123 Main Street, New York, NY 10001

Submit Your BOM

Comparator Propagation Delay vs Overdrive: Timing & Jitter

Comparator Propagation Delay vs Overdrive: Timing & Jitter

← Back to:Comparators & Schmitt Triggers

Propagation delay is not a constant: it depends on effective overdrive and the input slew rate at the threshold crossing. Make worst-case timing predictable by translating the real waveform into VOD(t), aligning datasheet conditions, and budgeting delay + jitter with measured distributions.

What this page solves: delay is not a constant

Propagation delay (tPD) changes with overdrive (VOD) and input slew rate (dVIN/dt). A single “typical tPD” number can be misleading, especially near small overdrive where delay increases and timing spread widens.

Outcomes this page delivers
  • Read datasheet tPD–VOD curves and identify the “small overdrive” risk region.
  • Map a real input waveform into effective VOD(t) near the threshold crossing.
  • Budget system timing and trigger uncertainty using worst-case delay + timing spread.
Scope guard (to avoid cross-page overlap)

This page focuses on tPD ↔ VOD ↔ dVIN/dt and timing/jitter budgeting. Hysteresis sizing, output-stage details, and offset/drift accuracy are referenced only briefly and are handled on their dedicated pages.

Overview: propagation delay vs overdrive and waveform mapping Two-panel diagram: left shows a typical and worst-case propagation delay versus overdrive curve; right shows a threshold crossing waveform and the corresponding overdrive versus time near zero crossing. tPD vs VOD (datasheet curve) Waveform → VOD(t) near crossing VOD (mV) tPD small VOD typ max spread widens spec point VTH VIN(t) cross VOD(t) ↑ low-VOD zone (most sensitive) Small overdrive + slow slew rate → larger delay and wider timing spread.

The left panel illustrates why small overdrive is a timing-risk region (delay grows and worst-case spread widens). The right panel shows how a real waveform maps to effective overdrive near the threshold crossing.

Definitions: overdrive, threshold crossing, and how tPD is measured

Consistent definitions are required before comparing datasheet numbers or building a timing budget. The same comparator can appear “fast” or “slow” depending on how overdrive, output threshold, load, and input slew rate are defined.

Compact glossary (engineering meaning)
VOD (Overdrive)
Instantaneous input difference beyond the threshold/reference after crossing (single-ended or differential).
tPD (Propagation delay)
Time from input threshold crossing to output reaching a defined level (often 50% of VOUT swing).
tPLH / tPHL
Rising vs falling delay can differ; system timing may be sensitive to only one direction.
Input slew rate (dVIN/dt)
Determines how voltage noise translates into timing uncertainty at the crossing.
Datasheet “tPD” numbers are comparable only when conditions match
  • Overdrive method: fixed step vs ramp vs sine crossing.
  • Output threshold: 50% swing vs logic threshold vs VOH/VOL limit.
  • Load: CL/RL and output swing strongly affect observed delay.
  • Supply & temperature: use the same VDD and corner for comparisons.
  • Input common-mode: behavior can shift near rails; note the test VICM.
Do / Don’t (quick guardrails)
  • Do: compare tPD under matched VOD, load, VDD, and output-threshold definitions.
  • Do: track tPLH and tPHL separately if edge direction matters.
  • Don’t: treat “typ tPD” as a worst-case timing guarantee.
  • Don’t: ignore slew rate when estimating trigger uncertainty.
Propagation delay measurement definition Waveform diagram defining input threshold crossing, overdrive after crossing, and propagation delay measured to a defined output threshold such as 50 percent. How tPD is defined: crossing → output threshold time VTH 50% VIN VOUT crossing tPD VOD tPLH / tPHL may differ track both if edge direction matters Define thresholds and test conditions first, then compare tPD numbers.

Propagation delay is measured from the input threshold crossing to a defined output level (commonly 50%). Overdrive is the input difference beyond the threshold after the crossing, and it must match the datasheet test definition when comparing devices.

Mechanism: why small overdrive makes delay longer (regeneration time)

A comparator does not “instantly decide.” A tiny input difference must be amplified until the internal regenerative stage commits to one side and the output buffer can switch. When overdrive is small, that regeneration takes longer, and the observed propagation delay grows.

