Ultrasound / Acoustic AFE: Low-Noise TIA + Sync ADC Arrays
← Back to:Analog-to-Digital Converters (ADCs)
Ultrasound / acoustic AFE design is won by recovery + coherence: fast return to linear after Tx leakage and tightly matched gain/phase/delay across channels. When those budgets are measurable and closed, beamformer-ready data follows; chasing “a better ADC SNR” alone rarely fixes sidelobes, ghosts, or drift.
What this page solves
This page targets ultrasound / acoustic receive AFEs where image quality is limited by front-end noise, overload recovery, and array coherence. It provides an engineering path from symptoms to measurable budgets and design levers, without turning into a beamforming-algorithm tutorial or a generic ADC catalog.
The three hard constraints (and what must be measured)
- Noise → input-referred noise over the echo band, plus code-domain noise after gain and filtering.
- Recovery → overload behavior (clipping/rail hit), release time, and the resulting near-field “blind zone”.
- Coherence → channel-to-channel gain, phase, and delay match versus frequency and temperature.
Symptom → likely owner (fast triage)
- High sidelobes / ghost features → coherence (delay/phase), crosstalk, or gain tracking errors.
- Long blind zone / missing near-field → overload recovery (front-end, clamp, gain control strategy).
- “Snowy” texture / lost fine detail → noise budget (front-end dominates before enough gain).
- Fixed stripes / tones → gain-step spurs, switching feedthrough, or coupling into the band.
Beamforming details are intentionally out of scope. Only the hardware requirements needed to keep channels time-aligned (synchronization, match, recovery, and spur control) are covered.
Definition
An ultrasound / acoustic AFE is the receive signal chain that converts transducer echoes into digitized channels with controlled noise, controlled overload behavior, and controlled channel-to-channel alignment. A “working” AFE is defined by what it can prove on the bench: noise in-band, recovery time after overload, and coherence across channels.
Typical blocks (what each must “own”)
- Clamp / protection → limits overload energy and protects the first stage.
- Damping / matching → controls ringing that looks like false echoes.
- TIA / LNA → sets input-referred noise and stability under pulse conditions.
- VGA / TGC → manages dynamic range while minimizing gain-step spurs.
- AAF → band-limits and stabilizes the driver into the ADC input network.
- ADC array + sync hooks → preserves coherence (gain/phase/delay) and enables alignment tests.
Key judging metrics (engineering meaning)
- Coherence: when the same input is applied to multiple channels, the measured gain, phase, and delay stay consistent across the signal band (and stay stable with temperature).
- Recovery: after a controlled overload pulse, the output returns to linear behavior within the allowed time window (blind-zone requirement).
- Crosstalk: coupling between channels remains below the artifact threshold during echo windows.
Principle: burst/echo timing, TGC artifacts, and array synchronization
Ultrasound and many acoustic arrays operate in burst/echo cycles. A high-energy transmit burst is followed by a recovery interval where the receive chain must return to linear behavior, and only then does the echo window become usable for imaging. The time spent recovering sets the near-field “blind zone”.
Time-gain control (TGC) is required because echoes decay rapidly with time. However, gain changes are not free: step-like gain updates can modulate the signal chain and create in-band spurs, and gain-dependent phase/delay shifts can break channel-to-channel coherence. The same single-channel SNR can still produce poor images if coherence and recovery are not controlled.
Symptom → mechanism (fast mapping)
- Long blind zone / missing near-field → overload recovery time and clamp release behavior.
- Fixed stripes / tones in the echo band → TGC steps, switching feedthrough, or coupling during echo windows.
- High sidelobes / blur / position drift → channel delay/phase mismatch and temperature drift.
Requirements & budgets: make “clearer images” measurable
Requirements should be written as measurable budgets, then verified on the bench. Budget ownership prevents “invisible” failure modes where each block looks acceptable in isolation but the array image still degrades.
Budget fields to write down (application-focused)
- Noise budget: input-referred noise over the echo band → code-domain noise after gain/filtering.
- Dynamic range: Tx leakage versus smallest echo; clipping margin; recovery time to linear behavior.
