Photodiode/PMT Front-End: Ultra-Low-Noise TIA & Auto-Range
← Back to:Operational Amplifiers (Op Amps)
A photodiode/PMT front-end succeeds when noise, bandwidth, stability, and dynamic range are designed as one budget. This page turns those trade-offs into measurable rules—so the TIA stays stable, the dark floor stays low, and auto-range/calibration stays predictable.
What this page solves (Photodiode/PMT front-end scope)
This page focuses on ultra-low-noise photodiode and PMT analog front-ends: where femto/pico currents become a stable, measurable voltage or code without losing bandwidth, stability, or dynamic range. The goal is to make noise, drift, and overload behavior predictable by mapping each failure symptom to a small set of controllable design variables.
Symptom map (what is actually failing)
- Noise worse than expected → Rf thermal noise, op-amp en/in, total input capacitance (C_total), leakage paths, and shot-noise floors.
- Rise time too slow / bandwidth short → C_total, GBW/slew limits, compensation (Cf), and any range-shaping filters.
- Ringing or oscillation → phase margin collapse caused by C_total changes (diode, package, PCB, protection, cable, probing).
- Dark reading drifts or never settles → humidity/contamination leakage, bias network drift, and improper dark-current trimming windows.
- Saturation at bright levels → insufficient range planning (Rf/gain/time-constant), slow overload recovery, and missing auto-range hysteresis.
- PMT HV coupling spurs → unintended return paths, shielding gaps, and common-mode injection from high-voltage supply structures.
What this page delivers (engineering outputs)
- Noise budget you can close → identify the dominant noise term and express it in a consistent domain (input current noise or output voltage noise) over a defined bandwidth.
- Stability/compensation playbook → make the front-end stable across diode/cable/probe capacitance variation without destroying bandwidth.
- Dark-current trim patterns → remove DC/slow drift using safe trim mechanisms that do not inject extra noise or destabilize the TIA.
- Auto-range blueprint → prevent saturation while controlling switching charge, settling time, and range “hunting”.
- Verification checklist → test hooks to validate integrated noise, impulse/step response, overload recovery, and temperature/humidity sensitivity.
Scope guardrail (to avoid cross-page overlap)
- This page does not teach general ADC-driver/FDA filter design; it only states interface requirements (range, common-mode, settling expectations).
- This page does not cover non-optical TIA applications; examples and constraints stay within photodiode/PMT front-ends.
- This page does not act as a generic op-amp encyclopedia; only the traits that directly move TIA noise, stability, drift, and overload behavior are used.
The main reason photodiode and PMT front-ends fail is not a single “bad op-amp part number”, but an unowned constraint: total input capacitance (diode + package + PCB + protection + cable), leakage/contamination paths, and overload recovery. The following sections turn these into owned variables with measurable pass/fail checks.
System architecture options (where the noise and range are won)
Architecture selection should start from constraints, not from a device list. For photodiodes and PMTs, the “shape” of the input (capacitance, leakage, and current span) determines which front-end path can meet the noise and bandwidth targets while staying stable across real-world variation.
Input constraints that drive the architecture
- Photodiode PV vs PC mode: PV favors low bias and low leakage sensitivity; PC (reverse bias) reduces effective capacitance for speed but increases leakage and protection demands.
- Total input capacitance (C_total): diode + package + PCB + protection + cable + probing. Stability must hold across this range, not just at a nominal value.
- Current span: if the expected signal spans decades, range planning must be owned (range steps, hysteresis, and recovery budget) before component selection.
- PMT high-voltage coupling: treat HV as a coupling aggressor; shielding and return paths must be designed so that CM injection does not look like signal.
Practical decision rules (pick a route in minutes)
- If dark accuracy matters, leakage control and a safe dark-current trim strategy come first; no amount of filtering fixes drift from contamination paths.
- If bandwidth is high and C_total is large, prioritize compensation robustness (stable step response across C variation) before chasing the lowest spot-noise numbers.
- If the input spans decades, define range steps, hysteresis, and overload recovery limits before choosing Rf and gain, otherwise range hunting and dead time will dominate.