Cause → effect (minimal model)
  • Smaller VOD means a smaller “starting push” at the input pair.
  • The regenerative (positive-feedback) stage needs more time to build a decisive output.
  • As VOD approaches zero, decision time can become long and variable, widening timing spread.
Near-zero overdrive: what breaks first
  • Output transition occurs later and with larger part-to-part and cycle-to-cycle variation.
  • Small noise or ringing near the crossing can shift the decision moment noticeably.
  • Timing windows and gated sampling become harder to guarantee with “typical” numbers.
Note: overdrive is not “the bigger the better”

Very large input excursions can introduce non-ideal behaviors (clamping, recovery effects, overshoot and EMI). For timing design, the key is to guarantee a minimum effective overdrive near the crossing, under the same load and corner conditions used for the budget.

Regeneration model and why small overdrive increases delay Block diagram of input pair, regenerative latch, and output buffer, plus two output transition curves showing faster switching for larger overdrive and slower switching for smaller overdrive. Minimal regeneration model: small VOD → longer decision time Comparator internal path (simplified) VIN+ VIN− Input pair ΔVIN = VOD Regenerative latch regeneration Output buffer logic drive VOUT small VOD region longer decision time wider timing spread Output transition (concept) time VOUT threshold large VOD small VOD tPD grows Regeneration needs time; small overdrive delays commitment and increases timing variation.

The regenerative stage must amplify a small input difference into a decisive internal state before the output can switch. With smaller overdrive, that build-up takes longer, increasing both propagation delay and timing spread.

Datasheet reading: the 6 fields that actually matter for tPD–VOD

Propagation delay numbers are only meaningful when the test setup is understood. A fast-looking “typical tPD” can hide a poor worst-case delay at small overdrive, or a measurement condition that does not match the real system waveform and load.

The 6 fields (read in this order)
  1. tPD vs overdrive curve (typ): locate the small-overdrive region and its slope.
  2. tPD max condition: note the exact overdrive used for the guaranteed limit.
  3. Input stimulus & source impedance: step amplitude and source-R change effective VOD(t).
  4. Load condition: CL/RL and output swing/threshold definitions change observed delay.
  5. Supply & temperature corners: worst-case timing must follow the intended corners.
  6. Input common-mode and near-rail behavior: confirm the test VICM and any crossover notes.
What to record for comparisons (minimal “timing sheet”)
  • VOD definition: fixed step, ramp crossing, or sine; include the numeric value used.
  • Output threshold: 50% swing vs logic threshold; include VOH/VOL limits if used.
  • Load: CL/RL and output swing; include cable/probe capacitance assumptions.
  • Corners: VDD(min/typ/max) and temperature range for the budget.
  • VICM: where the input sits relative to rails during the test condition.
Datasheet callouts for propagation delay vs overdrive A training-style virtual datasheet page with six highlighted boxes calling out the key fields needed to interpret and compare propagation delay: tPD-vs-VOD curve, max condition, stimulus/source impedance, load and output threshold, supply/temperature corners, and input common-mode notes. Virtual datasheet page: six callouts that control tPD meaning Comparator Timing Characteristics (excerpt) tPD vs VOD 1) tPD–VOD curve tPD max condition VOD = 20 mV / 100 mV · CL = … · VDD = … 2) tPD max @ VOD stimulus / source-R Rs DUT step / ramp matters 3) Step & source-R load / output level CL RL threshold definition 4) CL/RL + threshold VDD / Temp corners use worst-case for budget 5) Corners VICM / near-rail note: behavior can shift 6) VICM near rails Align definitions and conditions first; only then compare “tPD” numbers across parts.

Treat propagation delay as a conditional specification. The six highlighted areas capture what most frequently changes the meaning of “tPD” and determines whether a comparison will match real waveform and load conditions.

From voltage noise to time jitter: σt ≈ σv / (dVIN/dt)

Timing uncertainty at a comparator threshold is often driven by voltage noise and interference. A practical mapping is that voltage variation at the crossing becomes time variation through the local slope: σt ≈ σv / (dVIN/dt). Slower slope means the same noise produces larger time jitter.

