123 Main Street, New York, NY 10001

Submit Your BOM

Linearity & Distortion: THD/SFDR in Active Filters

Linearity & Distortion: THD/SFDR in Active Filters

← Back to:Active Filters & Signal Conditioning

Linearity & distortion in active filters are not “fixed numbers”—they are condition-driven results. This page shows how swing, load/drive stability, and differential common-mode control create THD/SFDR spurs, and how to diagnose and fix them with measurable pass criteria.

What “Linearity & Distortion” Means in Active Filters & Conditioning

Core idea (the page’s answer)

Distortion is conditional, not a part number. THD/SFDR is usually set by the operating conditions—output swing, drive/load, and common-mode control—before “typical” specs matter.

What “linearity” means here

In this signal-conditioning chain, linearity means the filter/gain/convert function stays predictable and repeatable under a defined condition set: target amplitude, frequency, source impedance, load, supply rails, and output common-mode.

What “distortion” means here

Distortion is the appearance of harmonics, intermodulation, and spurs not present in the input—typically caused by active-device nonlinearity, component voltage coefficients, common-mode modulation, and output-stage drive limits.

The 3-knob model (what usually moves THD/SFDR)
Swing (headroom)
Near-rail operation or internal node “wall hits” create a THD knee even when small-signal specs look fine.
Load / Drive
RL/CL and sampled or switching loads change stability and output-stage nonlinearity, often turning “clean” into “spur-limited.”
CM control (differential stages)
VOCM/CMFB injection and dynamic common-mode shifts can create or amplify HD2/spurs even when differential gain is correct.
Page boundary (to avoid content overlap)

Focus stays on distortion created inside the analog conditioning chain. ADC quantization/clock-jitter theory and RF PA linearization are out of scope for this page.

Distortion is conditional: swing, load, and common-mode control Block diagram from input signal to conditioning chain to FFT spectrum, highlighting three main knobs that determine THD and SFDR. “Linearity & Distortion” = predictable chain + spectral evidence Input A f Rs Define conditions before trusting THD/SFDR. Signal Conditioning Chain Filter PGA FDA TIA Clamp … FFT F HD2 HD3 spur The knobs that move THD/SFDR: Swing Load CM Control
Use the same device under different swing, load, or VOCM/CMFB conditions and the spectrum can change dramatically. Treat distortion as a function of conditions, then design and verify those conditions.

Metrics That Actually Drive Decisions: THD, HD2/HD3, SFDR, IMD

Each metric answers a different question. Use the one that matches the system failure mode, then record results with the full condition set so measurements remain reproducible.

THD
Captures total harmonic energy. Use when “overall coloration” or total harmonic contamination matters (audio, precision low-distortion measurement paths).
Typical trap: quoting THD without amplitude, load, and bandwidth makes comparisons meaningless.
SFDR
Captures the worst single spur. Use when one spur can break detection, mask small signals, or violate spectral masks (comms baseband, multi-tone systems).
Typical trap: the “largest spur” can be a measurement artifact if probing or grounding changes stability.
HD2 / HD3 (diagnostic fingerprints)
Use harmonic balance as a fingerprint: dominant HD2 often points to symmetry/CM modulation issues; dominant HD3 often points to output-stage or loop-gain limits.
Quick rule: changing VOCM / symmetry tends to move HD2; changing load / swing tends to expose HD3.
IMD (two-tone)
Use two-tone IMD to reveal nonlinearity mechanisms and predict multi-tone pollution. Often more sensitive than THD for real systems.
Typical trap: two-tone spacing and analyzer setup can hide or exaggerate intermod products—always log the setup.
Condition set template (always record these)
Signal
fin, amplitude (Vpp/Vrms), tones (1-tone / 2-tone)
Load / Drive
RL, CL, sampled/switching load, any Riso
Rails / CM
supply rails, output VOCM, CM loop settings (if any)
Measurement
bandwidth/filtering, FFT length/window, averaging, grounding/probing
Results
THD, SFDR, HD2, HD3, IMD3 (all in dBc)
THD, HD2/HD3, SFDR, and IMD on a spectrum An FFT spectrum labeled with fundamental, HD2, HD3, largest spur for SFDR, plus metric dictionary cards and a condition bar including amplitude, frequency, load, and VOCM. Spectrum view (labels show what each metric “looks at”) f A F HD2 HD3 spur SFDR THD = Σ harmonics Metric dictionary THD sum harmonics SFDR worst spur HD2 / HD3 fingerprint IMD two-tone Always log conditions: A f Load VOCM
THD summarizes harmonic energy, SFDR is the worst spur, HD2/HD3 act as fingerprints, and IMD predicts multi-tone pollution. Always compare results only under matching conditions (amplitude, frequency, load, and VOCM).

A Distortion “Ownership Map”: Where Spurs Are Actually Created

A spur is a symptom. Ownership is proven by a small set of condition sweeps (amplitude, load, and VOCM) and simple isolation moves (bypass, swap load, or remove a clamp under safe limits).