What “winning” looks like for each route
Route A — Low-BW precision (dark accuracy first)
- Define an explicit “dark” operating window and treat leakage/humidity as first-class requirements.
- Own the DC error stack: leakage paths, bias-network drift, and trim method noise injection.
- Noise is evaluated as integrated noise over the measurement bandwidth (not a single spot-noise number).
Route B — Wide-BW fast (capacitance and phase margin first)
- Stability is proven across the full expected C_total range (diode + cable + probing), not at a nominal bench setup.
- Compensation choices are validated by step response, overshoot limits, and repeatable settling under real sensor loading.
- Bandwidth targets are enforced with controlled shaping rather than “hoping the op-amp is fast enough”.
Route C — Ultra-wide range (range plan and recovery first)
- Range steps, hysteresis, and holdoff time are defined from measurement throughput needs and overload recovery limits.
- Switching transients are treated as part of the signal chain (charge injection, feedthrough, and post-switch settling).
- Cross-range consistency is verified so that the same optical level produces consistent codes across adjacent ranges.
With the architecture route selected, the rest of the design becomes a controlled closure exercise: own C_total, leakage, and overload recovery, then close the noise budget and stability margins with measurable checks. The next sections drill into device modeling, noise budgeting, compensation, dark-current trim, and auto-ranging in detail.
Photodiode & PMT device model (the constraints you cannot escape)
Front-end performance is set by a small set of unavoidable device terms: current span, dark/leakage behavior, and the total capacitance that lands on the inverting node. A stable, low-noise design starts by turning those terms into explicit design inputs rather than treating them as “unknown parasitics”.
Photodiode model (current source + capacitance + leakage)
- Signal current (Iph) sets the usable transimpedance gain and the no-saturation requirement.
- Junction capacitance (Cj) is part of the inverting-node load; it directly shifts stability and bandwidth limits.
- Shunt/leakage paths (Rsh + board leakage) set dark offset and drift sensitivity, especially under humidity and contamination.
- Shot-noise floor rises with current; filtering can shape bandwidth but cannot remove the underlying shot-noise physics.
PMT model (current output + cable capacitance + HV coupling)
- Output current behaves like a source feeding the TIA input; dynamic range planning is driven by expected peak events.
- Cable capacitance (Cable C) is usually dominant and variable; stability must hold across cable length and routing changes.
- High-voltage structures act as coupling aggressors; return paths, shielding, and separation control whether HV ripple looks like signal.
Minimum design inputs (treat these as required fields)
Current span
I_min / I_max (photocurrent or PMT current). Range planning and Rf selection must satisfy no-saturation at I_max.
Dark/leakage
Idark (device) + leakage (board). Dark stability is limited by leakage control before any trim algorithm is applied.
Total input capacitance
C_total = Cj + C_pkg + C_pcb + C_protection + C_cable + C_probe. Compensation must be robust across the full expected C_total range.
Bandwidth / time window
Target bandwidth (or integration time / sampling window). Noise is evaluated by integrating over this defined bandwidth.
The most common root cause behind “it only oscillates with certain diodes/cables” is an unbounded C_total assumption. Treat the full C_total range and the leakage environment as requirements, then select transimpedance gain and compensation to remain stable across those variations.
TIA topology & component selection (Rf, Cf, op-amp traits)
Component selection becomes repeatable when each element has an owner: range prevents saturation, stability guarantees controlled settling across C_total variation, and noise closes the budget over the defined bandwidth. The rules below keep these owners explicit instead of relying on “best-effort tuning”.
Rf selection (range first, then noise)
- Start from no-saturation: choose Rf so that I_max does not drive the output into its swing limit (including headroom).
- Range is not optional: if bright events saturate the output, overload recovery time becomes the real throughput limit.
- Noise trade: larger Rf increases signal gain but also sets a resistor thermal-noise term; noise should be judged as integrated noise over the target bandwidth.
Cf selection (stability and bandwidth shaping)
- C_total drives the problem: the inverting node sees diode + PCB + protection + cable capacitance, so compensation must hold across that variation.