What creates time jitter at the crossing
  • Voltage noise / interference shifts the instant when VIN intersects VTH.
  • Lower dVIN/dt magnifies that shift into a larger Δt at the same σv.
  • Small overdrive region typically increases both decision sensitivity and timing spread.
Practical actions to reduce σt (prioritized)
  • Raise dVIN/dt near VTH: avoid slow ramps at the decision point.
  • Reduce RC drag: minimize series resistance and effective input capacitance that flatten slope at VTH.
  • Lower source impedance: improve drive so the crossing is less affected by loading.
  • Use edge shaping where needed: buffer/limit/amplify to create a cleaner transition (details belong to application pages).
Quick sanity check

If the local slope at the threshold is cut in half, time jitter roughly doubles for the same voltage noise. Controlling the crossing slope is often the fastest way to stabilize timing.

Same noise, different slope: different timing spread A plot showing three noisy input ramps crossing the same threshold line. The fast ramp has a small time spread at the crossing, while the slow ramp shows a larger time spread, illustrating sigma t proportional to sigma v divided by slope. Same noise (σv), different slope (dVIN/dt) → different timing spread (Δt) time VIN VTH fast slope mid slope slow slope Δt small Δt large The same σv produces more timing spread when dVIN/dt is smaller at the crossing.

With the same noise level, the fast-slope crossing produces a tight timing cluster, while the slow-slope crossing produces a much wider Δt. Controlling slope near the threshold is a direct lever on trigger stability.

Timing budget for high-speed chains: trigger, latch, and sampling windows

A robust high-speed chain treats comparator delay as a budgeted uncertainty, not a single number. The budget must include worst-case propagation delay, timing jitter at the crossing, channel/route skew, and the downstream capture window.

Generic timing template (structure)
tPD(max) + Δt(jitter) + skew + downstream capture window + margin

The same definitions must be used across all terms: VOD test condition, output threshold definition, load (CL/RL), and the intended VDD/temperature corners.

Worst-case rules (non-negotiable)
  • Use max + explicit margin for guarantees; typical is only for early exploration.
  • Align conditions (VOD, CL/RL, VDD, Temp, threshold definition) before comparing parts.
  • Budget the small-VOD region if the real waveform crosses slowly or near zero overdrive.
What fails first when overdrive is insufficient
  • Trigger time shifts early/late and becomes less repeatable.
  • Gate/capture windows miss edges or capture the wrong cycle.
  • Sampling alignment drifts, degrading timing-based measurements.
Swimlane timing budget: crossing, comparator output, capture window A three-lane timeline: input crossing event, comparator output edge with jitter spread, and a downstream latch/FPGA capture window. Boxes mark tPD, jitter, skew, and margin. Timing budget swimlane: tPD + jitter + skew + window + margin VIN crossing Comparator output Latch / capture window time cross Δt tPD (max) skew capture window margin margin Budget the worst case: align definitions, then add explicit margin around the capture window.

The crossing event is followed by a delayed output edge with timing spread. A downstream latch/window must be placed with enough explicit margin to cover worst-case delay, jitter, and skew under matched test conditions.

Translate your input waveform into effective overdrive

Datasheet curves are plotted versus overdrive, but real systems provide a waveform that only slowly builds overdrive after the threshold crossing. Converting the local slope near VTH into VOD(t) is the fastest way to avoid paper-only selection and to predict worst-case delay and jitter in a real chain.

3-step recipe (waveform → VOD(t) → required overdrive)
  1. Local linear view near VTH: approximate the crossing region as VIN(t) ≈ VTH + S·t, where S = dVIN/dt at the crossing.
  2. Convert to overdrive growth: VOD(t) = VIN(t) − VTH ≈ S·t (use the differential form for VIN+−VIN− when applicable).
  3. Enforce the datasheet condition at tPD: if a worst-case delay is specified at VOD = VODreq, then the system must reach VOD(tPD(max)) ≥ VODreq under matched load and corners.
Minimal fields to record (so comparisons stay honest)
  • S = dVIN/dt at the threshold crossing (measured or simulated).
  • VTH and the intended crossing region (where the decision happens).
  • tPD(max) target for the budget, at matched VDD/Temp corners.
  • VODreq used for the datasheet guarantee (or the relevant curve region).
  • Load and threshold definition used for the delay measurement (CL/RL and output level).
Common causes that shrink effective overdrive growth
  • RC filtering: flattens the slope near VTH.
  • High source impedance: adds extra RC with input capacitance.
  • Series limiting resistor: slows the crossing under capacitive loading.
  • ESD / clamp structures: can reshape transients (especially with large excursions).
  • Cable capacitance: increases effective C and reduces dVIN/dt at the decision point.
Waveform to effective overdrive mapping Left shows an input ramp crossing a threshold. Right shows overdrive VOD(t) growing from zero after the crossing. A marker at tPD indicates the required overdrive VODreq that must be reached for a datasheet condition. Map VIN(t) near VTH into VOD(t), then check VOD at tPD(max) VIN(t) VOD(t) time VIN VTH local S t = 0 VIN → VOD time VOD 0 tPD(max) VODreq need VOD at tPD Evaluate VOD growth after crossing; verify VOD(tPD(max)) meets the curve/guarantee condition.

