123 Main Street, New York, NY 10001

Submit Your BOM

CMRR & PSRR in Instrumentation Amplifiers (INA)

CMRR & PSRR in Instrumentation Amplifiers (INA)

← Back to:Instrumentation Amplifiers (INA)

CMRR/PSRR only matter as effective performance in real wiring: frequency, source-impedance mismatch, and parasitic imbalance often dominate the datasheet number. This page shows how to model, measure, and budget CM and supply disturbances into input-referred error, with clear pass criteria for release.

What CMRR/PSRR Really Means in Field Wiring

CMRR/PSRR are not “one number.” The usable value in real wiring is a frequency-dependent system result dominated by asymmetry (ΔR, unequal parasitics) and coupling paths—often more than the amplifier’s intrinsic spec.

Reality check: what the spec does (and does not) guarantee

DC CMRR mainly reflects input-stage mismatch and biasing behavior at low frequency.

AC CMRR is frequently limited by unequal parasitics (cable/PCB capacitance imbalance) that convert common-mode into differential error as frequency rises.

PSRR is a rail-to-output coupling spec that changes with gain, output load/swing, internal bias paths, and frequency.

Dominant mechanisms in the field: “effective” CMRR/PSRR

The board-level result is an effective rejection ratio that combines intrinsic amplifier limits with wiring asymmetry and parasitic coupling. Two practical conversions keep the discussion measurable:

Common-mode to differential error (input-referred): Verr,in ≈ VCM · 10−CMRR/20

Supply ripple as an input-equivalent error: Verr,in ≈ (Vripple,out/G)

If measured rejection changes dramatically when a cable is moved, a probe is swapped, or one input has “extra” series resistance/capacitance, the dominant limiter is typically asymmetry rather than the intrinsic amplifier CMRR.

Quick sanity checks (before deep measurement)

Swap wiring conditions: compare short, symmetric leads vs long cable/fixture. Large deltas indicate wiring-dominated effective CMRR.

Sweep frequency bands: low-frequency stability with high-frequency collapse strongly suggests parasitic-capacitance imbalance.

Move/perturb the cable: if residual output changes with touch/motion, coupling geometry (Cp imbalance, return path) is participating.

Minimum record fields for repeatability: gain, common-mode level, common-mode amplitude, frequency range, cable/fixture type, output bandwidth setting, and load.

Acceptance framing: what “good” looks like

  • Within the target band, rejection should not hinge on incidental setup changes (probe ground lead, minor cable motion).
  • Improving symmetry (matched series elements, mirrored routing) should measurably improve the residual—otherwise intrinsic limits or operating point constraints are likely dominant.
  • Results must be reported as a function of frequency and stated conditions (gain, common-mode level, load, bandwidth).
Field wiring chain: effective CMRR and PSRR paths Block diagram with sensor, long cable, input network, INA, ADC, and supply ripple showing two main coupling paths: common-mode to differential and supply ripple to output. Field Chain (Reality): Effective CMRR/PSRR Is Set by Asymmetry + Frequency Two dominant paths: CM→DM (wiring imbalance) and Vsup→Vout (rail coupling) Sensor Bridge / Diff Long cable Rlead / Cp Input network Protection / RC INA ADC CM → DM (ΔR, Cp imbalance) Supply ripple Vsup noise Vsup → Vout Key idea Datasheet CMRR/PSRR + wiring asymmetry → effective rejection vs frequency. Always report gain, CM level, load, bandwidth, and fixture/cable.
Diagram: The field result is dominated by asymmetry (ΔR, unequal Cp) and frequency—more than a single “CMRR/PSRR” number.

Datasheet Reading: The 6 Traps That Make CMRR/PSRR Look Better Than They Are

CMRR/PSRR must be read together with test conditions. Missing curves, “typical-only” numbers, and idealized fixtures can produce optimistic expectations that collapse on real wiring.

The correct reading frame: bind the spec to conditions

Any CMRR/PSRR claim should be interpreted as a function of four condition groups:

Gain (G), frequency (f)

Source impedance and symmetry (Rs+, Rs−, ΔR, Cp mismatch)

Operating point (input CM level, output swing)

Load and measurement bandwidth (R/C load, filtering/averaging)

The 6 traps (each with a failure mode and a quick identification)

1) Single-point or DC-only CMRR

Failure: AC common-mode converts to differential above a few kHz. Check: a “CMRR vs frequency” curve must exist (preferably at multiple gains).

2) Ideal symmetric source impedance

Failure: real ΔR (leads/connectors/series parts) dominates effective CMRR. Check: the test circuit should state Rs+/Rs− or fixture symmetry assumptions.

3) “Typical only” with no guaranteed minimum

Failure: worst-case units/temperature corners break accuracy budgets. Check: confirm min values and the temperature range tied to the spec.

4) PSRR not separated by rail/domain

Failure: one rail injects ripple even when another looks clean. Check: PSRR+ / PSRR− (or AVDD/DVDD) should be specified independently.

5) Missing load/swing conditions

Failure: PSRR shifts with output drive demands and stability regions. Check: look for R/C load, output swing, and bandwidth conditions.

6) Measurement bandwidth/fixture not disclosed

Failure: lab readings are dominated by the instrument, grounding, or fixture parasitics. Check: the method must declare bandwidth limits and fixture symmetry.

When curves are missing: the minimum request set

For production-ready selection, request the following (or treat the risk as unbudgetable):

  • CMRR vs frequency across the target band, at the intended gain(s), with stated source/fixture assumptions.
  • PSRR vs frequency for each relevant rail/domain, at intended gain(s), with stated output load/swing conditions.
  • Guaranteed limits (min) and the temperature range tied to those limits.
  • Test conditions: common-mode level and amplitude, measurement bandwidth/averaging, and fixture topology.

