Spectrum Analyzer IF Chain (Mixers, RBW, Detectors, ADC)
← Back to: Test & Measurement / Instrumentation
A spectrum analyzer’s “trustable numbers” come from its IF chain: mixers/LO planning, RBW filtering, gain distribution, detector/VBW choices, and ADC capture work together to control spurs, noise floor, and amplitude accuracy. This page explains how each IF block shapes what you see on the screen and how to verify it with practical checks.
H2-1 · IF Chain at a glance: from RF input to “trustable numbers”
A spectrum analyzer display is a measurement pipeline. The IF chain converts RF into a controlled bandwidth, controlled level, and controlled statistics so that the screen shows repeatable, calibratable numbers instead of “pretty traces”.
What “trustable” means in practice
- Repeatable: same input → same reading (small drift across temperature and time after warm-up).
- Setting-consistent: RBW/VBW/detector changes behave as physics predicts (noise and peaks move in explainable ways).
- Calibratable: amplitude error sources have hooks (gain steps, detector/log linearization, RBW shape, frequency response).
Final specs ↔ IF chain responsibility map
1) DANL / noise floor
Set by early-stage noise + RBW equivalent noise bandwidth + detector/averaging statistics. RBW defines how much noise power
is admitted; gain staging decides whether later blocks add noticeable noise.
2) TOI / dynamic range
Governed by mixer linearity and “who compresses first” in the IF gain chain (step attenuators, PGAs, IF amps).
A good design keeps the detector/ADC inside a safe level window while preventing early compression from generating IMD spurs.
3) RBW / VBW behavior
RBW comes from IF selectivity (filter bank or digital equivalent). VBW comes after detection (video filtering / smoothing).
RBW controls “separability” and noise admission; VBW controls trace stability versus response time.
4) Amplitude linearity and accuracy
Determined by gain-step accuracy, frequency response flatness, and detector/log linearization.
The “dBm on screen” must be referenced to known internal levels and corrected for RBW shape and detector transfer nonlinearity.
5) Spur-free display (no fake lines)
A joint outcome of LO leakage / mixer products, insufficient IF rejection, detector overload, and ADC alias/leakage.
Each stage must avoid creating stable “ghost tones” that survive smoothing and averaging.
Common user knobs ↔ where they land in the IF chain
- RBW: selectivity stage (IF filter bank or digital equivalent) → affects noise admission and separability.
- Detector mode: detector/log stage → changes how peaks/noise are summarized (peak, sample, average, RMS).
- VBW / smoothing: post-detection video filter → trades stability for response time and peak fidelity.
- Ref level / attenuation: gain staging window → shifts where compression happens and what spurs dominate.
Scope boundary (to avoid topic overlap)
This page focuses only on the IF chain: mixing/LO injection point (as it affects IF), RBW/VBW selectivity, IF gain staging, detector/log behavior, and ADC/FFT capture as it impacts amplitude and spurs. RF front-end protection, trigger routing, system I/O, and metrology traceability are referenced only at a high level.
H2-2 · Architecture choices that change the IF chain (superhet vs direct-sampling vs hybrid)
Architecture selection is mainly about where selectivity (RBW), level management, and spur control are implemented. The IF chain still exists in all cases, but the “work split” between analog stages and digital capture changes the dominant error mechanisms.
Compare by three IF-centric questions
- Where is RBW realized? (analog filter bank vs digital equivalent filtering/FFT)
- Where is the ADC placed, and what burden does it carry? (mostly display capture vs core selectivity/DR)
- What artifacts dominate? (mixer/LO spurs vs alias/leakage vs mixed causes)
IF chain behavior by architecture
A) Multi-conversion Superhet
- RBW location: analog IF filter bank is the primary selectivity; filter shape factor and switching consistency matter.
- ADC role: often captures detected/video or narrow IF; it usually does not carry the main selectivity burden.
- Dominant artifacts: LO leakage, mixer products, and compression-driven IMD if gain staging is wrong.
- IF design focus: clean frequency plan + stable gain steps + calibrated detector/log linearity.
B) Direct RF sampling (digital IF)
- RBW location: mostly digital (DDC/decimation/FFT) but must respect analog anti-aliasing and sampling-rate constraints.
- ADC role: carries more of the dynamic range and selectivity responsibility; full-scale management becomes critical.