Convert the local crossing slope into overdrive growth. If the worst-case delay guarantee is tied to a specific overdrive, the waveform must reach that overdrive by the time the budgeted tPD is expected.

Design knobs to increase overdrive (without breaking other specs)

When the effective overdrive at the decision point is insufficient, the fix is rarely “pick a faster number.” The fastest path is choosing the right lever that increases VOD or dVIN/dt near the crossing while respecting the system’s limits.

Four levers (speed-focused view)
Gain / edge shaping
Raises effective signal swing and slope at VTH to reduce delay spread.
Threshold placement
Moves the decision point to a region of the waveform with higher dVIN/dt.
Source & RC control
Reduces slope loss from series-R and effective input capacitance near the crossing.
Comparator class
Match speed needs to family behavior (HS/latched vs low-power) using aligned conditions.
How to choose the lever (decision hints)
  • If dVIN/dt is slow at VTH, start with RC control and threshold placement.
  • If small-signal swing is the bottleneck, consider gain / edge shaping.
  • If worst-case tPD or jitter budget is still not met, reconsider comparator class under matched conditions.
Design lever map: gain, threshold, RC, comparator class A lever map with four cards labeled Gain, Threshold, RC, and Comparator class, each pointing to outcomes: VOD up, dV/dt up, tPD down, jitter down. Lever map: raise effective VOD and slope at crossing to reduce delay and jitter Gain AMP Threshold RC R C Comparator class LP HS Outcomes VOD ↑ dV/dt ↑ tPD ↓ jitter ↓ Pick the lever that increases effective overdrive or slope at the decision point, then re-check the worst-case budget.

The four levers map to the same outcomes: higher effective overdrive and steeper crossing slope reduce delay and jitter, but each lever must be validated under the same corner and load assumptions used for the timing budget.

Edge cases & traps: zero-cross, slow ramps, and multi-toggling

Most “comparator is slow” failures are waveform problems that force operation near zero overdrive, near-zero slope, or repeated threshold crossings. The goal here is fast diagnosis: use clear criteria on the bench and apply the minimum corrective action without drifting into unrelated design topics.

A) Zero-cross / tiny signal (VOD ≈ 0)
Criterion
The input stays near VTH for a long time and the output edge time spreads wide across repeated shots.
Action
Move the decision point to a steeper region or increase the local slope; then re-check worst-case timing budget.
B) Slow ramps (dVIN/dt too small)
Criterion
Small noise causes large timing drift at VTH; changing RC or probe loading changes jitter significantly.
Action
Reduce slope loss from source impedance and input networks; validate with a faster stimulus to separate device limits from waveform limits.
C) Ringing / noise (multiple crossings)
Criterion
VIN crosses VTH more than once and the output shows extra toggles or short pulses aligned to those crossings.
Action
Confirm the multi-crossing is real (not probe/trigger artifact), then apply damping/filtering/hysteresis strategies in the dedicated pages.
Three traps: zero-cross, slow ramp, ringing multi-cross A three-panel diagram. Each panel shows VIN versus time with a fixed VTH line. The zero-cross case stays near VTH, the slow ramp case has shallow slope causing large time spread, and the ringing case crosses VTH multiple times. Three common traps that explode delay/jitter Zero-cross Slow ramp Ringing VTH VTH VTH VOD ≈ 0 → spread ↑ Δt large multi-cross Diagnose at the waveform: VOD≈0, dV/dt≈0, or repeated crossings are the usual root causes.

These three cases are visible directly on the input waveform: lingering near VTH, shallow slope at VTH, or repeated crossings. Confirm the mechanism first, then apply the smallest correction and re-run the worst-case timing budget.

Verification: how to measure tPD–VOD and jitter on the bench

Credible verification aligns definitions and conditions: the same threshold definition, the same load, and the same supply/temperature corners. A useful bench method sweeps overdrive with a controlled fast edge, then measures the distribution of output timing rather than a single “pretty” capture.