Segment the chain before blaming any one IC
Input protection / clamp
Clamp C(V) and leakage paths can create AM-PM and low-order harmonics.
Filter network
Capacitor voltage coefficient (Cap VC) and resistor self-heating can seed harmonic energy.
Amplifier core
Open-loop nonlinearity leaks through as loop gain drops with frequency and signal swing.
Output stage
Headroom and current limits create a “knee” long before obvious clipping.
Load / interface
RL/CL and sampled loads can shift stability, turning ringing into discrete spurs.
Common-mode control (diff)
VOCM/CMFB injection and dynamic CM shifts often show up as HD2 or spur growth.
Minimal isolation moves (fast, low-risk, high-signal)
Amplitude sweep
Find the knee. A sharp bend points to headroom/output-stage limits.
Load sweep
Change RL/CL. Strong movement suggests drive/stability/interface ownership.
VOCM sweep (diff)
If HD2/spurs track VOCM/decoupling, suspect CM injection/CMFB dynamics.
Bypass / isolate
Temporarily bypass a filter cell or replace a clamp under safe limits to test ownership.
Spur ownership tree: mechanism, evidence, and fix direction A decision tree that maps an observed spur to likely mechanisms and the fastest evidence sweeps: amplitude, load, VOCM, and bypass. Leaves include headroom, clamp C(V), capacitor voltage coefficient, loop gain, stability, and CM injection. Spur → Mechanism → Evidence → Fix direction (ownership proof) Spur observed FFT Moves with Swing A-sweep knee Headroom margin SR / drive f-sweep Moves with Load RL/CL Riso Output stage I-limit Stability ring Moves with VOCM VOCM HD2 decoupling / symmetry CM injection VOCM Clamp C(V) bypass Evidence sweeps: A-sweep • RL/CL • VOCM • bypass → prove ownership before redesign
Treat ownership as a testable claim: if a spur tracks amplitude, load, or VOCM, the mechanism is usually predictable and the fix direction becomes obvious.

Output Swing & Headroom: Why THD Collapses Near the Rails

THD often degrades before visible clipping. The practical target is a linear swing window—a range of Vout where THD/SFDR stays within limits under the real supply, load, VOCM, and temperature.

Rail / headroom nonlinearity
Input/output stages become nonlinear near rails and internal nodes can “hit the wall” first, creating a THD knee.
Quick check: run an amplitude sweep and locate the knee; increasing rail margin or shifting VOCM should move it.
SR / dynamic drive limit (large-signal, high-f)
Slew-rate or charge/discharge current limits show up as strong frequency dependence: THD rises rapidly at higher f for a fixed swing.
Quick check: hold Vout constant and sweep frequency; improvement from lowering f indicates SR/loop-gain stress.
CM range / VOCM-induced wall hits (differential)
Limited common-mode range or CMFB dynamics can push internal nodes nonlinear even when differential swing looks reasonable.
Quick check: sweep VOCM or change VOCM decoupling; HD2/spurs that track VOCM point to CM ownership.
Engineering definition: linear swing window

Vout_linear is not a single number. Treat it as a function: Vout_linear = f(rails, load, VOCM, temperature). Verify the window with an amplitude sweep under real RL/CL and VOCM settings.

THD versus output swing: linear window, knee, and clipping zone A THD versus Vout peak-to-peak chart highlighting the linear swing window, a knee point where THD starts rising, and a clipping zone. Labels indicate that load and VOCM shift the knee. THD vs Vout swing: keep operation inside the linear window THD (dB) Vout_pp Linear window knee clipping THD rises before clipping Knee shifts Load heavier → knee left VOCM range → knee moves Rails more → knee right
The “linear swing window” is where THD/SFDR stays within target limits. The knee can shift significantly with load and VOCM/common-mode range, even if the circuit is unchanged.

Drive Conditions: How Load, Capacitance, and Source Impedance Break THD/SFDR

A “great” datasheet THD can collapse on a board because the output is not driving the same thing. Drive conditions change output current stress and stability margin, and both can create discrete spurs that dominate SFDR.

Resistive load (RL): current stress → output-stage nonlinearity
As RL decreases, output current increases. Output-stage nonlinearity rises and THD (often HD3) worsens even if the swing stays the same.
Fingerprint: THD tracks RL strongly; changing CL has a smaller effect.
First check: hold frequency and swing constant; sweep RL and log THD/HD3 in dBc.
Capacitive load (CL): phase margin ↓ → ringing → spurs (SFDR first)
CL (including cable/connector capacitance) reduces phase margin. Even “small” ringing can appear as discrete spurs in the FFT and destroy SFDR.
Fingerprint: SFDR collapses due to one or a few spurs; time-domain shows mild overshoot/ringing.
First check: keep RL fixed; add/remove CL (or swap cable) and observe spur movement and ringing.
Sampled / switching load: periodic charge draw → “walking” spurs
A sampled input (switch + Cs) looks like a dynamic load that periodically pulls charge. This can create spurs tied to switching activity even when the average load seems light.
Fingerprint: spurs shift or re-space when sampling rate / switching pattern changes.
First check: change fs (or gate/switching mode) and confirm whether the dominant spur “moves.”
Source impedance interaction: topology can amplify drive sensitivity
Source impedance and filter topology interact. Some structures are more sensitive to input current nonlinearity or impedance modulation, changing distortion unexpectedly.
First check: vary source impedance (or add a buffer) and see whether THD/SFDR changes without touching the output load.
Fix priority ladder (do this order)
1) Stabilize first
Add Riso, reduce effective CL, or insert a buffer/driver. Recover phase margin before optimizing THD.
2) Reduce output stress
Increase RL where possible; avoid heavy DC current and large transient charge demand at the output node.
3) Re-check spectrum
Confirm whether the limit is THD (harmonics) or SFDR (single spur). Optimize for the system failure mode.
Drive condition equivalent load model: RL, CL, sampled load, and Riso A driver block feeding an output node with parallel loads RL, CL, and a sampled load SW+Cs at fs. A series isolation resistor Riso is shown. A stability-to-spur arrow highlights that phase margin affects SFDR. Output sees an equivalent load: RL + CL + (SW+Cs at fs) → stability drives spurs Driver Op-Amp / FDA Vout Riso output node RL current stress CL phase margin SW + Cs fs Stability Spurs / SFDR PM SFDR
Model the output as an equivalent load (RL, CL, and sampled load). Use Riso and buffering to recover stability first—spurs often disappear when phase margin returns.