- Coherence: gain mismatch, phase mismatch, and delay mismatch versus frequency and temperature.
- Spurs: allowable in-band spurs from TGC steps, switching, or digital feedthrough during echo windows.
- Throughput: channels × sample rate × bits + framing overhead and buffering constraints.
Measurement-first acceptance (what to prove)
- Noise: integrated in-band noise at the ADC output under a representative gain plan.
- Recovery: controlled overload pulse → time-to-linear within the allowed blind-zone budget.
- Coherence: coherent stimulus across channels → gain/phase/delay mismatch versus frequency.
- Spurs: observe echo-window spectra while stepping gain and switching system rails/clocks.
Principle: burst/echo timing, TGC artifacts, and array synchronization
Ultrasound and many acoustic arrays operate in burst/echo cycles. A high-energy transmit burst is followed by a recovery interval where the receive chain must return to linear behavior, and only then does the echo window become usable for imaging. The time spent recovering sets the near-field “blind zone”.
Time-gain control (TGC) is required because echoes decay rapidly with time. However, gain changes are not free: step-like gain updates can modulate the signal chain and create in-band spurs, and gain-dependent phase/delay shifts can break channel-to-channel coherence. The same single-channel SNR can still produce poor images if coherence and recovery are not controlled.
Symptom → mechanism (fast mapping)
- Long blind zone / missing near-field → overload recovery time and clamp release behavior.
- Fixed stripes / tones in the echo band → TGC steps, switching feedthrough, or coupling during echo windows.
- High sidelobes / blur / position drift → channel delay/phase mismatch and temperature drift.
Requirements & budgets: make “clearer images” measurable
Requirements should be written as measurable budgets, then verified on the bench. Budget ownership prevents “invisible” failure modes where each block looks acceptable in isolation but the array image still degrades.
Budget fields to write down (application-focused)
- Noise budget: input-referred noise over the echo band → code-domain noise after gain/filtering.
- Dynamic range: Tx leakage versus smallest echo; clipping margin; recovery time to linear behavior.
- Coherence: gain mismatch, phase mismatch, and delay mismatch versus frequency and temperature.
- Spurs: allowable in-band spurs from TGC steps, switching, or digital feedthrough during echo windows.
- Throughput: channels × sample rate × bits + framing overhead and buffering constraints.
Measurement-first acceptance (what to prove)
- Noise: integrated in-band noise at the ADC output under a representative gain plan.
- Recovery: controlled overload pulse → time-to-linear within the allowed blind-zone budget.
- Coherence: coherent stimulus across channels → gain/phase/delay mismatch versus frequency.
- Spurs: observe echo-window spectra while stepping gain and switching system rails/clocks.
Design: low-noise TIA/LNA and stability (ultrasound-focused)
The first receive stage often dominates ultrasound image quality because it sees the weakest echoes while also being exposed to transmit leakage and large near-field returns. A usable front end must control in-band input-referred noise, ringing/settling, and overload recovery under burst conditions.
Which noise path tends to dominate (decision logic)
- If the transducer presents a higher effective impedance in the echo band, current-noise paths become more visible.
- If the effective impedance is lower, voltage-noise paths tend to set the floor and the driver/TIA choice is decisive.
- Noise should be judged as integrated in-band noise after the planned gain distribution, not as a single catalog number.
Protection and damping are part of the receive design, not add-ons. Clamp behavior and damping networks control the energy that reaches the first stage and the amount of ringing that can masquerade as false echoes. Those same elements also change the noise floor and the time constants that define overload release.
Why stability and recovery are tied to the blind zone
- Burst conditions force large steps and clamp transitions; marginal stability shows up as ringing and long tails.
- Slow release after overload extends the time-to-linear interval, directly increasing the near-field blind zone.
- A stable, well-damped front end reduces both false-echo risk and recovery time variance.
Executable verification (bench-friendly)
- Shorted-input noise: confirms the intrinsic receive-chain noise floor.
- In-band noise spectrum: integrates noise over the echo band under the planned gain distribution.
- Pulse/step response: checks ringing amplitude and settling time versus false-echo thresholds.