Rule for decision-making: if the necessary curves/conditions are unavailable, the design must plan board-level verification before committing to volume.

Pass criteria: documentation completeness for predictable outcomes

  • CMRR/PSRR are provided as frequency-dependent data tied to gain and stated test conditions.
  • Values include guaranteed minimums (not only typical) over the required temperature range.
  • Output load/swing and measurement bandwidth are explicitly stated, enabling apples-to-apples comparison.
Datasheet setup vs field setup for CMRR and PSRR Two card panels compare ideal datasheet conditions to real field wiring conditions, highlighting asymmetry, parasitics, and missing curves that affect effective CMRR and PSRR. Specs Look Better When Conditions Are Idealized Always bind CMRR/PSRR to gain, frequency, symmetry, operating point, and load/bandwidth. Datasheet setup (ideal) Short leads Matched Rs Single tone Symmetric fixture Often reported as a single dB number Field wiring (reality) Long cable ΔR (mismatch) Cp imbalance Load / bandwidth shifts Effective rejection varies with frequency & setup optimism gap Must-have: CMRR vs f (multi-gain) + PSRR vs f (per rail) + stated conditions (Rs symmetry, CM level, load, bandwidth).
Diagram: Without frequency curves and stated conditions, comparing CMRR/PSRR numbers is not predictive of field performance.

Core Error Model: How Common-Mode Becomes Differential Error

This section converts a CMRR “dB number” into a measurable, budgetable differential error. The model below describes intrinsic (amplifier-limited) leakage only; external asymmetry is treated separately.

Step 1: Convert CMRR(dB) into a linear CM→DM leakage gain

CMRR is a frequency-dependent spec. Use the value at the disturbance frequency band (CMRR(f)), not a DC-only number.

Linear leakage gain: Acm→dm = 10−CMRR/20

Step 2–3: Turn common-mode disturbance into input- and output-domain error

Use a single, consistent definition for disturbance amplitude (RMS or peak-to-peak) across the entire chain. Then apply the three-step mapping:

Input-referred differential error: Verr,in ≈ VCM · Acm→dm

Output error after gain: Verr,out = G · Verr,in

ADC budgeting hook: if comparing to digital resolution, translate Verr,out into LSB using the ADC full-scale and resolution. Keep bandwidth consistent with the measurement method.

Interpretation: if this intrinsic model predicts an error far below the observed residual, the dominant limiter is typically external asymmetry (ΔR / ΔZ / Cp mismatch), not intrinsic CMRR.

Align the model with real measurements (avoid apples-to-oranges)

Frequency alignment: use CMRR(f) at the disturbance band. Switching ripple, mains pickup, and RF ingress live at very different frequencies.

Amplitude alignment: keep RMS vs p-p consistent for VCM and residual readout; do not mix “scope p-p” with “RMS noise” numbers.

Bandwidth alignment: measurement filtering/averaging changes the reported residual. Record output bandwidth and apply the same band when interpreting Verr,out.

Pass criteria template (intrinsic CMRR contribution)

  • For the target frequency band and maximum expected VCM, the predicted Verr,in must remain below the input error budget allocation.
  • With stated gain G, the resulting Verr,out must not consume an unacceptable share of ADC full-scale or resolution (expressed in LSB or %FS).
  • Results are reported with conditions: gain, CM level, CM amplitude definition (RMS/p-p), and measurement bandwidth.
CM to DM error mapping using CMRR and gain Block diagram shows Vcm disturbance passing through a CMRR leakage block to become input-referred differential error, then amplified by gain G to output error; notes that external asymmetry is not included. Intrinsic Model: Vcm → Verr,in → Verr,out Convert CMRR(dB) to Acm→dm, then propagate through gain G Vcm Disturbance CMRR block Acm→dm = 10^(−CMRR/20) Verr,in ≈ Vcm · Acm→dm Gain G Verr,out = G · Verr,in Scope of this model Intrinsic leakage only. External asymmetry (ΔR / Cp mismatch / fixture imbalance) can dominate effective CMRR in real wiring. If predicted error is far below measured residual, inspect asymmetry before blaming the amplifier.
Diagram: Convert CMRR(dB) to a linear CM→DM leakage gain and propagate the error through gain G (intrinsic model only).

The Dominant Reality: Source Impedance Mismatch and Lead Resistance

In bridges and remote sensors, effective CMRR is often limited by asymmetry outside the INA: unequal lead resistance, connector contact resistance, unmatched series parts, and leakage paths that turn common-mode into differential error.

Why small ΔR matters: asymmetry is a CM→DM converter

Two input paths that are “almost the same” are still different impedances. Any common-mode current that flows through those paths produces unequal drops, which appear as a differential error at the INA inputs.

Two dominant mechanisms (engineering first-order)

DC / low-frequency: ΔR × effective input current

Any effective input current (bias return, leakage through protection, contamination) creates unequal drops: Vdm,err ≈ Iin,eff · ΔR

Higher frequency: common-mode current × impedance imbalance

Parasitic coupling creates common-mode current; impedance imbalance turns it into differential error: Vdm,err ≈ Icm · ΔZ

Typical asymmetry sources: unequal lead resistance, connector contact variation, unmatched series resistors/RC parts, sensor output impedance differences, and leakage paths that change with humidity and contamination.