Frequency Dependence: Loop Gain, SR, and High-Q Node Swing Effects

Distortion often worsens with frequency because loop gain drops, dynamic limits appear (SR/drive), and high-Q sections amplify internal node swing beyond the output swing.

Loop gain falls with f → correction weakens
As frequency increases, loop gain decreases and the closed-loop path cancels less of the device’s inherent nonlinearity. Residual distortion rises.
Quick check: at fixed swing, sweep frequency and watch whether harmonics climb smoothly with f.
SR / dynamic drive limit → sudden high-f degradation
Large-signal high-frequency operation stresses slew rate and output charge/discharge capability. Distortion can worsen abruptly when dynamic limits are reached.
Quick check: keep frequency fixed; reduce swing and confirm whether harmonics/spurs drop sharply.
High-Q sections: internal nodes swing > Vout
In a high-Q biquad or state-variable section, internal nodes (A1/A2) can have larger swing than the output. Internal nodes may hit nonlinear regions first.
Quick check: reduce Q (or change response shape) and see whether the distortion knee moves substantially.
Practical diagnosis rules (use slopes)
Frequency-sensitive
Fixed swing, sweep f → large change suggests SR/loop-gain stress dominates.
Amplitude-sensitive
Fixed f, sweep swing → a clear knee suggests headroom/internal wall hits dominate.
Q-sensitive
Only Q changes, distortion changes strongly → internal node swing amplification is likely the driver.
High-Q internal node swing amplification and loop gain roll-off A high-Q second-order section block shows internal nodes A1 and A2 with larger swing than Vout. A loop gain versus frequency sketch shows decreasing correction at high frequency. Bottom buttons indicate amplitude A, frequency f, and Q sweeps. High-Q can amplify internal swing; loop gain drops with f → distortion rises High-Q 2nd-order section Vin Biquad / SVF A1 A2 swing ↑ swing ↑ Vout swing High Q Loop gain vs f f LG correction ↓ nonlinearity leaks Diagnose with: A f Q
High-Q sections can make internal nodes (A1/A2) swing larger than Vout, while loop gain decreases with frequency. Use A, f, and Q sweeps to identify the dominant mechanism.

Component Nonlinearity: R/C Voltage Coefficient, DA, and Switching/Clamp Injection

Distortion is not only an amplifier problem. In active filters and signal conditioning, the dominant source can be passive and protection parts whose behavior changes with voltage: capacitor VC/DA, resistor self-heating, and clamp/switch C(V) or charge injection. Treat distortion as a node-level budget, not a single IC spec.

Capacitor nonlinearity (Voltage Coefficient + DA)
Symptom
Harmonics rise with AC swing even when the amplifier is well within its linear window; unit-to-unit spread can be large with high-K MLCC.
Quick check
Swap only the highest-ΔV capacitor to NP0/C0G (same nominal value if possible) and repeat the amplitude sweep.
Fix direction
Use NP0/C0G on high-swing nodes; reduce capacitor ΔV via topology or gain distribution; avoid high-K MLCC where distortion dominates.
Resistor nonlinearity (self-heating / power modulation)
Symptom
Distortion correlates with dissipated power and waveform crest factor; changing package size or resistor technology changes THD noticeably.
Quick check
Keep frequency and gain constant; reduce signal level (power) and confirm whether harmonics drop more than expected from small-signal behavior.
Fix direction
Prefer thin-film where distortion matters; lower dissipation by adjusting impedance levels; use larger packages or split power across resistors.
Clamp/switch effects (junction C(V) + injection)
Symptom
HD2 and discrete spurs appear or worsen with protection parts, clamps, or nearby switching activity; small layout or part changes cause large SFDR shifts.
Quick check
Under safe limits, compare with/without the clamp path (or alternate low-C protection) and observe whether HD2/spurs drop sharply.
Fix direction
Choose low-capacitance protection; limit clamp voltage swing; keep switching nodes away from sensitive high-impedance points; control return paths.
Minimal proof set (fast ownership)
1) Amplitude sweep
Fixed f; sweep level. Strong nonlinearity with level often indicates VC/self-heating/clamp C(V).
2) Single-part swap
Swap the highest-ΔV capacitor or highest-power resistor first; confirm whether THD/SFDR shifts immediately.
3) Reduce node ΔV
Re-distribute gain or topology so the critical part sees smaller swing; node-level improvement confirms passive ownership.
Component nonlinearity cards: Cap VC/DA, resistor self-heating, clamp C(V)/injection Three large cards illustrate common non-IC distortion sources: capacitor voltage coefficient and dielectric absorption, resistor self-heating, and clamp/switch junction capacitance and injection. Each card shows a small spectrum icon and short symptom/fix tags. Passive & protection parts can dominate distortion (node-level ownership) Cap VC / DA ΔV HD2/3 C0G reduce ΔV batch spread Res self-heat R THD thin-film lower P power-related Clamp C(V) C(V) spur low-C limit ΔV
Prioritize the highest-swing nodes: capacitor VC/DA, resistor self-heating, and clamp/switch C(V)/injection can set the real THD/SFDR limit.