Recommended measurement method
  • Fast edge + controlled amplitude: sweep input amplitude or bias to cover the relevant overdrive region.
  • Measure time distribution: collect repeated timestamps and build a histogram for jitter and spread.
  • Keep conditions fixed: VDD, temperature, load (CL/RL), and threshold definition must not drift during the sweep.
Bench traps that create “fake” delay/jitter
  • Probe capacitance: changes dVIN/dt and can dominate the result at small overdrive.
  • Trigger artifacts: trigger source/bandwidth can introduce apparent timing noise.
  • Output load/reflection: cabling and termination can shift the output threshold crossing time.
Minimal measurement fields (record every sweep)
  • VDD, temperature point, and any supply/thermal limits used.
  • Input stimulus: amplitude, slew, source impedance, and any series/RC elements present.
  • Output load: CL/RL, cable length/termination, and intended logic threshold.
  • Threshold definition: VIN crossing reference and VOUT measurement level (e.g., 50%).
  • Acquisition setup: instrument, bandwidth limit, and sampling settings used.
Bench setup for tPD-VOD sweep and jitter histogram A block diagram showing Pulse Generator, Attenuator/Bias, Comparator DUT, Output Load, and Scope or Time Interval Analyzer. Badges label VOD, slew, CL, and threshold definition. Bench architecture: sweep VOD, hold conditions constant, measure time distribution Pulse gen Atten / bias VOD control Comparator VIN VTH Load CL / cable Scope / Time interval analyzer tPD + histogram slew VOD CL threshold definition Sweep VOD with a controlled edge, keep load and definitions fixed, then quantify timing spread with repeated captures.

The key to a trustworthy tPD–VOD result is holding all conditions constant while sweeping overdrive and collecting enough repeated events to describe the timing distribution. Record the minimal fields so bench results remain comparable across parts and revisions.

Engineering checklist: make worst-case delay predictable

Worst-case propagation delay is predictable when the decision-point waveform is defined, translated into effective overdrive growth, aligned to datasheet conditions, and verified with repeated timing statistics (not a single capture). The checklist below is prioritized so fixes start at the real bottlenecks.

P0

Must-pass: align worst-case conditions (VOD, dV/dt, definition, corners)

P0-1 · Minimum effective overdrive at the decision time
Check: At t = tPD(max), does the waveform reach the overdrive region used by the guarantee/curve condition?
Pass: VOD(tPD(max)) meets or exceeds the required overdrive condition with margin under worst-case corners.
Action: Recompute VOD(t) at the crossing region using the worst-case waveform; if insufficient, apply a speed lever (gain/threshold/RC/class) and re-check.
P0-2 · Minimum slope at crossing (jitter control)
Check: Is dVIN/dt at VTH large enough that timing spread does not dominate the capture window?
Pass: Measured/simulated crossing slope supports the timing budget with margin (distribution is stable across repeated shots).
Action: Identify whether slope loss is caused by RC/source impedance/probe loading; then improve slope and re-run the histogram.
P0-3 · Delay definition alignment (input and output thresholds)
Check: Are the input crossing reference and output decision level aligned with the datasheet definition (e.g., VOUT@50% or a logic threshold)?
Pass: Delay numbers are compared only under identical threshold definitions (otherwise comparisons are invalid).
Action: Record threshold definition explicitly in the bench log; lock trigger/reference choices before sweeping VOD.
P0-4 · Output load alignment (CL, cable, termination)
Check: Is the output load (CL/RL/cable/termination) fixed and representative of the real receiver?
Pass: Changing cable/termination does not create “new” delay/jitter artifacts (no reflection-driven crossing shifts).
Action: Fix the load and cable as part of the test setup; avoid ad-hoc wiring during VOD sweeps.
P0-5 · Worst-case corners (VDD/Temp) and budgeting rule
Check: Does the budget use tPD(max) under the relevant VDD/Temp range (not typical-only)?
Pass: Worst-case timing is built from max + margin; typical is used only for early estimates.
Action: Lock the corner set (VDD/Temp/load/definition) and re-verify distributions at that corner.
P1

System knobs: restore effective VOD(t) and slope (RC, source impedance, protection effects)