- Dominant artifacts: alias lines from insufficient anti-alias filtering, clipping “skirts”, and spectral leakage that mimics spurs.
- IF design focus: clean analog anti-alias path + stable capture window + amplitude correction consistent with RBW math.
C) Hybrid (downconvert then sample)
- RBW location: shared: analog IF band-pass provides preselection; digital stages refine RBW and statistics.
- ADC role: less stressed than direct RF sampling, but still sensitive to anti-aliasing and level windowing.
- Dominant artifacts: both mixer/LO spurs and sampling/processing leakage; root-cause isolation must be layered.
- IF design focus: balance analog rejection with digital flexibility; ensure calibration ties both halves together.
“Real-time spectrum” constraints (IF/ADC only)
- No transient loss: capture must be effectively continuous for the analysis bandwidth, or short events will never appear.
- Throughput + memory: analysis bandwidth and record depth determine minimum observable event duration.
- Trace credibility: peak/hold and persistence views require detector + VBW/averaging rules that do not fabricate “stable” tones.
H2-3 · Mixer & LO planning for low-spur operation (what actually sets spur map)
A stable spur map is rarely “random.” Most persistent ghost lines can be predicted from frequency planning and coupling paths. The IF chain becomes spur-clean when (1) the RF/LO/IF plan avoids known collision zones, (2) the mixer is not driven into compression-induced intermodulation, and (3) LO leakage and reference-related spurs are prevented from reaching the detector/ADC path.
A minimal math backbone (enough to back-solve ghost lines)
- Downconversion identity: fIF = | fRF − fLO |.
- Image response: a different RF frequency can satisfy the same IF relation (depends on high-side vs low-side LO injection).
- Harmonic mixing paths: strong interferers or LO harmonics can create tones near IF (often appears only in specific bands/plans).
- Reference / fractional spurs: commonly show a comb pattern with near-constant spacing around certain LO-related regions.
- IMD products: with two strong tones, IM3 often lands near 2f1−f2 and 2f2−f1 and grows rapidly with level.
Mixer selection: choose by failure mode (not by buzzwords)
- Linearity (IMD control): higher IP3 and healthy compression margin reduce “input-dependent” spur growth near strong carriers.
- Conversion loss / gain: sets how much IF gain is needed later; too much downstream gain can amplify leakage and switching artifacts.
- Port isolation: poor LO↔RF/IF isolation creates “always-there” tones (LO feedthrough and self-mixing signatures).
- LO drive sensitivity: LO amplitude variation can move operating point and change spur levels; stable LO buffering matters.
- Match & reflections: poor match can create standing-wave interactions and band-specific spur hot spots.
LO chain: keep “frequency plan” and “coupling plan” consistent
- Buffer & isolation: LO buffers/isolators reduce load pulling and keep LO leakage from reaching IF nodes.
- Injection hygiene: avoid LO routing near detector/video/ADC traces; treat IF/LO as separate zones.
- Leakage paths to watch: direct coupling, power-rail coupling, ground return coupling, and “shield gaps” that behave like antennas.
- Observation rule: leakage-driven spurs tend to be frequency-stable and persist across RBW/VBW changes.
Quick triage: “ghost line” fingerprinting
- Fixed tone, barely reacts to input level: likely LO feedthrough / reference spur → inspect LO isolation and coupling paths.
- Comb with constant spacing: likely reference/fractional-related spur → check spur spacing consistency across bands/plans.
- Spur rises sharply with strong input: likely compression/IMD → add attenuation or reduce early IF gain; re-check IM3 slope.
- Only appears in certain LO plans/bands: likely frequency-plan collision (image/harmonic) → back-solve with IF relation.
- Moves with input tones in two-tone test: likely IMD → confirm by moving f1/f2 and observing the predicted shift.
H2-4 · IF gain distribution: step attenuators, PGAs, and AGC that protects linearity
Dynamic range is an operating-point strategy. The goal is to keep every stage inside a safe headroom band while ensuring the detector/ADC sees a consistent level window. When the window is respected, spurs do not “explode” with input level, and weak-signal noise floor is not dominated by the wrong stage.
Define the internal level window (what “good” looks like)
- Lower bound: above the stage noise/quantization so noise floor tracks RBW math, not gain artifacts.