- Cf is a stability tool: it shapes noise gain and prevents phase-margin collapse; choose the smallest Cf that produces a controlled step response across the expected C_total range.
- Verification hook: evaluate overshoot and settling with diode/cable variations and realistic probing; a stable design should not “change personality” when the cable changes.
Op-amp traits that matter in a TIA (mapping to outcomes)
Noise
en/in determine whether voltage noise or current noise dominates; bias/leakage sets the low-frequency dark floor.
Stability
Cin adds directly to C_total; GBW and phase behavior determine whether Cf can stabilize the design without unacceptable bandwidth loss.
Range & recovery
output swing/headroom and overload recovery set the real usable range; slow recovery turns brief saturation into long dead time.
A reliable selection workflow is: lock range (no-saturation at I_max), then lock stability across the full C_total range, then close the noise budget over the target bandwidth. This order prevents chasing low spot-noise numbers while the real limiter is overload recovery or capacitance-driven instability.
Noise modeling & budgeting (make the target measurable)
A noise target is only useful when it can be computed and verified. The most reliable workflow is to express every contributor in a single domain (input-referred current noise or output-referred voltage noise), then integrate over a defined bandwidth using ENBW so the prediction can be checked against measurements.
Noise sources to include (keep one consistent domain)
Rf thermal noise
A controllable floor term. It converts cleanly to input-referred current noise and often dominates at low photocurrent and low bandwidth.
Op-amp voltage noise (en)
en becomes input-referred current noise through the noise-gain path. When C_total and bandwidth push noise gain up, en can become the dominant term.
Op-amp current noise (in) + bias/leakage
in acts like an extra input current source. It becomes more painful as Rf and source impedance rise, and it is closely tied to dark stability requirements.
Shot noise (signal and dark)
Photocurrent shot noise rises with current; dark-current shot noise sets a hard floor for dark operation. Filtering shapes bandwidth but does not remove shot-noise physics.
Backend gain + ADC/DAQ noise (interface-only)
Treat the backend as a single equivalent noise term at the TIA output (or fold it back to input-referred). No ADC-driver details are required here—only the interface budget.
Budget workflow (compute → integrate → verify)
- Pick a reference domain: input-referred current noise is often the most intuitive for TIAs.
- Define bandwidth: use the actual measurement bandwidth or the effective time window (integration time implies an effective bandwidth).
- Use ENBW: convert “filter shape” into an equivalent noise bandwidth so integrated RMS noise can be compared across designs.
- Normalize each source: fold voltage-noise terms and backend noise into the same domain (input current or output voltage).
- RSS and close the loop: square-sum the contributors, then compare the predicted integrated RMS noise to the measured integrated noise over the same ENBW.
Key conclusions that drive design choices
- C_total + bandwidth decide whether en is fatal: higher noise gain at higher frequency can make op-amp voltage noise dominate.
- Rf sets a controllable floor: it defines transimpedance and a thermal-noise term, but it must still satisfy no-saturation and recovery constraints.
- ENBW decides integrated noise: spot-noise numbers are not enough; integrated RMS noise over ENBW is what matches measurements.
The fastest way to identify the dominant term is to deliberately change only one lever at a time: bandwidth (ENBW), Rf, or C_total. The measured integrated noise should shift in the same direction as the predicted budget; otherwise a missing term (leakage, coupling, or recovery behavior) is likely present.
Stability & compensation (how to stop oscillation without killing BW)
A TIA does not become unstable because “the op-amp is fast”; it becomes unstable because the inverting node capacitance pushes noise gain up while the open-loop phase at the crossover is no longer safe. The cure is to own the full C_total range and apply compensation that produces a controlled step response across diode, cable, and probing variations.
Root cause to keep in mind
- C_total is the destabilizer: Cj, package/PCB, protection capacitance, cable, and probe capacitance all land on the inverting node.
- Loop crossover is where failures happen: the loop crosses unity gain with insufficient phase margin, producing ringing or sustained oscillation.
- “It changes when the cable moves” is a C_total and coupling problem, not a mystery behavior.