P1-1 · Input RC flattening near VTH
Check: Does the filter/network reduce local slope at VTH compared with the expected stimulus?
Pass: A/B comparison shows the delay/jitter distribution is not dominated by network-induced slope loss.
Action: Make RC elements switchable for bench isolation; re-run the same VOD sweep with conditions unchanged.
P1-2 · Source impedance / series limiting resistors
Check: Is effective Rsource (including protection resistors) forming extra RC with input capacitance and wiring?
Pass: Changing drive strength or series resistance does not dramatically change crossing slope and timing spread.
Action: Reduce effective source impedance or re-shape the edge; then re-check VOD(tPD(max)).
P1-3 · ESD/clamps/protection reshaping transients
Check: Is the waveform near VTH distorted by clamp engagement, recovery, or cable-driven excursions?
Pass: Limiting excursions or changing protection path does not create a new “linger near VTH” behavior.
Action: Treat protection as part of the timing path; verify the decision-point waveform after protection, not before it.
P2

Board-level contributors (brief): coupling and ground-bounce that inflate dispersion

P2-1 · Output-to-input coupling (parasitic feedback)
Check: Does VIN show a synchronous “kick” when VOUT switches (measured close to the pin)?
Pass: The kick is small enough that it does not create extra crossings or widen the histogram.
Action: Reduce coupling loops and keep input nodes isolated from fast output routing (detail lives in layout guidance pages).
P2-2 · Ground bounce / supply transient shifting the effective threshold
Check: Does the reference/ground move at the switching instant and correlate with delay spread?
Pass: Improving return paths/decoupling tightens the distribution rather than changing only the average delay.
Action: Treat supply/ground stability as a timing condition; record it and verify at the same corner used for budgeting.
Checklist flow: define waveform, map to VOD(t), choose class, verify histogram A left-to-right flow with four boxes: Define worst-case waveform, Map to VOD(t), Choose comparator class, Verify histogram. Three arrows connect the boxes. Checklist flow to make worst-case delay predictable P0 P0 P1 P2 Define worst-case waveform min VOD / min dV/dt Map to VOD(t) check at tPD(max) Choose comparator class HS / latched / LP Verify histogram tPD + jitter Lock definitions and loads before sweeping VOD; quantify distributions, not single captures.

Reference part numbers (starting points only)

These part numbers are provided to speed up datasheet lookup and bench replication. Selection should follow the checklist above (aligned conditions and worst-case verification), not a single typical delay number.

High-speed / low-dispersion families (bench-friendly for VOD sweeps)
  • Analog Devices: ADCMP600 / ADCMP601 / ADCMP602
  • Texas Instruments: TLV3501 / TLV3502 (and automotive variants where applicable)
Low-power / slow-edge reality checks (shows why “small VOD → slower” matters)
  • Texas Instruments: TLV3691 (ultra-low power comparator class)
  • Analog Devices (LT): LT6700 / LT6700HV (reference-comparator style devices)
How to use these references (fast workflow)
  1. Find the tPD vs overdrive information and note the exact test conditions (VDD/Temp/CL/definition).
  2. Translate the real waveform into VOD(t) and verify VOD at tPD(max).
  3. Replicate conditions on the bench and capture a timing histogram to quantify spread.

Request a Quote

Accepted Formats

pdf, csv, xls, xlsx, zip

Attachment

Drag & drop files here or use the button below.

FAQs: propagation delay vs overdrive (tPD–VOD), slew, jitter, and verification

Short, actionable answers for timing predictability. Each FAQ uses the same data structure: Symptom / Check / Threshold / Action. No cross-topic expansion.