- Upper bound: below compression (P1dB) so IMD spurs do not rise nonlinearly.
- Stability requirement: gain steps and temperature drift should not push operating point across the window edges.
Role split: who handles range, who handles finesse, who closes the loop
- Step attenuator: coarse range + early-stage protection. Prevents mixer/early IF from becoming the first compression point.
- PGA/VGA: fine positioning into the detector/ADC sweet spot. Trades noise floor versus headroom in small, controlled steps.
- AGC: maintains a stable internal level. Loop speed must match measurement mode (too fast “chases noise”; too slow gets clipped).
Selection checklist (turn specs into decisions)
- P1dB margin: sets maximum safe internal level before spurs “jump.”
- IP3 ranking by stage: identifies which block will dominate IMD when strong carriers exist.
- Noise factor (NF): ensure early stages do not force excessive downstream gain that amplifies leakage/switching artifacts.
- Gain-step error & repeatability: impacts amplitude accuracy and calibration table complexity.
- Temperature drift: affects long-term repeatability and calibration cadence.
- Switching transients: can create sweep-time ghosts; ensure settling/blanking strategy around gain/attenuation steps.
Common symptoms → likely causes → first fixes
- Spurs multiply near strong signals: early compression/IMD → add attenuation or reduce early IF gain; re-check two-tone IM3.
- Noise floor “jumps” when level changes: gain step at threshold → add hysteresis or lock gain for critical measurements.
- Amplitude offset depends on gain state: gain-step error/drift → calibrate per state and verify repeatability across temperature.
- Ghosts appear during sweeps only: switching transients/settling → enforce blanking/settle time around state changes.
- Detector/ADC overload signatures: too much late gain → pull down VGA, keep detector input inside its linear region.
H2-5 · RBW implementation: filter banks, tunable filters, and what “RBW accuracy” really means
Resolution bandwidth (RBW) is not just a UI button; it is a physical (or digitally-assisted) band-limiting function that determines what energy is admitted into the detector/ADC. “RBW accuracy” therefore includes more than a -3 dB width: it also covers shape factor (skirts/side-lobes), equivalent noise bandwidth behavior for noise readings, and repeatability across switching states and temperature.
What RBW “really” constrains
- Separation of close tones: the skirt steepness decides whether a nearby large signal leaks into a small neighbor.
- Noise floor reading: admitted noise power scales with bandwidth; stable RBW behavior yields stable noise trends.
- Sweep stability: filter group delay and switching transients affect how quickly the trace settles after a change.
Analog RBW filter banks: crystal / SAW / LC / active
- Crystal / SAW: sharp skirts and strong shape factor; watch insertion loss, group delay, and temperature drift.
- LC / active filters: flexible center/BW options; watch large-signal distortion, stability, and state repeatability.
- Switching matrix impact: relay/switch on-resistance and leakage change insertion loss and can add small state-dependent offsets.
A practical rule: higher insertion loss pushes the IF gain distribution harder downstream, which can amplify leakage signatures and make gain-state artifacts more visible. RBW design and gain staging should therefore be treated as a coupled system.
Tunable / tracking filters: where they help, where they bite
- Why they help: limit out-of-band energy before the detector/ADC, improving usable dynamic range near strong carriers.
- Typical risks: tuning nonlinearity and control settling can create sweep-only “hair” and state-dependent spurs.
- Repeatability check: repeated RBW/center switching should return to the same amplitude and center response (no drift, no hysteresis).
RBW accuracy checklist (what should be verified)
- -3 dB width consistency: the nominal RBW should match the realized response within specified tolerance.
- Shape factor stability: skirt behavior (e.g., -60 dB vs -3 dB width ratio) should not vary wildly across states.
- Noise trend consistency: noise floor should move in the expected direction as RBW changes (a sanity check for bandwidth behavior).
- Temperature & switching repeatability: RBW state should return to the same response after switching and over temperature.
Calibration typically applies per-RBW-state correction for gain/offset and may include response shaping compensation when skirt behavior affects amplitude readings near strong neighbors.
H2-6 · Detector chain: log amp vs envelope vs RMS, and why the number matches intuition
The displayed number is only meaningful when the detector chain’s measurement semantics match the intent. A log amplifier and detector do not “invent” power; they decide which statistic is reported (peak, sample, average, RMS) and how quickly it is stabilized by the video filter (VBW). Once the statistic is understood, a reading such as -63 dBm becomes predictable and consistent across modes.