Compensation toolbox (prioritized)
- Cf (primary): stabilizes the loop by shaping noise gain; choose the minimum Cf that yields a controlled step response across the expected C_total range.
- Input isolation (use cautiously): small series resistance may help in extreme capacitance scenarios, but it adds noise and can degrade dark accuracy.
- Output isolation: prevents oscillation driven by capacitive loads at the output interface (wiring, sampling capacitance); keep it as a load-handling tool.
- Bandwidth defining RC: explicitly define the system bandwidth to make both stability and ENBW-based noise budgeting more predictable.
- Staged gain: avoid forcing one TIA stage to meet both extreme gain and extreme bandwidth when the physics fights back.
How to prove stability (repeatable checks)
- Step response first: overshoot, ringing duration, and settling time are the most practical stability evidence.
- Sweep C_total: validate with different diodes, added input capacitance, different cable lengths, and realistic probing.
- Look for “personality changes”: a robust design does not change from stable to unstable with small capacitance or fixture changes.
Stable TIAs are validated by behavior, not by claims: measure step response and settling, then repeat across diode/cable/probe variations to confirm robustness. Choose Cf as a controlled trade between stability and bandwidth so the final ENBW used in the noise budget remains aligned with real hardware behavior.
Biasing, protection, and leakage control (dark current starts here)
Dark accuracy is decided at the input node long before any calibration is applied. Bias mode, protection parts, and board leakage determine the true dark offset, the stability margin through C_total, and whether “dark” remains repeatable across humidity, handling, and time.
Bias mode (PV vs PC) and reverse-bias tradeoffs
- PV mode (zero bias): often simpler and can be more forgiving to leakage, but Cj can be larger, pushing compensation harder for bandwidth.
- PC mode (reverse bias): reduces junction capacitance for speed and bandwidth, but can raise dark/leakage sensitivity and increases the impact of protection capacitance.
- Reverse-bias is a system input: define the operating bias point and the expected Cj(Vbias) range, then validate stability and noise across that range.
Protection and clamps (protection is not free)
- TVS and clamp diodes: add parasitic capacitance (Cpar) into C_total, which can reduce phase margin and force larger Cf or lower bandwidth.
- Leakage side effects: protection devices can introduce temperature-dependent leakage that looks like dark current and can drift with environment.
- Series resistance networks: can limit fault current, but may introduce extra noise and additional poles that alter settling and stability.
Any protection part must be budgeted as Cpar + leakage, not as a schematic symbol. If Cpar and leakage are not bounded, calibration becomes fragile.
Leakage control checklist (make dark repeatable)
- Cleanliness: flux residue and contamination create humidity-sensitive leakage paths that change with handling and time.
- High-impedance routing: keep the inverting node small, short, and away from fast-switching nets; avoid geometry that traps moisture.
- Guarding: a guard ring around the high-Z node can shunt leakage away from the summing node when driven to a similar potential.
- Verification hook: compare dark offset and drift before/after cleaning, and across controlled humidity or handling conditions.
If dark offset shifts with humidity, handling, or protection part changes, the primary issue is typically leakage and parasitic capacitance at the summing node. Fixing these at the input prevents calibration from chasing an unstable baseline.
Dark-current trim & calibration (remove DC error without injecting noise)
Dark trim should remove DC offset and slow drift without raising the noise floor or destabilizing the TIA. The safest approach is to separate what can be corrected (repeatable offset terms) from what must be controlled (humidity-driven leakage and coupling).
Trim objective (and the boundary)
- Correct: DC offset and slow drift that stays correlated to temperature and time.
- Do not “correct”: shot noise and wideband noise; calibration cannot remove random noise.
- Do not depend on: humidity-sensitive leakage as a stable term; leakage must be controlled at the board level.
Path A — Analog current injection (Itrim cancellation)
- What it does: inject a small controlled current to cancel dark offset at the input.
- Main risk: injection circuitry can add noise directly in the input domain; injection noise must be budgeted like any other current-noise source.
- Temperature behavior: the cancellation term must track temperature or it becomes a new drift source.