- Controlled overload recovery: injects a repeatable overload pulse and measures time-to-linear (blind-zone budget).
Design: VGA/TGC for dynamic range and artifact control
TGC is required to track echo decay, but gain changes can create artifacts. The design target is not only “avoid clipping”: TGC must also respect a spur budget and a coherence budget inside echo windows.
System goals for the TGC curve
- Headroom: prevent clipping under leakage and strong near-field returns.
- Low artifacts: gain transitions must not create in-band spurs during echo windows.
- Coherence: gain state must keep phase/delay consistent across channels.
Analog VGA vs digital compensation (risk lens)
- Analog VGA: improves effective noise floor early, but gain-step settling and phase/delay shifts can create spurs and mismatch.
- Digital compensation: avoids analog gain-state phase shifts, but cannot undo analog clipping and may require more ADC dynamic range.
Overload recovery and TGC are coupled: aggressive early gain can cause secondary overload and extend the blind zone. Gain transitions should be planned with explicit settling windows and verified under the real burst/echo timing.
ADC array & channel matching: mismatch → imaging artifacts
Array imaging quality is often limited by structured channel mismatch rather than by single-channel SNR alone. Beamforming and coherent DSP amplify any channel-to-channel differences, turning small mismatches into visible artifacts.
Mismatch types (engineering view)
- Offset mismatch: baseline differences that can bias echo windows and create fixed-pattern texture.
- Gain mismatch: amplitude scaling differences that distort shading and effective aperture weighting.
- Timing (delay) mismatch: sampling skew that degrades coherence and raises sidelobes.
- Phase mismatch: frequency-dependent phase/group-delay differences that smear focus and shift apparent position.
- Crosstalk: coupling between channels that creates ghost features and fixed patterns.
Typical mapping: error → consequence
- Timing / phase mismatch → sidelobes ↑, blur ↑, and position drift.
- Gain mismatch → amplitude shading errors and stripe-like fixed patterns.
- Crosstalk → ghost features and spurious echoes correlated across channels.
Calibration hooks (no algorithm details)
- Coherent stimulus injection: same input into multiple channels for alignment characterization.
- Tone / multi-tone modes: sweep phase difference versus frequency to expose group-delay mismatch.
- Coefficient storage hooks: per-channel gain/offset/phase tables with temperature bins if needed.
Verification methods (what the measurement outputs)
- Two-channel coherent input → phase difference vs frequency (Δφ(f)).
- Impulse/edge stimulus → delay mismatch via cross-correlation (Δt).
- Repeat across temperature/time → drift in Δφ(f) and Δt.
Synchronization: what must match and how to prove it
For array receive systems, the key is not whether a clock is “good” in isolation. What matters is whether channel-to-channel sampling skew and group-delay mismatch remain controlled, repeatable, and verifiable across operating conditions.
Quantities to control (measurement outputs)
- Skew: channel-to-channel sampling time offset (Δt).
- Group-delay mismatch: phase difference versus frequency (Δφ(f)) and its slope.
- Drift: change of Δt and Δφ(f) over temperature/time.
- Deterministic latency: repeatable, fixed alignment from trigger/SYNC to sampled data.
Synchronization methods (classification only)
- Shared clock: common timebase distribution to all channels.
- Synchronous sampling: aligned start and consistent sampling phase across channels.
- Sync trigger: deterministic alignment of echo windows and gain steps.
- Timestamp hooks: markers that make system-level alignment observable and repeatable.
Verification workflow (repeatable)
- Apply coherent stimulus to multiple channels (same phase reference).
- Sweep frequency points → record Δφ(f) for group-delay mismatch.
- Apply impulse/edge → cross-correlate to extract Δt (skew).
- Repeat over temperature/time → quantify drift and repeatability.
Engineering checklist: from specs to sign-off
Use this checklist to turn burst/echo requirements, noise and recovery budgets, and multi-channel coherence into measurable sign-off items. Each group is written as check → how → output → pass, so results remain traceable from bench data to production readiness.