Why does my measured propagation delay vary a lot at small signals?
Symptom
Delay spread is wide and the output edge time “walks around” from shot to shot when the input is near the threshold.
Check
  • Inspect the crossing region: does VIN linger near VTH (VOD ≈ 0 for “too long”)?
  • Look for repeated crossings (ringing/noise) that can re-trigger.
  • Swap probe/ground method: does the histogram change (probe loading)?
Threshold
If the input spends a noticeable fraction of the expected delay window within a very small band around VTH, delay dispersion will dominate and “typical tPD” stops being predictive.
Action
  1. Lock delay definitions and output load first (same VTH/VOUT threshold, same CL/cable).
  2. Increase crossing slope or effective VOD growth near VTH (remove slope-killing RC/source resistance first).
  3. Re-verify using a timing histogram at the worst-case corner, then budget with max + margin.
Which datasheet conditions make tPD numbers incomparable across parts?
Symptom
Two parts show similar “prop delay” on paper, but behave very differently in the same circuit.
Check
  • Overdrive point used (fixed VOD? curve? “typ only”?)
  • Input stimulus (edge rate, step size, source impedance)
  • Output load (CL/RL/cable/termination) and output swing
  • Delay definition (VIN crossing reference; VOUT level such as 50% or logic threshold)
  • VDD and temperature corner
Threshold
If any one of these conditions differs, tPD numbers should be treated as non-comparable and must be re-validated under aligned conditions.
Action
  1. Build a one-page “condition alignment” table before comparing parts.
  2. Budget using tPD(max) only after matching VOD, definitions, load, and corners.
  3. Confirm with the same lab setup and a histogram at the worst-case point.
How do I estimate jitter from input noise and slew rate?
Symptom
Timing uncertainty grows as edges get slower or the environment gets noisier, even when average delay looks fine.
Check
  • Measure or simulate dVIN/dt exactly at the crossing region.
  • Estimate the effective noise/perturbation at VIN near VTH (including pickup and measurement loading).
  • Confirm the result is stable across trigger/probe changes (avoid “instrument jitter”).
Threshold
As a first-order rule, timing spread scales roughly in proportion to noise amplitude and inversely with crossing slew rate (σt ≈ σv / (dVIN/dt)).
Action
  1. Increase crossing slew rate first (remove slope-killing RC and reduce effective source impedance).
  2. Reduce noise coupled into the crossing region (shorten loops, clean reference/ground, stabilize loads).
  3. Budget timing using distribution metrics (spread) at the worst-case waveform and corner.
What overdrive should be used when budgeting worst-case timing?
Symptom
Timing budgets pass on paper but fail in the field because the real waveform provides less effective overdrive than assumed.
Check
  • Translate the real crossing waveform into VOD(t) (effective overdrive growth).
  • Read VOD at the relevant decision time (near the worst-case delay window).
  • Confirm the datasheet tPD(max) condition uses the same VOD definition and load.
Threshold
Worst-case timing should use the minimum effective VOD delivered by the worst-case waveform and corner, not a “typical” input amplitude number.
Action
  1. Budget with tPD(max) + margin at the smallest effective VOD region.
  2. If the waveform operates near zero-cross / tiny signal, include distribution spread (jitter/dispersion) explicitly.
  3. Verify with a histogram under the same VDD/Temp/load/definition used in the budget.
Why is tPLH different from tPHL and when does it matter?
Symptom
Rising-edge timing meets spec but falling-edge timing fails (or the opposite), even with the same input.
Check
  • Measure both tPLH and tPHL with identical definitions and load.
  • Confirm output swing and threshold level are the same for both directions.
  • Check if the output structure (push-pull vs open-drain) makes one direction load-dependent.
Threshold
If the system captures only one edge (or has a tighter window on one edge), the slower edge must dominate the budget regardless of average delay.
Action
  1. Budget using the slower direction (max + margin) for the edge the system actually uses.
  2. Align output load and termination to the real receiver.
  3. Re-verify with a histogram for that edge at the smallest effective overdrive condition.
My delay is fine on the bench but worse on the PCB—what changed first?
Symptom
Bench measurements look good, but in-circuit timing is slower and/or has larger spread.
Check
  • Compare effective VOD(t) and slope at VTH: source impedance, RC, cabling, protection parts.
  • Confirm output load on PCB matches bench load (CL/termination/receiver threshold).
  • Probe close to the pins and change probe grounding: does the story change (coupling/ground bounce)?
Threshold
If small changes in probing, cabling, or termination significantly change delay/spread, the measurement chain or board coupling is dominating.
Action
  1. Recreate PCB load and definitions on the bench before comparing results.
  2. Measure the decision-point waveform on PCB (at the input pin) and recompute VOD(t).
  3. Only after waveform and load are aligned, tune networks or choose a faster comparator class.
How does input RC protection affect effective overdrive?