Path B — Digital offset and LUT (dark window + temperature correlation)
- What it does: measure a dark/occluded baseline and store an offset (or LUT vs temperature) for subtraction.
- Requirements: a repeatable dark window, consistent bias state, and temperature coverage aligned with real operating gradients.
- Failure mode: if measurement uncertainty approaches the target noise floor, coefficient updates can inject step artifacts or “coefficient noise”.
Path C — Modulation / synchronous detection (move DC away)
- What it does: shift DC drift into a controlled band and recover it with synchronous processing.
- Tradeoffs: added complexity and potential ripple/spurs that must remain out of the measurement band.
- Best fit: extremely drift-sensitive systems that can tolerate a defined modulation strategy.
Calibration window rules (avoid injecting noise)
- Separate slow drift from random noise: the window must be long enough to estimate DC, but not so long that environment changes contaminate the estimate.
- Update rate must match drift: frequent updates can inject coefficient noise; infrequent updates can miss true drift.
- State consistency: calibrate under the same bias, cable/C_total condition, and thermal state used for measurement.
The most robust trims remove repeatable DC terms while keeping the bandwidth, stability, and ENBW unchanged. Always validate trim with the same bias state and C_total environment used in measurement, and confirm that updates do not introduce step artifacts or extra noise.
Auto-range design (dynamic range without saturating)
Auto-ranging prevents saturation when optical intensity spans decades, but it must not create false pulses or long dead time. A robust design uses hysteresis, debounce, and holdoff to avoid range-chatter, and it validates switching by settling time and range-to-range continuity.
Trigger logic (thresholds, hysteresis, debounce, holdoff)
- Upper/lower thresholds: use V_hi to step down gain before hard clipping, and V_lo to step up gain only after margin returns.
- Hysteresis window: the gap between V_hi and V_lo must exceed the expected noise and switching transient amplitude.
- Debounce: require N consecutive samples (or a minimum time) beyond the threshold before switching.
- Holdoff: after a switch, block new decisions until the analog path has settled (T_holdoff ≥ worst-case T_settle).
- Direction bias: up-range decisions can be more conservative to avoid hunting when signal is near a boundary.
Range switching methods (what changes, and what it breaks)
Switch Rf (transimpedance)
Directly changes gain at the summing node. Must control charge injection and C_total shifts that can alter stability and settling.
Switch post gain (keep node quiet)
Reduces disturbance at the inverting node, but overload recovery and bandwidth/noise changes can create discontinuities across ranges.
Switch integration time (ENBW control)
Changes effective bandwidth and noise integration without touching hardware, but it trades response time for sensitivity and requires ENBW consistency.
Parallel TIAs (multi-range lanes)
Avoids switch dead-time by measuring multiple ranges simultaneously, but requires careful isolation to prevent crosstalk and input loading.
Critical pitfalls (why auto-range “lies”)
- Switching transients: charge injection and digital feedthrough can look like real pulses at the input.
- Summing-node disturbance: changing Rf/Cf or connecting switches at the node can create steps and ringing.
- Overload recovery: saturation creates a dead window; recovery time can be range-dependent and must be measured.
Verification metrics (measurable pass/fail)
- T_settle after switching: time to return within the allowed error band after a range transition.
- Range-to-range continuity: the same stimulus should not jump when represented in adjacent ranges.
- Noise floor continuity: the noise floor should not show large “steps” across ranges unless bandwidth is intentionally changed.
Auto-range quality is defined by what happens at the transition: switching should not inject false pulses, and data should become valid after a measured settling time. Always validate continuity by sweeping a controlled stimulus across boundaries and confirming consistent readings in adjacent ranges.
Layout & EMC for femto/pico signals (guarding, shielding, parasitics)
Femto/pico front-ends fail more often from parasitics and coupling than from wrong equations. The layout goal is to keep the summing node small, clean, and shielded from fast edges and high-voltage dv/dt, while ensuring a controlled return path that does not carry interference through the sensitive reference.
High-impedance summing node rules (short, guarded, isolated)
- Shortest path: place the sensor and feedback network to minimize the -IN node area and trace length.