Differential Stages: How VOCM/CMFB Injection Can Create or Amplify Distortion

In differential signal chains, distortion is not purely a differential-path property. VOCM noise/ripple and CMFB dynamics can move the output common-mode, and any asymmetry converts CM motion into DM error—often seen as HD2 rise or dominant spurs.

VOCM injection → output CM ripple → CM-to-DM conversion
VOCM is a signal path. Source impedance, filtering, and local decoupling determine how much ripple/noise reaches the output common-mode. With mismatch, that CM ripple becomes DM distortion and spurs.
CMFB bandwidth/phase → dynamic CM shift under large-signal / load
CMFB is a loop with finite bandwidth and phase margin. Under load or large signal, CM can move dynamically. Any imbalance translates that motion into differential error components.
HD2 fingerprint: VOCM/symmetry changes move HD2
If HD2 changes noticeably with VOCM level, VOCM filtering/decoupling, or symmetry (matching/layout), CM injection is likely dominating.
Practical countermeasures (high leverage)
VOCM source + impedance
Keep VOCM low-impedance; avoid long high-Z routes; do not “share” VOCM with noisy references.
Local decoupling
Decouple VOCM at the pin with a small loop area; place return near the CM reference path.
Matching & symmetry
Match resistors and parasitics; route differential paths symmetrically; keep both loads balanced.
CMFB loop area control
Keep CM sensing and VOCM return compact; avoid coupling from switching nodes into the CM path.
Minimal proof set (VOCM ownership)
VOCM sweep
Sweep VOCM within the valid range; HD2/spurs that track VOCM implicate CM injection or CMFB dynamics.
Decoupling change
Change VOCM decoupling placement/value and observe whether HD2/spur levels shift with the CM path.
Symmetry toggle
Reduce mismatch (or temporarily add known imbalance) and see whether HD2 changes strongly—confirm CM-to-DM conversion.
VOCM injection path and HD2 fingerprint in a differential stage An FDA block with VOCM input and a CMFB loop shows how VOCM ripple creates output common-mode motion. A mismatch block converts CM to DM spurs. A small FFT plot highlights HD2 rise and its sensitivity to VOCM and symmetry. VOCM ripple → CM motion → mismatch converts CM to DM → HD2 / spur rise FDA INP INN diff core OUTP OUTN VOCM ripple CM CMFB mismatch CM → DM FFT HD2 Probe with: VOCM Decouple Symmetry
If HD2/spurs move with VOCM and symmetry changes, suspect CM injection and CM-to-DM conversion before changing the main amplifier.

Topology Sensitivity: Which Structures Expose Distortion First (Within This Page’s Scope)

This section compares topologies only through distortion-sensitive points (not a topology handbook). Use the table to find the most likely sensitivity dimension—Headroom, Drive, Component, or CM control—then validate with a targeted sweep.

Sallen-Key
Phenomenon: distortion often tracks output drive and large-signal stress directly.
Action: verify load and swing first, then evaluate parts.
Link: Sallen-Key LP/HP
MFB
Phenomenon: a single high-ΔV R/C can dominate THD due to VC/self-heating.
Action: identify the highest-ΔV element and swap to linear parts first.
Link: MFB LP/HP
SVF / Cascaded Biquads
Phenomenon: internal node swing can exceed Vout (especially at high Q), exposing headroom and passive VC early.
Action: reduce internal ΔV (gain distribution / Q / implementation) before changing the amplifier.
Link: State-Variable / Biquads
Fully-Differential
Phenomenon: VOCM/CMFB injection and asymmetry convert CM motion into DM distortion—often visible as HD2/spurs.
Action: control VOCM impedance/decoupling and symmetry before swapping the FDA.
Link: Fully-Differential Filters
Topology versus distortion sensitivity dimensions A table-style diagram compares common topologies across four sensitivity dimensions: headroom, drive, component nonlinearity, and common-mode control. Dots indicate low, medium, or high sensitivity. Topology sensitivity map (distortion view): find the first place to suspect Topology Headroom Drive RL CL Component CM control VOCM Sallen-Key MFB SVF / Biquads Fully-diff Legend: Low Mid High Use sweeps to confirm ownership
The same THD/SFDR target can fail for different reasons depending on topology: headroom, drive, passive nonlinearity, or CM control.

Measurement & Debug: How to Measure THD/SFDR Without Being Misled

If the measurement chain floor is unknown, DUT conclusions are not reliable. Start with conditions and floors, then insert blocks and sweep one variable at a time to identify ownership.