Log amp operating region: where dB readings stay honest
- Linear-in-dB region: the slope is stable and calibration is valid; gain staging should keep typical signals here.
- Compression/overload: readings flatten and apparent “skirts” and spurs can inflate; reduce internal level to recover.
- Temperature drift: without compensation, the same input can read differently; stable instruments correct per state and temperature.
Detector modes in practical terms (what changes on screen)
- Sample: closest to instantaneous values; noise appears highly “alive” and the trace moves more.
- Peak: captures short transients but biases noise upward because random noise has peaks by nature.
- Average: stabilizes readings; best for steady carriers and for repeatable “typical” levels.
- RMS: aligns with power intuition across multi-tone and noisy signals; good when energy content is the goal.
This section intentionally avoids EMC quasi-peak standards; the focus is spectrum-instrument readout semantics.
Noise display “intuition”: RBW vs detector vs VBW
- RBW: decides how much noise power is admitted into the chain (band-limiting).
- Detector: decides which statistic is reported (peak vs average vs RMS).
- VBW: decides how quickly the displayed trace is smoothed and how responsive it is to changes.
When these three are consistent with the measurement goal, noise-related differences between modes become expected behavior, not surprises.
Common symptoms → likely causes → first fixes
- Noise floor much higher in Peak than Average: statistical peak bias → use Average/RMS for noise-floor comparisons.
- Carrier level stops rising with input: overload in log/detector path → reduce internal level (attenuation or IF gain).
- Trace looks “hairy” only during fast sweeps: settling/VBW mismatch → increase settle time or reduce VBW aggressiveness.
- Reading drifts with temperature: insufficient compensation → verify state-based correction and thermal stability.
- Mode-to-mode disagreement feels too large: different statistics → choose the mode that matches the measurement intent and keep settings consistent.
H2-7 · Video bandwidth (VBW) & smoothing: stable traces without lying
VBW is a post-detector low-pass / smoothing stage. It does not change what the IF chain captured; it changes how quickly the displayed number follows changes. The trade-off is simple: more smoothing improves visual stability, but too much smoothing can under-report peaks and distort short-duration events.
Where VBW sits and what it controls
- Location: after the detector (log/envelope/RMS) and before the final display/ADC statistics.
- Controls: trace jitter, settle time per sweep point, and responsiveness to bursts or level steps.
- Does not replace RBW: RBW sets spectral resolution; VBW only smooths the detected output.
Typical symptoms (what “lying” looks like)
- VBW too low: peaks look smaller, fast bursts flatten, and sweep transitions “lag” behind real changes.
- VBW too high: noise looks rougher and the trace feels unstable, even though the underlying noise is unchanged.
- Mismatch to sweep speed: when VBW time constant is long, fast sweeps show drag/settling artifacts.
A practical setting flow (fast and repeatable)
- Pick RBW first to meet frequency separation and skirt/leakage needs.
- Pick detector mode to match intent (Peak for bursts, Average/RMS for stable level comparisons).
- Use VBW last to stabilize the trace without changing the measurement meaning.
- Sanity check: if the conclusion changes when VBW changes, the setting is likely masking peaks or transient behavior.
H2-8 · ADC capture & digital IF: anti-aliasing, DDC/decimation, FFT windowing (only what IF needs)
Digital IF becomes trustworthy only when it does not create “fake lines.” That means: (1) a real analog anti-alias filter in front of the ADC, (2) an internal level window that avoids clipping and avoids quantization-noise dominance, and (3) correct ordering of DDC, filtering, and decimation so foldback does not occur. FFT windowing is then used to control spectral leakage, with amplitude correction applied so the reported level stays consistent.
Anti-aliasing: what must be done before sampling
- Why it is mandatory: out-of-band energy will fold into the band of interest and appear as in-band “tones.”
- Where it sits: analog filter right before the ADC (after the last gain/level-control point).
- Relation to RBW: RBW sets measurement resolution; anti-aliasing sets sampling legality. RBW cannot undo aliasing.
ADC level window: avoid clipping and avoid “too small” signals
- Clipping: creates harmonic-like lines and broad spectral trash; the fix is lowering internal level.