- Guard ring: surround the high-Z node with a guard structure to intercept leakage and reduce coupling.
- Partition: keep digital edges, switching supplies, and high dv/dt nodes physically away from the summing node region.
Parasitic capacitance control (own C_total)
- Input geometry: pad size, trace width, and proximity to reference planes directly change C_total.
- Protection parts: TVS/clamps and connectors can dominate Cpar; select and place them to minimize node capacitance.
- Cable entry: treat cable and connector capacitance as part of the budget; validate stability for the worst-case configuration.
PMT high-voltage coupling (shielding + return-path control)
- Capacitive coupling: high dv/dt nodes can inject current into the summing node through stray capacitance.
- Isolation features: use controlled spacing, shielding structures, and isolation gaps to reduce direct coupling paths.
- Return path: define a single, predictable return path so shield currents do not flow through the sensitive reference region.
Measurement traps (probe capacitance creates fake failures)
- Probe capacitance: increases C_total and can turn a stable TIA into a ringing or oscillating one.
- Ground clip loops: long ground leads form loops that add ringing and inject coupled noise.
- Better practice: short ground springs, controlled fixtures, or differential probing for stability evaluation.
Layout success is measured by repeatability: stable behavior should remain stable across cables, fixtures, and realistic probing. Treat every added component, connector, and measurement tool as part of the C_total and coupling budget, then validate stability and noise in the worst-case configuration.
Verification & production checklist (engineering checklist + failure analysis)
This section is the reusable “engineering close-out” pack: review checklist, verification hooks, quick triage for failures, and a minimal production log schema. The goal is repeatable dark performance and stable bandwidth/settling across the worst-case input capacitance, cabling, probing, temperature, and humidity.
A) Design review checklist (before committing to a PCB spin)
Sensor + input model
- I-range and duty (min/typ/max), including dark current and expected drift regime.
- C_total budget: diode Cj (bias-dependent) + package + pads + protection + connector/cable worst-case.
- Bias mode documented (PV/PC + Vbias range) and verified in worst-case C_total.
TIA core (Rf/Cf/op-amp traits)
- Rf chosen for output swing headroom + overload recovery margin; resistor technology and tempco reviewed.
- Cf matched to C_total worst-case for stable step response; tolerance impact considered.
- Op-amp selected by en/in vs source-Z, input bias/leakage, Cin, GBW/slew, overload recovery, and output drive.
Protection + leakage
- Protection devices owned as Cpar + leakage in the budget; placement keeps Cpar away from the summing node.
- High-Z cleanliness plan: guarding strategy, keep-out areas, contamination control, humidity sensitivity expectations.
Auto-range strategy (if used)
- Thresholds (V_hi/V_lo), hysteresis window, debounce (N samples), and holdoff time defined.
- Switch method reviewed for transient injection risk (charge injection, feedthrough, node disturbance).
- Settling gate defined (data-valid only after T_settle is met).
Layout + EMC
- Summing node area minimized; guard ring implemented; sensitive routing isolated from fast edges and HV dv/dt.
- Defined return path: shield/connector currents do not traverse the sensitive reference region.
- Measurement plan documented (probe capacitance and ground loop control for stability testing).
B) Verification tests (repeatable evidence, not opinions)
- Dark zero: occluded input baseline under controlled bias state; record mean + drift vs time.
- Noise PSD + integrated noise: measure noise spectrum; integrate over the defined ENBW to compare with the budget.
- Bandwidth & step response: step/impulse stimulus; verify overshoot/ringing and time-to-settle within error band.
- Auto-range transition (if used): switching transient amplitude, T_settle after each transition, and dead-window duration.
- Temperature sweep: multiple temperature points; record offset and gain drift and recovery after thermal soak.
- Humidity/leakage stress: compare dark offset before/after cleaning and under humidity exposure; confirm guarding effectiveness.
A “pass” requires the measured evidence to stay stable when C_total changes (cable/probe), and when environment shifts (temperature/humidity).