Condition template (write these down for every plot)
Signal
Frequency, amplitude, tones (single / two-tone), crest factor.
Load & bandwidth
RL/CL (or dynamic load), filter settings, analysis bandwidth.
Operating point
Supply, VOCM (if differential), warm-up / thermal state.
Debug flow (repeatable)
Step 1 — Define
Freeze frequency, level, RL/CL, VOCM, and bandwidth. Make plots comparable.
Step 2 — Floor
Measure loopback (source → analyzer) and confirm the chain floor under the intended load.
Step 3 — Short chain
Use the shortest DUT configuration; establish “best-case” performance before adding blocks.
Step 4 — Insert & sweep
Insert one block at a time (protection → filter → amplifier → load/CM). Sweep level and RL/CL; for differential, sweep VOCM and symmetry.
Common pitfalls (fast checks)
Probe capacitance
Probe C can change stability and create spurs. Validate with lower-C probing or isolation.
Ground / return loops
Long return paths inject hum and switching spurs. Use compact loops and consistent reference points.
FFT coherence / leakage
Non-coherent captures can show fake spurs. Keep capture settings consistent and verify with a second configuration.
Instrument or source overload
Overloading the source/analyzer produces “DUT distortion.” Confirm headroom on every device in the chain.
Bandwidth / filtering differences
Different bandwidth modes change measured SFDR/THD. Record bandwidth and filtering with each plot.
Thermal settling
Warm-up drift can look like instability or spurs. Define a stabilization criterion before final captures.
THD/SFDR measurement setup and debug flow A block diagram shows source, DUT, load, and analyzer. A four-step flow defines conditions, measures floor, establishes a short-chain baseline, then inserts blocks and sweeps variables. Icons indicate probe capacitance, ground loops, and overload risk. Measurement chain + repeatable debug flow (avoid “false DUT distortion”) Source level DUT config Load RL CL Analyzer FFT / THD / SFDR Probe C Return loop Overload ! Flow Define f / level / RL / CL Floor loopback Short baseline Insert sweep Sweep knobs: Level RL / CL VOCM Symmetry
Measure the floor first, then insert blocks and sweep one knob at a time (level, RL/CL, VOCM, symmetry) to avoid false conclusions.

Engineering Checklist: Layout Priorities + Bring-Up Test Hooks (Linearity & Distortion)

Goal: make distortion ownership measurable on real hardware. Priorities align to the three main failure knobs: Swing, Drive, and CM control. Each item includes what to check, how to verify, and a pass criterion.

A) Symmetry & return paths (diff routing, VOCM/CMFB loop, matching, thermal coupling)
What to check
  • Diff pair symmetry: same length, same reference plane, same via pattern.
  • CM loop containment: VOCM/CMFB components placed to minimize loop area and keep returns local.
  • Critical resistor matching + thermal coupling: feedback and input networks kept close and equally heated.
Quick verification
  • VOCM sweep: step VOCM within the allowed window; observe HD2 and dominant spur movement.
  • Symmetry toggle: swap/rotate matched parts or use a temporary “mirror” patch; check HD2 sensitivity.
  • Heat poke: gentle airflow/heat on one side; check if even-order distortion changes disproportionately.
Pass criteria (engineering)
  • HD2 and the dominant spur change by ≤ 3 dB for small VOCM steps inside the operating window.
  • No “one-side-only” thermal sensitivity: HD2 does not jump with minor asymmetric heating.
Example BOM part numbers (matching & symmetry)
  • Precision thin-film resistors (low VCR): Vishay TNPW0805 / TNPW0603 series.
  • Matched resistor networks (ratio + thermal tracking): Vishay ACAS 0606 / ACAS 0612 series.
  • C0G/NP0 caps for signal path: Murata GRM1885C1H101JA01 (100 pF), GRM1885C1H102JA01 (1 nF), GRM1885C1H103JA01 (10 nF).
B) Decoupling & reference integrity (VOCM + supplies + return routing)
What to check
  • VOCM decoupler placed at the pin; return path stays local (no long shared returns).
  • Supply decoupling follows the current loop: pin → capacitor → plane → pin.
  • Reference/CM nodes never share high-di/dt return with output drive or digital clocks.
Quick verification
  • Measure VOCM pin ripple under large-signal operation; correlate ripple with HD2/spur changes.
  • Add/remove a small parallel C at VOCM (same footprint) and observe HD2/spur sensitivity.
  • Inject a tiny supply ripple (within safe limits) and check if spurs track the injection frequency.
Pass criteria (engineering)
  • VOCM ripple stays below the project budget (start point: < 5 mVpp at the pin for audio/baseband).
  • HD2 does not improve dramatically by “random” extra decoupling (indicates VOCM/return was the root cause).
Example BOM part numbers (decoupling & ESD with low parasitics)
  • Small C0G “pin-tight” caps: Murata GRM1885C1H103JA01 (10 nF), GRM1885C1H104JA01 (100 nF where available).
  • Bulk local decoupling (X7R): Murata GRM21BR60J106ME19 (10 µF, 6.3 V) / GRM31CR60J226ME19 (22 µF, 6.3 V).
  • Low-cap ESD for sensitive inputs (examples): TI TPD1E10B06, Nexperia PESD5V0S1UL.
C) Output drive readiness (Riso pads, swappable loads, measurement points)
What to check
  • Series isolation resistor footprint (Riso) placed at the driver output before any large CL/connector.
  • Provision for at least two RL options and optional CL (jumpers or solder-select).
  • Dedicated measurement interface: coax/diff probe pads located to avoid “probe-adds-CL” surprises.
Quick verification
  • Sweep RL/CL and Riso; watch if a “single dominant spur” appears (often stability/drive-related).
  • Compare plots with/without the intended connector/cable; cable capacitance often shifts spurs.
Pass criteria (engineering)
  • No narrow spurs that “turn on” only for certain CL/connector/probe combinations.
  • Stable spectrum across reasonable Riso values (start point: 5–49.9 Ω options).
Example BOM part numbers (Riso + connectors + test hardware)
  • Series resistors for isolation (thin-film): Vishay TNPW0603 5.1 Ω / 10 Ω / 22 Ω / 49.9 Ω.
  • SMA edge-launch connector (example): Johnson 142-0701-801 (SMA edge launch).
  • Scope/diff probe test pads (example): Keystone 5015 (test point) or dedicated coax pad footprint.
D) Bring-up tests (do / look / pass)
Level sweep (swing)
Do: sweep amplitude across at least 3 points up to the target swing.
Look: THD knee, HD2/HD3 slope vs level, dominant spur onset.
Pass: knee occurs beyond the target swing; spurs scale predictably without sudden jumps.
Load sweep (drive)
Do: test two RL points and optional CL / cable connection.
Look: “single spur dominates SFDR” under specific load/CL (often stability or drive current).
Pass: spectrum shape does not collapse when moving between intended loads.
VOCM sweep (CM control)
Do: step VOCM across the allowed range while holding level and load constant.
Look: HD2 sensitivity, spurs that track VOCM ripple or CMFB dynamics.
Pass: HD2 and dominant spur change remains small (start point: ≤ 3 dB) for small VOCM steps.
Temperature sweep (repeatability)
Do: repeat the same plot at two stabilized temperatures.
Look: spur stability vs temperature (passive VC/self-heating/CM drift often shows up here).
Pass: performance stays within the system budget; no new spurs appear only at one temperature point.
Layout review hotspot map for distortion-critical paths A board-level schematic-like hotspot map highlights VOCM decoupling, CMFB loop containment, symmetry and matching zone, series isolation resistor pads, and measurement points. Blue boxes indicate priority review areas. Layout review hotspots (distortion view): CM control, symmetry, drive readiness, measurement points Board-level concept (not PCB detail) Filter / Gain R/C network FDA / Driver CMFB + VOCM Output / Load RL / CL / Cable VOCM decap CMFB loop Symmetry + match Riso pads Probe points Legend CM Match Measure
A board-level hotspot map to keep VOCM/CMFB clean, preserve symmetry, and make drive-related spurs measurable (Riso + probe points).