1) Specs / requirements (write measurable fields)
- Check: fs, echo-bandwidth, Nch, echo window, min echo, max leakage, target blind zone.
- How: define burst conditions and the exact echo windows used for measurements.
- Output: one-page specs table with units and test windows.
- Pass: every field has a unit, window definition, and operating condition tag.
2) Noise measurements (bench-first, band-integrated)
- Check: shorted-input noise, in-band integrated noise, code-domain RMS noise under the gain plan.
- How: measure spectra inside the defined echo band and integrate over the same windows used in the system.
- Output: noise spectra + integrated results + code-domain summary table.
- Pass: noise budget closes with measured data (not only catalog numbers).
3) Recovery / blind zone (time-to-linear under burst conditions)
- Check: overload recovery time, clamp/protection action, early TGC policy, secondary overload risk.
- How: inject a controlled overload pulse; measure time-to-linear and verify inside real burst/echo timing.
- Output: recovery plots + time-to-linear numbers with conditions (gain/state/temp).
- Pass: blind-zone budget is met with margin and remains stable across conditions.
4) Coherence / matching (gain/phase/delay versus frequency)
- Check: gain mismatch, Δφ(f), Δt, and repeatability over temperature/time.
- How: coherent stimulus + multi-point sweep; use cross-correlation for delay extraction.
- Output: Δφ(f) curves, Δt distribution, drift comparison plots.
- Pass: coherence budget closes and drift stays within the system’s focusing tolerance.
5) Crosstalk / EMI (ghost and spur control)
- Check: channel-to-channel crosstalk, switching spurs in echo windows, supply-noise coupling.
- How: drive one channel while observing others; toggle switching states and record echo-band spectra.
- Output: crosstalk matrix (simplified), spur list, sensitive switch/source map.
- Pass: spurs stay below the artifact threshold and crosstalk does not form stable ghosts.
6) Production / calibration (repeatable and traceable)
- Check: calibration hooks, coefficient storage, BIST/self-test path, version traceability.
- How: inject coherent tones/edges; store and verify coefficients; run BIST coverage on key paths.
- Output: calibration flow, coefficient format, BIST report and logs.
- Pass: calibration is repeatable, coefficients are traceable, and production time is acceptable.
Applications: ultrasound/acoustic AFE use-cases (within this topic)
These use-cases stay inside ultrasound/acoustic AFEs. The key difference across applications is the relative weight of noise, recovery severity, coherence, and interference budgets.
Medical ultrasound
- Typical: large arrays and wide echo bands with near-field sensitivity.
- Dominant: coherence + blind-zone control + wide dynamic range.
- Emphasis: stable TGC behavior and repeatable Δφ(f)/Δt across conditions.
Industrial NDT
- Typical: higher energy bursts and stronger leakage/overload events.
- Dominant: protection behavior + recovery time + artifact (spur/false-echo) control.
- Emphasis: controlled overload recovery and echo-window spur verification.
Sonar / underwater acoustic
- Typical: lower bands, long observation windows, and strong interference environments.
- Dominant: drift control + sync expansion + interference/spur robustness.
- Emphasis: repeatability across temperature/time and deterministic trigger alignment.
Microphone arrays / beamforming audio
- Typical: much lower bands, but coherence and crosstalk remain critical for beam patterns.
- Dominant: matching + crosstalk/EMI + stable gain states.
- Emphasis: crosstalk matrix checks and stable Δt under system activity.
IC selection logic (before FAQ): fields → risk map → inquiry template
This section turns ultrasound/acoustic AFE requirements into vendor-answerable measurable fields, maps each missing field to a known imaging risk, and provides a copy-ready inquiry template. The goal is to prevent “generic ADC shopping” and force quantified answers under stated conditions.
Parameter fields (by module)
For each module, the table below shows what to ask, under which conditions, and what evidence is required. Conditions must explicitly include echo band/window, gain plan (TGC states), and temperature points when drift matters.