Practical troubleshooting: make asymmetry visible

Measure ΔR end-to-end: include connectors, cable, series parts, and any “one-sided” protection resistor.

Enforce symmetry: mirror the component count, placement, and values on both inputs (including test fixtures).

Swap +/− inputs: if the error follows the physical path rather than polarity, external asymmetry is dominant.

Short-lead baseline: compare to a short, symmetric setup to isolate the contribution from cable/fixture parasitics.

Leakage sensitivity check: if low-frequency drift correlates with humidity/handling, model leakage as Iin,eff.

Pass criteria template (external asymmetry contribution)

  • ΔR (end-to-end) is controlled by design and verified by measurement; both inputs contain mirrored series elements.
  • With the specified cable/fixture, residual error does not change abruptly with minor mechanical perturbations (touch/motion), indicating controlled coupling geometry.
  • Swapping +/− inputs does not cause an unexplained change in error magnitude beyond the model’s expected range.
Lead resistance mismatch: ΔR converts common-mode to differential error Two input paths with series lead resistances Rlead+ and Rlead- show mismatch ΔR; common-mode disturbance and effective input/leakage currents create unequal drops leading to Vdm error at INA input. External Asymmetry Model: ΔR and Leakage Turn CM into DM Rlead mismatch (connectors / cable / series parts) is often the dominant limiter Vcm Common-mode Rlead+ Rlead− ΔR Cp+ Cp− INA Ib Leak Vdm err Key takeaway Effective CMRR often tracks ΔR/ΔZ and leakage sensitivity more than intrinsic CMRR; enforce symmetry and verify end-to-end resistance.
Diagram: Lead resistance mismatch (ΔR) plus bias/leakage currents converts common-mode into a differential error at the INA input.

Why CMRR Collapses with Frequency: Parasitic Capacitance Imbalance

High-frequency CMRR is often limited by geometry, not by the INA core. Unequal parasitic capacitance (Cp+ ≠ Cp−) creates an impedance imbalance that converts common-mode disturbance into a differential residual.

Core mechanism: Cp imbalance turns CM current into DM voltage

Capacitive impedance drops with frequency: |ZC| = 1/(ωC) so parasitic paths become “stronger” as frequency increases.

If Cp+ ≠ Cp−, then Z+ ≠ Z−. Common-mode current flowing through unequal impedances creates an unequal drop that appears as a differential error at the INA input.

Practical causes of Cp imbalance (field reality)

Unequal proximity to noise sources: one input line runs closer to switching nodes, digital clocks, or high dv/dt nets.

Different return paths: the two inputs “see” different ground/chassis coupling due to routing, layer changes, or plane gaps.

Shield/fixture asymmetry: cable shield geometry, connector layout, and test fixtures often introduce one-sided capacitance.

Why touching a cable or changing a probe can shift the residual: it changes Cp (and the return geometry) on one side more than the other, increasing ΔZ and the CM→DM conversion.

Fast verification moves (keep it measurable)

Baseline with symmetry: short, symmetric twisted pair and mirrored fixture routing should improve stability of CMRR(f).

Probe-geometry A/B: compare “long ground lead” vs “short ground spring” at the same node; large changes indicate coupling geometry participation.

Gentle perturbation test: move the cable relative to chassis/metal; if the residual tracks position, Cp imbalance dominates.

Swap physical paths: if the error follows the physical routing rather than polarity, external imbalance is the limiter.

Pass criteria template (AC CMRR robustness)

  • Within the target frequency band, the residual does not change dramatically with minor probe or cable-geometry changes.
  • Symmetry improvements (mirrored routing/fixture) produce a clear, repeatable reduction in the CM→DM residual.
  • Reported results include frequency band, cable/fixture geometry, probe method, gain, and output bandwidth.
Cp imbalance: AC CMRR drops with frequency Two input lines each have parasitic capacitance to chassis; Cp+ differs from Cp- creating impedance imbalance and CM to DM conversion; includes a simplified CMRR vs frequency curve indicating drop region. Cp+ ≠ Cp− → ΔZ → AC CMRR drop Geometry and return paths control effective CMRR(f) Vcm Disturbance Cp+ Cp− ΔZ Icm INA Vdm err CMRR(f) f drop
Diagram: Unequal parasitic capacitance (Cp+ ≠ Cp−) creates an impedance imbalance that worsens AC CMRR as frequency increases.

PSRR Deep Dive: Supply Ripple Paths and Input-Referred Interpretation

PSRR is not a single “dB score.” It is rail- and frequency-dependent, and it reaches the output through multiple internal paths. Converting output ripple to input-referred error (/G) makes budgeting and comparisons meaningful.

Read PSRR correctly: per rail / per domain / vs frequency

Separate rails: PSRR+ / PSRR− (or AVDD / DVDD) can differ significantly and must be evaluated independently.

Use PSRR(f): supply ripple is often concentrated in switching bands; a DC-only PSRR value is not predictive.

Two internal coupling paths (keep the model budgetable)

Path A: input-stage / bias coupling (input-referred behavior)

Rail ripple perturbs internal bias/reference nodes and appears as an input-equivalent error that is then multiplied by gain.

Path B: output-stage / drive coupling (rail-to-output behavior)

Rail ripple couples through output drive, headroom, and load-dependent behavior and shows up directly at the output.

Linear mapping: Asup→out = 10−PSRR/20 Vripple,out ≈ Vsup,ripple · Asup→out

Input-referred interpretation (the budgeting-friendly view)

Convert output ripple into an input-equivalent error: Verr,in ≈ Vripple,out / G

Use consistent bandwidth: the reported ripple depends on measurement bandwidth and filtering. Record bandwidth and keep it constant for comparisons.

