorrected error remains necessary for evaluating the actual mechanism.

Confirm normal performance after the interference correction

After the selected mitigation, repeat ordinary gain, offset, input range, transient response and fault-visibility checks relevant to the current channel. A quieter output is not sufficient if a filter now hides a required current event or creates excessive delay. Compare both interference-present and interference-absent conditions without changing the definition of the reported current.

Retain the shunt and network drawings, active-device identities, layout, wiring, filter values, exposure method and raw analog results with the decision. ChipSimple can review the drawing-defined passive circuit contribution; the equipment owner retains EMC and safety acceptance. The completed record should identify whether the correction reduced coupling, avoided rectification or changed another error source, rather than attributing every improvement to resistor precision.