- Too low level: quantization noise dominates and weak signals vanish; the fix is raising level within headroom.
- Practical goal: keep typical measurements in a calibrated region where scaling is stable across gain states.
Clock jitter and reference quality also influence noise, but the clock-tree specifics belong to a dedicated ADC/clocking topic.
Digital IF pipeline: DDC → filtering → decimation → FFT
- DDC: frequency-translate the band of interest to baseband/low-IF so narrower processing is possible.
- Filtering before decimation: mandatory to prevent foldback when reducing the sample rate.
- Decimation: reduces data rate and sets effective bandwidth; incorrect settings can “move” lines by folding.
- FFT windowing: controls leakage when tones are not aligned to FFT bins; window choice changes skirt shape.
- Amplitude correction: compensates window gain so reported levels remain consistent across window choices.
Common “fake lines” checklist (first things to check)
- Aliasing fold-in: verify analog anti-alias filter and sampling rate margin.
- Clipping harmonics: reduce internal level; confirm lines collapse when headroom is restored.
- Decimation foldback: confirm filter order and decimation settings; see if lines “shift” when decimation changes.
- FFT leakage skirts: change window; confirm leakage shape changes while the corrected peak level remains consistent.
- Calibration mismatch: check gain-state and window correction tables for the selected measurement path.
H2-9 · Error budget: what limits DANL, TOI, and amplitude accuracy in practice
High-level specs become actionable only when they are split into a budget. This section maps three headline outcomes (DANL, TOI, and amplitude accuracy) to a small set of dominant contributors that can be isolated by changing RBW/Detector/VBW states and internal level/gain states. The goal is fast troubleshooting: identify the bottleneck stage before chasing “mystery” problems.
DANL budget: noise sources + RBW ENBW + detector statistics
- Front-end/early-IF noise contribution: the first gain elements after mixing strongly set the effective noise floor.
- RBW equivalent noise bandwidth (ENBW): admitted noise power scales with effective bandwidth, not only the nominal RBW label.
- Detector/averaging statistical bias: Peak vs Average vs RMS changes the reported noise statistic.
- ADC quantization dominance: if the internal level is too low, the display floor can be limited by quantization scaling.
- Temperature & calibration residual: drift and incomplete state correction can shift the displayed floor over time.
A quick sanity check: with detector fixed, changing RBW should move the noise floor in the expected direction; if it does not, suspect ENBW/state correction or a level-window issue that is dominating the budget.
TOI/IP3 budget: the weakest linearity stage wins
- Mixer linearity and isolation: IP3 and leakage paths determine how easily IMD and spurs enter the IF chain.
- IF amplifier/PGA compression: an aggressive gain state can push one stage close to P1dB and inflate IMD.
- Working-point (leveling) choice: internal gain distribution decides which stage becomes the bottleneck at high input levels.
- Filter visibility effect: RBW state can place IMD/spur products inside or outside the visible passband.
- ADC clipping/saturation: creates “sudden” harmonic/IMD-like growth that strongly tracks internal level changes.
A practical isolation test: adjust internal attenuation/gain while holding input fixed. If IMD collapses with small level changes, the limitation is likely a working-point/compression issue rather than a fixed spur source.
Amplitude accuracy budget: state errors + filter shape + detector linearization
- Gain-step and attenuator errors: finite step accuracy and repeatability create state-dependent offsets.
- IF gain flatness and drift: frequency response and temperature drift affect the same tone differently across band.
- RBW insertion loss & shape factor: state-to-state loss and skirt behavior can bias readings near strong neighbors.
- Detector/log nonlinearity: residual error after linearization becomes visible across wide dynamic range.
- Calibration residual: the remaining mismatch after applying state tables sets the long-term floor for accuracy.
A fast consistency check: a stable CW tone should read the same after switching gain/RBW states and returning back. If it does not return, suspect step repeatability or missing state correction.
H2-10 · Calibration hooks: leveling, detector linearization, RBW correction, self-test injection points
High-end instruments stay trustworthy because calibration is designed into the IF chain as explicit hooks: controlled injection points, repeatable pickoff points, and state-based correction tables. This section focuses on internal IF calibration (leveling, detector linearization, and RBW/VBW correction) without expanding into full system-level BIST strategies.