Rail attribution matters: if multiple rails inject simultaneously, evaluate sensitivity per rail/domain to find the dominant contributor.

Interpretation: if Vripple,out scales strongly with gain, Path A is likely dominant; if it tracks load/swing changes more than gain, Path B is likely dominant.

Pass criteria template (PSRR contribution)

  • PSRR is evaluated per rail/domain over the target frequency band with stated gain, output load, output swing, and measurement bandwidth.
  • For each rail, the maximum expected Vsup,ripple translates to Verr,in that stays below the allocated input error budget.
  • Results are reported as both Vripple,out and input-referred Verr,in to support system-level budgeting.
PSRR paths: Vsup ripple to Vout ripple and input-referred error Supply ripple enters separate blocks for input stage and output stage coupling, both contributing to output ripple; output ripple is converted to input-referred error by dividing by gain G. PSRR Model: Rail ripple → internal paths → Vout ripple → /G Evaluate per rail (PSRR+ / PSRR− or AVDD / DVDD) and convert to input-referred error Vsup ripple PSRR+ PSRR− Input stage bias / reference Output stage drive / load Vout ripple /G Verr,in Budgeting view: measure Vout ripple in a stated bandwidth, then convert to input-referred by /G; compare per rail and vs frequency.
Diagram: Supply ripple couples through both input-stage (bias/reference) and output-stage (drive/load) paths; convert output ripple to input-referred error by dividing by gain G.

How to Measure CMRR vs Frequency Without Lying to Yourself

A usable CMRR(f) measurement is a repeatable procedure: true common-mode injection, residual extraction in the output domain, baseline subtraction, and a strict record format. If any of those steps are skipped, the curve often becomes a measurement of probes and fixtures instead of the DUT.

Define the measurement: what is injected, what is measured, what is computed

Injection (true common-mode): apply the same signal in-phase to IN+ and IN−, with matched path impedance and the same reference.

Measurement: extract the residual differential component at the output (preferably at the injection tone in the frequency domain).

Compute: Acm→out(f) = Vout,res(f) / VCM,inj(f)

Input-referred view: Verr,in(f) ≈ Vout,res(f) / G

CMRR(f): 20·log10( VCM,inj(f) / Verr,in(f) )

Bandwidth rule: Vout,res depends on measurement bandwidth (RBW / averaging / filtering). Record bandwidth and keep it constant across sweeps and boards.

True common-mode injection: three non-negotiable requirements

Same source node: IN+ and IN− must be driven from the same injection node so they are phase-locked and amplitude-matched.

Matched path impedance: equal cable length, equal series parts, mirrored fixture geometry. Any ΔR/ΔCp becomes a CM→DM converter.

Controlled reference: define what the injected common-mode is referenced to (board ground or chassis) and keep it consistent.

Extract the residual correctly (tone-based and baseline-subtracted)

Prefer frequency-domain readout: measure the magnitude at the injection frequency bin (FFT / spectrum analyzer) rather than time-domain peak-to-peak.

Baseline subtraction: capture a “noise floor” spectrum with injection off (or input shorted to the same CM level) using identical settings.

Stay in linear operation: avoid output clipping and headroom limits; distortion can create spurs that masquerade as “poor CMRR.”

A quick sanity check: if swapping the physical IN+ and IN− paths changes the curve significantly, the fixture dominates the result.

The 6 most common “fake CMRR” failure modes (and the fast fixes)

Ground loop

Source ground + scope ground + DUT ground form a loop. Fix: isolate the source or eliminate multi-point ground references.

Generator return coupling

The generator return forces unequal current paths. Fix: use a defined injection node and symmetrical routing.

Probe ground lead

Long probe ground adds inductance and picks up fields. Fix: use a short ground spring and keep loops tiny.

Fixture asymmetry

Small ΔR/ΔCp converts CM to DM. Fix: mirror both input paths and verify end-to-end symmetry.

Cable geometry drift

Touching/moving a cable changes Cp. Fix: constrain geometry and keep distance to chassis consistent.

Bandwidth mismatch

Different RBW/filters change the measured residual. Fix: lock bandwidth, windowing, averaging, and report them.

Mandatory record format (so others can reproduce the curve)

  • Sweep plan: frequency points / spacing, dwell time, averaging count.
  • Injection: VCM amplitude definition (RMS or p-p), injection reference, source impedance, injection node description.
  • DUT: gain G, input common-mode level, output load, supply rails, output headroom status.
  • Measurement: instrument type, bandwidth/RBW, windowing, probe method (short ground vs lead), cable lengths.
  • Fixture: geometry notes (symmetry, shielding, distance to chassis), and revision/version of the fixture.
CMRR(f) test fixture: true common-mode injection and residual measurement Signal source feeds an isolation and split block into two symmetric paths to IN+ and IN- of DUT; output goes to FFT/scope; labels indicate guard, short leads, and fixture symmetry. CMRR(f) test fixture: inject CM, measure residual Isolation + symmetry prevent “fake CMRR” from ground loops and fixtures Signal source Vcm inj Isolate break loops Split matched paths Z Z DUT INA IN+ IN− Measure FFT / RBW Guard Short lead Symmetry Log: f, Vcm, CM level, gain, load, bandwidth, probe method, and fixture geometry.
Diagram: A repeatable CMRR(f) setup uses isolation and symmetric split paths to ensure true common-mode injection and valid residual extraction.