IC Selection Logic: What to Ask Vendors + How to Filter Candidates (THD/SFDR Under Real Conditions)

Selection must be bound to operating conditions. Ask for plots under matching swing, load, frequency, VOCM, and supply. The decision order is fixed: Headroom → Drive/Stability → CM control → THD/SFDR.

Condition lock (fill before comparing parts)
Signal
f, tone plan (single / two-tone), Vout (Vrms or Vpp), crest factor.
Load
RL, CL (including cable/connector), and the measurement interface.
Operating point
Supply, VOCM (if differential), temperature, warm-up state, bandwidth settings.
Vendor questions (ask for plots, not slogans)
THD/SFDR vs swing
Provide at least 3 amplitude points at the target frequency and intended load.
THD/SFDR vs load
Provide at least 2 load points and one case with representative CL/cable.
Differential CM control
VOCM injection sensitivity (VOCM ripple → HD2/spurs), CM behavior vs frequency, and allowed CM swing window.
Drive & stability guidance
Recommended Riso range for CL/cable stability and output current margin at the target swing/f.
Selection decision tree and request-field checklist A left-side yes/no decision tree prioritizes headroom, drive/stability, common-mode control, then THD/SFDR. A right-side card lists copy-ready vendor request fields for swing sweep, load sweep, and VOCM sensitivity. Selection order: Headroom → Drive/Stability → CM control → THD/SFDR Decision tree Headroom OK? Yes No: reduce swing / raise rails Drive & stability OK? Yes No: add Riso / change load model CM control OK? Yes No: fix VOCM impedance / symmetry THD/SFDR OK (under locked conditions)? Request fields (copy-ready) Conditions f, Vout, RL/CL, supply, VOCM, BW Swing sweep THD/SFDR @ 3 levels same f, same RL/CL Load sweep THD/SFDR @ 2 loads include representative CL/cable CM sensitivity VOCM ripple → HD2/spurs + CM window
Ask for plots under locked conditions. Do not accept “typical THD” without swing/load/VOCM context.
Reference example part numbers (starting points only; verify under your locked conditions)
Fully-differential amplifiers (FDA / diff driver)
  • TI OPA1632 (low-distortion FDA class, diff chains).
  • TI THS4551 (diff driver; check VOCM/CM behavior vs load).
  • TI THS4561 (diff driver; evaluate Riso guidance for CL).
  • ADI ADA4940-1 (diff ADC driver family; verify CM injection fingerprints).
  • ADI ADA4945-1 (high-speed diff driver class; confirm headroom window).
Single-ended op-amps (active filters / gain blocks)
  • TI OPA1612 (low distortion audio/AF; validate drive vs RL/CL).
  • TI OPA1656 (low distortion; verify swing window on single supply).
  • ADI ADA4625-1 (precision low distortion class; confirm frequency-dependence under load).
  • TI OPA192 / OPA197 (precision general use; validate THD at target frequency).
Analog switches (cal hooks / reconfig) and why they matter
  • ADI ADG1419 (SPDT switch class; characterize C(V) / injection at signal amplitude).
  • ADI ADG1219 (low leakage variant class; still validate distortion at target swing).
  • TI TS5A23157 (SPDT class; verify charge injection and capacitance in your band).
Passives (often the hidden THD limiters)
  • C0G/NP0 caps: Murata GRM1885C1H101JA01, GRM1885C1H102JA01, GRM1885C1H103JA01.
  • Thin-film resistors: Vishay TNPW0603 / TNPW0805 series.
  • Matched networks: Vishay ACAS 0606 / ACAS 0612 series.
Use rule
These part numbers are provided to speed up datasheet lookup and bench verification. Final selection must be driven by the locked condition template above and confirmed by swing/load/VOCM sweeps on the target PCB.