Symptom
Adding series resistance or RC “protection” makes delay longer and increases timing spread, especially for small signals.
Check
  • Measure slope at VTH with RC enabled vs bypassed (same stimulus, same load).
  • Look for a “flattened” crossing region where VOD(t) grows slowly.
  • Confirm the RC does not introduce ringing that creates multiple crossings.
Threshold
If bypassing the RC noticeably tightens the histogram (same VOD setting and definitions), the network is dominating timing uncertainty.
Action
  1. Keep RC elements switchable for bench isolation and regression testing.
  2. Budget using the worst-case slope delivered with protection enabled.
  3. Verify across the smallest effective overdrive and worst corners (VDD/Temp/load).
Can adding hysteresis “speed up” the delay at small overdrive?
Symptom
Multiple output toggles or unstable trigger timing appear near the threshold; adding hysteresis seems to “clean it up.”
Check
  • Verify the real issue is repeated crossings (ringing/noise) rather than slow regeneration alone.
  • Check whether the system can tolerate a shifted effective switching threshold (accuracy impact).
  • Confirm the load and threshold definition remain unchanged when comparing results.
Threshold
Hysteresis most reliably improves timing stability when the dominant failure is multi-crossing/false toggling; it is not a universal cure for small-VOD regeneration delay.
Action
  1. Use hysteresis to eliminate repeated crossings, then re-check delay distribution at the worst-case waveform.
  2. Record the new effective switching point and re-run the system timing budget.
  3. If small-VOD dispersion remains large, improve slew/effective VOD growth instead of increasing hysteresis further.
What’s the quickest way to sweep tPD vs VOD in the lab?
Symptom
Sweeps are slow and results are not reproducible across days, scopes, or setups.
Check
  • Use a fast, clean edge and vary amplitude/bias to sweep effective overdrive.
  • Keep VDD/Temp/output load/threshold definition fixed throughout the sweep.
  • Collect repeated timestamps per point (histogram), not single screenshots.
Threshold
If changing scope triggering/bandwidth or cabling changes the result more than changing VOD, the sweep is not condition-locked and must be fixed first.
Action
  1. Fix the load and threshold definition; lock trigger source and bandwidth settings.
  2. Sweep VOD using amplitude/bias changes while keeping edge rate constant.
  3. Report each point as a pair: (tPD metric) + (spread metric) at the worst-case corner.
Why does the delay depend on output load and output swing?
Symptom
Delay changes when cable length, termination, or load capacitance changes—even though the input looks the same.
Check
  • Repeat the same VOD point while varying CL/cable/termination and log the shift.
  • Confirm the output measurement level (50% or logic threshold) is consistent.
  • Check for reflection-induced ringing near the output threshold.
Threshold
If load changes shift the output threshold crossing time measurably, the output stage plus load must be treated as part of the timing path.
Action
  1. Budget delay using the real receiver load and swing, not a lab-only setup.
  2. Fix cable/termination during sweeps and record them in the measurement fields.
  3. If the system needs tight timing, choose output types and loads that minimize threshold-crossing variability.
How to avoid false jitter introduced by oscilloscope triggering?
Symptom
Jitter numbers change a lot when trigger source, trigger level, or bandwidth settings are changed.
Check
  • Repeat the same point with different trigger sources (input reference vs output) and compare histograms.
  • Toggle bandwidth limits/noise filters and observe whether spread changes more than it should.
  • Verify probe grounding and noise pickup are controlled (short ground, minimal loop).
Threshold
If the histogram width shifts materially when only triggering/bandwidth settings change, the measurement chain is contributing “fake jitter.”
Action
  1. Lock a single trigger definition and keep it unchanged during sweeps.
  2. Minimize probe loading and pickup (short ground, consistent probing point near pins).
  3. Validate jitter with a second method (time-interval style measurement) before final budgeting.
For zero-cross detection, how to keep timing stable near 0VOD?
Symptom
Timing becomes unstable near the crossing point: edges drift and the distribution widens around 0VOD.
Check
  • Confirm the crossing region is not a slow ramp (small dV/dt) or a ringing region (multi-cross).
  • Verify the decision point is measured close to the input pin (avoid cable/probe artifacts).
  • Check the output load/threshold definition is fixed across measurements.
Threshold
Near 0VOD, delay dispersion and noise-to-time conversion dominate; stability requires controlling slope and avoiding repeated crossings more than chasing a single “fast tPD” number.
Action
  1. Increase crossing slew rate at the decision point (remove slope killers first: RC/source resistance/probe loading).
  2. Eliminate multi-crossing mechanisms (damping or hysteresis trade-offs) and then re-check the histogram.
  3. Budget timing using max + margin plus distribution spread at the worst-case waveform and corner.
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.