How to Measure PSRR Across Real Rails (Including Switching Ripple)

PSRR measurements fail most often due to uncontrolled variables: rail/domain ambiguity, injection that shifts the DC operating point, and output noise that hides the injected component. The procedure below isolates the injected tone and keeps input and operating conditions constant.

Define the PSRR measurement: injected rail ripple vs output response

Inject: a small AC ripple Vsup,inj(f) onto one rail at a defined injection point (one rail at a time).

Measure: the net output component at the same frequency (tone-based readout), separated from baseline noise.

Compute: PSRR(f) = 20·log10( Vsup,inj(f) / Vout,inj(f) )

Injection approach (principles, not complex circuits)

Series injection

Insert a small controlled injection impedance so the AC ripple appears on the rail without large DC shift.

Coupled injection

Use coupling (e.g., transformer/AC coupling concept) to superimpose ripple while keeping DC supply regulation intact.

Injection network

Control the AC path and keep the injection local to a known point (module output vs IC pin are different tests).

Constraint: injection must be small enough that the DC operating point does not shift (no clipping, no headroom change, no load-step behavior).

Measure the injected component (separate it from baseline noise)

Tone readout: use FFT / analyzer to read the amplitude at the injection frequency. This avoids “noise floor confusion.”

Baseline subtraction: record output spectrum with injection OFF using identical RBW/bandwidth and averaging; subtract at the tone bin.

Switching ripple band: include the converter ripple band and its vicinity; PSRR(f) can differ dramatically there.

Control variables (PSRR changes with real operating conditions)

Gain and bandwidth: keep gain G and measurement bandwidth constant across rails and frequencies.

Load and output swing: keep output load and swing fixed; near-rail operation can change the coupling path.

Input quiet + constant CM: keep input differential quiet (zero) and hold the same input common-mode level while injecting on the rail.

One-rail-at-a-time rule: inject on one rail while keeping the other rails as clean and constant as possible, otherwise rail cross-coupling hides the true sensitivity.

Mandatory record format (PSRR is not comparable without this)

  • Which rail/domain: PSRR+ / PSRR− (or AVDD / DVDD); injection point location (module output vs IC pin).
  • Injection: method category, Vsup,inj amplitude definition (RMS or p-p), sweep band (include switching band), dwell and averaging.
  • DUT: gain G, input CM level, output load, output swing/headroom state, supply voltages.
  • Measurement: instrument, RBW/bandwidth, windowing, averaging, probe method and wiring.
  • Outputs: PSRR(f) curve plus optional input-referred Verr,in(f) = Vout,inj/G for budgeting.
PSRR test: inject ripple on a rail and measure output response Injection source connects to a marked injection point on the supply rail feeding the DUT; output goes to measurement; side labels indicate keeping input quiet and constant common-mode level. PSRR test: rail injection and tone extraction Mark the injection point and keep input conditions constant Rails AVDD DVDD Inject Ripple injector tone / sweep DUT INA Keep input quiet constant CM Measure FFT / RBW One rail at a time Fixed load & swing Record bandwidth Report: rail, injection point, Vsup,inj, f-band (include switching), gain, CM, load, swing, RBW/bandwidth, and baseline method.
Diagram: Inject a small ripple at a defined rail point and extract the output tone while keeping input quiet and operating conditions fixed.

Turning dB into µV/LSB: Budgeting CMRR/PSRR for Precision

CMRR and PSRR only become actionable when they are translated into input-referred error (µV) and mapped to ADC LSB and budget share. The workflow below turns field interference and rail ripple into a comparable, decision-ready number.

Step 1 — Convert dB specs to linear coupling coefficients

CMRR(f): Acm→dm(f) = 10−CMRR(f)/20

Use CMRR(f) at the frequency of the disturbance, not a single DC number.

PSRR(f): Asup→out(f) = 10−PSRR(f)/20

Treat rails separately (AVDD/DVDD or +/−). Different domains often have different PSRR.

Consistency rule: keep amplitude definition consistent (RMS or p-p) across Vcm, Vripple, and measurement readouts.

Step 2 — Compute input-referred error from field Vcm and rail ripple

CMRR path (common-mode → input-referred differential error)

Verr,in,CM(f) ≈ VCM(f) · Acm→dm(f)

This isolates the CMRR-related mechanism into a single input-referred number in µV.

PSRR path (rail ripple → output ripple → input-referred)

Vout,ripple(f) ≈ Vsup(f) · Asup→out(f)

Verr,in,PS(f) ≈ Vout,ripple(f) / G

Converting to input-referred simplifies budgeting and avoids gain confusion.

Compare apples-to-apples: evaluate Verr,in,CM(f) and Verr,in,PS(f) in the same frequency band and the same measurement bandwidth.

Step 3 — Map µV to ADC LSB and budget share (mapping only)

ADC LSB: LSB = VFS / 2N

Use the ADC’s actual full-scale definition and input range in the same units as Verr,in.

Error in LSB: ErrorLSB = Verr,in / LSB

This directly connects CMRR/PSRR to resolution and decision thresholds.

Budget share: Share = Verr,in / Vbudget,in

Vbudget,in is the allowed input-referred contribution assigned to CMRR/PSRR-related mechanisms.

Spur vs wideband note: a single injected tone behaves like a discrete spur, while wideband ripple/noise must be interpreted under a declared measurement bandwidth.

Step 4 — Decide which dominates (conditions, not adjectives)

CMRR / asymmetry dominates when

  • Verr,in,CM(f) exceeds Verr,in,PS(f) across most of the band.
  • Results are highly sensitive to cable geometry, fixture symmetry, or swapping IN+ and IN− paths.
  • The dominant residual tracks “path imbalance” rather than rail ripple amplitude.