Transducer interface (Tx leakage survivability)
TIA/LNA (noise + stability + overload recovery)
VGA/TGC (gain steps → spurs → artifacts)
ADC array (target band performance + channel-to-channel specs)
Sync (clock + trigger alignment + deterministic latency)
Power / EMI (spur immunity inside echo windows)
Test / calibration (production readiness)
Risk map (field gap → artifact → evidence)
Recovery time not sufficient
Risk: blind zone expands → near-field loss.
Evidence: time-to-linear plot under stated Tx leakage conditions.
Skew/drift not controlled
Risk: sidelobes rise → focus/position drifts.
Evidence: Δt distribution + drift vs temperature/time (repeatability).
TGC gain steps create in-band spurs
Risk: fixed stripes / ghost patterns.
Evidence: echo-window spur list vs gain state + settling time after step.
Crosstalk not specified/tested
Risk: false targets / coherent ghosts.
Evidence: crosstalk matrix + single-channel excitation observation.
Inquiry template (copy-ready)
Paste the template below into an email to a distributor/vendor. Replace bracketed fields. The questions force quantified answers and evidence.
Subject: Ultrasound/AffE multi-channel AFE inquiry (must quantify)
Project (one line)
- Application: [Medical / NDT / Sonar / Mic array]
- Array: [Nch] channels, target echo band: [fL–fH] MHz, fs: [fs] MSPS
- Tx leakage at Rx input: [Vpk] V, [pulse width] ns/us, PRF: [kHz], source Z: [Ω]
- Target blind zone (time-to-linear): ≤ [T] us
- Coherence target: channel skew Δt ≤ [X] ps/ns; drift over temperature: ≤ [Y] ps/ns
- Operating temperature: [Tmin–Tmax] °C
Key specs (please fill with conditions)
1) Recovery time (time-to-linear) under stated Tx leakage conditions:
- Value:
- Threshold definition:
- Evidence: recovery plot + measurement method
2) In-band noise (echo band integrated) under stated gain plan:
- Integrated noise:
- Noise spectrum:
- Evidence: spectra + integration method
3) TGC / gain changes:
- Gain range and update method:
- In-band step spur level (echo band) and settling time:
- Evidence: spur list vs gain state + transient plot
4) Multi-channel matching:
- Channel-to-channel skew (typ/max) and deterministic latency:
- Drift vs temperature/time:
- Crosstalk spec and test method:
- Evidence: Δt/Δφ(f) data + crosstalk matrix
5) Sync/trigger:
- Supported clock topology and SYNC/trigger alignment method:
- Deterministic latency statement:
- Evidence: application note or characterization results
6) Test/calibration hooks:
- BIST / stimulus injection modes:
- Coefficient storage/retention method:
- Evidence: flow + outputs + traceability approach
Must-quantify questions (please answer with numbers + evidence)
- What is time-to-linear recovery under the stated Tx leakage? Provide plot + method.
- What is channel-to-channel skew and its drift over temperature/time? Provide Δt and repeatability.
- What is TGC gain-step spur level inside the echo band? Provide spur list + settling time.
- What is worst-case crosstalk and how is it measured? Provide method + conditions.
- What is deterministic latency behavior across power cycles and resets? Provide statement + test notes.
Benchmark part numbers (for equivalence class)
- Ultrasound AFE: TI AFE5808A / AFE5804; ADI AD9670 / AD9271
- T/R & Tx platform: TI TX810; TI TX7332
- VGA/TGC reference: ADI AD8331 / AD8334
- Clock/sync: TI LMK04828; ADI AD9528
- Low-noise rails: TI TPS7A47 / TPS7A49; ADI LT3042
Tip for faster vendor responses: require a reply format of number + condition + evidence (plot/table/test method) for each must-quantify item.
Request a Quote
FAQ (ultrasound / acoustic AFE): artifacts, recovery, coherence, EMI, and selection
These FAQs capture long-tail issues without expanding the main text. Each answer follows a measurable format: What it means → Fast checks → Measurements (data) → Fix levers.
Why are sidelobes high even when single-channel SNR looks good?
What it usually means: The array is limited by coherence (gain/phase/delay mismatch), not by single-channel noise. Small channel-to-channel timing/phase errors raise sidelobes and blur focus even if one channel’s SNR is excellent.