Amplitude leveling & flatness correction (inside IF)
- Internal reference tone: a stable cal tone provides a repeatable amplitude anchor across gain/RBW states.
- Injection + pickoff: coupling the tone into the IF chain and measuring at a known point closes the correction loop.
- State tables: per-gain and per-RBW correction compensates step errors, insertion loss, and frequency response residue.
Detector linearization (log/envelope/RMS consistency)
- Why it exists: detector and log-amplifier behavior is not perfectly linear across wide dynamic range.
- How it is corrected: segmented calibration or lookup-table correction vs input level and temperature.
- How it is validated: a stepped-level reference should produce a stable dB/dB slope after correction.
RBW/VBW correction (keeping readings consistent across states)
- RBW bank correction: per-filter-state insertion loss and response differences are compensated by state tables.
- Repeatability tracking: switching hysteresis and temperature drift are handled by periodic self-checks.
- VBW consistency goal: VBW should stabilize the trace without changing the meaning of the chosen detector statistic.
Self-test hooks (IF-only): injection and pickoff points
- Cal tone injection: validates gain state, RBW loss, and detector chain stability under controlled input.
- RBW loopback compare: distinguishes “RBW bank state problem” from “downstream detector/ADC problem.”
- Detector reference pickoff: checks detector linearization and temperature compensation quickly.
- Digital path check: verifies DDC/decimation/FFT amplitude correction consistency without external gear.
H2-11 · Validation checklist: spur map verification, RBW/VBW accuracy, linearity sweeps, repeatability
“IF chain done” means the displayed spectrum is repeatable, state-consistent, and artifact-controlled. This checklist proves it across three layers: R&D validation (deep, root-cause capable), production test (fast, high coverage), and field self-test (trend and drift surveillance).
Spur map
Two-tone IMD
RBW 3 dB BW
Shape factor
VBW bias
Detector linearity
Gain repeat-back
Drift trend
R&D validation (deep coverage, bottleneck identification)
- Spur map verification (frequency sweep “ghost map”): sweep a single tone across band and record spur frequency/level vs input level and IF gain state. Pass = spur families follow consistent slopes and state signatures; Fail = spurs that do not track level often indicate leakage/pickoff issues inside the IF path.
- Two-tone / multi-tone IMD sweep: fix tone spacing, step input level, and locate where IMD grows quickly. Pass = IMD onset aligns with expected working-point limits; Fail = sudden “IMD explosion” that collapses with small level changes often signals compression at a single stage or ADC clipping.
- RBW accuracy & shape factor: measure RBW 3 dB bandwidth and a skirt ratio (e.g., 60 dB/3 dB or 40 dB/3 dB) across key RBW states. Pass = bandwidth and skirt remain within the intended family; Fail = state-to-state outliers often point to RBW bank insertion-loss/shape correction gaps.
- Amplitude linearity (level sweep + state sweep + return-to-state): sweep CW level through the usable window, then switch key attenuator/PGA states and return back. Pass = readings are monotonic and “return-to-state” is tight; Fail = non-return suggests step repeatability or missing per-state correction.
- VBW bias check (stable traces without peak loss): keep RBW and detector fixed, compare VBW high/mid/low on a burst/step signal. Pass = smoothing changes jitter but preserves the measurement meaning; Fail = systematically reduced peaks or delayed response indicates VBW time constant is masking the event.
- Repeatability & drift: repeat the same setup across time (and warm-up/temperature points). Pass = consistent statistics and stable offsets; Fail = drift beyond the correction envelope indicates temperature compensation or state tables need adjustment.
Troubleshooting rule: change one control at a time (RBW, detector, VBW, gain state, level window). If the outcome flips with a single control, the dominant contributor is usually within that control’s path.
Production test (fast, minimal fixtures, high coverage)
- Gain-step quick check (30–60 s): apply 1–2 known input levels, switch 2–3 key gain/attenuation states, confirm state-to-state deltas match the correction table and return-to-state is tight.
- Detector linearity spot check (30–60 s): use a stepped level (or internal cal tone) and confirm the detector/log output remains consistent across the intended working window.
- RBW bank switching consistency (60–90 s): compare two RBW states on the same CW tone; verify expected insertion-loss compensation and no abnormal skirt artifacts.