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

FAQs: Linearity & Distortion (THD/SFDR, swing/drive/CM control)

Answers are short and actionable. Each item follows the same 4-line structure: Likely cause / Quick check / Fix / Pass criteria.

THD looks great at 1 Vrms but collapses at 2 Vrms—what is the first headroom check?
Likely cause: Output or internal nodes are hitting a headroom/linear-swing knee (often before visible clipping), especially under real RL/CL and VOCM.
Quick check: Hold frequency+load constant and run a 3-point level sweep (−6 dB, nominal, +6 dB). If THD knees near the higher level, check VOUT_peak + VOCM vs rails (start point: ≥200 mV margin to each rail).
Fix: Increase supply or shift VOCM to re-center swing; reduce gain in the stage that saturates first; choose higher-swing / higher-output-current parts (examples: TI OPA1656, TI OPA1612; differential: TI OPA1632).
Pass criteria: THD degrades by ≤3 dB between 1 Vrms and 2 Vrms at locked conditions, and the THD knee occurs above the target maximum swing.
SFDR is limited by one spur that moves with VOCM—how to confirm CM injection?
Likely cause: VOCM ripple/impedance or CMFB dynamics are converting common-mode modulation into differential spurs (often visible as HD2 or a moving dominant spur).
Quick check: Step VOCM by ±50–100 mV (within allowed range) at fixed level+load. If spur level changes ≥3 dB or tracks injected VOCM ripple frequency, CM injection is confirmed. Measure VOCM ripple (start target: <5 mVpp at the pin).
Fix: Lower VOCM source impedance (buffer: TI OPA192); add pin-tight VOCM decoupling C0G/NP0 (Murata GRM1885C1H104JA01 as a common 0.1 µF example); enforce symmetry/matching (Vishay ACAS 0606); consider FDAs with robust CM behavior (TI THS4551, TI OPA1632, ADI ADA4940-1).
Pass criteria: Dominant spur changes ≤3 dB for ±50 mV VOCM steps, and VOCM ripple stays below budget (start point <5 mVpp) under large-signal operation.
Why does adding a small Riso improve both stability and SFDR?
Likely cause: Effective capacitive load (CL/cable/probe/switch-cap input) reduces phase margin; light ringing becomes discrete spurs that limit SFDR.
Quick check: Compare 0 Ω vs 22 Ω vs 49.9 Ω in series at the driver output; observe step/settling and FFT. If a single dominant spur collapses as Riso increases, it was stability/drive-interaction driven.
Fix: Add output isolation at the driver pin (start range 5–49.9 Ω). Example parts: Vishay TNPW0603 10 Ω / 22 Ω / 49.9 Ω (thin-film). If needed, add an RC snubber at the far load (tune on PCB).
Pass criteria: SFDR improves by ≥6–10 dB (or meets target), and ringing is controlled (start point: <10% overshoot) across intended RL/CL/cable cases.
Load is “high-Z” but THD is still bad—could probe capacitance be the cause?
Likely cause: Probe/cable capacitance (and long ground lead inductance) creates hidden CL, destabilizing the driver or increasing current spikes.
Quick check: Re-test with (a) no probe, (b) 10× probe + ground spring, (c) coax/SMA into a proper analyzer input. If THD/SFDR shifts ≥6 dB across setups, probing is part of the problem.
Fix: Add Riso (start 10–49.9 Ω; Vishay TNPW0603) and provide controlled measurement access (SMA edge-launch example: Johnson 142-0701-801). Use low-C probing for final numbers.
Pass criteria: THD/SFDR stays within ±3 dB across intended measurement methods, and no “probe-only” spur appears.
HD2 dominates—what asymmetry (layout/VOCM/CMFB) should I check first?
Likely cause: Even-order distortion is driven/unmasked by asymmetry: mismatch, unequal parasitics, CM→DM conversion, or VOCM/CMFB injection.
Quick check: VOCM step (±50 mV) + symmetry swap (swap diff legs in measurement or swap matched R/C pairs). If HD2 shifts ≥6–10 dB, asymmetry/CM conversion is the lever.
Fix: Use matched networks (Vishay ACAS 0606/0612) or tight thin-film resistors (Vishay TNPW); keep routing symmetric; decouple VOCM at the pin (Murata GRM1885C1H104JA01); consider FDAs with clean CM behavior (TI OPA1632, TI THS4551).
Pass criteria: HD2 drops ≥10 dB (or meets spec) after symmetry/CM fixes, and HD2 sensitivity to small VOCM steps is ≤3 dB.
HD3 dominates only at high frequency—how to tell SR limit vs loop-gain limit quickly?
Likely cause: Either slew-rate limiting at large V·f, or loop-gain roll-off leaving more nonlinear residue at high frequency.
Quick check: At the same frequency, drop amplitude by 6 dB. If HD3 improves >12 dB, SR is likely dominant; if improvement is modest (≈6 dB or less) and small-signal THD is already worsening with frequency, loop-gain is likely dominant. Also halve frequency at same amplitude; SR-driven distortion should improve strongly.
Fix: SR-limited: reduce swing, lower Q/internal swing, or use higher SR/GBW (TI OPA1612, TI OPA1656; differential: TI THS4551). Loop-gain-limited: increase GBW at required closed-loop gain and reduce capacitive loading (use Riso).
Pass criteria: HD3 meets the budget at target frequency+swing and shows no sharp “turn-on” knee within the operating band.
Swapping X7R to C0G improves THD—how to choose capacitor type/voltage rating systematically?