PSRR dominates when

  • Verr,in,PS(f) is larger and correlates with rail ripple magnitude and switching bands.
  • Changing load or output swing changes the measured injected component (under fixed injection).
  • Injecting one rail at a time reveals a clear rail-specific sensitivity.

Step 5 — Combine components correctly (tone vs multiple spurs vs bandwidth)

Single tone: use the magnitude at the injection frequency bin (FFT/analyzer) and map it through the workflow above.

Multiple discrete spurs: combine input-referred terms by RSS if they are uncorrelated: Verr,in,total ≈ √(Σ Verr,in,i2)

Wideband content: always declare the measurement bandwidth and keep it fixed; “µV” without bandwidth is not comparable.

Budget flow: dB to µV to LSB Two inputs Vcm and rail ripple feed dB-to-linear conversion blocks, combine into input-referred error in microvolts, then map to ADC LSB and budget share. Budget flow: translate specs into system error Vcm / Vripple → linear coupling → input-referred µV → LSB → budget share Vcm(f) field CM Vripple(f) rail ripple CMRR(f) 10^(−dB/20) PSRR(f) 10^(−dB/20) Verr,in (µV) compare paths input-referred ADC /LSB LSB Budget share Verr,in / Vbudget,in Keep units consistent and declare bandwidth for any wideband measurement.
Diagram: A budgeting workflow that converts CMRR/PSRR from dB into input-referred µV, maps to ADC LSB, and reports budget share.

Design Hooks to Maximize Effective CMRR/PSRR (Only symmetry, parasitics, and verification points that directly impact CMRR/PSRR)

Effective CMRR/PSRR is usually limited by external asymmetry and operating conditions, not by the headline datasheet dB. The hooks below focus on mirrored impedance, mirrored parasitics, and controlled reference consistency—without expanding into full EMI grounding or power filter design.

Symmetry first: keep geometry, coupling, and return paths matched

Matched length and spacing: route IN+ and IN− as a pair (equal length, consistent spacing, same layer, same reference plane).

Matched coupling: keep both inputs at similar distance to aggressors and chassis; avoid “one side closer” parasitic capacitance.

Matched return: ensure both paths see the same return geometry; asymmetry in return paths behaves like CM→DM conversion.

Source impedance matching: avoid “accidental ΔZ” in front of the INA

Mirror front-end parts: any series R, RC, clamp network, or connector transition must be mirrored and value-matched on both inputs.

Keep the part count symmetric: one extra via, one extra pad, or one extra “optional jumper” can dominate high-frequency effective CMRR.

Fixture and cable symmetry: if measurement results change with cable touch or movement, Cp mismatch is likely the limiter.

Reference consistency: keep what the INA “sees” consistent (pointed guidance only)

Same reference environment: both inputs should couple similarly to the reference node (board ground or chassis) to avoid asymmetry-driven residuals.

Do not create hidden return paths: any additional return path on one side (through shielding, mounting, or test leads) can convert CM movement into DM error.

Verification hint: if swapping only the input leads changes the result significantly, the limiting factor is likely external asymmetry rather than intrinsic device CMRR.

PSRR reality check: load and swing can change rail-to-output sensitivity

Same injection, different load: if the injected tone at Vout increases with heavier load, the effective PSRR is load-dependent and must be budgeted at worst load.

Same injection, different swing: if the injected tone grows near output headroom limits, budget PSRR under worst swing and worst rail headroom.

One rail at a time: rail cross-coupling can mask the true sensitivity. Inject and record rails independently.

Production-ready checklist (effective CMRR/PSRR focused)

  • IN+ and IN− have mirrored components (same count, same values, same placement).
  • Same number of vias and transitions; pair routing is symmetric and consistent.
  • Distance to chassis/aggressors is matched; Cp markers are symmetric.
  • CMRR(f) validation uses a symmetric fixture and a fixed bandwidth record format.
  • PSRR(f) validation fixes load and swing and injects one rail at a time.
Layout symmetry checklist: maximize effective CMRR/PSRR Two mirrored input paths show connector, series resistor, filter capacitor, clamp, and INA pins; symmetric Cp markers to chassis/ground are shown; short labels indicate match Z, match Cp, and match return. Effective CMRR/PSRR: symmetry and parasitic control Mirror Z, Cp, and return paths to avoid CM→DM conversion ≡ IN+ path IN− path Conn R C Clamp INA pin Conn R C Clamp INA pin Cp Cp Match Z Match Cp Match return
Diagram: A mirrored front-end (same parts, placement, geometry, and parasitics) is the most reliable way to maximize effective CMRR and preserve real PSRR behavior.

Selection Notes: What to Ask Vendors and What Curves Must Exist

For CMRR/PSRR-driven precision work, curves without conditions are not budgetable. The checklist below forces “field-usable” answers: required curves, required test conditions, and clear reject clauses.

A) Must-have curves (missing curve = uncontrolled risk)

1) CMRR vs frequency (CMRR(f))

  • At least three gain points: G=1, G=10, G=100 (or closest available gain settings).
  • Must state minimum / guaranteed behavior (not typical-only).
  • Must state the input common-mode level used for the curve.

2) PSRR vs frequency (PSRR(f)) by rail / domain

  • Separate curves for PSRR+ and PSRR−, or AVDD/DVDD if rails are domain-split.
  • At least two gain points: G=1 and one higher gain point (e.g., G=10 or G=100).
  • Must state output load and output swing/headroom conditions.