- IF hook health (30–60 s): run an internal injection/pickoff loop check (cal tone inject → detector/ADC read) as a go/no-go for the IF calibration hooks.
Production philosophy: prioritize state consistency (gain steps, RBW switching, detector behavior) over ultimate limits. A unit that is state-consistent calibrates well; a unit that is not state-consistent will never be trustworthy in the field.
Field self-test (no external gear, drift and repeatability trends)
- Internal cal tone check: inject a known internal tone at a fixed state (RBW/VBW/detector/gain) and verify the reading stays within a defined window.
- Temperature-compensation consistency: compare compensated readings across temperature points (or warm-up) and confirm offsets remain bounded and monotonic.
- Repeat-back drift trend: periodically re-run a short “golden setup” sequence and log the delta from baseline to catch slow degradation early.
Suggested log fields: timestamp, temperature, RBW/VBW, detector mode, gain/atten state, cal tone level, noise-floor proxy, pass/fail.
Recommended fixtures & example part numbers (for validation and production coverage)
The following are common, practical examples. Choose equivalents that match frequency range, power, and connector standards.
| Category | Example part number | Why it is useful here |
|---|---|---|
| Fixed attenuators (SMA) | Mini-Circuits VAT-10+ / VAT-20+ / VAT-30+ | Stable level stepping for gain-state repeat-back and detector linearity checks. |
| Power splitter/combiner | Mini-Circuits ZX10-2-12-S+ | Two-tone/multi-tone IMD generation and repeatable stimulus distribution. |
| Directional coupler | Mini-Circuits ZFDC-20-5+ | Level monitoring / pickoff for leveling verification and injection-path checks. |
| 50 Ω termination | Mini-Circuits ANNE-50X+ | Clean noise-floor and spur-map baselines; prevents reflections during repeatability testing. |
| Log / power detector ICs | ADI AD8318 / ADL5513 / AD8361 | Low-cost production fixtures for detector linearity cross-checks and amplitude sanity checks. |
| Step attenuator / RF switch examples | pSemi PE43711 / PE4259 | Useful references when validating gain-step repeatability and switch-state consistency mechanisms. |
| IF chain examples (for bottleneck discussion) | ADI ADL5801 / AD8338 / ADA4899-1 | Concrete examples when linking IMD onset or level-window issues to a likely stage type in the IF chain. |
| Reference & temperature sensing examples | ADI ADR4550 / TI TMP117 | Anchors for calibration stability and temperature-compensation consistency checks. |
Request a Quote
H2-12 · FAQs (Spectrum Analyzer IF Chain)
Practical questions engineers ask when verifying spur behavior, RBW/VBW settings, detector meaning, and ADC/digital-IF artifacts.
1) Why does the displayed noise floor drop when RBW is reduced?
Reducing RBW reduces the filter’s equivalent noise bandwidth, so less integrated noise power reaches the detector. The
displayed floor typically moves in a predictable direction as RBW changes, but the exact shift also depends on detector
statistics (peak vs average) and state corrections (RBW insertion loss/ENBW tables). Smaller RBW also slows sweeps and can mask short bursts.
2) Why do fixed “ghost lines” appear at certain frequencies even with a clean input?
Fixed ghost lines often come from internal spur paths: LO leakage, mixer harmonic mixing, or unintended coupling into the IF
chain. A key clue is “movement.” If changing the LO plan or IF settings shifts the line’s position, the source is usually
LO/mixer-related. If the line stays pinned to the same display frequency across input level changes and terminations, suspect internal leakage or a reference-related spur in the IF path.
3) Should IF gain be placed earlier or later to avoid compression?
The goal is a stable “level window”: keep early stages (mixer/early IF amps) comfortably below compression while keeping the
detector/ADC high enough above noise and quantization. Early gain helps DANL but risks overload on strong inputs; later gain
preserves headroom but can make noise dominant. A practical method is to lock input level, switch gain/atten states, and watch how IMD and the noise floor change—large IMD sensitivity usually points to a bottleneck stage near compression.
4) When do log detectors and RMS detectors disagree the most?
The largest differences appear on signals with high crest factor or time variation: noise-like spectra, bursty modulation,
multi-tone composites, or intermittently appearing spurs. A log/envelope detector emphasizes instantaneous amplitude, while an
RMS detector estimates true power over its integration window. Disagreement also increases near the ends of the detector’s
linear range or when temperature drift/linearization residuals are significant. Use a known CW tone as a sanity baseline before comparing complex waveforms.