Likely cause: X7R voltage coefficient + dielectric absorption create harmonics under AC swing; C0G/NP0 is far more linear.
Quick check: Replace only signal-path caps with C0G of the same value and re-measure. If THD improves ≥6–10 dB, cap nonlinearity was limiting. Secondary check: keep X7R but increase voltage rating/size; distortion often improves as electric field reduces.
Fix: Use C0G/NP0 for distortion-critical filter caps (Murata GRM1885C1H101JA01 / GRM1885C1H102JA01 / GRM1885C1H103JA01). If X7R must be used, derate aggressively (start point: rated voltage ≥10× AC peak) and verify across temperature.
Pass criteria: THD meets target across temperature, and unit-to-unit THD variation stays tight (start point ≤2 dB at locked conditions).
Why does distortion worsen after a notch/biquad stage even when gain is unity?
Likely cause: Internal node swing in high-Q sections can be much larger than the stage output, pushing the amplifier or passives into nonlinearity at unity overall gain.
Quick check: Reduce Q/depth by 20–30% and re-measure. If THD/SFDR improves disproportionately, internal swing amplification is the culprit. Also scale impedance (reduce R, increase C proportionally) to reduce capacitor voltage stress.
Fix: Ensure headroom for internal nodes; use C0G caps (Murata GRM1885C1H*** family) + thin-film resistors (Vishay TNPW) in the high-Q section; distribute gain so peaking is not concentrated inside the notch section.
Pass criteria: Adding the notch/biquad worsens THD by ≤3 dB at locked conditions, and no new dominant spur appears near corner/notch regions.
FDA output CM shifts with temperature—how does that translate into spurs?
Likely cause: CMFB operating point drifts with temperature; CM shift/ripple converts to differential error via mismatch/parasitics, raising HD2 or discrete spurs.
Quick check: Log output common-mode vs temperature (two stabilized points) at fixed tone level; correlate ΔCM with ΔHD2. If HD2 tracks CM shift, CM→DM conversion is active.
Fix: Stabilize VOCM reference/decoupling (Murata GRM1885C1H104JA01 at VOCM; buffer: TI OPA192). Improve symmetry/matching (Vishay ACAS). If needed, choose FDA with better CM stability (TI OPA1632 / THS4551; ADI ADA4940-1).
Pass criteria: Output CM drift stays within budget (start point <10 mV over the temperature span), and HD2/spur change is ≤3 dB between temperature points at locked conditions.
Analyzer shows worse SFDR than expected—what is the fastest source-DUT-analyzer isolation method?
Likely cause: The measurement chain sets the spur floor (source distortion, ground coupling, termination errors, or analyzer overload).
Quick check: Capture (1) source→analyzer loopback, (2) loopback + attenuator, (3) DUT inserted—same level/BW/window. If the dominant spur exists in (1) or disappears in (2), the source/analyzer is limiting. Use correct terminations (50 Ω where appropriate).
Fix: Add attenuation/isolation and correct terminations (fixed attenuator examples: Mini-Circuits VAT-3+ / VAT-6+), break ground loops, keep analyzer input out of overload. For differential DUTs, measure differentially (examples: TI THS4551, ADI ADA4940-1 as diff interface references).
Pass criteria: Loopback SFDR floor is ≥10 dB better than the target DUT SFDR, and DUT SFDR differs by ≤3 dB across repeatable setups.
Distortion varies lot-to-lot—what component nonlinearity or thermal coupling is most suspicious?
Likely cause: Passive nonlinearity (X7R, thick-film resistors), clamp/switch C(V), or uneven self-heating/thermal coupling shifting ratios under signal.
Quick check: Cross-swap the most distortion-sensitive passives between “good” and “bad” units (high-swing caps, feedback resistors, clamps). If THD follows a swap by ≥6 dB, that component class is dominant.
Fix: Replace thick-film with thin-film (Vishay TNPW), use matched networks (Vishay ACAS), swap X7R to C0G (Murata GRM1885C1H***), and use low-cap protection where needed (TI TPD1E10B06).
Pass criteria: Unit-to-unit THD variation is ≤2 dB at locked conditions, and the dominant spur pattern is consistent across lots.
“Unity-gain stable” op-amp still shows spurs when driving filter+load—what stability check is mandatory?
Likely cause: Unity-gain stable does not guarantee stability with capacitive/complex loads; marginal phase margin creates ringing that shows up as spurs.
Quick check: Measure time-domain step response with the real filter+load attached. If overshoot/ringing exists, sweep Riso (0→22→49.9 Ω) and CL; spurs that change sharply with Riso/CL indicate stability interaction.
Fix: Add series isolation (Vishay TNPW0603 10–49.9 Ω), minimize CL at the driver pin, and select parts suited for the intended load class (examples: TI OPA1656 for SE AF; TI THS4551 / TI OPA1632 for differential). Keep probe capacitance out of the stability loop (use SMA/controlled probing).
Pass criteria: Overshoot <10% with clean settling, no sustained ringing, and SFDR/THD stays within ±3 dB across intended RL/CL and probing conditions.
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.