Curves are only budgetable when gain, Vcm, source impedance model, load, and measurement bandwidth are explicitly declared.

B) Required test conditions to request (so results are reproducible)

CMRR(f) curve must include

  • Source impedance model: symmetrical Z? any series R / protection included?
  • Fixture symmetry statement: cable length, routing symmetry, ground reference handling.
  • Injected common-mode amplitude: RMS or p-p (must match the curve definition).
  • Measurement bandwidth / settings: analyzer RBW, FFT windowing, filtering.
  • Gain setting definition: RG value or internal gain code.
  • Temperature range: what “min” guarantee covers (ambient and/or junction assumptions).

PSRR(f) curve must include

  • Which rail is injected: rail/domain identity and injection reference point.
  • Output load: resistive and capacitive load conditions.
  • Output swing/headroom: output level relative to rails during the measurement.
  • Input condition: input quiet and common-mode fixed during injection.
  • Measurement bandwidth / settings: same reproducibility requirements as CMRR.

Practical RFQ rule: a curve without declared conditions cannot be used for worst-case budgeting.

C) Reject clauses (copy/paste for RFQ)

  • Missing CMRR vs frequency with at least G=1 and one higher gain point (G=10 or G=100) → risk not controllable.
  • Missing PSRR+ vs frequency and PSRR− vs frequency (or AVDD/DVDD domains) → risk not controllable.
  • Curves missing Vcm, source impedance model, load/swing, or measurement bandwidth → not reproducible, not budgetable.
  • Typical-only curves without minimum / guaranteed values or stated worst-case methodology → not acceptable for production budgeting.
  • No statement on condition relevance to real wiring (fixture symmetry or equivalent guidance) → effective CMRR/PSRR cannot be validated.

D) Reference part numbers (benchmark targets for RFQ completeness)

These examples are provided to speed up datasheet lookup and to benchmark curve/condition completeness. Final selection must be driven by the same curve-and-conditions checklist above.

General precision INA (bridges / industrial measurement)

  • Texas Instruments: INA826
  • Texas Instruments: INA818
  • Analog Devices: AD620

AC CMRR emphasis (useful for CMRR(f) scrutiny)

  • Analog Devices: AD8221

Higher bandwidth / faster recovery (dynamic measurement)

  • Analog Devices: AD8421
  • Texas Instruments: INA828

Zero-drift / low drift (weak DC signals)

  • Texas Instruments: INA333
  • Analog Devices: AD8237
RFQ checklist: curves and conditions required for CMRR/PSRR budgeting Eight stacked cards list required curves and test conditions: CMRR(f), PSRR+(f), PSRR-(f), Vcm condition, source impedance model, load and swing, bandwidth/settings, and temperature range. A reject rule is shown at the bottom. RFQ checklist (CMRR/PSRR): curves + conditions Request these fields in writing. Missing items cannot be budgeted. CMRR vs f (G=1, G=10, G=100; minimum/guaranteed) PSRR+ vs f (rail/domain stated; minimum/guaranteed) PSRR− vs f (or DVDD if domain-split; minimum/guaranteed) Vcm condition (common-mode level used for curves) Source impedance model (symmetry, series R, protection included?) Load & swing (output load, output headroom during PSRR tests) Bandwidth / settings (RBW, filtering, FFT/windowing, fixture notes) Temperature range (min guarantee coverage and assumptions)

Reject rule: missing curve or missing conditions = risk not controllable.

Diagram: An RFQ-ready checklist that forces curve completeness and declared test conditions for budgetable CMRR/PSRR decisions.

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

FAQs: CMRR & PSRR (Field Reality, Measurement, Budgeting)

These FAQs close long-tail issues without expanding the main text. Each answer is fixed to four executable lines: Likely cause → Quick check → Fix → Pass criteria.