5) What can an overly small VBW hide, even if the trace looks “clean”?
VBW is post-detection smoothing (analog or digital). Making VBW very small reduces trace variance, but it can also smear or
attenuate short-lived peaks and intermittently appearing spurs—turning real events into “nothing happened.” It also increases
effective response time, so the display may lag frequency changes in swept measurements. A safe workflow is: set RBW for
resolution first, then reduce VBW only enough to stabilize readings without changing the measurement meaning.
6) A comb-like spur pattern appears—check LO planning first or ADC/digital capture first?
Use “what does the spacing lock to?” If spur spacing tracks PLL reference settings, LO step size, or divider changes, the
pattern is often LO/PLL-related. If the pattern shifts with sample rate, decimation, FFT length, or window settings, it is
more likely tied to ADC/digital IF processing (bin leakage, scaling, or clock-related sampling artifacts). Changing one knob
at a time (LO ref vs Fs/FFT) quickly separates the two families without needing system-level debugging.
7) Why can peak readings differ dramatically between detector modes for the same signal?
Detector modes answer different questions: peak finds the maximum within a detector window; sample captures a single instant;
average reports a mean; RMS estimates power. For noise-like signals, the “peak” is a statistical extreme and can be many dB
above the average depending on bandwidth and time. To compare meaningfully, keep VBW fixed and note the integration behavior
of each detector. For stable CW tones, detector modes should converge closely; large gaps on CW often indicate overload,
linearization errors, or an unintended smoothing state.
8) How to tell whether amplitude error comes from RBW shape factor or gain-step error?
Run a small matrix with a CW tone: (A) hold gain state constant and switch RBW states; (B) hold RBW constant and switch
gain/attenuation steps, then switch back (return-to-state). If the reading shifts mainly with RBW, suspect RBW insertion-loss
correction or shape-factor effects (especially near strong neighbors and filter skirts). If the reading shifts mainly with
gain steps and does not return cleanly, suspect step repeatability, state table gaps, or switching transients inside the IF gain path.
9) How do ADC clipping and quantization noise look different on a spectrum?
Clipping is abrupt and “nonlinear-looking”: harmonics, intermod-like products, and broadband splatter often appear suddenly
once the internal level crosses a threshold, and they change strongly with a 1–2 dB level adjustment. Quantization noise is
more “linear-looking”: a relatively flat noise-floor contribution that improves when the signal uses more of the ADC full
scale (until clipping). A quick test is to reduce IF level slightly; clipping artifacts collapse quickly, while quantization-limited floors improve only when gain is increased and headroom remains.
10) Two-tone IM3 results are unstable—what are common IF-chain causes?
Instability is often caused by IF-state changes or borderline compression: AGC or gain-step logic toggling, thermal drift
near a stage’s P1dB, or ADC headroom changes across sweeps. Fix the measurement first: lock attenuation/gain states, keep the
internal level window constant, and use a detector/VBW setting that does not change the meaning between captures. If IM3
collapses when the IF level is reduced slightly, a single stage is likely operating too close to compression or clipping.
11) How can calibration injection points be designed for self-test without degrading IF performance?
Good injection hooks are switchable, weakly coupled, and highly isolated when “off.” Common approaches include a small
directional coupler or a resistive adder at a controlled IF node, plus a pickoff after RBW or at the detector input to close
the loop. The key pass criteria are: minimal added insertion loss/mismatch, off-state leakage below the noise floor, and
repeatable coupling so per-state correction tables remain valid. Hooks should validate RBW switching, gain states, and detector linearization without creating new spur paths.
12) During spur-map testing, which settings most often turn false issues into “real” ones?
The top traps are state choices that change visibility or meaning: RBW too narrow can exaggerate skirt interactions and make
neighbors look like spurs; peak detector plus heavy VBW smoothing can elevate random extremes into “persistent” peaks; and
insufficient attenuation/too much IF gain can push a stage into compression or ADC clipping, creating products that look like
real spurs. A reliable workflow uses a “golden setup”: fixed gain state, moderate RBW, VBW chosen for stability (not peak suppression), and repeat-back checks to confirm state consistency.