C) 10-minute triage (fast fault localization)
Noise got worse
- Check ENBW changes (filter/integration time/range change): noise floor can shift when bandwidth shifts.
- Check C_total growth (cable, connector, protection, probe): in-noise and stability margin can degrade abruptly.
- Check coupling/return paths (shield currents, ground bounce, nearby switching): interference can masquerade as noise.
Oscillation or ringing appeared
- Remove/replace the probe first (C_probe and ground lead loops are common false-failure sources).
- Re-check worst-case C_total and Cf margin (including protection and cable variants).
- Check output loading and isolation (long cables/cap loads can trigger secondary instability).
Drift or dark offset moves
- Leakage first: contamination/humidity effects; verify guard ring and cleanliness controls.
- Thermal gradients: confirm soak time and consistent bias state during “dark” measurement.
- Protection/switch leakage: evaluate temperature dependence of clamp/switch leakage at the input node.
D) Minimal production log schema (enough to close the loop)
- Identity: Serial number, lot/batch, PCB revision, key BOM revision.
- Conditions: Temperature point(s), supply voltage, bias state (PV/PC + Vbias), range state.
- Metrics: Dark offset, drift slope, integrated noise over defined ENBW, bandwidth/step metrics, auto-range T_settle (if used).
- Calibration: Calibration version, coefficient CRC, timestamp (and any LUT temperature table ID).
E) Example parts (material numbers by role)
Part numbers below are common reference points for each role. Verify the latest datasheets for limits, packages, and leakage specifications.
Electrometer / ultra-low-bias front ends
ADA4530-1 · LMP7721
Wideband photodiode TIAs
OPA657 · LTC6268 · LTC6269 · ADA4817-1
GHz-class pulse / fast optical TIAs
OPA855 · OPA858 · LTC6268-10
Auto-range / switching (low charge injection, low leakage)
ADG1219 · ADG1209 · TMUX1136 · TMUX6111
Applications (Photodiode + PMT front-end patterns)
These application patterns help map use-cases to constraints without expanding into system algorithms. Each card provides a practical input current range, bandwidth intent, a typical range method, and example material numbers.
Photodiode — Optical power meter / laser monitor (DC-leaning)
- I-range: pA → µA (wide span with strong dark sensitivity)
- BW: low to mid (drift/leakage often dominate)
- Range method: fixed Rf or slow auto-range with long holdoff
- Example parts: ADA4530-1 · LMP7721 · ADG1219 (for switching)
Photodiode — Fluorescence / weak-light analog readout
- I-range: pA → nA
- BW: mid (balance noise and responsiveness)
- Range method: staged gain + dark offset baseline update
- Example parts: LMP7721 · LTC6268 · TMUX1136
Photodiode — Spectrometer / scanning measurement
- I-range: decade-spanning, often sensor- and optics-dependent
- BW: mid (set by scan speed and settling requirements)
- Range method: auto-range with strict settling gate + continuity checks
- Example parts: OPA657 · LTC6269 · ADG1219
Photodiode — Pulsed / fast optical monitor
- I-range: nA → mA peak (pulse-dependent)
- BW: high (rise time and distortion matter)
- Range method: fixed range preferred; avoid switching artifacts in-band
- Example parts: OPA657 · LTC6268-10 · OPA858
PMT — Photon counting / time-resolved pulses
- I-range: very small pulses with fast edges
- BW: very high (bandwidth and overload recovery define usable rate)
- Range method: fixed range + recovery management; HV coupling must be controlled
- Example parts: OPA855 · OPA858 · LTC6268-10
PMT — Weak-light analog measurement (wide dynamic range)
- I-range: pA → µA (gain-dependent)
- BW: mid to high
- Range method: auto-range with verified settling + continuity; leakage and HV dv/dt coupling must be owned
- Example parts: OPA657 · LMP7721 · TMUX6111
Request a Quote
FAQs (Photodiode/PMT front-end) — short answers + troubleshooting fields
These FAQs close long-tail questions without expanding the main content boundary. Each item includes a short answer (2–4 sentences) plus a structured troubleshooting set: symptom → likely cause → checks → fixes.