Why does measured CMRR look great at DC but terrible above a few kHz? ▾
Likely cause: AC CMRR is dominated by external asymmetry (Cp mismatch / return-path mismatch), not DC intrinsic CMRR.
Quick check: Sweep frequency and repeat after swapping IN+ and IN− wiring; if the curve shifts, fixture/cable asymmetry is limiting.
Fix: Enforce mirror symmetry (same components/placement/vias/geometry) and use a symmetric fixture with fixed bandwidth settings.
Pass criteria: In the target band, residual at the injection frequency meets the budget and changes ≤ 2–3 dB after IN+/IN− swap or cable re-routing.
Why does touching/moving the cable change the “CMRR” result? ▾
Likely cause: Hand/cable motion changes parasitic capacitance to chassis/ground unequally (Cp+ ≠ Cp−), converting CM into DM.
Quick check: Hold the cable in a fixed symmetric shape and compare to a moved/touched state at the same frequency point.
Fix: Use a rigid symmetric fixture and keep both input paths equally coupled to their environment (distance to chassis/aggressors matched).
Pass criteria: Cable touch/motion changes the measured residual by ≤ 2 dB (or within the defined repeatability window) across the target band.
How can 1% resistor mismatch outside the INA dominate the spec? ▾
Likely cause: External ΔZ (series R/contact R/protection R) converts common-mode movement into a differential error (effective CMRR collapses).
Quick check: Measure DC resistance of both input paths end-to-end (including connectors) and quantify ΔR.
Fix: Mirror the networks (same values/placements), tighten tolerance where it matters, and remove “one-sided” series elements.
Pass criteria: After symmetry fixes, effective CMRR improves by ≥ 10 dB (or meets the system budget) and stays stable under input-path swapping.
Is AC CMRR limited by the INA or by parasitic capacitance imbalance? ▾
Likely cause: Low frequency is often intrinsic-limited; higher frequency is often Cp/geometry-limited (fixture and layout dominate).
Quick check: Keep the same INA and change fixture/cable symmetry; if high-frequency CMRR shifts, parasitics dominate.
Fix: Fix Cp/return-path symmetry first; only then compare INAs using the same controlled fixture and record format.
Pass criteria: High-frequency residual improves primarily with symmetry fixes; part-to-part changes are consistent and repeatable within ≤ 2–3 dB.
How to separate true CMRR limitation from ground-loop artifacts? ▾
Likely cause: Instrument grounding and return paths create unintended DM injection (generator ground, scope ground clip, asymmetric fixture return).
Quick check: Repeat with a controlled symmetric injection/distribution and re-route measurement grounds; large swings indicate artifacts.
Fix: Use a symmetric fixture, keep lead lengths short, and keep the measurement reference consistent (no “one-side-only” ground paths).
Pass criteria: Changing instrument grounding or probe attachment shifts the measured residual by ≤ 2 dB at each test point.
What common-mode amplitude should be used for a realistic CMRR test? ▾
Likely cause: Too small amplitude is noise-floor limited; too large amplitude triggers headroom/nonlinearity and invalidates the result.
Quick check: Sweep Vcm amplitude and verify the residual scales linearly (in the valid region) without clipping or settling anomalies.
Fix: Choose an amplitude that makes the residual ≥ 10–20 dB above the noise floor while keeping the INA in linear range.
Pass criteria: Residual vs Vcm is proportional (linearity holds) and repeated runs agree within ≤ 2 dB using fixed bandwidth settings.
Why does PSRR differ a lot between AVDD and DVDD (or +/− rails)? ▾
Likely cause: Different internal coupling paths exist per rail/domain; one rail may modulate bias/output more strongly than the other.
Quick check: Inject ripple into one rail at a time (same amplitude and frequency) and measure the output component at that frequency.
Fix: Budget PSRR per rail/domain and verify under the same declared conditions (gain, Vcm, load, bandwidth).
Pass criteria: For each rail, input-referred ripple contribution ≤ Vbudget,in across the target frequency band.
Why does PSRR look worse at high gain (or with heavy output load)? ▾
Likely cause: Rail-to-output sensitivity depends on operating point (headroom, output stage stress, load), and can vary with gain and load.
Quick check: Repeat PSRR at the same injection with different gains and loads; if the injected tone changes, PSRR is condition-dependent.
Fix: Verify PSRR under worst-case gain/load/swing, then convert to input-referred by dividing by gain for a comparable budget metric.
Pass criteria: Worst-case condition meets ≤ Vbudget,in; changing gain/load within the allowed envelope stays within ≤ 3 dB.
How to convert CMRR/PSRR dB into an input-referred error quickly? ▾
Likely cause: Budget mistakes come from using a single DC dB value or mixing RMS/p-p and bandwidth assumptions.
Quick check: Use frequency-specific values: A = 10^(−dB/20); then compute with the same amplitude definition (RMS or p-p).
Fix: CMRR: Verr,in ≈ Vcm·A. PSRR: Vout,ripple ≈ Vsup·A, then Verr,in ≈ Vout,ripple/G.
Pass criteria: Computed Verr,in matches injected measurement within ≤ 3 dB using the same bandwidth and gain.
When does PSRR dominate over CMRR in low-frequency precision systems? ▾
Likely cause: PSRR dominates when rail ripple is large in-band and the system common-mode disturbance is small or well-controlled.
Quick check: Compute both in the same band: Verr,in,CM ≈ Vcm·10^(−CMRR/20) and Verr,in,PS ≈ (Vsup·10^(−PSRR/20))/G.
Fix: Make the dominant path the release gate (measure and budget it at worst-case rail/load/swing and declared bandwidth).
Pass criteria: max(Verr,in,CM, V err,in,PS) ≤ Vbudget,in across the required low-frequency band.
Why does adding “a small RC” on only one input kill CMRR? ▾
Likely cause: One-sided RC creates ΔZ and ΔCp, converting common-mode content into a differential component (effective CMRR collapse).
Quick check: Add the same RC on the other input (same placement) and compare the residual at the injection frequency.
Fix: Mirror both impedance and parasitics: same component count/value/placement, same vias, same routing and return geometry.
Pass criteria: Symmetric networks restore the measured CMRR(f) to within ≤ 3 dB of the symmetric-fixture baseline across the target band.
What are pass criteria for CMRR/PSRR verification before release? ▾
Likely cause: “Great results” often come from undeclared conditions; without a record format, the numbers are not reproducible.
Quick check: Run a fixed matrix (frequency points, injection amplitude, Vcm, gain, load, bandwidth) and repeat ≥ 3 times.
Fix: Gate on worst-case points and sensitivity tests (swap IN+/IN−, cable re-route, rail-by-rail injection) under declared settings.
Pass criteria: Worst-case input-referred error ≤ Vbudget,in; repeatability ≤ 2–3 dB; sensitivity tests shift ≤ 2–3 dB.
icnavigator

ICNavigator helps engineers find verified ICs, alternatives, and fast quotes.

Explore
Categories
  • Power Management
  • MCU
  • Automotive ICs
  • Sensors
  • Interface & Transceivers
  • Analog Front-End
Get in Touch
  • Email: hello@icnavigator.com
  • WeChat/WhatsApp: +86-xxxx

© 2025 ICNavigator. All rights reserved.