- Feed a coherent stimulus to 2 channels; compare phase and delay.
- Repeat after power-cycle; look for deterministic vs random offset.
- Compare results across gain states (TGC).
| Δt (skew) | ps/ns, distribution + drift vs temperature |
| Δφ(f) | phase mismatch vs frequency over echo band |
| Gain mismatch | dB vs gain state, repeatability |
Tighten sync tree (shared clock/trigger), reduce deterministic skew, control Δφ(f) via consistent analog paths, avoid gain-state dependent phase steps, and add calibration hooks for gain/phase/delay.
What causes “ghost targets” or duplicated features in the image?
What it usually means: A coherent unwanted path exists: channel crosstalk, TGC step spurs, or clock/trigger-coupled spurs. These can look like stable duplicates because they are phase-related across channels.
- Excite one channel; watch other channels for correlated replicas.
- Toggle gain-step timing; see if ghosts move/appear with steps.
- Change clock/trigger edge rate or routing; compare spur behavior.
| Crosstalk matrix | dB, per channel pair, worst-case condition |
| Echo-window spur list | freq + amplitude vs gain state and timing |
| Correlation evidence | coherent relationship across channels (yes/no) |
Improve isolation and return paths, separate sensitive analog from fast digital edges, control gain-step timing and settling, reduce clock/trigger coupling, and validate crosstalk with a consistent stimulus/fixture.
Why do I see fixed stripes / banding patterns?
What it usually means: A periodic spur falls inside the echo band or echo window and repeats frame-to-frame. Common sources are TGC steps, power ripple coupling, or clock-related spurs.
- Capture spectra inside the echo window; identify persistent tones.
- Shift switching frequencies (if possible) and see whether the stripe moves.
- Compare with gain fixed vs gain stepping.
| Spur list | (f, amplitude) in echo window + gain state |
| Rail-noise correlation | spur amplitude vs rail ripple level |
| Edge-coupling sensitivity | spur change vs clock/trigger edge behavior |
Smooth or re-time gain steps away from echo windows, reduce rail ripple and coupling paths, improve grounding/partitioning, and isolate clock/trigger routing from sensitive analog nodes.
Why does the focus/position drift over time or temperature?
What it usually means: Channel-to-channel delay/phase is not stable with temperature/time. Drift can come from clock distribution, analog group delay changes, or gain-state dependent phase steps.
- Repeat Δt and Δφ(f) at two temperatures (cold/hot points).
- Hold gain constant; compare drift vs normal gain plan.
- Track drift after warm-up to separate thermal settling vs random drift.
| Δt drift | ps/ns vs temperature and time |
| Δφ(f) drift | phase mismatch vs frequency at temp points |
| Gain-state sensitivity | Δt/Δφ change vs TGC states |
Use a coherent clock/trigger distribution with controlled skew, reduce temperature gradients, avoid gain-state phase steps, and add temperature-aware calibration/verification at sign-off.
Why is the blind zone longer than expected?
What it usually means: The receive chain is overloaded by Tx leakage and takes too long to return to linear operation. Protection/clamp choices can trade survivability for recovery speed, expanding the blind zone.
- Measure recovery under the true Tx leakage pulse (Vpk/pw/PRF).
- Compare clamp enabled/disabled (or alternative network) recovery time.
- Check early TGC strategy for secondary overload.
| Time-to-linear | µs under Vpk/pw/PRF; define threshold |
| Headroom margin | clipping margin vs leakage envelope |
| Clamp impact | recovery delta with clamp network variants |
Increase headroom where needed, choose clamp/protection for fast release, reduce leakage coupling, validate early gain plan, and measure recovery using real burst timing and thresholds.
What does “slow recovery after saturation” look like on the bench and how to measure it?
What it usually means: After an overload pulse, the chain remains non-linear (distorted, clamped, or biased) long enough to corrupt the early echo window. The key metric is time-to-linear, not just “recovery time” wording.
- Use a controlled overload pulse; capture output settling and distortion tail.
- Define a linear threshold (error/THD/step residual) and apply consistently.
- Repeat across gain states and temperatures.
| Overload profile | Vpk, pulse width, PRF, source impedance |
| Time-to-linear | µs to meet defined threshold |
| State dependence | time-to-linear vs gain state and temperature |
Reduce overload energy at the input, use clamps with fast release, avoid saturating internal nodes, and validate recovery with a consistent threshold and burst timing.
How should early TGC be set to avoid secondary overload and artifacts?
What it usually means: Early gain is a trade between capturing weak near echoes and preventing a second overload event. Gain steps can also inject spurs; step timing and settling must be managed relative to the echo window.
- Freeze gain early; check if artifacts disappear.
- Move gain-step timing away from the echo window; compare spur lists.
- Verify headroom margin under worst-case leakage and early echoes.
| Gain plan | time vs gain table + update timing method |
| Step settling | µs to settle after a gain change |
| In-window spurs | spur list (f, amplitude) vs step timing/state |
Use smaller/less frequent gain steps near the early window, enforce settling before capture, keep early headroom margin, and validate with in-window spur measurements.
How to measure channel-to-channel skew (Δt) quickly?
What it usually means: A practical Δt test must be coherent, repeatable, and tied to the same echo band conditions used in the system. Δt should be measured as both a distribution and a drift over temperature/time.
- Drive two channels with the same coherent source and identical paths.
- Use cross-correlation to estimate Δt.
- Repeat across gain states and power cycles.
| Δt per pair | ps/ns, typ/max + histogram |
| Repeatability | Δt change after power cycle/reset |
| Temp drift | Δt vs temperature points |
Use deterministic sync mechanisms, reduce path asymmetry, control routing and fanout skew, and add calibration hooks to store and apply Δt corrections.
How to measure phase mismatch versus frequency (Δφ(f)) for coherence?
What it usually means: Phase mismatch that varies with frequency indicates group-delay mismatch and analog path differences. The required output is a Δφ(f) curve over the echo band, not a single number.
- Sweep multiple tones across the echo band using the same coherent source.
- Compute phase difference per frequency point and plot Δφ(f).
- Repeat at a second temperature point to reveal drift.
| Δφ(f) curve | degrees vs frequency over echo band |
| Derived group delay | equivalent delay mismatch vs frequency |
| Drift | Δφ(f) change vs temperature/time |
Match analog path components and layouts, control filter group delay variation, keep gain-state changes phase-consistent, and use calibration data to compensate residual Δφ(f).
“Changing the ADC didn’t help” — how to locate the spur/EMI source?
What it usually means: The dominant issue is not the converter core, but a system coupling path: gain switching, clock/trigger feedthrough, power/ground return coupling, or channel-to-channel interference.
- Freeze gain plan (no steps); compare in-window spur list.
- Change clock/trigger routing or edge behavior; observe spur change.
- Turn off nearby switching loads one by one; record differences.
| Spur delta table | spur amplitude change vs each toggle |
| Rail correlation | spur vs rail ripple / load switching |
| Coupling hypothesis | mapped source → victim path (short statement) |
Reduce fast-edge coupling, strengthen partitioning and return paths, clean rails and references, re-time or smooth gain steps, and validate with a controlled toggle-and-measure plan until one source dominates.
Is SAR vs Pipeline the key choice for ultrasound/acoustic AFE?
What it usually means: Converter architecture is rarely the only root cause. For ultrasound/acoustic AFEs, the primary sign-off order is typically recovery/blind zone → coherence → spurs/crosstalk → in-band ENOB/SNR.
- Rank issues by measured evidence (recovery, Δt/Δφ, spurs, crosstalk).
- Only compare architectures after coherence and recovery meet the budget.
- Use in-band metrics (echo band), not generic full-band summaries.
| Time-to-linear | µs under stated leakage conditions |
| Coherence | Δt, Δφ(f), gain mismatch + drift |
| Artifacts drivers | spur list + crosstalk matrix in echo windows |
Choose architectures based on measured budgets and system constraints, then lock coherence and recovery first. After that, use in-band ENOB/SNR and power/throughput to select the best-fit implementation.
